Retro SSTV anyone?

[Klaas Robers]

Steve, it is already for quite some time that I wonder if an FM demodulator or discriminator shouldn't have an S-curve that is linear over the whole bandwidth of the FM spectrum. That is for SSTV it should be straight from 300 Hz (1200 - 900) up to 3200 Hz (2300 + 900). For the lower limit it could be 600 Hz (1500-900) because I allow the demodulated sync pulses to be distorted, as the sync is extracted in a different way.

Although, one might think that the frequency never goes lower than 1200 Hz (1500 for the video content) and higher than 2300 Hz. But the side bands are not in the signal just for fun. If you demodulate the FM with two tuned circuits, one at 1200 Hz and one on 2300 Hz, as ROBOT did, then you mis the side bands between 2300 and 3200 Hz, isn't it? At least they are attenuated in stead of boosted.

If I watch in this thread the page 10 then I see that for your demodulated "Blokgolf" signal, that in the upgoing black to white transients there is no ringing of the low pass filter, but a somewhat rounded shoulder, while at the white to black transients there is ringing and it looks as if the transients are steeper. In the same page a photo of an oscilloscope screen shows ringing up as well as down from a grey scale signal.

However, video signals are strange things. A picture has other requirements than audio.

[Steve Anderson]

I guess it depends on your modus operandi. If you're someone digging through all the dirt on the HF bands you'll want to keep out as much of the garbage as possible, even if it means a loss of definition. A picture is better than no picture, even if imperfect.

"The more you open the windows the more dirt blows in."

If like myself you're working with almost perfect conditions (closed circuit) then yes, there is a case for opening up the demodulators bandwidth.

I chose to copy/improve(?) on the ROBOT design as it represented what seems to have been the better of demodulators in the 70s. It would seem that the others were based on slope demodulation from FM to AM then rectification to baseband or zero-crossing techniques. All valid, all worked, to varying degrees.

So the 'improved' (dare I say) ROBOT demodulator was/is a reference, the best of what the 70s customer/price ratio could be tolerated. Can this be improved upon? Of course it can.

Now we could simply widen the peaks of the ROBOT filters design and fudge any non-linearity that came along, which isn't that hard. In fact I may have a go at it and see what improvement there may be - though it won't be tomorrow.

What is your thinking on the matter Klaas?

Steve A.

[Harry Dalek]

Klass i found this reminds me of your monitor rebuild Heathkit case look and all .

[Klaas Robers]

Well Steve, my thinking is this:

- When receiving on HF the X-tal filter of my receiver suppresses everything that is outside the pass band of say 2.4 kHz width. That should be sufficient. So most of the upper side band of the FM-signal is heavily attenuated by that action.

- But then the audio is amplified and limited. The limited audio must follow the mathematical rules of an FM-signal. So an upper side band, 2300 - 3200 Hz, is restored again. The information for that is extracted from the remaining part of the signal, say 1400 - 2300 Hz. And yes, that implies that those side band signals will be somewhat lower in amplitude. I don't know if that implies that the side bands are attenuated by 3 dB or by 6 dB compared to originally, but they are both present. This trick is also used in video tape recording, where the spectrum is sloping down in the carrier part of the signal as well. When you do this in an apropriate way (-6 dB per octave?) then the spectrum of the FM-signal is restored completely.

- So bandwidth limitation should be done in the domain where the audio is not yet limited.

- In the first designs you see indeed that FM to AM conversion is done by one single tuned circuit. The drawback is that you cannot influence the shape or linearity of the detection curve. In general this will be non linear, but that introduces only a slight change in the gamma. We know that slight changes in gamma are invisible.

- Robot uses two tuned circuits, one AM signal is rectified positively, the other negatively. Then, by fiddling with the Q of both tuned circuits you can end up with a part of the resulting slope that is rather well linear. But I still think that the linear part should be larger than what they did. At least it should be 600 to 3200 Hz. It would be nice if for an experiment you could extend this bandwidth and see if the ringing close to 2300Hz is present again.

[Klaas Robers]

Very good Harry.

Now that I had a type number I could find the extensive description in the archives of 73 magazine of January 1973. I wasn't aware that by that time some US based company had made already a Heathkit-outlined SSTV monitor.

The strange thing is that this monitor has just 4 controls, while mine has 9 of them. However two are to select the 72 built in test pictures and one is for the volume control of my built in loudspeaker with amplifier. But then still the 6 remain that my original 1973 "wooden" monitor had already. But I read that the sync settings (frequency tuning and tripping level setting) are internally presetted in this monitor and then indeed the 4 controls are left. Unhappily the function of them wasn't mentioned in the longer article.

Thanks for attentioning me onto this product.

[Steve Anderson]

Given the time I'll have a go at re-positioning the frequencies of the two tuned circuits within the Robot demodulator. Unfortunately it's not quite as simple as a LC tuned circuit and potentially several passives in each one will require changing - definitely a breadboard exercise! At least at first.

The original is surprisingly linear, the Robot guys must have fiddled around with values for a long time given the fact that they had no simulation software in those days. Although I replaced the passive rectifiers with active ones, the filters/tuned circuits are exact copies of the Robot version. The slight deviation in the centre portion of the chart below is probably down to my measurement errors.

I have also included the 'as built' version of the circuit, there were one or two minor changes at the end.

Steve A.

[Harry Dalek]

Very good Harry.

Now that I had a type number I could find the extensive description in the archives of 73 magazine of January 1973. I wasn't aware that by that time some US based company had made already a Heathkit-outlined SSTV monitor.

The strange thing is that this monitor has just 4 controls, while mine has 9 of them. However two are to select the 72 built in test pictures and one is for the volume control of my built in loudspeaker with amplifier. But then still the 6 remain that my original 1973 "wooden" monitor had already. But I read that the sync settings (frequency tuning and tripping level setting) are internally presetted in this monitor and then indeed the 4 controls are left. Unhappily the function of them wasn't mentioned in the longer article.

Thanks for attentioning me onto this product.

NOt much on this monitor about ....found this if its of any help ...

Bear in mind that 'MXV monitors' noise immunity circuits are only open to vertical sync pulses ever/ picture "bottom" and thus it will not reset to the "top" every 2 seconds. Both units are superb, however.

June 75 73 mag has a mention in the sstv monitor build .

[Steve Anderson]

After being away for longer than I had hoped, I've had a chance to try out the slightly different manner of SSTV demodulation. And lo and behold - some encouraging results.

The chart below shows what was the expected output voltage (dark blue) and the measured voltage (magenta). I think I know what the slight difference is but I need to look into it first. It could also be my DVM needs calibrating, it's been many years since the last time!

These are static spot-frequency measurements, I need to asses it yet under dynamic conditions, i.e. with modulation.

X=Frequency (Hz).
Y=Voltage (V).

Steve A.

Harry, I will get back onto your SSTV monitor.

[Steve Anderson]

Well, I've thrown a few previous files I generated for the SSTV up-converter at it. Results - no overshoot or ringing, it resolves the 900Hz in the multiburst file to about 90%.

At the moment the signal is a bit noisy probably because it's roughly built on a breadboard. I will build a 'cleaner' version which I hope will get rid of the noise.

More to follow.

Steve A.

[Harry Dalek]

OH Steve its very similar to Klass's results he posted on my SSTV posting in a way ,i suppose you have both made different demodulators with a good 40 or so years between yours and Klass ....well it works which i am sure your happy about .

I knew you were busy thats fine always just grateful for the help when you have time .

In the mean time i was testing my little modules on a cold cathode tube the one i am using looks weak against a fly back i have tried before but a much cleaner DC than both a fly back and my other module which was just AC even trying that one with a micro wave HV diode i was not getting results as clean as the monitors Module which must have at least a capacitor if not better rectification ...sure showed up visually between the cathode and anode what was good and what wasn't .

OH well its clean just see how it go's when needed ,if its not up to the job i can always make a flyback circuit for it .

[Steve Anderson]

I've found half of the error in the last chart. The output uses the 5V rail as a reference which measures 5.05V on the same DVM and the same range. Using the value of 5.05 in the chart cancels out any meter or supply error, what is left is the usual component tolerences etc..

There is one other possible error source, I will try a fix today but I doubt it will make any visible difference on the chart.

Steve A.

[Steve Anderson]

The 'one other possible' error source is the culprit. I didn't think it would make any noticeable difference, but as you can see below - it did.

There are truly two lines on this chart, colours as before.

Now that's done, time to move onto sync detection.

The source of noise isn't the lash-up breadboard arrangement, but once again the PC.

Steve A.

[Harry Dalek]

Spot on overlapping what was expected and real world measurement ,does not get better !

[Klaas Robers]

Very good Steve !!
I like those straight lines. What I don't understand is why the line starts at zero volts and 500 Hz. And what happens with the output voltage at input frequencies below 500 Hz? Goes it negative?

In fact I am curious what you did. It is possible that you introduced an offset to force 500 Hz to an output of zero volts, but you could also have used a Physical principle that has a certain bottom limitation.

And it is good that you have seen that the demodulator detects the 900 Hz multi burst bar with 90% amplitude. This strongens my idea about the need for the S-curve to include the first order upper and lower side bands of the FM signal.

With this demodulator you will also see steeper upgoing edges on the grey scale video, and constant overshoots, not dependant from the grey level. And as said, don't be affraid for overshoots, they give a welcome sensation of sharpness. However in principal the overshoots should be symmetrical. That is an undershoot before the transient and an equally large overshoot after the transient. With a transversal FIR filter this is easily possible. Much better than the all pass that I made following the 7th order butterworth low pass.

[Steve Anderson]

Yes, I have forced the output at 500Hz to 0V and at just under 3200Hz to 5V. Originally the origin of the line was 0Hz/0.0V, I just added an offset so that over the frequencies of interest I get the largest signal. The calculated output signal for SSTV should be 2.054V (syncs to white), measured with tones it's 2.06V and with a real SSTV signal it's also 2.06V. (Syncs at +1.29V, white at +3.35V).

Though I haven't checked it yet, below 500Hz it should stay at 0.0V, above 3200Hz I'll probably get it to do the same - or perhaps to stick at 5V...not decided yet. It's a nicety I'll fiddle with later.

There are no overshoots at all...but I take your point about apparent sharpening. That can also be considered later, at the moment I'm more concerned with proof of concept.

I've made a start on sync detection and I'm thinking about sync detection bandwidth. I've seem many and various figures thrown around, anything from +/-50Hz (1150-1250Hz) to +/-200Hz (1000Hz to 1400Hz). somewhere between those two arrangements is where I'll aim for, but I'm open to suggestions.

Klaas, I notice in your recently re-built monitor that you can also can tune the sync frequency, I'm wondering is that a useful addition? If so I'll consider that too. Though if the sync frequency is off it's probable that the video portion will be too.

Yes Harry...it's nice when a plan comes together!

More to come.

Steve A.