Author Topic: How fast can Morse code be sent  (Read 4128 times)

0 Members and 1 Guest are viewing this topic.

Offline jonovidTopic starter

  • Super Contributor
  • ***
  • Posts: 1425
  • Country: au
    • JONOVID
How fast can Morse code be sent
« on: April 15, 2022, 07:17:53 pm »
I am asking when does chopping up a radio frequency carrier by a lower frequency become radio interference?
if the lower frequency been Morse code.
proficiency in Morse code is measured in how many words per minute that is humanly possible to send or receive.
now add digital technology capabilities to Morse code proficiency and crank up the words per __ speed.
How fast can Morse code be sent over the electromagnetic spectrum?
when does the harmonics become radio interference?
in Morse burst transmissions.
why - more data sent in less time.
Hobbyist with a basic knowledge of electronics
 

Offline CatalinaWOW

  • Super Contributor
  • ***
  • Posts: 5170
  • Country: us
Re: How fast can Morse code be sent
« Reply #1 on: April 15, 2022, 07:57:53 pm »
The answer is straight forward, but I am away from my reference books so can't give a numerical answer, which in any case depends on application.

The key to the problem is the authorized bandwidth.  Various radio services have limits on this bandwidth, 5kHz is a common one, but there are many others.  It depends on the frequency band, modulation method (AM, SSB, FM and so on) and possibly other factors.  You can calculate the spectrum of a Morse signal (simple Fourier transform), use your proposed modulation method to see how this translates into real world bandwidth and compare it to the authorized use.. 
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 21606
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: How fast can Morse code be sent
« Reply #2 on: April 15, 2022, 08:33:40 pm »
Simple Fourier analysis of the waveforms gives the answer.

"Morse" code can't really be called such, beyond some data rate where its pulsation isn't even perceptible; probably on the order of say 20ms bit time, or 50 baud.  Even then I don't know that you could understand it at that rate, and almost impossible to send.  But as human feats go, quite a lot is possible with enough practice, so, I wouldn't rule it out that someone can do this, or maybe even a bit more.

If it's being registered with a tone, then the tone itself becomes meaningless when the tone burst is shorter than a few cycles.  The cycles, or lack thereof, overlap into a jumble of lumps.  Not that detection is impossible at this point, given a suitable receiver, but it gets a lot harder than, say for example: bandpass filter --> diode detector --> lowpass filter --> discriminator (say schmitt trigger).  The bandwidth of such a signal chain is too low to reliably receive bits.

Notice that using an audio carrier is no different than the on-air signal itself.  Just at RF rather than AF.  So the same argument applies: when the signal is starting and stopping every couple of cycles, it's a jumble and hard to make any sense of at all.

Applying Fourier analysis, we can see that the frequency response corresponding to a single rectangular pulse, is a sinc envelope with bandwidth inversely proportional to pulse width.  This is true at baseband, and it's also true at RF.  This relation holds true: multiplication and convolution are exchanged under the Fourier transform.  If we multiply a signal rect(xt) with a carrier sin(w0 t), it's equivalent to convolving F(rect(xt)) with F(sin(w0 t)).  Convolution isn't always the simplest operation, but we have a special case here, that F(sin(w0 t)) = delta(w - w0) (that is, the Dirac delta function, meaning, one spike on the spectrum*), and convolution with a delta (impulse, spike) is identical to shifting the other argument over to that center frequency.

*Well, two because it's time-symmetric, so +/- w0.
*And yes, it's "not a function". But also we can define it in such a way that it is.  It's fine.

So a rectangular window / gated / tone-burst sine wave, has a frequency response (spectrum) of a sinc function, centered at +/-w0, with a width (-3dB bandwidth) of pi/x or something like that (I have to look up the exact F(rect(xt)), but it is what it is).

If we send multiple rect at various times, their spectra superimpose and we get new peaks and valleys.  For example, an alternating pulse train as a square wave, everything but the harmonics cancel out, so we would get a peak at w0 (carrier), and secondary peaks at w0 +/-(1, 3, 5, ...) * (2pi)/T.  All the odd harmonics, up to whatever bandwidth the square wave happens to have.  Or if we send some other duty cycle, then even harmonics show up as well, and their amplitudes take on a different distribution; or if we send random data (like a PRBS (pseudorandom bit stream) test, which, generally, something like Morse code will approximate over time), then again a sinc sort of envelope but with random peaks and dips as the data comes and goes.

For accurate detection of data, a sinc spectrum can be bandpass filtered to the first zero, which corresponds to a bandwidth of two bit times (so, if the bits are alternating 1/0, that's a square wave of that frequency, so, needs at least that much BW to see).

The term for these extra tones/spectra is sidebands.

Radio channels are allocated with fixed (maximum) bandwidth.  Morse code channels for example every 500Hz or so, SSB a couple 3kHz or more (I forget what exactly), AM, NB FM a bit more, WB FM (commercial broadcast FM) 200kHz, television some MHz, etc.  Notice these are all quite narrow in relation to their center frequencies: even a code channel at 143kHz has very little fractional bandwidth (143/0.5), and amateur channels even at quite high center frequencies (say 144MHz, or heck anything up into the GHz with the same modulation) even less.

So from the standpoint of not stomping on neighboring channels, exceeding that bandwidth, which we know is also the symbol rate, would be considered interference.

More data can also be sent by using more levels/states per symbol, but this is only feasible when the symbols can be discriminated from each other reliably, i.e. the signal to noise ratio is adequate.  Like how WiFi negotiates anything from BPSK (couple megabits) with poorest reception, up to, I forget what, but some dense QAM constellation for >60MBps, or even more on the 5GHz band I think?  (And, notice again, these are fairly narrow channels, compared to the 2.45GHz center frequency.)

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline RoGeorge

  • Super Contributor
  • ***
  • Posts: 6145
  • Country: ro
Re: How fast can Morse code be sent
« Reply #3 on: April 15, 2022, 09:58:23 pm »
...
How fast can Morse code be sent over the electromagnetic spectrum?
...
why - more data sent in less time.

Not Morse, but the maximum possible capacity (bits/second) depends only of the channel's bandwidth and the signal to noise ratio:



https://en.wikipedia.org/wiki/Shannon%E2%80%93Hartley_theorem
 
The following users thanked this post: boB, radiolistener

Offline emece67

  • Frequent Contributor
  • **
  • !
  • Posts: 614
  • Country: 00
Re: How fast can Morse code be sent
« Reply #4 on: April 16, 2022, 07:43:53 am »
.
« Last Edit: August 19, 2022, 05:23:47 pm by emece67 »
 
The following users thanked this post: boB

Offline El Rubio

  • Regular Contributor
  • *
  • Posts: 74
  • Country: us
Re: How fast can Morse code be sent
« Reply #5 on: April 17, 2022, 11:15:17 pm »
How small can you cut the time units down to? Morse code is dits and dahs. The dah has to be longer than the dit and time between dit and dah’s, longer time between characters, even longer time between words. Obviously some words are longer than others and some characters have more dit’s and dah’s. It’s CW. How fast can a transmitter switch on and off? How small of a time unit can you slice into what makes up a dit and what makes up a dah and still be decoded?
 

Offline joeqsmith

  • Super Contributor
  • ***
  • Posts: 11630
  • Country: us
Re: How fast can Morse code be sent
« Reply #6 on: April 18, 2022, 12:25:44 am »
McElroy licensed with Creed & Company, built in Littleton MA.   Looks like they could send at up to 500 words per minute.  Their Roman character printer could run at 100 WPM. 

https://en.wikipedia.org/wiki/Theodore_Roosevelt_McElroy

***
A few photos from Creed & Co
http://www.museumoftechnology.org.uk/objects/telegraphy.php 
« Last Edit: April 18, 2022, 12:28:18 am by joeqsmith »
 

Offline TheUnnamedNewbie

  • Super Contributor
  • ***
  • Posts: 1208
  • Country: 00
  • mmwave RFIC/antenna designer
Re: How fast can Morse code be sent
« Reply #7 on: April 19, 2022, 11:08:28 am »
Depends on what you mean with 'morse code being sent'. do you mean by humans? No clue. And what are your limitations? Human? Legal? Electrical?

By electronic devices: What is morse but on-off keying with a line code that also happens to include your encoding of letters onto the data. We've made millimeter-wave on-off-keying transmitters that achieve in excess of 30 gbit/second in CMOS, and I'm sure you can do better in something like SiGe or InP (though bandwidth in InP might be harder)
The best part about magic is when it stops being magic and becomes science instead

"There was no road, but the people walked on it, and the road came to be, and the people followed it, for the road took the path of least resistance"
 

Offline cdev

  • Super Contributor
  • ***
  • !
  • Posts: 7350
  • Country: 00
Re: How fast can Morse code be sent
« Reply #8 on: April 19, 2022, 12:45:59 pm »
...
How fast can Morse code be sent over the electromagnetic spectrum?
...
why - more data sent in less time.

Not Morse, but the maximum possible capacity (bits/second) depends only of the channel's bandwidth and the signal to noise ratio:



https://en.wikipedia.org/wiki/Shannon%E2%80%93Hartley_theorem

I think a better question is howfast is AI (cw) modulation with a human at both ends? I would guess its veryt fast compared to most of us, especially old guys like me. However, a digital keyer and digital CW decoder coud handle it. How fast?

Good question!~ How fast is clear reception practical? It certainly would help to be able to hear between the dits and dahs to be able to tell when the channel was clear.

IMHO. Other than that I dont know.
"What the large print giveth, the small print taketh away."
 

Online Bud

  • Super Contributor
  • ***
  • Posts: 6877
  • Country: ca
Re: How fast can Morse code be sent
« Reply #9 on: April 19, 2022, 02:08:31 pm »
Still faster than texting  8)



Facebook-free life and Rigol-free shack.
 

Offline mag_therm

  • Frequent Contributor
  • **
  • Posts: 633
  • Country: us
Re: How fast can Morse code be sent
« Reply #10 on: June 05, 2022, 09:54:22 pm »
FCC and ITU standards:

Ref: qsl.net/broward/part97a.htm#97.3
The FCC defines an emission band width in S97.3(a) as the frequency difference between the points when the signal has dropped to 26dB
below the average signal power across that bandwidth.

Ref: life.itu.int/radioclub/rr/R-REC-SM.1138-2-200810-I!!PDF-E.pdf
ITU in ITU-R SM1138-2 publishes simplified functions for BW.
For CW (Morse) emission designator A1A:
BW=B * K   : B in Baud, using B = 0.8 * word_per_minute   ;  K is a keying envelope factor, suggested to =5 for a smooth rise and fall .

So for 13 word_per_minute:
BW = 13 * 0.8 * 5
BW = 52 Hz

Note Morse is a varicode, and so is the BFSK31 mode, FSK and AFSK are baudot.

FCC places, on the amateur digital modes,  Baud rate limits  and Bandwith limits in S97.305(c) and 97.307(f), regardless of  digital mode.
example 160m to 12 m 300 Baud and 1 kHz.
example 70 cm 56 kBaud and 100 kHz

New and experimental  modes are allowed to be introduced by amateurs as long as they publish the details first.
 
The following users thanked this post: boB

Online fourfathom

  • Super Contributor
  • ***
  • Posts: 1851
  • Country: us
Re: How fast can Morse code be sent
« Reply #11 on: June 06, 2022, 12:23:24 am »
So a rectangular window / gated / tone-burst sine wave, has a frequency response (spectrum) of a sinc function, centered at +/-w0, with a width (-3dB bandwidth) of pi/x or something like that (I have to look up the exact F(rect(xt)), but it is what it is).

If we send multiple rect at various times, their spectra superimpose and we get new peaks and valleys.  For example, an alternating pulse train as a square wave, everything but the harmonics cancel out, so we would get a peak at w0 (carrier), and secondary peaks at w0 +/-(1, 3, 5, ...) * (2pi)/T.  All the odd harmonics, up to whatever bandwidth the square wave happens to have.  Or if we send some other duty cycle, then even harmonics show up as well, and their amplitudes take on a different distribution; or if we send random data (like a PRBS (pseudorandom bit stream) test, which, generally, something like Morse code will approximate over time), then again a sinc sort of envelope but with random peaks and dips as the data comes and goes.

For accurate detection of data, a sinc spectrum can be bandpass filtered to the first zero, which corresponds to a bandwidth of two bit times (so, if the bits are alternating 1/0, that's a square wave of that frequency, so, needs at least that much BW to see).

To complicate the analysis, morse code is typically sent with amplitude shaping to avoid the square-edge "key clicks".  This filtering/shaping isn't well-defined, but the pulse rise and fall times are usually (eyeballing it) perhaps a few milliseconds. This significantly reduces the sidebands beyond the first few harmonics of the keying rate.

I was recently playing with Bessel filtering of FSK frequency transitions, and the effect of even a little filtering can be dramatic when compared to square-edge transitions.  That was frequency-domain rather than amplitude-domain, but the principals still apply.
We'll search out every place a sick, twisted, solitary misfit might run to! -- I'll start with Radio Shack.
 

Offline boB

  • Frequent Contributor
  • **
  • Posts: 307
  • Country: us
    • my work www
Re: How fast can Morse code be sent
« Reply #12 on: June 06, 2022, 12:55:32 am »

I used to know a guy that could do 80 WPM with Morse code here in the PNW area.  That is about the fastest I have heard him or others be able to do.

That Shannon equation regarding bandwidth and S/N ratio is the real theoretical answer of course.

I had not heard of the 3000 WPM  EME contacts but it sure sounds right.

I would think that CW carrier on-off with NO modulation would be the way to go as well as the machine knowing ahead of time what the "baud" rate would be. That way you don't have to worry about extra modulated tones having to be decoded as well.  The sample rate of an RF CW carrier On/Off keying should be plenty high enough for really fast Morse Code I would think.

I have heard of people being able to decode slow-ish RTTY by ear which I would sort of believe is possible.

I'm good for mid 30 WPM Morse myself but that's about it for this old man.

boB
K7IQ
 

Offline radiolistener

  • Super Contributor
  • ***
  • Posts: 3282
  • Country: ua
Re: How fast can Morse code be sent
« Reply #13 on: June 06, 2022, 05:12:36 am »
Not Morse, but the maximum possible capacity (bits/second) depends only of the channel's bandwidth and the signal to noise ratio:



https://en.wikipedia.org/wiki/Shannon%E2%80%93Hartley_theorem

Since standard CW bandwidth is 500 Hz and usable signal to noise ratio for a human ear is about SNR = 10 dB = 10, then we get:

C = 500 * log2( 1 + 10 ) = 500 * log2( 11 ) = 1729 bps

Typically character size is about 10 bits per character and word size is about 5 chars per word.
So we can get the max CW speed as:

CPS = BPS / 10 = 1729 / 10 = 172 characters per second
or
WPM = ((BPS / 10) / 5) * 60 = BPS * 1.2 = 1729 * 1.2 = 2074 words per minute.

As I know, there is another limitation for machine decoding of CW transmitted by people. Since people cannot keep stable bps speed and cannot keep stable dot/dash duration it may vary in time, so it leads to a trouble for machine CW decode. But another human still can decode it. :)
« Last Edit: June 06, 2022, 05:39:12 am by radiolistener »
 

Online fourfathom

  • Super Contributor
  • ***
  • Posts: 1851
  • Country: us
Re: How fast can Morse code be sent
« Reply #14 on: June 06, 2022, 05:42:21 am »
Not Morse, but the maximum possible capacity (bits/second) depends only of the channel's bandwidth and the signal to noise ratio:



https://en.wikipedia.org/wiki/Shannon%E2%80%93Hartley_theorem

Since standard CW bandwidth is 500 Hz and usable signal to noise ratio for a human ear is about SNR = 10 dB = 10, then we get:

C = 500 * log2( 1 + 10 ) = 500 * log2( 11 ) = 1729 bps

Which is about 1729 / 10 = 172 characters per second or 1729 * 1.2 = 2074 words per minute :)

No, the SNR in Shannon's theorem is the SNR in the channel, not the human threshold for detection.  In the equation the channel capacity can be increased by increasing the transmit power (thus improving the SNR) -- so by that measure there is no limit.  But the channel capacity assumes some form of ideal modulation (appropriate for the channel bandwidth and SNR), and that is definitely not the case with morse code.

We could always look at that 500Hz filter and assume that the symbol rate must be slow enough that a human can discern the dits, dahs, and spaces after filtering.  Perhaps a 100 Hz symbol rate, which ends up being 120 WPM (morse symbol rate = 1.2 / WPM, "PARIS " is the standard 50-unit word).  You would need a wider filter for faster speed.
We'll search out every place a sick, twisted, solitary misfit might run to! -- I'll start with Radio Shack.
 
The following users thanked this post: boB

Online Berni

  • Super Contributor
  • ***
  • Posts: 4922
  • Country: si
Re: How fast can Morse code be sent
« Reply #15 on: June 06, 2022, 05:58:59 am »
As others have said the speed is mostly limited by the radio channel bandwidth you are allowed to use.

So if you want to take it to extremes you could use a laser to transmit radio waves at about 600 000 GHz (light is radio waves after all). There are also no bandwidth limits or license requirements on this part of the spectrum. So you could modulate the carrier with a bandwidth of 1000 GHz. Plenty of room to transmit morse code at a a rate of >10G characters per second.

That being said morose code is actually a very inefficient modulation and coding method. The on-off keying modulation occupies a pretty wide bandwidth while the morse code encoding wastes fair bit of time.



Tho could squeeze more bandwith out of morose code by introducing an additional encoding. From the international morse code the shortest letter is E and the second shortest is T. If we only send these two then we can set the space to be 0 seconds long while setting the character spacing to be the same length as the dot. Now we can transmit 1 bit of information in 2 or 3 clock cycles (so average 1 bit per 2.5 baud). The result can still be received by a morse code receiver, just needs additional decoding to reveal the message.

Tho at this point we have just re-invented digital pulse length encoding (that is used in simple data RF links, IR remotes, OneWire chips etc..)

Sending ASCII text trough this does throw away morse codes inherent western language compression capabilities, but we can get that back by adding huffman tree encoding to compress down the common characters. We can even do one better and apply the usual repetition based compression methods (like ZIP files do) and compress longer messages even more effectively.

However by now the data is completely unreadable by a human, no matter how much training. The whole point of morse code was to be a binary code that is easy to read/write by a human while still being reasonably efficient on the equipment of the era.
 

Offline RoGeorge

  • Super Contributor
  • ***
  • Posts: 6145
  • Country: ro
Re: How fast can Morse code be sent
« Reply #16 on: June 06, 2022, 06:59:55 am »
So a rectangular window / gated / tone-burst sine wave, has a frequency response (spectrum) of a sinc function, centered at +/-w0, with a width (-3dB bandwidth) of pi/x or something like that (I have to look up the exact F(rect(xt)), but it is what it is).

If we send multiple rect at various times, their spectra superimpose and we get new peaks and valleys.  For example, an alternating pulse train as a square wave, everything but the harmonics cancel out, so we would get a peak at w0 (carrier), and secondary peaks at w0 +/-(1, 3, 5, ...) * (2pi)/T.  All the odd harmonics, up to whatever bandwidth the square wave happens to have.  Or if we send some other duty cycle, then even harmonics show up as well, and their amplitudes take on a different distribution; or if we send random data (like a PRBS (pseudorandom bit stream) test, which, generally, something like Morse code will approximate over time), then again a sinc sort of envelope but with random peaks and dips as the data comes and goes.

For accurate detection of data, a sinc spectrum can be bandpass filtered to the first zero, which corresponds to a bandwidth of two bit times (so, if the bits are alternating 1/0, that's a square wave of that frequency, so, needs at least that much BW to see).

To complicate the analysis, morse code is typically sent with amplitude shaping to avoid the square-edge "key clicks".  This filtering/shaping isn't well-defined, but the pulse rise and fall times are usually (eyeballing it) perhaps a few milliseconds. This significantly reduces the sidebands beyond the first few harmonics of the keying rate.

I was recently playing with Bessel filtering of FSK frequency transitions, and the effect of even a little filtering can be dramatic when compared to square-edge transitions.  That was frequency-domain rather than amplitude-domain, but the principals still apply.

In practice there is yet another limitation caused by "ringing".  I'm not a ham, so not sure about the name, Romanian hams call it "the bell effect (as in ringing)", in Ro spelled "efectul de clopot".  It can be hear when one applies a very narrow (high Q) bandpass filter, with the reason of reducing the noise level, but then, if the bandpass filter is too narow, a single pulse/blip sounds like a reverbing/ringing tone instead of a clean beep, spilling into the following simbols because of the sin(x)/x shaped response of the very narrow bandpass filter.

I think it's the same kind of problem that is causing intersymbol interference in digital transmission https://en.wikipedia.org/wiki/Intersymbol_interference, and AFAIK this is inescapable, will appear anytime one tries to cram too faster data speeds into a too narrow band.

The attachments were made with Audacium (thanks cirip for the idea), by recording some random YT morse sample, then applying a very narrow bandpass filter with 55dB attenuation.

Note how after filtering it almost sounds like a continuous beep, and this is without noise.  With noise will be even worst.  This is to illustrate what happens in the time domain, and how ringing affects the signal when the band is too narrow, or when the symbols speed is too fast for a given band.
« Last Edit: June 06, 2022, 07:03:47 am by RoGeorge »
 

Offline m k

  • Super Contributor
  • ***
  • Posts: 1924
  • Country: fi
Re: How fast can Morse code be sent
« Reply #17 on: June 06, 2022, 09:13:37 am »

I used to know a guy that could do 80 WPM with Morse code here in the PNW area.  That is about the fastest I have heard him or others be able to do.

That Shannon equation regarding bandwidth and S/N ratio is the real theoretical answer of course.

I had not heard of the 3000 WPM  EME contacts but it sure sounds right.

I would think that CW carrier on-off with NO modulation would be the way to go as well as the machine knowing ahead of time what the "baud" rate would be. That way you don't have to worry about extra modulated tones having to be decoded as well.  The sample rate of an RF CW carrier On/Off keying should be plenty high enough for really fast Morse Code I would think.

I have heard of people being able to decode slow-ish RTTY by ear which I would sort of believe is possible.

I'm good for mid 30 WPM Morse myself but that's about it for this old man.

boB

Back in the day here army had 40 minimum.

I guess 250 straight to the ear is pretty much.
But if standard five letter crypts should be reliable decodable I'd say half of that is already too much.

Other side is also interesting.
How bad can still go through, negative with just a tangent can be "very" clear.
Advance-Aneng-Appa-AVO-Data Tech-Fluke-General Radio-H. W. Sullivan-Heathkit-HP-Kaise-Kyoritsu-Leeds & Northrup-Mastech-REO-Simpson-Sinclair-Tektronix-Triplett-YFE
(plus lesser brands from the work shop of the world)
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 21606
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: How fast can Morse code be sent
« Reply #18 on: June 06, 2022, 01:21:01 pm »
No, the SNR in Shannon's theorem is the SNR in the channel, not the human threshold for detection.  In the equation the channel capacity can be increased by increasing the transmit power (thus improving the SNR) -- so by that measure there is no limit.  But the channel capacity assumes some form of ideal modulation (appropriate for the channel bandwidth and SNR), and that is definitely not the case with morse code.

That said, suppose we took that as a figure for human detection / parsing -- voice has been compressed to around 500bps with okay fidelity, and perhaps we can assert something about the redundancy of spoken language in that and say that the human limit is somewhat higher -- it's in the right ballpark, at least.  Or maybe these two figures are a mere coincidence and such a connection isn't justified, who knows. ;D

In practice there is yet another limitation caused by "ringing".  I'm not a ham, so not sure about the name, Romanian hams call it "the bell effect (as in ringing)", in Ro spelled "efectul de clopot".  It can be hear when one applies a very narrow (high Q) bandpass filter, with the reason of reducing the noise level, but then, if the bandpass filter is too narow, a single pulse/blip sounds like a reverbing/ringing tone instead of a clean beep, spilling into the following simbols because of the sin(x)/x shaped response of the very narrow bandpass filter.

I think it's the same kind of problem that is causing intersymbol interference in digital transmission https://en.wikipedia.org/wiki/Intersymbol_interference, and AFAIK this is inescapable, will appear anytime one tries to cram too faster data speeds into a too narrow band.

Indeed.  You need channel bandwidth at least half the symbol rate (i.e., not much below -3dB for a perfectly alternating signal), and if you're detecting multiple levels per symbol, you generally want to minimize bleed from one to another -- ISI.  For single bits, this just closes the eye diagram a bit, not too big a deal, but obviously it's quite important the more levels / points in the constellation you're using.

Certain modulation schemes use orthogonal code sequences to implement CDM, and require a root-cosine filter on the rx/tx, so that their product has a cosine (Hann) window per symbol, the value of which goes to zero at exactly the next bit time, minimizing ISI.  (Your basic IIR or analog filter does not do this so well.  You can, of course, tune it so the impulse response crosses zero -- but not so it goes directly to zero and stays there, it will oscillate some.  If the zeroes of the oscillation all occur at bit times, you've still got it though.)

Tim
« Last Edit: June 06, 2022, 01:30:24 pm by T3sl4co1l »
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline gf

  • Super Contributor
  • ***
  • Posts: 1132
  • Country: de
Re: How fast can Morse code be sent
« Reply #19 on: June 06, 2022, 02:04:52 pm »
So a rectangular window / gated / tone-burst sine wave, has a frequency response (spectrum) of a sinc function, centered at +/-w0, with a width (-3dB bandwidth) of pi/x or something like that (I have to look up the exact F(rect(xt)), but it is what it is).

If we send multiple rect at various times, their spectra superimpose and we get new peaks and valleys.  For example, an alternating pulse train as a square wave, everything but the harmonics cancel out, so we would get a peak at w0 (carrier), and secondary peaks at w0 +/-(1, 3, 5, ...) * (2pi)/T.  All the odd harmonics, up to whatever bandwidth the square wave happens to have.  Or if we send some other duty cycle, then even harmonics show up as well, and their amplitudes take on a different distribution; or if we send random data (like a PRBS (pseudorandom bit stream) test, which, generally, something like Morse code will approximate over time), then again a sinc sort of envelope but with random peaks and dips as the data comes and goes.

For accurate detection of data, a sinc spectrum can be bandpass filtered to the first zero, which corresponds to a bandwidth of two bit times (so, if the bits are alternating 1/0, that's a square wave of that frequency, so, needs at least that much BW to see).

To complicate the analysis, morse code is typically sent with amplitude shaping to avoid the square-edge "key clicks".  This filtering/shaping isn't well-defined, but the pulse rise and fall times are usually (eyeballing it) perhaps a few milliseconds. This significantly reduces the sidebands beyond the first few harmonics of the keying rate.

I was recently playing with Bessel filtering of FSK frequency transitions, and the effect of even a little filtering can be dramatic when compared to square-edge transitions.  That was frequency-domain rather than amplitude-domain, but the principals still apply.

In practice there is yet another limitation caused by "ringing".  I'm not a ham, so not sure about the name, Romanian hams call it "the bell effect (as in ringing)", in Ro spelled "efectul de clopot".  It can be hear when one applies a very narrow (high Q) bandpass filter, with the reason of reducing the noise level, but then, if the bandpass filter is too narow, a single pulse/blip sounds like a reverbing/ringing tone instead of a clean beep, spilling into the following simbols because of the sin(x)/x shaped response of the very narrow bandpass filter.

I think it's the same kind of problem that is causing intersymbol interference in digital transmission https://en.wikipedia.org/wiki/Intersymbol_interference, and AFAIK this is inescapable, will appear anytime one tries to cram too faster data speeds into a too narrow band.

The attachments were made with Audacium (thanks cirip for the idea), by recording some random YT morse sample, then applying a very narrow bandpass filter with 55dB attenuation.

Note how after filtering it almost sounds like a continuous beep, and this is without noise.  With noise will be even worst.  This is to illustrate what happens in the time domain, and how ringing affects the signal when the band is too narrow, or when the symbols speed is too fast for a given band.

A band-limited signal cannot be time-limited, so any attempt to limit the bandwidth of (the time-limited) symbols of the NRZ signal will unavoidably stretch the symbols in the time domain (leading to overlapping symbols in the time domain then). However, for transmission over a band-limited channel you have to limit the bandwidth, and you have to live/deal with the resulting ISI. The keyword is "pulse shaping", see https://web.stanford.edu/class/ee179/lectures/notes15.pdf.

If the binary NRZ signal should be recoverable from the pulse-shaped signal in a simple way (e.g. with a comparator), then I would not push the bitrate beyond BW/2 (where BW is the total available bandwidth for lower and upper sideband together). The attached image shows a (random) binary NRZ signal, with raised cosine shaping applied, using a roll-off factor of 1. Obviously the binary NRZ signal is well recoverable with a comparator with threshold=0.5. The amplitude of the eyes is only 60% of the maximum, but the threshold crossings show almost no jitter. Ringing is low, due to the conservative roll-off factor.

EDIT:

Also tried to push things to the limit, using (almost) sinc pulses, which doubles the bitrate/BW ratio. The eyes are still open, but the threshold crossings show significant jitter now, so that the timing of the edges of the original NRZ signal can no longer be reconstructed with a simple comparator, but a more sophisticated timing recovery is necessary in order to hit the eyes. Added spectrum and eye diagram for the simulated sinc pulse shaping.
« Last Edit: June 06, 2022, 05:42:27 pm by gf »
 

Offline Wallace Gasiewicz

  • Super Contributor
  • ***
  • Posts: 1160
  • Country: us
Re: How fast can Morse code be sent
« Reply #20 on: June 06, 2022, 08:36:27 pm »
This is from the ARRL book for the Extra Exam:
Bandwidth of CW signal:
Bw=B X K

B is speed in bauds, K is a factor in keying envelope. As the CW rise and fall times get more abrupt K gets larger., more harmonics.
One Baud is one element per sec.

Bw= WPM/1.2 times K                  approximate using PARIS as morse transmission for Baud Rate

If rise and fall times of 5 millisec K=4.8
So bandwidth of 13 wpm is about 52 Hz

I have known people who could understand Morse code at unbelievable speed, maybe like 100 wpm.
My father was a communications officer in WW2 and could "copy" at 35 WPM (more?) This means hand printing five letter nonsense code, no real words. He knew other experts who could do this at 65 WPM, but they could type, he could not type. Theoretically you could "copy" as fast as you could type. of course there were no electric typewriters for this, although there were RTTY machines, I suppose. I think 100 wpm is pretty fast typing although people have typed at 200 wpm. I certainly cannot.
He could understand Morse at maybe 60 to 100 wpm, but this was conversation, not letter for letter "copy". Remember "copying' meant no errors because after you interpreted the 5 letter "copy" you needed to look it up in a book or on a machine, I suppose you could bomb the wrong town!!!!  This would be even more important when you are flying over Japan in a B 29 with one big bomb.
I also new a guy who spent his weekends with the USAF flying over the north pole copying Morse signals from Russia in the 60s. He fully admitted he did not understand anything he received, he just typed it. He could send and receive at 100 wpm. Quite the speedy guy.. Very few people liked to communicate with him over CW.
There are types of signals with much more narrow bandwidths, but the original binary (digital) transmission is still alive.
 
The following users thanked this post: boB

Online fourfathom

  • Super Contributor
  • ***
  • Posts: 1851
  • Country: us
Re: How fast can Morse code be sent
« Reply #21 on: June 06, 2022, 09:09:23 pm »
Rather than how fast, improving the SNR is quite interesting.  Ham radio is now using modern coding, synchronization, demodulation, and decoding (including FEC) to be able to receive signals that would be well below the noise floor of a 500 Hz filter.  WSPR and FT8 are examples of this.  But way back in 1975, in the ham magazine QST, Raymond Petit described his "Coherent CW" system that used synchronization and matched digital filters to gain about 20 dB over regular CW.  In Coherent CW, decoding is still done by the human ear, but the filtering (10 Hz bandwidth) is done using digital timing and analog integrate-and-dump filters.  Once receive timing is aligned with the transmit timing (also digitally synch'd) there is no ringing even though the filter bandwidth is identical to the symbol rate.  I never built one of these systems, but I learned a lot by reading the articles:
http://www.arrl.org/files/file/Technology/tis/info/pdf/7509026.pdf
http://www.arrl.org/files/file/Technology/tis/info/pdf/8106018.pdf
We'll search out every place a sick, twisted, solitary misfit might run to! -- I'll start with Radio Shack.
 
The following users thanked this post: boB

Offline radiolistener

  • Super Contributor
  • ***
  • Posts: 3282
  • Country: ua
Re: How fast can Morse code be sent
« Reply #22 on: June 12, 2022, 06:11:44 am »
Rather than how fast, improving the SNR is quite interesting.  Ham radio is now using modern coding, synchronization, demodulation, and decoding (including FEC) to be able to receive signals that would be well below the noise floor of a 500 Hz filter.  WSPR and FT8 are examples of this.

it requires time synchronization. So it needs internet connection or GPS to synchronize local clock. Using very low bandwidth also requires precise frequency standard because very low bandwidth needs a precise tuning :)

By the way, machine decoding CW is a very interesting task. I've been wanting to try something like this for a long time. But this is not an easy task because hand-keyed manual morse CW has a variable speed, so for machine algorithm it's hard to distinguish dots, dashes and pauses. Such a task needs to analyze overall context.

Recently I found interesting research paper about machine decoding CW:
https://apps.dtic.mil/sti/pdfs/ADA046503.pdf

Here is also article about machine decoding CW that I found:
https://www.pa3fwm.nl/software/rscw/algorithm.html

And here is very interesting project on github for a Morse decoder based on the code orinally created by Dr. E. L. Bell in 1977:
https://github.com/ag1le/morse-wip
« Last Edit: June 12, 2022, 06:34:53 am by radiolistener »
 

Online fourfathom

  • Super Contributor
  • ***
  • Posts: 1851
  • Country: us
Re: How fast can Morse code be sent
« Reply #23 on: June 12, 2022, 03:13:47 pm »
By the way, machine decoding CW is a very interesting task. I've been wanting to try something like this for a long time. But this is not an easy task because hand-keyed manual morse CW has a variable speed, so for machine algorithm it's hard to distinguish dots, dashes and pauses. Such a task needs to analyze overall context.

I wrote a couple of morse decoders in Pascal back in the early IBM-PC/DOS days, hooking into the BIOS timer interrupts and sampling one of the RS232 input pins (which was connected to a hardware tone decoder I built).  My program used an adaptive digital filter to track the dot and dash timing and it worked pretty well on hand-sent code.  No context analysis, just element timing, and pretty simple at that.  One could probably improve the performance by attacking it as a cryptographic problem, but my simple approach was adequate for my purposes.

Thanks for the links -- I will definitely look at them.
We'll search out every place a sick, twisted, solitary misfit might run to! -- I'll start with Radio Shack.
 

Online fourfathom

  • Super Contributor
  • ***
  • Posts: 1851
  • Country: us
Re: How fast can Morse code be sent
« Reply #24 on: June 12, 2022, 04:07:33 pm »
Recently I found interesting research paper about machine decoding CW:
https://apps.dtic.mil/sti/pdfs/ADA046503.pdf

Here is also article about machine decoding CW that I found:
https://www.pa3fwm.nl/software/rscw/algorithm.html

And here is very interesting project on github for a Morse decoder based on the code orinally created by Dr. E. L. Bell in 1977:
https://github.com/ag1le/morse-wip

Yes, interesting papers!  The first goes *way* over my head theoretically, but Table IV (page 35) is a chart that might give an answer to this threads original question.

The second uses a software decoder with a demodulator that (as the author points out)  is very similar to the Coherent CW system I mentioned previously.  He then uses some simple dot-space-dash analysis, followed by a parallel cross-correlator that detects potential characters (or phrases?).  I wonder how well this works in practice, as I don't think there's enough redundancy in the Morse character set for the correlator to add much value?  This program is designed for machine-sent telemetry code and can't track the human fist.

The third is a C++ port of a Fortran program and appears to use some heavy-duty DSP.  The source files are named "FFT", "Kalman", "Trellis", etc. 

My own program, lost in the mists of time, was quite naive, using simple single-stage IIR filters for debouncing and speed-tracking, and using the character-space as a delimiter for the pattern-to-character function.
We'll search out every place a sick, twisted, solitary misfit might run to! -- I'll start with Radio Shack.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf