CW Modem #modem


Mark Hubelbank
 

I have been thinking of taking on the challenge of writing a CW decoder having been in the signal processing world for a long time. If this works well, is there interest in this group of integrating it into FLdigi? I would define working well as it should decode 99% of what I can do by ear.


Gene Rozea
 

Fldigi’s integrated CW decoder works really well. Is another one needed?

 

 

From: Mark Hubelbank
Sent: Saturday, December 5, 2020 10:12 AM
To: winfldigi@groups.io
Subject: [winfldigi] CW Modem #modem

 

I have been thinking of taking on the challenge of writing a CW decoder having been in the signal processing world for a long time. If this works well, is there interest in this group of integrating it into FLdigi? I would define working well as it should decode 99% of what I can do by ear.

 


K3eui <K3eui@...>
 

When I listen to CW that is "well-timed"  (like W1AW CW bulletins) the FLDIGI, in CW op mode, copies 100%  on 3581.5 kHz in the evenings.

What more is needed?
Poorly sent CW is easy on my ears, but no program does well turning that into print on a screen.

SENDING  CW
Now FLDIGI can send CW (with my keyboard) by two different methods: each one works fine.  You can key the audio tones (like any sound card mode) and then have your rig in USB or USB-DATA and what comes out to the antenna is keyed RF, just like "real" CW.

Or, put your rig in CW mode, and use FLDIGI to send a HIGH signal to the DTR pin, which inside your (modern) radio, closes the CW keying port to SEND.  That works even better  (no need to monitor ALC).
I have an article in the que in QST that describes both methods, and shows SDR  RF spectra to prove the methods are equally clean, as long as your ALC reads zero in SSB mode.

De   K3eui  Barry



On Dec 5, 2020, at 10:12 AM, Mark Hubelbank <mhubel@...> wrote:

I have been thinking of taking on the challenge of writing a CW decoder having been in the signal processing world for a long time. If this works well, is there interest in this group of integrating it into FLdigi? I would define working well as it should decode 99% of what I can do by ear.


W8DU Arnie <w8du@...>
 

I have been playing with this for a few weeks now, more as something to do in COVID time than out of necessity. It is fun, but why re-invite the entire wheel?
This one has been fantastic and the author has been very responsive and available. Works and looks great:
(309) STM32F103 (BluePill) CW Decoder - YouTube

You might also want to check out Budd Churchward's CW decoder that uses a 567 tone decoder instead of the Goertzel algorithm. You can find his excellent
youtubes as well. Just search for morseduino or WB7FHC.
Good luck and let us know how you fare.
73 de Arnie W8DU


Bob McGraw - K4TAX <rmcgraw@...>
 

Barry:

There are efforts on-going by Dave, W1HKS, to provide other means to send CW via the keyboard.   The brands and models of radios is being expanded.   Presently testing is underway for this project.

While using a SSB mode does send CW simply by supplying an audio tone to the radio, one must be extremely careful so as not to have any humm or noise embedded with the tone.   Be sure the computer sounds are turned OFF and be sure the microphone is not hot to prevent room sounds in teh CW band.    The FCC frowns on this and the VM program will pick up on this and file a report on the station operation.   Also, the actual frequency transmitted will be carrier frequency plus or minus, depending USB or LSB mode, and the audio frequency.   Hence if the radio is on 7.025 and the tone is 1 kHz, the transmitted frequency will be either 7.024 or 7.026 depending sideband selection.  Also one must avoid any ALC action or speech processing or equalization in the audio path, either inside the radio or external to the radio. 

As to DATA modes, usually these are analog audio paths and usually the radio will remove any processing or EQ.  Those however do not relieve the operator from correct levels and avoiding any ALC action.   As as rule and convention, CW is considered a LSB mode while DATA modes are considered USB modes,  regardless of band convention. 

Sending true and pure CW from FLDIGI from the keyboard to the radio is very nice.   Timing is correct, character length is correct as well as spacing.   As to hand sent CW, some of my regulars I find with hand sent CW is near 100 % copy while others are very difficult to copy due to their "fist". 

73

Bob, K4TAX


On 12/5/2020 9:58 AM, K3eui via groups.io wrote:

When I listen to CW that is "well-timed"  (like W1AW CW bulletins) the FLDIGI, in CW op mode, copies 100%  on 3581.5 kHz in the evenings.

What more is needed?
Poorly sent CW is easy on my ears, but no program does well turning that into print on a screen.

SENDING  CW
Now FLDIGI can send CW (with my keyboard) by two different methods: each one works fine.  You can key the audio tones (like any sound card mode) and then have your rig in USB or USB-DATA and what comes out to the antenna is keyed RF, just like "real" CW.

Or, put your rig in CW mode, and use FLDIGI to send a HIGH signal to the DTR pin, which inside your (modern) radio, closes the CW keying port to SEND.  That works even better  (no need to monitor ALC).
I have an article in the que in QST that describes both methods, and shows SDR  RF spectra to prove the methods are equally clean, as long as your ALC reads zero in SSB mode.

De   K3eui  Barry



On Dec 5, 2020, at 10:12 AM, Mark Hubelbank <mhubel@...> wrote:

I have been thinking of taking on the challenge of writing a CW decoder having been in the signal processing world for a long time. If this works well, is there interest in this group of integrating it into FLdigi? I would define working well as it should decode 99% of what I can do by ear.


Mark Hubelbank
 

Well it seems that there is little interest. I find all the CW decoders out there lacking. Even the best make a lot more errors on real off the air poorly formed signals than I do and I am no so good anymore. I can't really keep up with any off the air signals for more than a few minutes. Machines don't get tired.
This effort is strictly for decoding, not sending.
If I do this, I will still post the results.

On 12/5/2020 10:12 AM, Mark Hubelbank wrote:
I have been thinking of taking on the challenge of writing a CW decoder having been in the signal processing world for a long time. If this works well, is there interest in this group of integrating it into FLdigi? I would define working well as it should decode 99% of what I can do by ear.

-- 
Mark Hubelbank
NorthEast Monitoring
141 Parker St
Suite 200
Maynard, MA, 01754 - USA
mhubel@...
978-443-3955


Joe Hutchens
 

I've heard a few stations sending "touch-tone" CW (multiple tones), and I suspect they're running tones directly into their mic jack.  I've sent a few of them a note asking how their station is configured for sending CW, but haven't received any replies. I've given up trying to find their trick...

I send Fldigi CW tones to a homebrew interface that works nicely.  If interested, look up:

( http://www.aj8mh.com/notebookind8.html ).

Somewhere on that page is "Notebook Series - Audio to PTT Converter (FLdigi)."

On 12/5/2020 12:35 PM, Bob McGraw - K4TAX wrote:

"While using a SSB mode does send CW simply by supplying an audio tone to the radio, one must be extremely careful so as not to have any humm or noise embedded with the tone.   Be sure the computer sounds are turned OFF and be sure the microphone is not hot to prevent room sounds in teh CW band.    The FCC frowns on this and the VM program will pick up on this and file a report on the station operation.   Also, the actual frequency transmitted will be carrier frequency plus or minus, depending USB or LSB mode, and the audio frequency.   Hence if the radio is on 7.025 and the tone is 1 kHz, the transmitted frequency will be either 7.024 or 7.026 depending sideband selection. Also one must avoid any ALC action or speech processing or equalization in the audio path, either inside the radio or external to the radio."

73

Bob, K4TAX

Joe Hutchens
AJ8MH - Radio
Marquette, MI


K3eui <K3eui@...>
 

Hi Bob  K4TAX

I think that sending CW by keying an audio pitch with the rig in USB (or LSB) will generate a perfectly adequate CW signal.  After all, this is not different from generating a  PSK31 signal (in either LSB or USB). You could have "other" audio superimposed like hum, or harmonics.  And, it is also no different from AFSK  RTTY, compared to FSK RTTY. It would be stupid to have a "live mic" when sending digi modes from your keyboard (as well as illegal).

No one should be so careless to allow Windows sounds (you've got mail) into their radio's sound card.  Or, if that happens once, you correct it.  Same with PSK or FT8 or any digi mode based on a sound card audio and a SSB radio.

I have an article in the que at QST that shows RF spectra of  CW keying at 30 wpm sent by keying audio, and then by keying the carrier. As long as the rig's  ALC  reads zero, there is no difference in the spectra of either method. The article shows the two spectra, side by side.

Of course, the actual RF signal being sent with the rig in USB and the audio at say 1 kHz audio pitch, is the sum of the two frequencies, where the RF lands. But everyone knows that.


If there is a NBEMS net on 3583 kHz (the PANBEMS net) and the audio center frequency is at 1500 Hz on the FLDIGI waterfall, then that 3583 is only the VFO dial (the suppressed carrier frequency) and there is NO RF there.  The RF signal is centered at  3583.0  +  1.500 kHz  =  3584.5 kHz. We all know that.  The frequency of the net is the suppressed carrier frequency. Well, it is the same with PHONE.  If my VFO dial reads 14.200 and I am on USB phone, the RF is in a band of frequencies starting at  14.200  +  500 Hz or so  up to  14.200 + 2500 Hz or so.

When I listen to W1AW on their frequency of 3581.5 kHz at night, I tune my radio VFO to 3580.0  in upper sideband mode, and then the keyed CW sounds are at 1500 Hz on the FLDIGI waterfall.  Perfect. All of the data filters are then centered on 1500 Hz.  If that pitch is too high, I just change the VFO dial a bit. FLDIGI copies 100% at any pitch.

I love operating CW by keying audio with the rig in USB-D mode on any of my HF rigs.  I can change the pitch of the audio by clicking a different CW station in the waterfall.  I hardly ever touch the VFO dial.... unless there is a contest (hi). But yes, there are more ways to send a BAD signal on CW if the rig is in SSB mode and you are keying audio. But it is legal and (usually) works well. There is never a "live microphone".

73
Barry  k3eui



On Dec 5, 2020, at 12:35 PM, Bob McGraw - K4TAX <rmcgraw@...> wrote:

Barry:

There are efforts on-going by Dave, W1HKS, to provide other means to send CW via the keyboard.   The brands and models of radios is being expanded.   Presently testing is underway for this project.

While using a SSB mode does send CW simply by supplying an audio tone to the radio, one must be extremely careful so as not to have any humm or noise embedded with the tone.   Be sure the computer sounds are turned OFF and be sure the microphone is not hot to prevent room sounds in teh CW band.    The FCC frowns on this and the VM program will pick up on this and file a report on the station operation.   Also, the actual frequency transmitted will be carrier frequency plus or minus, depending USB or LSB mode, and the audio frequency.   Hence if the radio is on 7.025 and the tone is 1 kHz, the transmitted frequency will be either 7.024 or 7.026 depending sideband selection.  Also one must avoid any ALC action or speech processing or equalization in the audio path, either inside the radio or external to the radio. 

As to DATA modes, usually these are analog audio paths and usually the radio will remove any processing or EQ.  Those however do not relieve the operator from correct levels and avoiding any ALC action.   As as rule and convention, CW is considered a LSB mode while DATA modes are considered USB modes,  regardless of band convention. 

Sending true and pure CW from FLDIGI from the keyboard to the radio is very nice.   Timing is correct, character length is correct as well as spacing.   As to hand sent CW, some of my regulars I find with hand sent CW is near 100 % copy while others are very difficult to copy due to their "fist". 

73

Bob, K4TAX


On 12/5/2020 9:58 AM, K3eui via groups.io wrote:
When I listen to CW that is "well-timed"  (like W1AW CW bulletins) the FLDIGI, in CW op mode, copies 100%  on 3581.5 kHz in the evenings.

What more is needed?
Poorly sent CW is easy on my ears, but no program does well turning that into print on a screen.

SENDING  CW
Now FLDIGI can send CW (with my keyboard) by two different methods: each one works fine.  You can key the audio tones (like any sound card mode) and then have your rig in USB or USB-DATA and what comes out to the antenna is keyed RF, just like "real" CW.

Or, put your rig in CW mode, and use FLDIGI to send a HIGH signal to the DTR pin, which inside your (modern) radio, closes the CW keying port to SEND.  That works even better  (no need to monitor ALC).
I have an article in the que in QST that describes both methods, and shows SDR  RF spectra to prove the methods are equally clean, as long as your ALC reads zero in SSB mode.

De   K3eui  Barry



On Dec 5, 2020, at 10:12 AM, Mark Hubelbank <mhubel@...> wrote:

I have been thinking of taking on the challenge of writing a CW decoder having been in the signal processing world for a long time. If this works well, is there interest in this group of integrating it into FLdigi? I would define working well as it should decode 99% of what I can do by ear.


K3eui <K3eui@...>
 

Now that is very strange... TOUCH-TONE CW

I've been operating CW for decades, and I have never heard on an HF band a "two-tone" CW signal. As far as I know, the left/right channel with FLDIGI must be the same audio pitch .... oh, or could the right channel is set up as a PTT channel or a QSK channel, and the audio is getting thru the rig as a legitimate second audio CW signal.

It must sound like CW sent with the rig in RTTY mode ???

Fascinating... maybe even legal, but an awful waste of spectrum.

K3eui

On Dec 5, 2020, at 3:57 PM, Joe Hutchens <aj8mh-radio@...> wrote:

I've heard a few stations sending "touch-tone" CW (multiple tones), and I suspect they're running tones directly into their mic jack. I've sent a few of them a note asking how their station is configured for sending CW, but haven't received any replies. I've given up trying to find their trick...

I send Fldigi CW tones to a homebrew interface that works nicely. If interested, look up:

( http://www.aj8mh.com/notebookind8.html ).

Somewhere on that page is "Notebook Series - Audio to PTT Converter (FLdigi)."

On 12/5/2020 12:35 PM, Bob McGraw - K4TAX wrote:

"While using a SSB mode does send CW simply by supplying an audio tone to the radio, one must be extremely careful so as not to have any humm or noise embedded with the tone. Be sure the computer sounds are turned OFF and be sure the microphone is not hot to prevent room sounds in teh CW band. The FCC frowns on this and the VM program will pick up on this and file a report on the station operation. Also, the actual frequency transmitted will be carrier frequency plus or minus, depending USB or LSB mode, and the audio frequency. Hence if the radio is on 7.025 and the tone is 1 kHz, the transmitted frequency will be either 7.024 or 7.026 depending sideband selection. Also one must avoid any ALC action or speech processing or equalization in the audio path, either inside the radio or external to the radio."

73

Bob, K4TAX

Joe Hutchens
AJ8MH - Radio
Marquette, MI






Bob McGraw - K4TAX <rmcgraw@...>
 

Barry et al:

Well some of that sort of works.   Since my radio, as my choice uses a 800 Hz sidetone which prefer for CW, if I should click on the waterfall at 1500 Hz, yes from the keyboard I can send and receive.  However, I occasionally need to send quickly something from my paddle connected to the radio and that would send at the carrier frequency, not 1500 Hz above the carrier frequency.   Case and point, if I am receiving CW traffic and need to acknowledge a word group I won't send a single RR from the keyboard.  Its two slow compared to the paddle.

I would question "perfectly adequate" as I am seeing the sidebands on some of the CW signals.  Agree, some of the radios even in CW mode have noticeable sideband energy, being clicks or chirps.  As to Windows sounds, these are heard regularly on some of the bands.   And in one instance, I've heard the same TV audio on the ham band in the CW portion along with keyboard CW, that I hear from my TV downstairs.   You are making too many generalities with regard to hams being careless and "they should know that".     Sorry, but I have a diminishing appreciation for general ham technical and operational knowledge these days, largely thanks to the VE program.  

In a typical transceiver in CW mode, the carrier frequency or RF transmitted frequency is the frequency displayed.  The receiver is then off-set in frequency by the selected sidetone frequency as determined by the user in order to produce a beat note from the other station.    I'm constantly amazed in monitoring the bands to see two stations in a QSO on somewhat different frequencies.  Usually not much difference with good operators but I've observed 250 to 500 Hz difference as not being unusual.  I guess this is a case where one operator has a preference for the CW tone to be different than the sidetone off-set of their radio.   Or they don't know or don't care to make the correct set-up parameters.   If the sidetone off-set is the same as the preferred received CW note, then both stations will be on the same frequency.    I really appreciate the automatic zero beat feature of my radio.  I can be assured I am on the other stations frequency and I have a CW note frequency which I prefer.

I look forward to your article in QST. 

73

Bob, K4TAX


On 12/5/2020 3:02 PM, K3eui via groups.io wrote:

Hi Bob  K4TAX

I think that sending CW by keying an audio pitch with the rig in USB (or LSB) will generate a perfectly adequate CW signal.  After all, this is not different from generating a  PSK31 signal (in either LSB or USB). You could have "other" audio superimposed like hum, or harmonics.  And, it is also no different from AFSK  RTTY, compared to FSK RTTY. It would be stupid to have a "live mic" when sending digi modes from your keyboard (as well as illegal).

No one should be so careless to allow Windows sounds (you've got mail) into their radio's sound card.  Or, if that happens once, you correct it.  Same with PSK or FT8 or any digi mode based on a sound card audio and a SSB radio.

I have an article in the que at QST that shows RF spectra of  CW keying at 30 wpm sent by keying audio, and then by keying the carrier. As long as the rig's  ALC  reads zero, there is no difference in the spectra of either method. The article shows the two spectra, side by side.

Of course, the actual RF signal being sent with the rig in USB and the audio at say 1 kHz audio pitch, is the sum of the two frequencies, where the RF lands. But everyone knows that.

If there is a NBEMS net on 3583 kHz (the PANBEMS net) and the audio center frequency is at 1500 Hz on the FLDIGI waterfall, then that 3583 is only the VFO dial (the suppressed carrier frequency) and there is NO RF there.  The RF signal is centered at  3583.0  +  1.500 kHz  =  3584.5 kHz. We all know that.  The frequency of the net is the suppressed carrier frequency. Well, it is the same with PHONE.  If my VFO dial reads 14.200 and I am on USB phone, the RF is in a band of frequencies starting at  14.200  +  500 Hz or so  up to  14.200 + 2500 Hz or so.

When I listen to W1AW on their frequency of 3581.5 kHz at night, I tune my radio VFO to 3580.0  in upper sideband mode, and then the keyed CW sounds are at 1500 Hz on the FLDIGI waterfall.  Perfect. All of the data filters are then centered on 1500 Hz.  If that pitch is too high, I just change the VFO dial a bit. FLDIGI copies 100% at any pitch.

I love operating CW by keying audio with the rig in USB-D mode on any of my HF rigs.  I can change the pitch of the audio by clicking a different CW station in the waterfall.  I hardly ever touch the VFO dial.... unless there is a contest (hi). But yes, there are more ways to send a BAD signal on CW if the rig is in SSB mode and you are keying audio. But it is legal and (usually) works well. There is never a "live microphone".

73
Barry  k3eui



On Dec 5, 2020, at 12:35 PM, Bob McGraw - K4TAX <rmcgraw@...> wrote:

Barry:

There are efforts on-going by Dave, W1HKS, to provide other means to send CW via the keyboard.   The brands and models of radios is being expanded.   Presently testing is underway for this project.

While using a SSB mode does send CW simply by supplying an audio tone to the radio, one must be extremely careful so as not to have any humm or noise embedded with the tone.   Be sure the computer sounds are turned OFF and be sure the microphone is not hot to prevent room sounds in teh CW band.    The FCC frowns on this and the VM program will pick up on this and file a report on the station operation.   Also, the actual frequency transmitted will be carrier frequency plus or minus, depending USB or LSB mode, and the audio frequency.   Hence if the radio is on 7.025 and the tone is 1 kHz, the transmitted frequency will be either 7.024 or 7.026 depending sideband selection.  Also one must avoid any ALC action or speech processing or equalization in the audio path, either inside the radio or external to the radio. 

As to DATA modes, usually these are analog audio paths and usually the radio will remove any processing or EQ.  Those however do not relieve the operator from correct levels and avoiding any ALC action.   As as rule and convention, CW is considered a LSB mode while DATA modes are considered USB modes,  regardless of band convention. 

Sending true and pure CW from FLDIGI from the keyboard to the radio is very nice.   Timing is correct, character length is correct as well as spacing.   As to hand sent CW, some of my regulars I find with hand sent CW is near 100 % copy while others are very difficult to copy due to their "fist". 

73

Bob, K4TAX


On 12/5/2020 9:58 AM, K3eui via groups.io wrote:
When I listen to CW that is "well-timed"  (like W1AW CW bulletins) the FLDIGI, in CW op mode, copies 100%  on 3581.5 kHz in the evenings.

What more is needed?
Poorly sent CW is easy on my ears, but no program does well turning that into print on a screen.

SENDING  CW
Now FLDIGI can send CW (with my keyboard) by two different methods: each one works fine.  You can key the audio tones (like any sound card mode) and then have your rig in USB or USB-DATA and what comes out to the antenna is keyed RF, just like "real" CW.

Or, put your rig in CW mode, and use FLDIGI to send a HIGH signal to the DTR pin, which inside your (modern) radio, closes the CW keying port to SEND.  That works even better  (no need to monitor ALC).
I have an article in the que in QST that describes both methods, and shows SDR  RF spectra to prove the methods are equally clean, as long as your ALC reads zero in SSB mode.

De   K3eui  Barry



On Dec 5, 2020, at 10:12 AM, Mark Hubelbank <mhubel@...> wrote:

I have been thinking of taking on the challenge of writing a CW decoder having been in the signal processing world for a long time. If this works well, is there interest in this group of integrating it into FLdigi? I would define working well as it should decode 99% of what I can do by ear.


Gene Rozea
 

I’ve found CWget to be the most accurate.

 

 

From: Mark Hubelbank
Sent: Saturday, December 5, 2020 3:46 PM
To: winfldigi@groups.io
Subject: Re: [winfldigi] CW Modem #modem

 

Well it seems that there is little interest. I find all the CW decoders out there lacking. Even the best make a lot more errors on real off the air poorly formed signals than I do and I am no so good anymore. I can't really keep up with any off the air signals for more than a few minutes. Machines don't get tired.
This effort is strictly for decoding, not sending.
If I do this, I will still post the results.

On 12/5/2020 10:12 AM, Mark Hubelbank wrote:

I have been thinking of taking on the challenge of writing a CW decoder having been in the signal processing world for a long time. If this works well, is there interest in this group of integrating it into FLdigi? I would define working well as it should decode 99% of what I can do by ear.



-- 
Mark Hubelbank
NorthEast Monitoring
141 Parker St
Suite 200
Maynard, MA, 01754 - USA
mhubel@...
978-443-3955

 


Chris Shaker, KJ7BLE
 

I haven't seen software yet that decodes morse code really well yet. I've paid for CW skimmer and CWGet, and they mostly work. Only for Windows 10.

There is definitely room for software that decodes Morse code from radio noise really well. And I don't know of much available for

Linux.

Chris Shaker, KJ7BLE


Chris Shaker, KJ7BLE
 

I have messed with FLDigi's CW decoder, and I can't get it to do much of anything correctly.

Chris Shaker, KJ7BLE


Dave
 

Mark,

All of the fldigi code is open source and available for download at http://www.w1hkj.com/files/fldigi/  and also at Source Forge.

Download the tar.gz source archive, uncompress and then look at the file:

src/cw_rtty/cw.cxx

Any decoder you develop would have to fit in to this C++ class structure.

73, David, W1HKJ


On 12/5/20 9:12 AM, Mark Hubelbank wrote:
I have been thinking of taking on the challenge of writing a CW decoder having been in the signal processing world for a long time. If this works well, is there interest in this group of integrating it into FLdigi? I would define working well as it should decode 99% of what I can do by ear.


Artie Langston <artielangston@...>
 

I use the fldigi CW modem all the time for traffic nets, and find it very very useful. Not the receive part, but if I'm sending a number of messages, it's really handy to be able to type the messages ahead of time, and then drop  them in the send window as needed. It also makes it easy if they ask for a fill, and I have hot keys set up for the various CW nets I check into.

I much prefer keying the IC-7300 with DTR  in the rig's regular CW mode.

The receive part is OK, but I haven't found anything that's terribly useful for receiving CW.

73 Artie


Nathan Rosenthal
 

Mark, you do realize that Morse code was never meant to be decoded by a machine. Characters have varying lengths and there are no start and stop signals. The ear and brain are much better at deciphering inconsistent code signals than any machine. 


Rory Bowers <k6cks01@...>
 

I agree Nathan... if CW is sent by a machine it has no "personality".  CW sent by key does.  That personality can prove difficult for a machine to recognize and decode correctly.
73,
Rory, K5CKS

On Sun, Dec 6, 2020 at 9:52 AM Nathan Rosenthal <nate05pa@...> wrote:
Mark, you do realize that Morse code was never meant to be decoded by a machine. Characters have varying lengths and there are no start and stop signals. The ear and brain are much better at deciphering inconsistent code signals than any machine. 


Gary Rondeau <grondeau@...>
 

I've been using FLDIGi's CW decoder for several years now.  It has been my "trainer" as I gradually really learn what CW sounds like.  There is no question that a trained ear is better than the decoder, but the issues are mostly on leading characters when the decoder is trying to track speeds.  I'm sure modern neural net approaches could eventually perform very well for even hand-sent code.  There could be improvements in handling over the pole flutter where maybe a neural net could "hear" the underlying message more like a human would.  

The present decoder serves very well, however, I've managed my CW DXCC using teh FLDIGI deocder and what my ear has learned along the way.  ITtprovides a very accessible entry point to the CW world that is a real benefit to the community.

These days I prefer to send using TTL serial logic using N1MM as the control program, but I keep FLDIGI on the RX signal.  Soon I may be able to dispense with the crutch...  but it has been a great aid getting to where I am.
Gary AF7NX


Artie Langston <artielangston@...>
 

I think the challenge is not actually in signal processing, but in some sort of artificial intelligence. Fldigi reads extremely well considering, when it's another machine doing the sending. With the peculiarities of various humans, on various keys, not so much.

This leaves the question, is designing an AI of that sophistication going to be worth the time and investment required?

I have noticed that my Digital Audio Workstation I use in my work as a musician can do amazing things with what is referred to as Quantizing a musical performance to various degrees. It can take a rhythmically  imprecise musical performance and then "line it up" correctly with the beats and time signature, so it sounds pretty perfect.

Perhaps once the WPM is detected by the software, it could then quantize the CW before decoding, thus delivering something like W1AW or keyboard generated copy even from a shaky fist.

73

Artie


Artie Langston <artielangston@...>
 

If 3:1 is the 'ideal' ratio when it comes to being able to hear and differentiate between dots and dashes for the average to good operator, then a dot is simply, say, a dotted 16th note to the dash's quarter note. I'm going to record some people on air, and then try to quantize them in the music software, and see how that turns out. Maybe it could be like a speech processor for your fist. Perfect sending from your straight key, all time.

Or maybe, much more likely,  a complete waste of time. It will be fun any way.

Artie


On Sun, Dec 6, 2020 at 3:25 PM Artie Langston via groups.io <artielangston=gmail.com@groups.io> wrote:
I think the challenge is not actually in signal processing, but in some sort of artificial intelligence. Fldigi reads extremely well considering, when it's another machine doing the sending. With the peculiarities of various humans, on various keys, not so much.

This leaves the question, is designing an AI of that sophistication going to be worth the time and investment required?

I have noticed that my Digital Audio Workstation I use in my work as a musician can do amazing things with what is referred to as Quantizing a musical performance to various degrees. It can take a rhythmically  imprecise musical performance and then "line it up" correctly with the beats and time signature, so it sounds pretty perfect.

Perhaps once the WPM is detected by the software, it could then quantize the CW before decoding, thus delivering something like W1AW or keyboard generated copy even from a shaky fist.

73

Artie