Discussion:
[Freetel-codec2] new FreeDV 2400A VHF Modem
David Rowe
2016-03-15 23:28:50 UTC
Permalink
Hello Lists,

We've just tested the new FreeDV 2400A mode designed for VHF and above.
Here's the scoop:

http://www.rowetel.com/blog/?p=5119

Thanks,

David
Steve
2016-03-16 08:42:06 UTC
Permalink
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this shows how wide
it is.

I noticed that the freedv_tx was outputting +/- 1.0 signals and a couple of
clips going on. Maybe need to knock it down to +/- .5 like the old FDM
modes.

FYI

Steve
glen english
2016-03-16 08:46:15 UTC
Permalink
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Steve
2016-03-16 08:50:45 UTC
Permalink
Just using the script:

freedv_tx 2400A ve9qrp_10s.raw modem.raw

Which produces a RAW audio file sampled at 48 kHz and importing that into
audacity.
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB compression?
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this shows how wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals and a couple
of clips going on. Maybe need to knock it down to +/- .5 like the old FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
glen english
2016-03-16 09:02:47 UTC
Permalink
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Steve
2016-03-16 09:10:18 UTC
Permalink
Ha, I wish it was through a radio!

I'm trying to think what I have laying around here that might make a good
transmitter for it.
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and importing that into
audacity.
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB compression?
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this shows how wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals and a couple
of clips going on. Maybe need to knock it down to +/- .5 like the old FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Adrian Musceac
2016-03-16 20:34:49 UTC
Permalink
Hi David,

Is it possible to get our hands on the Gnuradio code? I've been trying
to optimize modems running at 1200 bits/second and the best I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB above for
GMSK. Your modem seems to be 16-20 dB better than that! If I could see
the Gnuradio graph maybe I could figure out what I'm doing wrong.

Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here that might make a good
transmitter for it.
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and importing that into
audacity.
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB compression?
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this shows how wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals and a couple
of clips going on. Maybe need to knock it down to +/- .5 like the old FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Adrian Musceac
2016-03-16 22:28:36 UTC
Permalink
Also, for comparison purposes, here is my 2000 bit/s modem running
through a Yaesu FT60 handheld radio:

The filter in Gqrx is adjusted for wide deviation (5 kHz) but the
signal is only really about 3 kHz wide.

Please note how the signal is 15-16 dB smack on above the noise floor.
If I were to run your modem, it would probably decode even where my
SDR falls over. This is probably because the RTL-SDR has a much lower
noise figure than my USRP, although I will trade that any day for the
12 bit ADC.

73,
Adrian
Post by Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code? I've been trying
to optimize modems running at 1200 bits/second and the best I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB above for
GMSK. Your modem seems to be 16-20 dB better than that! If I could see
the Gnuradio graph maybe I could figure out what I'm doing wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here that might make a good
transmitter for it.
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and importing that into
audacity.
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB compression?
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this shows how wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals and a couple
of clips going on. Maybe need to knock it down to +/- .5 like the old FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
David Rowe
2016-03-16 23:53:20 UTC
Permalink
Hi Adrian,

I think Brady's going to get back you with his GR code. Not sure if the
FSK modem is still outboard of GR, or if he has a GR module for it.

Note we are not running our modems through the analog FM
modulator/demodulator, as that's how you throw perfectly good dB away.
I'm looking at you DSTAR, DRM, and C4FM.

Heaps of my blog posts on this topic over the last 18 months.

Cheers,

David
Post by Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code? I've been trying
to optimize modems running at 1200 bits/second and the best I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB above for
GMSK. Your modem seems to be 16-20 dB better than that! If I could see
the Gnuradio graph maybe I could figure out what I'm doing wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here that might make a good
transmitter for it.
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and importing that into
audacity.
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB compression?
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this shows how wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals and a couple
of clips going on. Maybe need to knock it down to +/- .5 like the old FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Brady O'Brien
2016-03-17 00:04:44 UTC
Permalink
Hey Adrian,

The gnuradio-companion pipelines I used for the test are attached. These
grc pipelines are only meant to get 16-bit 48k samples in and out of the
HackRF and rtlsdr for testing. The meat of the FSK is done with the C fsk
modem, used in freedv_rx, freedv_tx, fsk_mod, and fsk_demod. We are also
running our FSK modem directly from SDR samples instead of through FM
radios. We will be running 2400B over legacy FM radios, but don't yet have
the same degree of test coverage.

Thanks,

Brady O'Brien
Post by Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code? I've been trying
to optimize modems running at 1200 bits/second and the best I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB above for
GMSK. Your modem seems to be 16-20 dB better than that! If I could see
the Gnuradio graph maybe I could figure out what I'm doing wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here that might make a good
transmitter for it.
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and importing that
into
Post by Steve
audacity.
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB compression?
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this shows how wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals and a
couple
Post by Steve
of clips going on. Maybe need to knock it down to +/- .5 like the old FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Adrian Musceac
2016-03-17 12:41:58 UTC
Permalink
Hi Brady, David,

Thank you very much for sharing the GRC flowgraphs. This will help me
a lot setting up a similar test. Regarding the performance, I was just
surprised to see such a big difference between a modem which runs
through analog FM and your new modem. I would have expected something
in the range of 6-10 dB, not 16-20 dB. Maybe I need to calibrate my
signal again.
I enjoyed a lot reading about the experiment on your blog David. It
seems like a new revolution for VHF/UHF weak signal work.

73,
Adrian (YO8RZZ)
Post by Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are attached. These
grc pipelines are only meant to get 16-bit 48k samples in and out of the
HackRF and rtlsdr for testing. The meat of the FSK is done with the C fsk
modem, used in freedv_rx, freedv_tx, fsk_mod, and fsk_demod. We are also
running our FSK modem directly from SDR samples instead of through FM
radios. We will be running 2400B over legacy FM radios, but don't yet have
the same degree of test coverage.
Thanks,
Brady O'Brien
Post by Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code? I've been trying
to optimize modems running at 1200 bits/second and the best I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB above for
GMSK. Your modem seems to be 16-20 dB better than that! If I could see
the Gnuradio graph maybe I could figure out what I'm doing wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here that might make a good
transmitter for it.
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and importing that
into
Post by Steve
audacity.
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB compression?
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this shows how wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals and a
couple
Post by Steve
of clips going on. Maybe need to knock it down to +/- .5 like the old FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
David Rowe
2016-03-17 18:50:37 UTC
Permalink
Hi Adrian,

Getting the modem right can be really critical. It's good that we are
discovering this and talking about it in Ham and open source circles,
and getting high quality modem code out there.

I learned about modems in the sat-com industry. As the satellites are
expensive, we would try to get our modems to within 0.25dB of theory.
Otherwise we were just wasting mega-$.

It still puzzles me why something like DMR has deliberately chosen a
modem waveform (with less than Rs Hz spacing tones) that hobbles the
system by 6dB. The power being wasted is like throwing away 3/4 of your
HT battery.

Suspect of your 16dB loss is a measurement error or perhaps the FM modem
is not set up right. I found simulating analog FM and getting
performance like the text books to be surprisingly difficult:

http://www.rowetel.com/blog/?p=3799

Maybe you could first test your FM modem, see if you can get the C/No
input versus SNR out curve.

Looking forward, FreeDV 2400A is competitive with SSB for weak signal
work today, and I think we can push down another 10dB over time.

Cheers,

David
Post by Adrian Musceac
Hi Brady, David,
Thank you very much for sharing the GRC flowgraphs. This will help me
a lot setting up a similar test. Regarding the performance, I was just
surprised to see such a big difference between a modem which runs
through analog FM and your new modem. I would have expected something
in the range of 6-10 dB, not 16-20 dB. Maybe I need to calibrate my
signal again.
I enjoyed a lot reading about the experiment on your blog David. It
seems like a new revolution for VHF/UHF weak signal work.
73,
Adrian (YO8RZZ)
Post by Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are attached. These
grc pipelines are only meant to get 16-bit 48k samples in and out of the
HackRF and rtlsdr for testing. The meat of the FSK is done with the C fsk
modem, used in freedv_rx, freedv_tx, fsk_mod, and fsk_demod. We are also
running our FSK modem directly from SDR samples instead of through FM
radios. We will be running 2400B over legacy FM radios, but don't yet have
the same degree of test coverage.
Thanks,
Brady O'Brien
Post by Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code? I've been trying
to optimize modems running at 1200 bits/second and the best I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB above for
GMSK. Your modem seems to be 16-20 dB better than that! If I could see
the Gnuradio graph maybe I could figure out what I'm doing wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here that might make a good
transmitter for it.
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and importing that
into
Post by Steve
audacity.
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB compression?
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this shows how wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals and a
couple
Post by Steve
of clips going on. Maybe need to knock it down to +/- .5 like the old FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Bruce Perens
2016-03-17 20:48:21 UTC
Permalink
Did I hear that right? Works with non linear amplifiers, competitive with
SSB, replaces all your multi carrier modems?
Post by David Rowe
Hi Adrian,
Getting the modem right can be really critical. It's good that we are
discovering this and talking about it in Ham and open source circles,
and getting high quality modem code out there.
I learned about modems in the sat-com industry. As the satellites are
expensive, we would try to get our modems to within 0.25dB of theory.
Otherwise we were just wasting mega-$.
It still puzzles me why something like DMR has deliberately chosen a
modem waveform (with less than Rs Hz spacing tones) that hobbles the
system by 6dB. The power being wasted is like throwing away 3/4 of your
HT battery.
Suspect of your 16dB loss is a measurement error or perhaps the FM modem
is not set up right. I found simulating analog FM and getting
http://www.rowetel.com/blog/?p=3799
Maybe you could first test your FM modem, see if you can get the C/No
input versus SNR out curve.
Looking forward, FreeDV 2400A is competitive with SSB for weak signal
work today, and I think we can push down another 10dB over time.
Cheers,
David
Post by Adrian Musceac
Hi Brady, David,
Thank you very much for sharing the GRC flowgraphs. This will help me
a lot setting up a similar test. Regarding the performance, I was just
surprised to see such a big difference between a modem which runs
through analog FM and your new modem. I would have expected something
in the range of 6-10 dB, not 16-20 dB. Maybe I need to calibrate my
signal again.
I enjoyed a lot reading about the experiment on your blog David. It
seems like a new revolution for VHF/UHF weak signal work.
73,
Adrian (YO8RZZ)
Post by Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are attached. These
grc pipelines are only meant to get 16-bit 48k samples in and out of the
HackRF and rtlsdr for testing. The meat of the FSK is done with the C
fsk
Post by Adrian Musceac
Post by Brady O'Brien
modem, used in freedv_rx, freedv_tx, fsk_mod, and fsk_demod. We are also
running our FSK modem directly from SDR samples instead of through FM
radios. We will be running 2400B over legacy FM radios, but don't yet
have
Post by Adrian Musceac
Post by Brady O'Brien
the same degree of test coverage.
Thanks,
Brady O'Brien
Post by Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code? I've been trying
to optimize modems running at 1200 bits/second and the best I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB above for
GMSK. Your modem seems to be 16-20 dB better than that! If I could see
the Gnuradio graph maybe I could figure out what I'm doing wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here that might make a good
transmitter for it.
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and importing that
into
Post by Steve
audacity.
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB
compression?
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this shows
how
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals and a
couple
Post by Steve
of clips going on. Maybe need to knock it down to +/- .5 like the
old
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
David Rowe
2016-03-17 21:15:55 UTC
Permalink
Competitive with SSB on VHF (AWGN non fading) channels. This mode will
be error free when SSB is full of noise. Not a HF solution. But that's
coming.
Post by Bruce Perens
Did I hear that right? Works with non linear amplifiers, competitive
with SSB, replaces all your multi carrier modems?
Hi Adrian,
Getting the modem right can be really critical. It's good that we are
discovering this and talking about it in Ham and open source circles,
and getting high quality modem code out there.
I learned about modems in the sat-com industry. As the satellites are
expensive, we would try to get our modems to within 0.25dB of theory.
Otherwise we were just wasting mega-$.
It still puzzles me why something like DMR has deliberately chosen a
modem waveform (with less than Rs Hz spacing tones) that hobbles the
system by 6dB. The power being wasted is like throwing away 3/4 of your
HT battery.
Suspect of your 16dB loss is a measurement error or perhaps the FM modem
is not set up right. I found simulating analog FM and getting
http://www.rowetel.com/blog/?p=3799
Maybe you could first test your FM modem, see if you can get the C/No
input versus SNR out curve.
Looking forward, FreeDV 2400A is competitive with SSB for weak signal
work today, and I think we can push down another 10dB over time.
Cheers,
David
Post by Adrian Musceac
Hi Brady, David,
Thank you very much for sharing the GRC flowgraphs. This will help me
a lot setting up a similar test. Regarding the performance, I was
just
Post by Adrian Musceac
surprised to see such a big difference between a modem which runs
through analog FM and your new modem. I would have expected something
in the range of 6-10 dB, not 16-20 dB. Maybe I need to calibrate my
signal again.
I enjoyed a lot reading about the experiment on your blog David. It
seems like a new revolution for VHF/UHF weak signal work.
73,
Adrian (YO8RZZ)
Post by Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are
attached. These
Post by Adrian Musceac
Post by Brady O'Brien
grc pipelines are only meant to get 16-bit 48k samples in and
out of the
Post by Adrian Musceac
Post by Brady O'Brien
HackRF and rtlsdr for testing. The meat of the FSK is done with
the C fsk
Post by Adrian Musceac
Post by Brady O'Brien
modem, used in freedv_rx, freedv_tx, fsk_mod, and fsk_demod. We
are also
Post by Adrian Musceac
Post by Brady O'Brien
running our FSK modem directly from SDR samples instead of
through FM
Post by Adrian Musceac
Post by Brady O'Brien
radios. We will be running 2400B over legacy FM radios, but
don't yet have
Post by Adrian Musceac
Post by Brady O'Brien
the same degree of test coverage.
Thanks,
Brady O'Brien
On Wed, Mar 16, 2016 at 3:34 PM, Adrian Musceac
Post by Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code? I've been
trying
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
to optimize modems running at 1200 bits/second and the best I
could do
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB above for
GMSK. Your modem seems to be 16-20 dB better than that! If I
could see
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
the Gnuradio graph maybe I could figure out what I'm doing wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here that might
make a
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
good
transmitter for it.
On Wed, Mar 16, 2016 at 4:02 AM, glen english
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and
importing that
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
into
Post by Steve
audacity.
On Wed, Mar 16, 2016 at 3:46 AM, glen english
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB
compression?
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Thought I'd upload a spectrum snapshot. You can't really see the
carriers
until you drop the number of samples to about 256, but this
shows how
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals
and a
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
couple
Post by Steve
of clips going on. Maybe need to knock it down to +/- .5
like the old
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Steve
2016-03-17 22:10:02 UTC
Permalink
Well, it will be competitive when SDR transceivers are available for VHF
and up.

I don't see that day too far away (maybe 2020 when SDR rigs flood the Ham
market), but right now my 10 year old 2 metre SSB rig and FDM will have to
do, hee...

I suspect we will be buying modems as Apps for our Apple or Google Radios.

Steve
Post by Bruce Perens
Did I hear that right? Works with non linear amplifiers, competitive with
SSB, replaces all your multi carrier modems?
Adrian Musceac
2016-03-18 10:03:58 UTC
Permalink
Hi David,
You are right, it was measurement error. With the help of Brady's GRC flowgraphs, I was able to set up a more reliable test.
The chain was: gnuradio DBPSK modem on laptop -> Yaesu FT60 set on narrow deviation -> USRP B200 -> Gqrx
The USRP has fine adjustable gain, between 0 and 70 dB.
I was able to obtain error free copy at 18 dB above noise or 6 dB above (scratchy) analog FM. I am using a packet encoder so if 1 bit is lost the whole packet is dropped.
The interesting thing is that sometimes the modem would decode at barely 10 dB above noise. It was critical to adjust audio levels so that the radio does not over deviate.
Also, FM modulation does strange things to my phase, so I had to run a very tight FLL band edge loop to keep the constellation stable. I think there is still lots of room for improvement there.
Also, I could not make the MSK modem decode after Gqrx demodulation, although if I use two analog radios it passes just fine. So I can't provide measurements for that modem.
Hope this is useful,

Adrian (YO8RZZ)
Post by David Rowe
Hi Adrian,
Getting the modem right can be really critical. It's good that we are
discovering this and talking about it in Ham and open source circles,
and getting high quality modem code out there.
I learned about modems in the sat-com industry. As the satellites are
expensive, we would try to get our modems to within 0.25dB of theory.
Otherwise we were just wasting mega-$.
It still puzzles me why something like DMR has deliberately chosen a
modem waveform (with less than Rs Hz spacing tones) that hobbles the
system by 6dB. The power being wasted is like throwing away 3/4 of your
HT battery.
Suspect of your 16dB loss is a measurement error or perhaps the FM modem
is not set up right. I found simulating analog FM and getting
http://www.rowetel.com/blog/?p=3799
Maybe you could first test your FM modem, see if you can get the C/No
input versus SNR out curve.
Looking forward, FreeDV 2400A is competitive with SSB for weak signal
work today, and I think we can push down another 10dB over time.
Cheers,
David
Post by Adrian Musceac
Hi Brady, David,
Thank you very much for sharing the GRC flowgraphs. This will help me
a lot setting up a similar test. Regarding the performance, I was
just
Post by Adrian Musceac
surprised to see such a big difference between a modem which runs
through analog FM and your new modem. I would have expected something
in the range of 6-10 dB, not 16-20 dB. Maybe I need to calibrate my
signal again.
I enjoyed a lot reading about the experiment on your blog David. It
seems like a new revolution for VHF/UHF weak signal work.
73,
Adrian (YO8RZZ)
Post by Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are attached.
These
Post by Adrian Musceac
Post by Brady O'Brien
grc pipelines are only meant to get 16-bit 48k samples in and out of
the
Post by Adrian Musceac
Post by Brady O'Brien
HackRF and rtlsdr for testing. The meat of the FSK is done with the
C fsk
Post by Adrian Musceac
Post by Brady O'Brien
modem, used in freedv_rx, freedv_tx, fsk_mod, and fsk_demod. We are
also
Post by Adrian Musceac
Post by Brady O'Brien
running our FSK modem directly from SDR samples instead of through
FM
Post by Adrian Musceac
Post by Brady O'Brien
radios. We will be running 2400B over legacy FM radios, but don't
yet have
Post by Adrian Musceac
Post by Brady O'Brien
the same degree of test coverage.
Thanks,
Brady O'Brien
Post by Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code? I've been
trying
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
to optimize modems running at 1200 bits/second and the best I could
do
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB above for
GMSK. Your modem seems to be 16-20 dB better than that! If I could
see
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
the Gnuradio graph maybe I could figure out what I'm doing wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here that might make
a
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
good
transmitter for it.
On Wed, Mar 16, 2016 at 4:02 AM, glen english
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and importing
that
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
into
Post by Steve
audacity.
On Wed, Mar 16, 2016 at 3:46 AM, glen english
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB
compression?
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this
shows how
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals and
a
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
couple
Post by Steve
of clips going on. Maybe need to knock it down to +/- .5 like
the old
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Brady O'Brien
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
--
Sent from my Android device with K-9 Mail. Please excuse my brevity.
Adrian Musceac
2016-03-25 10:17:24 UTC
Permalink
Hi Brady, David,

I have started transitioning my GRC flowgraphs to C++ and I have put the
code up on Github, just in case you might find it useful in your tests for
the 2400B mode. My choice of a modem is AFSK, 2000 bit/sec (1200 bit voice
or 1300 bit data). It has no FEC, no scrambling and no interleaving (yet).
The audio is framed for the standard 40 msec Codec2. It works through a
handheld FM transceiver and can be decoded easily with Gqrx.

The perfomance is 5 dB worse than analog FM, measured for 12 dB SNR. I
would be very curious to see a comparison with 2400B. My next step is to
transition the DBPSK modem which is 3 dB worse than analog FM based on my
USRP captured data.

Link to the code repo: https://github.com/kantooon/gnuradio_audio_modems
Video of operation:


Cheers,
Adrian (YO8RZZ)
Post by Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are attached. These
grc pipelines are only meant to get 16-bit 48k samples in and out of the
HackRF and rtlsdr for testing. The meat of the FSK is done with the C fsk
modem, used in freedv_rx, freedv_tx, fsk_mod, and fsk_demod. We are also
running our FSK modem directly from SDR samples instead of through FM
radios. We will be running 2400B over legacy FM radios, but don't yet have
the same degree of test coverage.
Thanks,
Brady O'Brien
Post by Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code? I've been trying
to optimize modems running at 1200 bits/second and the best I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB above for
GMSK. Your modem seems to be 16-20 dB better than that! If I could see
the Gnuradio graph maybe I could figure out what I'm doing wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here that might make a
good
Post by Steve
transmitter for it.
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and importing that
into
Post by Steve
audacity.
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB compression?
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this shows how wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals and a
couple
Post by Steve
of clips going on. Maybe need to knock it down to +/- .5 like the old FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Adrian Musceac
2016-07-04 22:46:32 UTC
Permalink
Hi,

Some more news from the VHF modem front. First of all, I did some
performance comparisons between GMSK waveforms and regular AFSK as used by
a typical APRS modem. Obviously, there is a reason why Bell AFSK
was chosen: the Gaussian pulse shaping filter does introduce ISI as a cost
for reducing the signal bandwidth.
The next question was "how much". Luckily, using the excellent filter
design tool from Gnuradio, we can change the filter parameters and then
test the result against a channel with AWGN noise. The difference seems to
be around 2 dB in favour of AFSK.

However, my earlier experiments showed that DBPSK will do even better.
Unfortunately, sending PSK over a narrow FM channel requires some tuning of
the modem parameters. 3 dB of improvement at the same BER is not
surprising, what was surprising to me though was the fact that adding a
pulse shaping filter immediately before the frequency conversion improved
the signal decoding by another 1-2 dB. A 9 dB SNR signal has the same BER
as a 14 dB SNR GMSK modem. At this point we are almost doing better than
analog voice already.
To add further gain we can always decrease the bitrate to 1200 bit/s.

And since I was focused on filters, I had to take another look at the data.
What we want is a signal that can pass 1300 bit/s Codec2 through a narrow
mode FM transceiver (wide mode is beginning to be phased out at this
point). We still have bandwidth to spare, so it seemed to me that using the
same filter bandwidth for TX and RX is wasteful. Instead, the TX filter is
20% wider than the RX filter for (very little) added performance.

The bad news is that the new modem is computationally about twice more
expensive. Using it on my phone will be a big stretch. We will see if I can
make it work on a 1.2 GHz CPU.

Finally, there are some people out there who criticise Gnuradio for being
slow, bloated and too high level for this kind of stuff. I just want to say
that for me as a student, it is a very valuable tool for learning. Maybe
some day I will join the ranks of the Octave gurus, but not today.

A video and a screenshot of my flowgraph:


Cheers,
Adrian
Post by Adrian Musceac
Hi Brady, David,
I have started transitioning my GRC flowgraphs to C++ and I have put the
code up on Github, just in case you might find it useful in your tests for
the 2400B mode. My choice of a modem is AFSK, 2000 bit/sec (1200 bit voice
or 1300 bit data). It has no FEC, no scrambling and no interleaving (yet).
The audio is framed for the standard 40 msec Codec2. It works through a
handheld FM transceiver and can be decoded easily with Gqrx.
The perfomance is 5 dB worse than analog FM, measured for 12 dB SNR. I
would be very curious to see a comparison with 2400B. My next step is to
transition the DBPSK modem which is 3 dB worse than analog FM based on my
USRP captured data.
Link to the code repo: https://github.com/kantooon/gnuradio_audio_modems
Video of operation: http://youtu.be/2FZBQvVm5Co
Cheers,
Adrian (YO8RZZ)
Post by Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are attached. These
grc pipelines are only meant to get 16-bit 48k samples in and out of the
HackRF and rtlsdr for testing. The meat of the FSK is done with the C fsk
modem, used in freedv_rx, freedv_tx, fsk_mod, and fsk_demod. We are also
running our FSK modem directly from SDR samples instead of through FM
radios. We will be running 2400B over legacy FM radios, but don't yet have
the same degree of test coverage.
Thanks,
Brady O'Brien
Post by Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code? I've been trying
to optimize modems running at 1200 bits/second and the best I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB above for
GMSK. Your modem seems to be 16-20 dB better than that! If I could see
the Gnuradio graph maybe I could figure out what I'm doing wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here that might make a
good
Post by Steve
transmitter for it.
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and importing that
into
Post by Steve
audacity.
Hi Steve
Nice
is that through a PA- if so, linear or linear in say 3dB compression?
Thought I'd upload a spectrum snapshot. You can't really see the carriers
until you drop the number of samples to about 256, but this shows how wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0 signals and a
couple
Post by Steve
of clips going on. Maybe need to knock it down to +/- .5 like the old FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
David Rowe
2016-07-05 02:21:44 UTC
Permalink
Nice work Adrian. A good way to present this work is plots of BER
against Eb/No for the various modulation schemes you are experimenting
with. See my blog posts on FSK etc for examples.

It's useful to include the theoretical curves as comparisons to your
real world implementations, these are well known for BPSK, FSK, and GMSK.

You can also plot BER against SNR if the bit rate Rb and Noise BW N are
equal. Eb/No is convenient as everything is normalised.

Cheers,

David
Post by Adrian Musceac
Hi,
Some more news from the VHF modem front. First of all, I did some
performance comparisons between GMSK waveforms and regular AFSK as used
by a typical APRS modem. Obviously, there is a reason why Bell AFSK
was chosen: the Gaussian pulse shaping filter does introduce ISI as a
cost for reducing the signal bandwidth.
The next question was "how much". Luckily, using the excellent filter
design tool from Gnuradio, we can change the filter parameters and then
test the result against a channel with AWGN noise. The difference seems
to be around 2 dB in favour of AFSK.
However, my earlier experiments showed that DBPSK will do even better.
Unfortunately, sending PSK over a narrow FM channel requires some tuning
of the modem parameters. 3 dB of improvement at the same BER is not
surprising, what was surprising to me though was the fact that adding a
pulse shaping filter immediately before the frequency conversion
improved the signal decoding by another 1-2 dB. A 9 dB SNR signal has
the same BER as a 14 dB SNR GMSK modem. At this point we are almost
doing better than analog voice already.
To add further gain we can always decrease the bitrate to 1200 bit/s.
And since I was focused on filters, I had to take another look at the
data. What we want is a signal that can pass 1300 bit/s Codec2 through a
narrow mode FM transceiver (wide mode is beginning to be phased out at
this point). We still have bandwidth to spare, so it seemed to me that
using the same filter bandwidth for TX and RX is wasteful. Instead, the
TX filter is 20% wider than the RX filter for (very little) added
performance.
The bad news is that the new modem is computationally about twice more
expensive. Using it on my phone will be a big stretch. We will see if I
can make it work on a 1.2 GHz CPU.
Finally, there are some people out there who criticise Gnuradio for
being slow, bloated and too high level for this kind of stuff. I just
want to say that for me as a student, it is a very valuable tool for
learning. Maybe some day I will join the ranks of the Octave gurus, but
not today.
http://youtu.be/YOX7H3Fcxe4
Cheers,
Adrian
Hi Brady, David,
I have started transitioning my GRC flowgraphs to C++ and I have put
the code up on Github, just in case you might find it useful in your
tests for the 2400B mode. My choice of a modem is AFSK, 2000 bit/sec
(1200 bit voice or 1300 bit data). It has no FEC, no scrambling and
no interleaving (yet). The audio is framed for the standard 40 msec
Codec2. It works through a handheld FM transceiver and can be
decoded easily with Gqrx.
The perfomance is 5 dB worse than analog FM, measured for 12 dB SNR.
I would be very curious to see a comparison with 2400B. My next step
is to transition the DBPSK modem which is 3 dB worse than analog FM
based on my USRP captured data.
Link to the code repo: https://github.com/kantooon/gnuradio_audio_modems
Video of operation: http://youtu.be/2FZBQvVm5Co
Cheers,
Adrian (YO8RZZ)
On Thu, Mar 17, 2016 at 12:04 AM, Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are
attached. These grc pipelines are only meant to get 16-bit 48k
samples in and out of the HackRF and rtlsdr for testing. The
meat of the FSK is done with the C fsk modem, used in freedv_rx,
freedv_tx, fsk_mod, and fsk_demod. We are also running our FSK
modem directly from SDR samples instead of through FM radios. We
will be running 2400B over legacy FM radios, but don't yet have
the same degree of test coverage.
Thanks,
Brady O'Brien
On Wed, Mar 16, 2016 at 3:34 PM, Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code? I've
been trying
to optimize modems running at 1200 bits/second and the best
I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB above for
GMSK. Your modem seems to be 16-20 dB better than that! If I
could see
the Gnuradio graph maybe I could figure out what I'm doing wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here that
might make a good
Post by Steve
transmitter for it.
On Wed, Mar 16, 2016 at 4:02 AM, glen english
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and
importing that into
Post by Steve
audacity.
On Wed, Mar 16, 2016 at 3:46 AM, glen english
Hi Steve
Nice
is that through a PA- if so, linear or linear in say
3dB compression?
Post by Steve
Thought I'd upload a spectrum snapshot. You can't
really see the
Post by Steve
carriers
until you drop the number of samples to about 256, but
this shows how
Post by Steve
wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0
signals and a couple
Post by Steve
of clips going on. Maybe need to knock it down to +/-
.5 like the old
Post by Steve
FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Steve
_______________________________________________
Freetel-codec2 mailing
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
Post by Steve
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Steve
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Adrian Musceac
2016-07-05 13:32:59 UTC
Permalink
Hi David,

I wish I was able to generate BER plots similar to those in your images. Unfortunately, while Gnuradio has powerful real time plotting blocks, I haven't figured out how to use them yet.
For this reason, I have to use two flowgraphs, one for the analog radio and one for the USRP, get the SNR figures from one and use pen and paper to compare with the known bits received on the other. This allows me to change the block parameters live and watch the result. Hopefully some day I will be able to learn Octave.

For reference, the pulse shaping filters I am experimenting with are derived from this paper:
http://www.ochiailab.dnj.ynu.ac.jp/pdf/tanahashi_tcom06_final.pdf

Cheers,
Adrian
Post by David Rowe
Nice work Adrian. A good way to present this work is plots of BER
against Eb/No for the various modulation schemes you are experimenting
with. See my blog posts on FSK etc for examples.
It's useful to include the theoretical curves as comparisons to your
real world implementations, these are well known for BPSK, FSK, and GMSK.
You can also plot BER against SNR if the bit rate Rb and Noise BW N are
equal. Eb/No is convenient as everything is normalised.
Cheers,
David
Post by Adrian Musceac
Hi,
Some more news from the VHF modem front. First of all, I did some
performance comparisons between GMSK waveforms and regular AFSK as
used
Post by Adrian Musceac
by a typical APRS modem. Obviously, there is a reason why Bell AFSK
was chosen: the Gaussian pulse shaping filter does introduce ISI as
a
Post by Adrian Musceac
cost for reducing the signal bandwidth.
The next question was "how much". Luckily, using the excellent filter
design tool from Gnuradio, we can change the filter parameters and
then
Post by Adrian Musceac
test the result against a channel with AWGN noise. The difference
seems
Post by Adrian Musceac
to be around 2 dB in favour of AFSK.
However, my earlier experiments showed that DBPSK will do even
better.
Post by Adrian Musceac
Unfortunately, sending PSK over a narrow FM channel requires some
tuning
Post by Adrian Musceac
of the modem parameters. 3 dB of improvement at the same BER is not
surprising, what was surprising to me though was the fact that adding
a
Post by Adrian Musceac
pulse shaping filter immediately before the frequency conversion
improved the signal decoding by another 1-2 dB. A 9 dB SNR signal has
the same BER as a 14 dB SNR GMSK modem. At this point we are almost
doing better than analog voice already.
To add further gain we can always decrease the bitrate to 1200 bit/s.
And since I was focused on filters, I had to take another look at the
data. What we want is a signal that can pass 1300 bit/s Codec2
through a
Post by Adrian Musceac
narrow mode FM transceiver (wide mode is beginning to be phased out
at
Post by Adrian Musceac
this point). We still have bandwidth to spare, so it seemed to me
that
Post by Adrian Musceac
using the same filter bandwidth for TX and RX is wasteful. Instead,
the
Post by Adrian Musceac
TX filter is 20% wider than the RX filter for (very little) added
performance.
The bad news is that the new modem is computationally about twice
more
Post by Adrian Musceac
expensive. Using it on my phone will be a big stretch. We will see if
I
Post by Adrian Musceac
can make it work on a 1.2 GHz CPU.
Finally, there are some people out there who criticise Gnuradio for
being slow, bloated and too high level for this kind of stuff. I just
want to say that for me as a student, it is a very valuable tool for
learning. Maybe some day I will join the ranks of the Octave gurus,
but
Post by Adrian Musceac
not today.
http://youtu.be/YOX7H3Fcxe4
Cheers,
Adrian
Hi Brady, David,
I have started transitioning my GRC flowgraphs to C++ and I have
put
Post by Adrian Musceac
the code up on Github, just in case you might find it useful in
your
Post by Adrian Musceac
tests for the 2400B mode. My choice of a modem is AFSK, 2000
bit/sec
Post by Adrian Musceac
(1200 bit voice or 1300 bit data). It has no FEC, no scrambling
and
Post by Adrian Musceac
no interleaving (yet). The audio is framed for the standard 40
msec
Post by Adrian Musceac
Codec2. It works through a handheld FM transceiver and can be
decoded easily with Gqrx.
The perfomance is 5 dB worse than analog FM, measured for 12 dB
SNR.
Post by Adrian Musceac
I would be very curious to see a comparison with 2400B. My next
step
Post by Adrian Musceac
is to transition the DBPSK modem which is 3 dB worse than analog
FM
Post by Adrian Musceac
based on my USRP captured data.
https://github.com/kantooon/gnuradio_audio_modems
Post by Adrian Musceac
Video of operation: http://youtu.be/2FZBQvVm5Co
Cheers,
Adrian (YO8RZZ)
On Thu, Mar 17, 2016 at 12:04 AM, Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are
attached. These grc pipelines are only meant to get 16-bit
48k
Post by Adrian Musceac
samples in and out of the HackRF and rtlsdr for testing. The
meat of the FSK is done with the C fsk modem, used in
freedv_rx,
Post by Adrian Musceac
freedv_tx, fsk_mod, and fsk_demod. We are also running our
FSK
Post by Adrian Musceac
modem directly from SDR samples instead of through FM radios.
We
Post by Adrian Musceac
will be running 2400B over legacy FM radios, but don't yet
have
Post by Adrian Musceac
the same degree of test coverage.
Thanks,
Brady O'Brien
On Wed, Mar 16, 2016 at 3:34 PM, Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code?
I've
Post by Adrian Musceac
been trying
to optimize modems running at 1200 bits/second and the
best
Post by Adrian Musceac
I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB
above for
GMSK. Your modem seems to be 16-20 dB better than that!
If I
Post by Adrian Musceac
could see
the Gnuradio graph maybe I could figure out what I'm
doing
Post by Adrian Musceac
wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here
that
Post by Adrian Musceac
might make a good
Post by Steve
transmitter for it.
On Wed, Mar 16, 2016 at 4:02 AM, glen english
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and
importing that into
Post by Steve
audacity.
On Wed, Mar 16, 2016 at 3:46 AM, glen english
Hi Steve
Nice
is that through a PA- if so, linear or linear in say
3dB compression?
Post by Steve
Thought I'd upload a spectrum snapshot. You can't
really see the
Post by Steve
carriers
until you drop the number of samples to about 256,
but
Post by Adrian Musceac
this shows how
Post by Steve
wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0
signals and a couple
Post by Steve
of clips going on. Maybe need to knock it down to
+/-
Post by Adrian Musceac
.5 like the old
Post by Steve
FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in
San
Post by Adrian Musceac
Francisco, CA to explore cutting-edge tech and listen to tech
luminaries
Post by Adrian Musceac
present their vision of the future. This family event has something
for
Post by Adrian Musceac
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech
luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
--
Sent from my Android device with K-9 Mail. Please excuse my brevity.
Brady O'Brien
2016-07-05 13:42:33 UTC
Permalink
Hi Adrian,

I think you really ought to look into using Gnuradio with python directly
rather than through the gnuradio companion. That would allow you to build
up the entire modem and channel sim in a block, run it over a set of
parameters, and produce a graph of BER vs Eb/N0 or some other parameter
with pyqtgraph or matplotlib. The companion will even spit out a python
pipeline of whatever you've built up for you to start from. For much of my
2400A testing, I built a pipeline in GRC but ran and modified the python
directly in the field.

Thanks,

Brady O'Brien
Post by Adrian Musceac
Hi David,
I wish I was able to generate BER plots similar to those in your images.
Unfortunately, while Gnuradio has powerful real time plotting blocks, I
haven't figured out how to use them yet.
For this reason, I have to use two flowgraphs, one for the analog radio
and one for the USRP, get the SNR figures from one and use pen and paper to
compare with the known bits received on the other. This allows me to change
the block parameters live and watch the result. Hopefully some day I will
be able to learn Octave.
For reference, the pulse shaping filters I am experimenting with are
http://www.ochiailab.dnj.ynu.ac.jp/pdf/tanahashi_tcom06_final.pdf
Cheers,
Adrian
Post by David Rowe
Nice work Adrian. A good way to present this work is plots of BER
against Eb/No for the various modulation schemes you are experimenting
with. See my blog posts on FSK etc for examples.
It's useful to include the theoretical curves as comparisons to your
real world implementations, these are well known for BPSK, FSK, and GMSK.
You can also plot BER against SNR if the bit rate Rb and Noise BW N are
equal. Eb/No is convenient as everything is normalised.
Cheers,
David
Post by Adrian Musceac
Hi,
Some more news from the VHF modem front. First of all, I did some
performance comparisons between GMSK waveforms and regular AFSK as
used
Post by Adrian Musceac
by a typical APRS modem. Obviously, there is a reason why Bell AFSK
was chosen: the Gaussian pulse shaping filter does introduce ISI as
a
Post by Adrian Musceac
cost for reducing the signal bandwidth.
The next question was "how much". Luckily, using the excellent filter
design tool from Gnuradio, we can change the filter parameters and
then
Post by Adrian Musceac
test the result against a channel with AWGN noise. The difference
seems
Post by Adrian Musceac
to be around 2 dB in favour of AFSK.
However, my earlier experiments showed that DBPSK will do even
better.
Post by Adrian Musceac
Unfortunately, sending PSK over a narrow FM channel requires some
tuning
Post by Adrian Musceac
of the modem parameters. 3 dB of improvement at the same BER is not
surprising, what was surprising to me though was the fact that adding
a
Post by Adrian Musceac
pulse shaping filter immediately before the frequency conversion
improved the signal decoding by another 1-2 dB. A 9 dB SNR signal has
the same BER as a 14 dB SNR GMSK modem. At this point we are almost
doing better than analog voice already.
To add further gain we can always decrease the bitrate to 1200 bit/s.
And since I was focused on filters, I had to take another look at the
data. What we want is a signal that can pass 1300 bit/s Codec2
through a
Post by Adrian Musceac
narrow mode FM transceiver (wide mode is beginning to be phased out
at
Post by Adrian Musceac
this point). We still have bandwidth to spare, so it seemed to me
that
Post by Adrian Musceac
using the same filter bandwidth for TX and RX is wasteful. Instead,
the
Post by Adrian Musceac
TX filter is 20% wider than the RX filter for (very little) added
performance.
The bad news is that the new modem is computationally about twice
more
Post by Adrian Musceac
expensive. Using it on my phone will be a big stretch. We will see if
I
Post by Adrian Musceac
can make it work on a 1.2 GHz CPU.
Finally, there are some people out there who criticise Gnuradio for
being slow, bloated and too high level for this kind of stuff. I just
want to say that for me as a student, it is a very valuable tool for
learning. Maybe some day I will join the ranks of the Octave gurus,
but
Post by Adrian Musceac
not today.
http://youtu.be/YOX7H3Fcxe4
Cheers,
Adrian
Hi Brady, David,
I have started transitioning my GRC flowgraphs to C++ and I have
put
Post by Adrian Musceac
the code up on Github, just in case you might find it useful in
your
Post by Adrian Musceac
tests for the 2400B mode. My choice of a modem is AFSK, 2000
bit/sec
Post by Adrian Musceac
(1200 bit voice or 1300 bit data). It has no FEC, no scrambling
and
Post by Adrian Musceac
no interleaving (yet). The audio is framed for the standard 40
msec
Post by Adrian Musceac
Codec2. It works through a handheld FM transceiver and can be
decoded easily with Gqrx.
The perfomance is 5 dB worse than analog FM, measured for 12 dB
SNR.
Post by Adrian Musceac
I would be very curious to see a comparison with 2400B. My next
step
Post by Adrian Musceac
is to transition the DBPSK modem which is 3 dB worse than analog
FM
Post by Adrian Musceac
based on my USRP captured data.
https://github.com/kantooon/gnuradio_audio_modems
Post by Adrian Musceac
Video of operation: http://youtu.be/2FZBQvVm5Co
Cheers,
Adrian (YO8RZZ)
On Thu, Mar 17, 2016 at 12:04 AM, Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are
attached. These grc pipelines are only meant to get 16-bit
48k
Post by Adrian Musceac
samples in and out of the HackRF and rtlsdr for testing. The
meat of the FSK is done with the C fsk modem, used in
freedv_rx,
Post by Adrian Musceac
freedv_tx, fsk_mod, and fsk_demod. We are also running our
FSK
Post by Adrian Musceac
modem directly from SDR samples instead of through FM radios.
We
Post by Adrian Musceac
will be running 2400B over legacy FM radios, but don't yet
have
Post by Adrian Musceac
the same degree of test coverage.
Thanks,
Brady O'Brien
On Wed, Mar 16, 2016 at 3:34 PM, Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code?
I've
Post by Adrian Musceac
been trying
to optimize modems running at 1200 bits/second and the
best
Post by Adrian Musceac
I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB
above for
GMSK. Your modem seems to be 16-20 dB better than that!
If I
Post by Adrian Musceac
could see
the Gnuradio graph maybe I could figure out what I'm
doing
Post by Adrian Musceac
wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here
that
Post by Adrian Musceac
might make a good
Post by Steve
transmitter for it.
On Wed, Mar 16, 2016 at 4:02 AM, glen english
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and
importing that into
Post by Steve
audacity.
On Wed, Mar 16, 2016 at 3:46 AM, glen english
Hi Steve
Nice
is that through a PA- if so, linear or linear in say
3dB compression?
Post by Steve
Thought I'd upload a spectrum snapshot. You can't
really see the
Post by Steve
carriers
until you drop the number of samples to about 256,
but
Post by Adrian Musceac
this shows how
Post by Steve
wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0
signals and a couple
Post by Steve
of clips going on. Maybe need to knock it down to
+/-
Post by Adrian Musceac
.5 like the old
Post by Steve
FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
Post by David Rowe
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
Post by David Rowe
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in
San
Post by Adrian Musceac
Francisco, CA to explore cutting-edge tech and listen to tech
luminaries
Post by Adrian Musceac
present their vision of the future. This family event has something
for
Post by Adrian Musceac
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech
luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
--
Sent from my Android device with K-9 Mail. Please excuse my brevity.
------------------------------------------------------------------------------
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
David Rowe
2016-07-05 20:47:28 UTC
Permalink
Hi Adrian,

Sure, it's not always possible to get the output we need from the tools
we are using.

However the main point I was making is to compare your GR results to a
known baseline, like the theoretical Eb/No versus BER performance of the
modem you are simulating. OK to use another tool for that (e.g. a
spreadsheet or pencil and paper).

Re filter design it may affect performance of non-ideal modems, like
those that pass symbols through an analog modem (e.g. legacy FM radios),
but filter design won't impact ideal modem performance. Ideal modem
performance is set by the laws of physics. Fortunately we can get to
within a few fractions of a dB with real world implementations - if we
choose the right modem design.

Cheers,

David
Post by Adrian Musceac
Hi David,
I wish I was able to generate BER plots similar to those in your images. Unfortunately, while Gnuradio has powerful real time plotting blocks, I haven't figured out how to use them yet.
For this reason, I have to use two flowgraphs, one for the analog radio and one for the USRP, get the SNR figures from one and use pen and paper to compare with the known bits received on the other. This allows me to change the block parameters live and watch the result. Hopefully some day I will be able to learn Octave.
http://www.ochiailab.dnj.ynu.ac.jp/pdf/tanahashi_tcom06_final.pdf
Cheers,
Adrian
Post by David Rowe
Nice work Adrian. A good way to present this work is plots of BER
against Eb/No for the various modulation schemes you are experimenting
with. See my blog posts on FSK etc for examples.
It's useful to include the theoretical curves as comparisons to your
real world implementations, these are well known for BPSK, FSK, and GMSK.
You can also plot BER against SNR if the bit rate Rb and Noise BW N are
equal. Eb/No is convenient as everything is normalised.
Cheers,
David
Post by Adrian Musceac
Hi,
Some more news from the VHF modem front. First of all, I did some
performance comparisons between GMSK waveforms and regular AFSK as
used
Post by Adrian Musceac
by a typical APRS modem. Obviously, there is a reason why Bell AFSK
was chosen: the Gaussian pulse shaping filter does introduce ISI as
a
Post by Adrian Musceac
cost for reducing the signal bandwidth.
The next question was "how much". Luckily, using the excellent filter
design tool from Gnuradio, we can change the filter parameters and
then
Post by Adrian Musceac
test the result against a channel with AWGN noise. The difference
seems
Post by Adrian Musceac
to be around 2 dB in favour of AFSK.
However, my earlier experiments showed that DBPSK will do even
better.
Post by Adrian Musceac
Unfortunately, sending PSK over a narrow FM channel requires some
tuning
Post by Adrian Musceac
of the modem parameters. 3 dB of improvement at the same BER is not
surprising, what was surprising to me though was the fact that adding
a
Post by Adrian Musceac
pulse shaping filter immediately before the frequency conversion
improved the signal decoding by another 1-2 dB. A 9 dB SNR signal has
the same BER as a 14 dB SNR GMSK modem. At this point we are almost
doing better than analog voice already.
To add further gain we can always decrease the bitrate to 1200 bit/s.
And since I was focused on filters, I had to take another look at the
data. What we want is a signal that can pass 1300 bit/s Codec2
through a
Post by Adrian Musceac
narrow mode FM transceiver (wide mode is beginning to be phased out
at
Post by Adrian Musceac
this point). We still have bandwidth to spare, so it seemed to me
that
Post by Adrian Musceac
using the same filter bandwidth for TX and RX is wasteful. Instead,
the
Post by Adrian Musceac
TX filter is 20% wider than the RX filter for (very little) added
performance.
The bad news is that the new modem is computationally about twice
more
Post by Adrian Musceac
expensive. Using it on my phone will be a big stretch. We will see if
I
Post by Adrian Musceac
can make it work on a 1.2 GHz CPU.
Finally, there are some people out there who criticise Gnuradio for
being slow, bloated and too high level for this kind of stuff. I just
want to say that for me as a student, it is a very valuable tool for
learning. Maybe some day I will join the ranks of the Octave gurus,
but
Post by Adrian Musceac
not today.
http://youtu.be/YOX7H3Fcxe4
Cheers,
Adrian
Hi Brady, David,
I have started transitioning my GRC flowgraphs to C++ and I have
put
Post by Adrian Musceac
the code up on Github, just in case you might find it useful in
your
Post by Adrian Musceac
tests for the 2400B mode. My choice of a modem is AFSK, 2000
bit/sec
Post by Adrian Musceac
(1200 bit voice or 1300 bit data). It has no FEC, no scrambling
and
Post by Adrian Musceac
no interleaving (yet). The audio is framed for the standard 40
msec
Post by Adrian Musceac
Codec2. It works through a handheld FM transceiver and can be
decoded easily with Gqrx.
The perfomance is 5 dB worse than analog FM, measured for 12 dB
SNR.
Post by Adrian Musceac
I would be very curious to see a comparison with 2400B. My next
step
Post by Adrian Musceac
is to transition the DBPSK modem which is 3 dB worse than analog
FM
Post by Adrian Musceac
based on my USRP captured data.
https://github.com/kantooon/gnuradio_audio_modems
Post by Adrian Musceac
Video of operation: http://youtu.be/2FZBQvVm5Co
Cheers,
Adrian (YO8RZZ)
On Thu, Mar 17, 2016 at 12:04 AM, Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are
attached. These grc pipelines are only meant to get 16-bit
48k
Post by Adrian Musceac
samples in and out of the HackRF and rtlsdr for testing. The
meat of the FSK is done with the C fsk modem, used in
freedv_rx,
Post by Adrian Musceac
freedv_tx, fsk_mod, and fsk_demod. We are also running our
FSK
Post by Adrian Musceac
modem directly from SDR samples instead of through FM radios.
We
Post by Adrian Musceac
will be running 2400B over legacy FM radios, but don't yet
have
Post by Adrian Musceac
the same degree of test coverage.
Thanks,
Brady O'Brien
On Wed, Mar 16, 2016 at 3:34 PM, Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code?
I've
Post by Adrian Musceac
been trying
to optimize modems running at 1200 bits/second and the
best
Post by Adrian Musceac
I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB
above for
GMSK. Your modem seems to be 16-20 dB better than that!
If I
Post by Adrian Musceac
could see
the Gnuradio graph maybe I could figure out what I'm
doing
Post by Adrian Musceac
wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here
that
Post by Adrian Musceac
might make a good
Post by Steve
transmitter for it.
On Wed, Mar 16, 2016 at 4:02 AM, glen english
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and
importing that into
Post by Steve
audacity.
On Wed, Mar 16, 2016 at 3:46 AM, glen english
Hi Steve
Nice
is that through a PA- if so, linear or linear in say
3dB compression?
Post by Steve
Thought I'd upload a spectrum snapshot. You can't
really see the
Post by Steve
carriers
until you drop the number of samples to about 256,
but
Post by Adrian Musceac
this shows how
Post by Steve
wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0
signals and a couple
Post by Steve
of clips going on. Maybe need to knock it down to
+/-
Post by Adrian Musceac
.5 like the old
Post by Steve
FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in
San
Post by Adrian Musceac
Francisco, CA to explore cutting-edge tech and listen to tech
luminaries
Post by Adrian Musceac
present their vision of the future. This family event has something
for
Post by Adrian Musceac
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech
luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Adrian Musceac
2016-07-06 19:20:52 UTC
Permalink
Hi Brady,

Regarding using Python scripts instead of GRC, I am aware of the power of
Python, I use it daily, but I still find the "visual" design tools (like
the filter designer and the embedded docs) much more easy when prototyping.
My usual workflow is to come up with a flowgraph with labels, buttons,
sliders and GUI sinks using GRC, and when I am ready to test live (for
example on my phone) I translate the graph to C++ (by replacing certain
blocks with their C++ counterparts if necessary). It is very easy and
powerful. Also allows me to switch on the fly between an audio sink to a
USRP sink at the press of a button.

David, regarding the actual numbers, I think that I might create a "mega
graph" which calculates all the numbers and displays them along with BER
figures live, while you send data over radio and adjust sliders for various
values. I like seeing the numbers change even as I run the flowgraph (I
have two soundcards, one of them for the radio and the other for
monitoring) and the USRP and SDR stick are connected via USB but to a
different graph. I would like to unify that if possible, but I still
struggle with certain blocks, like the trellis-viterbi, FEC or BER tools.

Regarding theory versus practice, I am not trying to compare my numbers
with theory, because I can do that with a real SDR connected. Instead, I am
very interested in this idea of using a cheap FM handheld as a data modem.
There are various challenges which make it a completely different exercise
compared with SDRs.
For example, the pre-emphasis and de-emphasis slopes can sometimes be
different between radios and this will affect the bit stream, or you can
get frequency or phase errors introduced by various components of the
chain, radios might not be tuned to the exact same frequency, you symbol
rate might actually not be exactly the same due to CPU limitations and you
have to correct all that. I find this a very challenging exercise.

I have read your blog posts about the 2400B modem, but I would really like
to understand more, like what kind of filters do you use for pulse shaping
and how do they affect performance/bandwidth, how do you track the carrier
frequency and correct frequency and phase errors. This does not seem to be
explained anywhere (I know I could just look at the code, but reading an
explanation seems easier :) )
Also I have some questions regarding your numbers, like what figures do you
use for channel bandwidth (Eb/No), given that there are actually two
channels, one through the audio path of the radio and the other one
whatever width you set your filters to.

Thanks anyway for making all this possible. If I had not read your blog
posts, I probably would never have become interested in this subject.

Cheers,
Adrian
Post by David Rowe
Hi Adrian,
Sure, it's not always possible to get the output we need from the tools
we are using.
However the main point I was making is to compare your GR results to a
known baseline, like the theoretical Eb/No versus BER performance of the
modem you are simulating. OK to use another tool for that (e.g. a
spreadsheet or pencil and paper).
Re filter design it may affect performance of non-ideal modems, like
those that pass symbols through an analog modem (e.g. legacy FM radios),
but filter design won't impact ideal modem performance. Ideal modem
performance is set by the laws of physics. Fortunately we can get to
within a few fractions of a dB with real world implementations - if we
choose the right modem design.
Cheers,
David
Post by Adrian Musceac
Hi David,
I wish I was able to generate BER plots similar to those in your images.
Unfortunately, while Gnuradio has powerful real time plotting blocks, I
haven't figured out how to use them yet.
Post by Adrian Musceac
For this reason, I have to use two flowgraphs, one for the analog radio
and one for the USRP, get the SNR figures from one and use pen and paper to
compare with the known bits received on the other. This allows me to change
the block parameters live and watch the result. Hopefully some day I will
be able to learn Octave.
Post by Adrian Musceac
For reference, the pulse shaping filters I am experimenting with are
http://www.ochiailab.dnj.ynu.ac.jp/pdf/tanahashi_tcom06_final.pdf
Cheers,
Adrian
Post by David Rowe
Nice work Adrian. A good way to present this work is plots of BER
against Eb/No for the various modulation schemes you are experimenting
with. See my blog posts on FSK etc for examples.
It's useful to include the theoretical curves as comparisons to your
real world implementations, these are well known for BPSK, FSK, and GMSK.
You can also plot BER against SNR if the bit rate Rb and Noise BW N are
equal. Eb/No is convenient as everything is normalised.
Cheers,
David
Post by Adrian Musceac
Hi,
Some more news from the VHF modem front. First of all, I did some
performance comparisons between GMSK waveforms and regular AFSK as
used
Post by Adrian Musceac
by a typical APRS modem. Obviously, there is a reason why Bell AFSK
was chosen: the Gaussian pulse shaping filter does introduce ISI as
a
Post by Adrian Musceac
cost for reducing the signal bandwidth.
The next question was "how much". Luckily, using the excellent filter
design tool from Gnuradio, we can change the filter parameters and
then
Post by Adrian Musceac
test the result against a channel with AWGN noise. The difference
seems
Post by Adrian Musceac
to be around 2 dB in favour of AFSK.
However, my earlier experiments showed that DBPSK will do even
better.
Post by Adrian Musceac
Unfortunately, sending PSK over a narrow FM channel requires some
tuning
Post by Adrian Musceac
of the modem parameters. 3 dB of improvement at the same BER is not
surprising, what was surprising to me though was the fact that adding
a
Post by Adrian Musceac
pulse shaping filter immediately before the frequency conversion
improved the signal decoding by another 1-2 dB. A 9 dB SNR signal has
the same BER as a 14 dB SNR GMSK modem. At this point we are almost
doing better than analog voice already.
To add further gain we can always decrease the bitrate to 1200 bit/s.
And since I was focused on filters, I had to take another look at the
data. What we want is a signal that can pass 1300 bit/s Codec2
through a
Post by Adrian Musceac
narrow mode FM transceiver (wide mode is beginning to be phased out
at
Post by Adrian Musceac
this point). We still have bandwidth to spare, so it seemed to me
that
Post by Adrian Musceac
using the same filter bandwidth for TX and RX is wasteful. Instead,
the
Post by Adrian Musceac
TX filter is 20% wider than the RX filter for (very little) added
performance.
The bad news is that the new modem is computationally about twice
more
Post by Adrian Musceac
expensive. Using it on my phone will be a big stretch. We will see if
I
Post by Adrian Musceac
can make it work on a 1.2 GHz CPU.
Finally, there are some people out there who criticise Gnuradio for
being slow, bloated and too high level for this kind of stuff. I just
want to say that for me as a student, it is a very valuable tool for
learning. Maybe some day I will join the ranks of the Octave gurus,
but
Post by Adrian Musceac
not today.
http://youtu.be/YOX7H3Fcxe4
Cheers,
Adrian
Hi Brady, David,
I have started transitioning my GRC flowgraphs to C++ and I have
put
Post by Adrian Musceac
the code up on Github, just in case you might find it useful in
your
Post by Adrian Musceac
tests for the 2400B mode. My choice of a modem is AFSK, 2000
bit/sec
Post by Adrian Musceac
(1200 bit voice or 1300 bit data). It has no FEC, no scrambling
and
Post by Adrian Musceac
no interleaving (yet). The audio is framed for the standard 40
msec
Post by Adrian Musceac
Codec2. It works through a handheld FM transceiver and can be
decoded easily with Gqrx.
The perfomance is 5 dB worse than analog FM, measured for 12 dB
SNR.
Post by Adrian Musceac
I would be very curious to see a comparison with 2400B. My next
step
Post by Adrian Musceac
is to transition the DBPSK modem which is 3 dB worse than analog
FM
Post by Adrian Musceac
based on my USRP captured data.
https://github.com/kantooon/gnuradio_audio_modems
Post by Adrian Musceac
Video of operation: http://youtu.be/2FZBQvVm5Co
Cheers,
Adrian (YO8RZZ)
On Thu, Mar 17, 2016 at 12:04 AM, Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are
attached. These grc pipelines are only meant to get 16-bit
48k
Post by Adrian Musceac
samples in and out of the HackRF and rtlsdr for testing. The
meat of the FSK is done with the C fsk modem, used in
freedv_rx,
Post by Adrian Musceac
freedv_tx, fsk_mod, and fsk_demod. We are also running our
FSK
Post by Adrian Musceac
modem directly from SDR samples instead of through FM radios.
We
Post by Adrian Musceac
will be running 2400B over legacy FM radios, but don't yet
have
Post by Adrian Musceac
the same degree of test coverage.
Thanks,
Brady O'Brien
On Wed, Mar 16, 2016 at 3:34 PM, Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code?
I've
Post by Adrian Musceac
been trying
to optimize modems running at 1200 bits/second and the
best
Post by Adrian Musceac
I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB
above for
GMSK. Your modem seems to be 16-20 dB better than that!
If I
Post by Adrian Musceac
could see
the Gnuradio graph maybe I could figure out what I'm
doing
Post by Adrian Musceac
wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here
that
Post by Adrian Musceac
might make a good
Post by Steve
transmitter for it.
On Wed, Mar 16, 2016 at 4:02 AM, glen english
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and
importing that into
Post by Steve
audacity.
On Wed, Mar 16, 2016 at 3:46 AM, glen english
Hi Steve
Nice
is that through a PA- if so, linear or linear in say
3dB compression?
Post by Steve
Thought I'd upload a spectrum snapshot. You can't
really see the
Post by Steve
carriers
until you drop the number of samples to about 256,
but
Post by Adrian Musceac
this shows how
Post by Steve
wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0
signals and a couple
Post by Steve
of clips going on. Maybe need to knock it down to
+/-
Post by Adrian Musceac
.5 like the old
Post by Steve
FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
Post by Adrian Musceac
Post by David Rowe
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
Post by Adrian Musceac
Post by David Rowe
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in
San
Post by Adrian Musceac
Francisco, CA to explore cutting-edge tech and listen to tech
luminaries
Post by Adrian Musceac
present their vision of the future. This family event has something
for
Post by Adrian Musceac
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Jeroen Vreeken
2016-07-06 21:38:53 UTC
Permalink
Hi Adrian,
Post by Adrian Musceac
I have read your blog posts about the 2400B modem, but I would really
like to understand more, like what kind of filters do you use for
pulse shaping and how do they affect performance/bandwidth, how do you
track the carrier frequency and correct frequency and phase errors.
This does not seem to be explained anywhere (I know I could just look
at the code, but reading an explanation seems easier :) )
I can help you with that one (I was curious as well): The waveform is
just a square manchester encoding. Any filtering will come from the
transceivers.
I have not yet tested it, but my gut feeling says some pulse shaping
(even just a sine table) would probably improve the error rate a bit.

Regards,
Jeroen
David Rowe
2016-07-07 00:55:44 UTC
Permalink
Hi Jeroen,

Pls check that gut feeling abt pulse shaping very carefully - many ideal
demods use "integrate and dump" filters - and IIRC Brady obtained some
surprisingly good Eb/No versus BER results for the 2400B modem.

Indeed, the pulse shaping filters used on DMR/C4FM and friends seem to
be responisble for the drop in performance compared to ideal FSK modems.

Adrian - the 2400B modem is designed to deal with the issues you
mentioned - i.e. to run on $40 FM HTs. Could use some more
characterization I guess. Re one of your questions - the use of "No"
(spectral noise density) takes away the issue of "noise bandwidth" - one
neat reason for using Eb/No.

Cheers,

David
Post by David Rowe
Hi Adrian,
Post by Adrian Musceac
I have read your blog posts about the 2400B modem, but I would really
like to understand more, like what kind of filters do you use for
pulse shaping and how do they affect performance/bandwidth, how do you
track the carrier frequency and correct frequency and phase errors.
This does not seem to be explained anywhere (I know I could just look
at the code, but reading an explanation seems easier :) )
I can help you with that one (I was curious as well): The waveform is
just a square manchester encoding. Any filtering will come from the
transceivers.
I have not yet tested it, but my gut feeling says some pulse shaping
(even just a sine table) would probably improve the error rate a bit.
Regards,
Jeroen
------------------------------------------------------------------------------
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
glen english
2016-07-07 01:04:12 UTC
Permalink
agreed.
the only way these DMR/ C4FM mod schemes fit into a channel is because
they are aggressively filtered...
Post by David Rowe
Hi Jeroen,
Pls check that gut feeling abt pulse shaping very carefully - many ideal
demods use "integrate and dump" filters - and IIRC Brady obtained some
surprisingly good Eb/No versus BER results for the 2400B modem.
Indeed, the pulse shaping filters used on DMR/C4FM and friends seem to
be responisble for the drop in performance compared to ideal FSK modems.
Adrian - the 2400B modem is designed to deal with the issues you
mentioned - i.e. to run on $40 FM HTs. Could use some more
characterization I guess. Re one of your questions - the use of "No"
(spectral noise density) takes away the issue of "noise bandwidth" - one
neat reason for using Eb/No.
Cheers,
David
Post by David Rowe
Hi Adrian,
Post by Adrian Musceac
I have read your blog posts about the 2400B modem, but I would really
like to understand more, like what kind of filters do you use for
pulse shaping and how do they affect performance/bandwidth, how do you
track the carrier frequency and correct frequency and phase errors.
This does not seem to be explained anywhere (I know I could just look
at the code, but reading an explanation seems easier :) )
I can help you with that one (I was curious as well): The waveform is
just a square manchester encoding. Any filtering will come from the
transceivers.
I have not yet tested it, but my gut feeling says some pulse shaping
(even just a sine table) would probably improve the error rate a bit.
Regards,
Jeroen
------------------------------------------------------------------------------
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Brady O'Brien
2016-07-07 00:33:27 UTC
Permalink
Hi Adrian,

About 2400B:

-Pulse Shaping: None on the TX end, it's just a manchester encoded bit
stream
-RX Symbol Filtering: Just a simple integrate-and-dump aligned with the
symbols
-Frequency tracking: None.
-Phase tracking and clock offset correction: This is done by the fine
timing estimator

The job of the fine timing estimator is to estimate the location of symbols
in the incoming sample stream. It first takes the incoming sample stream
and applies a nonlinarity, in our case s^2. This causes a line to appear in
the frequency domain of the incoming signal. We then extract the average
phase of this nonlinearity over the buffer of samples. This phase is
proportional to the offset of the symbols within the sample stream. This is
all done on lines 186 to 218 of fmfsk.c . In order to handle clock offset,
the demod may consume a variable number of samples for a fixed number of
bits. A sample clock offset can be though of as a drift in the symbol
offset over time. To compensate for this, the demod tries to keep the
symbol offset within a window. If the symbol offset is outside of this
window, it will change the number of samples consumed in the next frame to
either skip or re-process a few samples. This is done in lines 233 to 242
of fmfsk.c.

Thanks,

Brady O'Brien
Post by Adrian Musceac
Hi Brady,
Regarding using Python scripts instead of GRC, I am aware of the power of
Python, I use it daily, but I still find the "visual" design tools (like
the filter designer and the embedded docs) much more easy when prototyping.
My usual workflow is to come up with a flowgraph with labels, buttons,
sliders and GUI sinks using GRC, and when I am ready to test live (for
example on my phone) I translate the graph to C++ (by replacing certain
blocks with their C++ counterparts if necessary). It is very easy and
powerful. Also allows me to switch on the fly between an audio sink to a
USRP sink at the press of a button.
David, regarding the actual numbers, I think that I might create a "mega
graph" which calculates all the numbers and displays them along with BER
figures live, while you send data over radio and adjust sliders for various
values. I like seeing the numbers change even as I run the flowgraph (I
have two soundcards, one of them for the radio and the other for
monitoring) and the USRP and SDR stick are connected via USB but to a
different graph. I would like to unify that if possible, but I still
struggle with certain blocks, like the trellis-viterbi, FEC or BER tools.
Regarding theory versus practice, I am not trying to compare my numbers
with theory, because I can do that with a real SDR connected. Instead, I am
very interested in this idea of using a cheap FM handheld as a data modem.
There are various challenges which make it a completely different exercise
compared with SDRs.
For example, the pre-emphasis and de-emphasis slopes can sometimes be
different between radios and this will affect the bit stream, or you can
get frequency or phase errors introduced by various components of the
chain, radios might not be tuned to the exact same frequency, you symbol
rate might actually not be exactly the same due to CPU limitations and you
have to correct all that. I find this a very challenging exercise.
I have read your blog posts about the 2400B modem, but I would really like
to understand more, like what kind of filters do you use for pulse shaping
and how do they affect performance/bandwidth, how do you track the carrier
frequency and correct frequency and phase errors. This does not seem to be
explained anywhere (I know I could just look at the code, but reading an
explanation seems easier :) )
Also I have some questions regarding your numbers, like what figures do
you use for channel bandwidth (Eb/No), given that there are actually two
channels, one through the audio path of the radio and the other one
whatever width you set your filters to.
Thanks anyway for making all this possible. If I had not read your blog
posts, I probably would never have become interested in this subject.
Cheers,
Adrian
Post by David Rowe
Hi Adrian,
Sure, it's not always possible to get the output we need from the tools
we are using.
However the main point I was making is to compare your GR results to a
known baseline, like the theoretical Eb/No versus BER performance of the
modem you are simulating. OK to use another tool for that (e.g. a
spreadsheet or pencil and paper).
Re filter design it may affect performance of non-ideal modems, like
those that pass symbols through an analog modem (e.g. legacy FM radios),
but filter design won't impact ideal modem performance. Ideal modem
performance is set by the laws of physics. Fortunately we can get to
within a few fractions of a dB with real world implementations - if we
choose the right modem design.
Cheers,
David
Post by Adrian Musceac
Hi David,
I wish I was able to generate BER plots similar to those in your
images. Unfortunately, while Gnuradio has powerful real time plotting
blocks, I haven't figured out how to use them yet.
Post by Adrian Musceac
For this reason, I have to use two flowgraphs, one for the analog radio
and one for the USRP, get the SNR figures from one and use pen and paper to
compare with the known bits received on the other. This allows me to change
the block parameters live and watch the result. Hopefully some day I will
be able to learn Octave.
Post by Adrian Musceac
For reference, the pulse shaping filters I am experimenting with are
http://www.ochiailab.dnj.ynu.ac.jp/pdf/tanahashi_tcom06_final.pdf
Cheers,
Adrian
Post by David Rowe
Nice work Adrian. A good way to present this work is plots of BER
against Eb/No for the various modulation schemes you are experimenting
with. See my blog posts on FSK etc for examples.
It's useful to include the theoretical curves as comparisons to your
real world implementations, these are well known for BPSK, FSK, and GMSK.
You can also plot BER against SNR if the bit rate Rb and Noise BW N are
equal. Eb/No is convenient as everything is normalised.
Cheers,
David
Post by Adrian Musceac
Hi,
Some more news from the VHF modem front. First of all, I did some
performance comparisons between GMSK waveforms and regular AFSK as
used
Post by Adrian Musceac
by a typical APRS modem. Obviously, there is a reason why Bell AFSK
was chosen: the Gaussian pulse shaping filter does introduce ISI as
a
Post by Adrian Musceac
cost for reducing the signal bandwidth.
The next question was "how much". Luckily, using the excellent filter
design tool from Gnuradio, we can change the filter parameters and
then
Post by Adrian Musceac
test the result against a channel with AWGN noise. The difference
seems
Post by Adrian Musceac
to be around 2 dB in favour of AFSK.
However, my earlier experiments showed that DBPSK will do even
better.
Post by Adrian Musceac
Unfortunately, sending PSK over a narrow FM channel requires some
tuning
Post by Adrian Musceac
of the modem parameters. 3 dB of improvement at the same BER is not
surprising, what was surprising to me though was the fact that adding
a
Post by Adrian Musceac
pulse shaping filter immediately before the frequency conversion
improved the signal decoding by another 1-2 dB. A 9 dB SNR signal has
the same BER as a 14 dB SNR GMSK modem. At this point we are almost
doing better than analog voice already.
To add further gain we can always decrease the bitrate to 1200 bit/s.
And since I was focused on filters, I had to take another look at the
data. What we want is a signal that can pass 1300 bit/s Codec2
through a
Post by Adrian Musceac
narrow mode FM transceiver (wide mode is beginning to be phased out
at
Post by Adrian Musceac
this point). We still have bandwidth to spare, so it seemed to me
that
Post by Adrian Musceac
using the same filter bandwidth for TX and RX is wasteful. Instead,
the
Post by Adrian Musceac
TX filter is 20% wider than the RX filter for (very little) added
performance.
The bad news is that the new modem is computationally about twice
more
Post by Adrian Musceac
expensive. Using it on my phone will be a big stretch. We will see if
I
Post by Adrian Musceac
can make it work on a 1.2 GHz CPU.
Finally, there are some people out there who criticise Gnuradio for
being slow, bloated and too high level for this kind of stuff. I just
want to say that for me as a student, it is a very valuable tool for
learning. Maybe some day I will join the ranks of the Octave gurus,
but
Post by Adrian Musceac
not today.
http://youtu.be/YOX7H3Fcxe4
Cheers,
Adrian
Hi Brady, David,
I have started transitioning my GRC flowgraphs to C++ and I have
put
Post by Adrian Musceac
the code up on Github, just in case you might find it useful in
your
Post by Adrian Musceac
tests for the 2400B mode. My choice of a modem is AFSK, 2000
bit/sec
Post by Adrian Musceac
(1200 bit voice or 1300 bit data). It has no FEC, no scrambling
and
Post by Adrian Musceac
no interleaving (yet). The audio is framed for the standard 40
msec
Post by Adrian Musceac
Codec2. It works through a handheld FM transceiver and can be
decoded easily with Gqrx.
The perfomance is 5 dB worse than analog FM, measured for 12 dB
SNR.
Post by Adrian Musceac
I would be very curious to see a comparison with 2400B. My next
step
Post by Adrian Musceac
is to transition the DBPSK modem which is 3 dB worse than analog
FM
Post by Adrian Musceac
based on my USRP captured data.
https://github.com/kantooon/gnuradio_audio_modems
Post by Adrian Musceac
Video of operation: http://youtu.be/2FZBQvVm5Co
Cheers,
Adrian (YO8RZZ)
On Thu, Mar 17, 2016 at 12:04 AM, Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test are
attached. These grc pipelines are only meant to get 16-bit
48k
Post by Adrian Musceac
samples in and out of the HackRF and rtlsdr for testing. The
meat of the FSK is done with the C fsk modem, used in
freedv_rx,
Post by Adrian Musceac
freedv_tx, fsk_mod, and fsk_demod. We are also running our
FSK
Post by Adrian Musceac
modem directly from SDR samples instead of through FM radios.
We
Post by Adrian Musceac
will be running 2400B over legacy FM radios, but don't yet
have
Post by Adrian Musceac
the same degree of test coverage.
Thanks,
Brady O'Brien
On Wed, Mar 16, 2016 at 3:34 PM, Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio code?
I've
Post by Adrian Musceac
been trying
to optimize modems running at 1200 bits/second and the
best
Post by Adrian Musceac
I could do
was 100% copy at 8 dB above analog FM for DBPSK and 12 dB
above for
GMSK. Your modem seems to be 16-20 dB better than that!
If I
Post by Adrian Musceac
could see
the Gnuradio graph maybe I could figure out what I'm
doing
Post by Adrian Musceac
wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around here
that
Post by Adrian Musceac
might make a good
Post by Steve
transmitter for it.
On Wed, Mar 16, 2016 at 4:02 AM, glen english
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48 kHz and
importing that into
Post by Steve
audacity.
On Wed, Mar 16, 2016 at 3:46 AM, glen english
Hi Steve
Nice
is that through a PA- if so, linear or linear in say
3dB compression?
Post by Steve
Thought I'd upload a spectrum snapshot. You can't
really see the
Post by Steve
carriers
until you drop the number of samples to about 256,
but
Post by Adrian Musceac
this shows how
Post by Steve
wide
it is.
I noticed that the freedv_tx was outputting +/- 1.0
signals and a couple
Post by Steve
of clips going on. Maybe need to knock it down to
+/-
Post by Adrian Musceac
.5 like the old
Post by Steve
FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
Post by Adrian Musceac
Post by David Rowe
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
Post by Adrian Musceac
Post by David Rowe
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in
San
Post by Adrian Musceac
Francisco, CA to explore cutting-edge tech and listen to tech
luminaries
Post by Adrian Musceac
present their vision of the future. This family event has something
for
Post by Adrian Musceac
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Adrian Musceac
2016-07-09 19:06:58 UTC
Permalink
Hi Brady,

Thank you very much for the detailed.description of the code. This was very helpful. It appears that there is no difference between 2400b and my modem except for the pulse shaping filters.
I think the performance should also.be very similar. Unfortunately I no longer have the time to keep playing with this.
Eagerly awaiting the sm2000, I keep hitting refresh every day on David's website :)

@chris ,
Adrian
Post by David Rowe
Hi Adrian,
-Pulse Shaping: None on the TX end, it's just a manchester encoded bit
stream
-RX Symbol Filtering: Just a simple integrate-and-dump aligned with the
symbols
-Frequency tracking: None.
-Phase tracking and clock offset correction: This is done by the fine
timing estimator
The job of the fine timing estimator is to estimate the location of symbols
in the incoming sample stream. It first takes the incoming sample stream
and applies a nonlinarity, in our case s^2. This causes a line to appear in
the frequency domain of the incoming signal. We then extract the average
phase of this nonlinearity over the buffer of samples. This phase is
proportional to the offset of the symbols within the sample stream. This is
all done on lines 186 to 218 of fmfsk.c . In order to handle clock offset,
the demod may consume a variable number of samples for a fixed number of
bits. A sample clock offset can be though of as a drift in the symbol
offset over time. To compensate for this, the demod tries to keep the
symbol offset within a window. If the symbol offset is outside of this
window, it will change the number of samples consumed in the next frame to
either skip or re-process a few samples. This is done in lines 233 to 242
of fmfsk.c.
Thanks,
Brady O'Brien
Post by Adrian Musceac
Hi Brady,
Regarding using Python scripts instead of GRC, I am aware of the
power of
Post by Adrian Musceac
Python, I use it daily, but I still find the "visual" design tools
(like
Post by Adrian Musceac
the filter designer and the embedded docs) much more easy when
prototyping.
Post by Adrian Musceac
My usual workflow is to come up with a flowgraph with labels,
buttons,
Post by Adrian Musceac
sliders and GUI sinks using GRC, and when I am ready to test live
(for
Post by Adrian Musceac
example on my phone) I translate the graph to C++ (by replacing
certain
Post by Adrian Musceac
blocks with their C++ counterparts if necessary). It is very easy and
powerful. Also allows me to switch on the fly between an audio sink
to a
Post by Adrian Musceac
USRP sink at the press of a button.
David, regarding the actual numbers, I think that I might create a
"mega
Post by Adrian Musceac
graph" which calculates all the numbers and displays them along with
BER
Post by Adrian Musceac
figures live, while you send data over radio and adjust sliders for
various
Post by Adrian Musceac
values. I like seeing the numbers change even as I run the flowgraph
(I
Post by Adrian Musceac
have two soundcards, one of them for the radio and the other for
monitoring) and the USRP and SDR stick are connected via USB but to a
different graph. I would like to unify that if possible, but I still
struggle with certain blocks, like the trellis-viterbi, FEC or BER
tools.
Post by Adrian Musceac
Regarding theory versus practice, I am not trying to compare my
numbers
Post by Adrian Musceac
with theory, because I can do that with a real SDR connected.
Instead, I am
Post by Adrian Musceac
very interested in this idea of using a cheap FM handheld as a data
modem.
Post by Adrian Musceac
There are various challenges which make it a completely different
exercise
Post by Adrian Musceac
compared with SDRs.
For example, the pre-emphasis and de-emphasis slopes can sometimes be
different between radios and this will affect the bit stream, or you
can
Post by Adrian Musceac
get frequency or phase errors introduced by various components of the
chain, radios might not be tuned to the exact same frequency, you
symbol
Post by Adrian Musceac
rate might actually not be exactly the same due to CPU limitations
and you
Post by Adrian Musceac
have to correct all that. I find this a very challenging exercise.
I have read your blog posts about the 2400B modem, but I would really
like
Post by Adrian Musceac
to understand more, like what kind of filters do you use for pulse
shaping
Post by Adrian Musceac
and how do they affect performance/bandwidth, how do you track the
carrier
Post by Adrian Musceac
frequency and correct frequency and phase errors. This does not seem
to be
Post by Adrian Musceac
explained anywhere (I know I could just look at the code, but reading
an
Post by Adrian Musceac
explanation seems easier :) )
Also I have some questions regarding your numbers, like what figures
do
Post by Adrian Musceac
you use for channel bandwidth (Eb/No), given that there are actually
two
Post by Adrian Musceac
channels, one through the audio path of the radio and the other one
whatever width you set your filters to.
Thanks anyway for making all this possible. If I had not read your
blog
Post by Adrian Musceac
posts, I probably would never have become interested in this subject.
Cheers,
Adrian
Post by David Rowe
Hi Adrian,
Sure, it's not always possible to get the output we need from the
tools
Post by Adrian Musceac
Post by David Rowe
we are using.
However the main point I was making is to compare your GR results to
a
Post by Adrian Musceac
Post by David Rowe
known baseline, like the theoretical Eb/No versus BER performance of
the
Post by Adrian Musceac
Post by David Rowe
modem you are simulating. OK to use another tool for that (e.g. a
spreadsheet or pencil and paper).
Re filter design it may affect performance of non-ideal modems, like
those that pass symbols through an analog modem (e.g. legacy FM
radios),
Post by Adrian Musceac
Post by David Rowe
but filter design won't impact ideal modem performance. Ideal modem
performance is set by the laws of physics. Fortunately we can get
to
Post by Adrian Musceac
Post by David Rowe
within a few fractions of a dB with real world implementations - if
we
Post by Adrian Musceac
Post by David Rowe
choose the right modem design.
Cheers,
David
Post by Adrian Musceac
Hi David,
I wish I was able to generate BER plots similar to those in your
images. Unfortunately, while Gnuradio has powerful real time
plotting
Post by Adrian Musceac
Post by David Rowe
blocks, I haven't figured out how to use them yet.
Post by Adrian Musceac
For this reason, I have to use two flowgraphs, one for the analog
radio
Post by Adrian Musceac
Post by David Rowe
and one for the USRP, get the SNR figures from one and use pen and
paper to
Post by Adrian Musceac
Post by David Rowe
compare with the known bits received on the other. This allows me to
change
Post by Adrian Musceac
Post by David Rowe
the block parameters live and watch the result. Hopefully some day I
will
Post by Adrian Musceac
Post by David Rowe
be able to learn Octave.
Post by Adrian Musceac
For reference, the pulse shaping filters I am experimenting with
are
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
http://www.ochiailab.dnj.ynu.ac.jp/pdf/tanahashi_tcom06_final.pdf
Cheers,
Adrian
Post by David Rowe
Nice work Adrian. A good way to present this work is plots of
BER
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
against Eb/No for the various modulation schemes you are
experimenting
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
with. See my blog posts on FSK etc for examples.
It's useful to include the theoretical curves as comparisons to
your
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
real world implementations, these are well known for BPSK, FSK,
and
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
GMSK.
You can also plot BER against SNR if the bit rate Rb and Noise BW
N are
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
equal. Eb/No is convenient as everything is normalised.
Cheers,
David
Post by Adrian Musceac
Hi,
Some more news from the VHF modem front. First of all, I did
some
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
performance comparisons between GMSK waveforms and regular AFSK
as
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
used
Post by Adrian Musceac
by a typical APRS modem. Obviously, there is a reason why Bell
AFSK
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
was chosen: the Gaussian pulse shaping filter does introduce
ISI as
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
a
Post by Adrian Musceac
cost for reducing the signal bandwidth.
The next question was "how much". Luckily, using the excellent
filter
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
design tool from Gnuradio, we can change the filter parameters
and
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
then
Post by Adrian Musceac
test the result against a channel with AWGN noise. The
difference
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
seems
Post by Adrian Musceac
to be around 2 dB in favour of AFSK.
However, my earlier experiments showed that DBPSK will do even
better.
Post by Adrian Musceac
Unfortunately, sending PSK over a narrow FM channel requires
some
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
tuning
Post by Adrian Musceac
of the modem parameters. 3 dB of improvement at the same BER is
not
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
surprising, what was surprising to me though was the fact that
adding
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
a
Post by Adrian Musceac
pulse shaping filter immediately before the frequency conversion
improved the signal decoding by another 1-2 dB. A 9 dB SNR
signal has
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
the same BER as a 14 dB SNR GMSK modem. At this point we are
almost
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
doing better than analog voice already.
To add further gain we can always decrease the bitrate to 1200
bit/s.
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
And since I was focused on filters, I had to take another look
at the
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
data. What we want is a signal that can pass 1300 bit/s Codec2
through a
Post by Adrian Musceac
narrow mode FM transceiver (wide mode is beginning to be phased
out
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
at
Post by Adrian Musceac
this point). We still have bandwidth to spare, so it seemed to
me
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
that
Post by Adrian Musceac
using the same filter bandwidth for TX and RX is wasteful.
Instead,
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
the
Post by Adrian Musceac
TX filter is 20% wider than the RX filter for (very little)
added
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
performance.
The bad news is that the new modem is computationally about
twice
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
more
Post by Adrian Musceac
expensive. Using it on my phone will be a big stretch. We will
see if
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
I
Post by Adrian Musceac
can make it work on a 1.2 GHz CPU.
Finally, there are some people out there who criticise Gnuradio
for
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
being slow, bloated and too high level for this kind of stuff. I
just
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
want to say that for me as a student, it is a very valuable tool
for
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
learning. Maybe some day I will join the ranks of the Octave
gurus,
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
but
Post by Adrian Musceac
not today.
http://youtu.be/YOX7H3Fcxe4
Cheers,
Adrian
On Fri, Mar 25, 2016 at 10:17 AM, Adrian Musceac
Hi Brady, David,
I have started transitioning my GRC flowgraphs to C++ and I
have
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
put
Post by Adrian Musceac
the code up on Github, just in case you might find it
useful in
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
your
Post by Adrian Musceac
tests for the 2400B mode. My choice of a modem is AFSK,
2000
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
bit/sec
Post by Adrian Musceac
(1200 bit voice or 1300 bit data). It has no FEC, no
scrambling
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
and
Post by Adrian Musceac
no interleaving (yet). The audio is framed for the standard
40
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
msec
Post by Adrian Musceac
Codec2. It works through a handheld FM transceiver and can
be
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
decoded easily with Gqrx.
The perfomance is 5 dB worse than analog FM, measured for
12 dB
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
SNR.
Post by Adrian Musceac
I would be very curious to see a comparison with 2400B. My
next
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
step
Post by Adrian Musceac
is to transition the DBPSK modem which is 3 dB worse than
analog
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
FM
Post by Adrian Musceac
based on my USRP captured data.
https://github.com/kantooon/gnuradio_audio_modems
http://youtu.be/2FZBQvVm5Co
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Cheers,
Adrian (YO8RZZ)
On Thu, Mar 17, 2016 at 12:04 AM, Brady O'Brien
Hey Adrian,
The gnuradio-companion pipelines I used for the test
are
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
attached. These grc pipelines are only meant to get
16-bit
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
48k
Post by Adrian Musceac
samples in and out of the HackRF and rtlsdr for
testing. The
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
meat of the FSK is done with the C fsk modem, used in
freedv_rx,
Post by Adrian Musceac
freedv_tx, fsk_mod, and fsk_demod. We are also running
our
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
FSK
Post by Adrian Musceac
modem directly from SDR samples instead of through FM
radios.
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
We
Post by Adrian Musceac
will be running 2400B over legacy FM radios, but don't
yet
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
have
Post by Adrian Musceac
the same degree of test coverage.
Thanks,
Brady O'Brien
On Wed, Mar 16, 2016 at 3:34 PM, Adrian Musceac
Hi David,
Is it possible to get our hands on the Gnuradio
code?
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
I've
Post by Adrian Musceac
been trying
to optimize modems running at 1200 bits/second and
the
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
best
Post by Adrian Musceac
I could do
was 100% copy at 8 dB above analog FM for DBPSK and
12 dB
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
above for
GMSK. Your modem seems to be 16-20 dB better than
that!
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
If I
Post by Adrian Musceac
could see
the Gnuradio graph maybe I could figure out what
I'm
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
doing
Post by Adrian Musceac
wrong.
Thanks,
Adrian
Post by Steve
Ha, I wish it was through a radio!
I'm trying to think what I have laying around
here
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
that
Post by Adrian Musceac
might make a good
Post by Steve
transmitter for it.
On Wed, Mar 16, 2016 at 4:02 AM, glen english
ah. good stuff. nice progress.
freedv_tx 2400A ve9qrp_10s.raw modem.raw
Which produces a RAW audio file sampled at 48
kHz and
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
importing that into
Post by Steve
audacity.
On Wed, Mar 16, 2016 at 3:46 AM, glen english
Hi Steve
Nice
is that through a PA- if so, linear or linear
in say
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
3dB compression?
Post by Steve
Thought I'd upload a spectrum snapshot. You
can't
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
really see the
Post by Steve
carriers
until you drop the number of samples to about
256,
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
but
Post by Adrian Musceac
this shows how
Post by Steve
wide
it is.
I noticed that the freedv_tx was outputting
+/- 1.0
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
signals and a couple
Post by Steve
of clips going on. Maybe need to knock it down
to
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
+/-
Post by Adrian Musceac
.5 like the old
Post by Steve
FDM
modes.
FYI
Steve
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications
with
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
Post by Adrian Musceac
Post by David Rowe
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications
with
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications
with
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Intel Data Analytics Acceleration Library.
Click to learn
more.http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing
lists.sourceforge.net/lists/listinfo/freetel-codec2
Post by Adrian Musceac
Post by David Rowe
<http://lists.sourceforge.net/lists/listinfo/freetel-codec2>
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Transform Data into Opportunity.
Accelerate data analysis in your applications
with
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by Steve
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
Post by Steve
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
Post by Adrian Musceac
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park
in
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
San
Post by Adrian Musceac
Francisco, CA to explore cutting-edge tech and listen to tech
luminaries
Post by Adrian Musceac
present their vision of the future. This family event has
something
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
for
Post by Adrian Musceac
everyone, including kids. Get more information and register
today.
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park
in San
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
Francisco, CA to explore cutting-edge tech and listen to tech luminaries
present their vision of the future. This family event has
something for
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
everyone, including kids. Get more information and register
today.
Post by Adrian Musceac
Post by David Rowe
Post by Adrian Musceac
Post by David Rowe
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Post by David Rowe
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in
San
Post by Adrian Musceac
Post by David Rowe
Francisco, CA to explore cutting-edge tech and listen to tech
luminaries
Post by Adrian Musceac
Post by David Rowe
present their vision of the future. This family event has something
for
Post by Adrian Musceac
Post by David Rowe
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------------
Post by Adrian Musceac
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in
San
Post by Adrian Musceac
Francisco, CA to explore cutting-edge tech and listen to tech
luminaries
Post by Adrian Musceac
present their vision of the future. This family event has something
for
Post by Adrian Musceac
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
------------------------------------------------------------------------
------------------------------------------------------------------------------
Attend Shape: An AT&T Tech Expo July 15-16. Meet us at AT&T Park in San
Francisco, CA to explore cutting-edge tech and listen to tech
luminaries
present their vision of the future. This family event has something for
everyone, including kids. Get more information and register today.
http://sdm.link/attshape
------------------------------------------------------------------------
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
--
Sent from my Android device with K-9 Mail. Please excuse my brevity.
Tomas Härdin
2016-03-16 08:47:35 UTC
Permalink
Cool. Perhaps something to put my HackRF to use for, a local FreeDV
beacon. OH6ESQ expressed some interest in FreeDV, perhaps he'll be able
to pick it up from across the Baltic sea

/Tomas, SA2TMS
Post by David Rowe
Hello Lists,
We've just tested the new FreeDV 2400A mode designed for VHF and above.
http://www.rowetel.com/blog/?p=5119
Thanks,
David
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Gregory Maxwell
2016-03-18 07:20:53 UTC
Permalink
Post by David Rowe
Hello Lists,
We've just tested the new FreeDV 2400A mode designed for VHF and above.
http://www.rowetel.com/blog/?p=5119
Very interesting, -- is there a reason this isn't using a coherent
demodulator beyond development time? It seems to me that its just
leaving performance on the floor.
glen english
2016-03-18 07:36:47 UTC
Permalink
Hi Greg,
I think one reason would be that a coherent demodulator in a rapid, flat
frequency fading environment is not easy .

especially when you don't have too much horsepower at your fingertips...
(microcontroller)

The coherent demod often will be worse than what you leave on the floor.
Post by Gregory Maxwell
Post by David Rowe
Hello Lists,
We've just tested the new FreeDV 2400A mode designed for VHF and above.
http://www.rowetel.com/blog/?p=5119
Very interesting, -- is there a reason this isn't using a coherent
demodulator beyond development time? It seems to me that its just
leaving performance on the floor.
---------------------------------------------------
Brady O'Brien
2016-03-18 07:41:04 UTC
Permalink
Coherent 4FSK only gives us about another 1 dB of performance in the ideal
case at the cost of increased implementation complexity and probably some
implementation loss.
Post by Gregory Maxwell
Post by David Rowe
Hello Lists,
We've just tested the new FreeDV 2400A mode designed for VHF and above.
http://www.rowetel.com/blog/?p=5119
Very interesting, -- is there a reason this isn't using a coherent
demodulator beyond development time? It seems to me that its just
leaving performance on the floor.
------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785231&iu=/4140
_______________________________________________
Freetel-codec2 mailing list
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Steve
2016-03-18 08:28:20 UTC
Permalink
I think your experiments are interesting, in that you've got data to show.
While we all can talk about theory, and leftovers on the table, it's no
substitute for creating something and documenting it.

It's like an engine though, in that more horsepower isn't always a good
thing.
Post by Brady O'Brien
Coherent 4FSK only gives us about another 1 dB of performance in the ideal
case at the cost of increased implementation complexity and probably some
implementation loss.
Brady O'Brien
2016-03-18 18:18:17 UTC
Permalink
I think in the case of coherent vs non-coherent for 4FSK, the plot of
EbN0(dB) vs BER for ideal coherent and non-coherent says it all. If we find
that we really need that extra dB, It'd be easier to switch to non-coherent
8FSK than write and test coherent 4FSK (and at 8FSK, coherent and
non-coherent are even closer).

​
Loading...