Good Morning All:
Is there any visible difference between the several available software sound-card solutions for decoding SSTV? Other than soundcard noise, I expect that the faster the soundcard samples, the better?
I've been using MMSTV and really like it. Wanted to also try HRD/DM780 but cant seem to get it to work just yet.
Having a great time with ARISS!
Thanks,
Curt KU8L EN82hi
Curt KU8L asked:
Is there any visible difference between the several available software sound-card solutions for decoding SSTV? Other than soundcard noise, I expect that the faster the soundcard samples, the better?
Even the crummiest sound card will record everything there is to record from a communications-grade audio channel.
In the case of SSTV you are doing frequency demodulation, and the more samples, the better. With modern computers the compute horsepower required is a minor consideration.
All the usual references...
Laura Halliday VE7LDH "Que les nuages soient notre Grid: CN89mg pied a terre..." ICBM: 49 16.05 N 122 56.92 W - Hospital/Shafte
_________________________________________________________________
Hi Laura
All the usual references...
Which are?
The only way I've been able to create a number of formats is from copyrighted texts. Do you know of detailed texts that are public domain? If so let's have 'em!
73, Howard G6LVB
Curt Nixon wrote:
I expect that the faster the soundcard samples, the better?
It's counterintuitive, but through the magic of sampling theory, once you sample at twice the maximum frequency of the signal you are interested in you know _everything_ about it! As a practical matter, most ham gear audio rolls off sharply above 3 kHz (and SSTV is lower than that) , so sampling at a measly 8 or 11.025kHz rate is plenty.
Howard Long wrote:
Do you know of detailed texts that are public domain? If so let's have 'em!
And it's not public domain, but it is available as free software: check out QSSTV for source code.
If you are interested in sampling theory, check out http://www.dspguide.com/ch3/2.htm
-Joe KM1P
Joe Fitzgerald wrote:
Curt Nixon wrote:
I expect that the faster the soundcard samples, the better?
It's counterintuitive, but through the magic of sampling theory, once you sample at twice the maximum frequency of the signal you are interested in you know _everything_ about it! As a practical matter, most ham gear audio rolls off sharply above 3 kHz (and SSTV is lower than that) , so sampling at a measly 8 or 11.025kHz rate is plenty.
Hmm. That's not quite true. Consider a signal at 4kHz, being sampled at 8kHz. What you'd see is a triangle wave, if you were sampling at the peaks of the incoming signal (if you were sampling at the crossover points, you'd see no signal at all!). How do you tell what the waveform of the signal originally was? You can't...
If you sample your 4kHz signal at 16kHz, you've got four points across each cycle, so at least you can start to get an approximation. If the input signal was a sinewave you might see a sample at a crossing point, then a sample at a peak, then a sample at the next crossing point, then a sample at the next peak. You'd get a roughly sinewave-y signal, if you squint a bit.
When I'm recording digital signals or SSTV I use 44.1kHz 16-bit mono .wav files. Disk is cheap, and the ISS doesn't fly over that often...
Howard Long wrote:
Do you know of detailed texts that are public domain? If so let's have 'em!
And it's not public domain, but it is available as free software: check out QSSTV for source code.
It's annoying how many really good pieces of software for amateur radio are closed-source Windows-only things. I don't know about anyone else, but I think that rather goes against the spirit of amateur radio. I'd rather play with something hacker-friendly that I can take apart and adapt than some big sealed-up box'o'tricks. I wrote a bunch of music synthesis plugins and made them available free (free as in beer and Free as in speech). Other people have sent me back little tweaks and improvements, that I've folded back into the code. That way, *everyone* gets a better toy to play with, and we all get to learn something.
If you are interested in sampling theory, check out http://www.dspguide.com/ch3/2.htm
At some point I must finish my rewrite (for copyright reasons) of the Ensoniq Mirage sampling guide. It's got a great explanation of it.
Gordon MM3YEQ
The filtering ensures that the shape is known. A triangle wave above half the cutoff frequency becomes a sine wave as filtering removes the harmonics.
73
John KD6OZH
Hmm. That's not quite true. Consider a signal at 4kHz, being sampled at 8kHz. What you'd see is a triangle wave, if you were sampling at the peaks of the incoming signal (if you were sampling at the crossover points, you'd see no signal at all!). How do you tell what the waveform of the signal originally was? You can't...
If you sample your 4kHz signal at 16kHz, you've got four points across each cycle, so at least you can start to get an approximation. If the input signal was a sinewave you might see a sample at a crossing point, then a sample at a peak, then a sample at the next crossing point, then a sample at the next peak. You'd get a roughly sinewave-y signal, if you squint a bit.
According to Mr Nyquist (I think that's who it was) you only need to sample at twice the highest frequency contained in the sampled waveform but remember that here we are talking about twice the frequency of the highest harmonic contained in the waveform so 8KHz wouldn't be sufficient for a 4KHz square wave but would be sufficient to reconstruct a 4KHz sine wave..... or is my theory screwed up as well?
Gordon JC Pearce MM3YEQ wrote:
Hmm. That's not quite true.
Nope...you got it just right!
As Simon pointed out, its all about knowing what information you want to get from the "re-constructed" wave. If you want all the detail, you have to sample at least 2f of the highest harmonic content. Square waves with fast rising leading edges are a real mutha!! Its why we sold alot of multi-gig one-shot samplers! For repetitive waveforms, averaging with a sample rate much slower and randomly distributed across the signal results in good reconstruction but is not "realtime" or one-shot capable.
Curt
Nigel Gunn G8IFF/W8IFF wrote:
According to Mr Nyquist (I think that's who it was) you only need to sample at twice the highest frequency contained in the sampled waveform but remember that here we are talking about twice the frequency of the highest harmonic contained in the waveform so 8KHz wouldn't be sufficient for a 4KHz square wave but would be sufficient to reconstruct a 4KHz sine wave..... or is my theory screwed up as well?
Gordon JC Pearce MM3YEQ wrote:
Hmm. That's not quite true.
Sent via [email protected]. Opinions expressed are those of the author. Not an AMSAT-NA member? Join now to support the amateur satellite program! Subscription settings: http://amsat.org/mailman/listinfo/amsat-bb
Hi Joe:
I beg to differ with you on the sampling information. Having been a digital signal acquisition and instrumentation specialist for a number of years, I'm quite familiar with the nyquist criteria and all. Sampling at 2f, you know the frequency and something about amplitude of the signal and thats about all. That stuff about knowing ALL about it is just not true--"if" you knew ahead of time, that the signal in question was nothing but sinusoidal, then yes, you are correct--but how often is that the case. That is what has driven highly oversampled AD conversion. If you meant twice the maximum frequency COMPONENT of the signal, then thats close, but is still based upon sin theory (fourier series) and leaves alot of detail out of non-sinusoidal parts of the signal unless one considers all of the harmonic content of the source. For investigative or diagnostic work, especially characterizing waveshape leading edges of clocks, etc, we used to recommend sample rates of 5-8 times the highest frequency component of interest.
I wasn't so interested in the actual SSTV algorithms as I was in seeing if there is any visible difference in the result of different algorithm use--or is everyone using the same decoding kernel?
Thank you Sir, I suspect you know all this anyway--thought I would clarify it as I came to understand it.
Curt
KU8L
Joe Fitzgerald wrote:
Curt Nixon wrote:
I expect that the faster the soundcard samples, the better?
It's counterintuitive, but through the magic of sampling theory, once you sample at twice the maximum frequency of the signal you are interested in you know _everything_ about it! As a practical matter, most ham gear audio rolls off sharply above 3 kHz (and SSTV is lower than that) , so sampling at a measly 8 or 11.025kHz rate is plenty. Howard Long wrote:
Do you know of detailed texts that are public domain? If so let's have 'em!
And it's not public domain, but it is available as free software: check out QSSTV for source code.
If you are interested in sampling theory, check out http://www.dspguide.com/ch3/2.htm
-Joe KM1P
Hi,
Briefly - the SSTV developers have their own algorithms more-or-less, what we are mostly interested in is the frequency at a point in time. Simplifying things further - we need to know frequency for each point in the line, the filtering we use to pre-process the incoming signal removes noise etc. After all - it's an analogue specification.
There would be no advantage to sampling at 48kHz instead of 8kHz, in fact it would use a lot more CPU to run the filters.
The filtering we use is very standard stuff, SSTV is a nice and simple concept, the big problem is combating noise, fading and multi-path interference.
Simon Brown, HB9DRV www.ham-radio-deluxe.com
----- Original Message ----- From: "Curt Nixon" [email protected]
I wasn't so interested in the actual SSTV algorithms as I was in seeing if there is any visible difference in the result of different algorithm use--or is everyone using the same decoding kernel?
Don't forget that we had perfectly functional SSTV software 10 years ago with JVFAX, which didn't even use a sound card. Just a 1-bit interface - an opamp tied to the control lines of a serial port on an 8 mHz 286 PC running MS-DOS. Scottie-1 on 20 meters, yeah! It didn't decode Robot-36, but I figured out that I could receive the SSTV pictures from MIR by using one of the B&W modes.
Greg KO6TH
----------------------------------------
From: [email protected] To: [email protected]; [email protected] Date: Fri, 17 Oct 2008 15:20:33 +0200 Subject: [amsat-bb] Re: SSTV Algorithms?
Hi,
Briefly - the SSTV developers have their own algorithms more-or-less, what we are mostly interested in is the frequency at a point in time. Simplifying things further - we need to know frequency for each point in the line, the filtering we use to pre-process the incoming signal removes noise etc. After all - it's an analogue specification.
There would be no advantage to sampling at 48kHz instead of 8kHz, in fact it would use a lot more CPU to run the filters.
The filtering we use is very standard stuff, SSTV is a nice and simple concept, the big problem is combating noise, fading and multi-path interference.
Simon Brown, HB9DRV www.ham-radio-deluxe.com
----- Original Message ----- From: "Curt Nixon"
I wasn't so interested in the actual SSTV algorithms as I was in seeing if there is any visible difference in the result of different algorithm use--or is everyone using the same decoding kernel?
Sent via [email protected]. Opinions expressed are those of the author. Not an AMSAT-NA member? Join now to support the amateur satellite program! Subscription settings: http://amsat.org/mailman/listinfo/amsat-bb
_________________________________________________________________ When your life is on the go—take your life with you. http://clk.atdmt.com/MRT/go/115298558/direct/01/
participants (9)
-
Curt Nixon -
Gordon JC Pearce MM3YEQ -
Greg D. -
Howard Long -
Joe Fitzgerald -
John B. Stephensen -
laura halliday -
Nigel Gunn G8IFF/W8IFF -
Simon (HB9DRV)