SSTV Up-Converter Mk VI

[Steve Anderson]

From the Mk I SSTV-to-625 SSTV up converter a jump now to the Mk VI version, Mk II-to-Mk V were design exercises that never made it to publication. This is a work-in-progress thread so it's not presented as a fait accompli, and probably will undergo modification as the design develops. This will not happen overnight, so some patience may be needed.

It is possible a PCB could be produced if there is sufficient interest shown once finalised, if ever that happens!

As previous, the input is a monochrome SSTV signal of 120/128 lines in a 1:1 aspect ratio at either a 15Hz (US) or 16.7Hz (EU) line-rate. The SSTV subcarrier is 1200Hz for Sync, 1500Hz for Black, and 2300Hz for white. Sync duration is 5ms for line-sync, and 30ms for frame-sync. This basically defines the original Cop McDonald SSTV standard in terms of frequencies and timing.

It is hoped to produce two versions, the first with a monochrome 625 output suitable for standard EU TVs/monitors. The second with a monochrome 800x600 60Hz VGA output. It may be possible to eventually produce a version switchable between the two output standards, though not both simultaneously.

The hoped-for simplification is physically smaller RAM and Read processor chips and a reduction in associated wiring. Also an improvement in demodulator filtering to reduce/eliminate overshoots with large/rapid transitions within the recovered baseband SSTV signal. This was a major flaw in the Mk I version to me...entirely my own fault!

Comments/feedback welcome at this early stage, easier to change things now rather than later...

Steve A.

For those wondering, no, I'm not interested in the more recent SSTV modes with higher resolutions and/or colour. I concede I'm a Luddite! I also should admit I have never put together a true P7-based CRT SSTV display. I did start on one many years ago, it's still sitting on a shelf here incomplete, covered in a thick layer of dust!

[acl]

Hi Steve,

Looking forward to your progress.

Regards Chris

[Steve Anderson]

Thanks Chris. One thing I should mention is regarding the 625-output version, I don't intend to do a 525 version, most 'reasonable' TVs from the 90s or later will correctly display 625 or 525 in PAL or NTSC (not that the colour standard matters here). The 120/128-line SSTV standard 'fits better' into a 625 raster rather than a 525 one.

VGA gets around that problem, 800x600/60Hz non-interlaced is (was?) a worldwide standard. The downside - greater bandwidth required but quite achievable even on veroboard/stripboard with a compact layout.

In addition VGA requires three video signals (R,G.B) plus separate Horizontal and Vertical syncs. The signal being monochrome. the three RGB video signals can be identical. Some monitors may accept 'syncs on green', but I wouldn't count on it.

Steve A.

[Dave Moll]

The signal being monochrome. the three RGB video signals can be identical.

Would the red and blue signals require suitable attenuation to achieve a neutral monochrome display?

[Steve Anderson]

Would the red and blue signals require suitable attenuation to achieve a neutral monochrome display?

Not within the VGA standard, or analogue RGB which VGA is a version of. All three signals are 700mV from blanking/black to full 100% amplitude, usually with a circuit impedance of 75 ohms.. This should produce a pure white raster when all three are at 700mV. Sync signals are usually, but not always separate, usually at TTL levels and of varying polarities.
Component video (YUV or Y, Cb, Cr) is different, the luminance Y is basically a full bandwidth B&W signal as per monochrome TV with syncs. The U signal is the bipolar (+/-) blue-difference signal, the V signal being the red-difference version. There's no syncs on the U & V signals. Mostly the U & V signals are of reduced bandwidth similar to composite PAL/NTSC/SECAM video, where the U and V signals are encoded within the subcarrier. This encoding allowed for compatability between colour and monochrome receivers...the trade-off was a reduction in resolution, especially in the colour difference channels and the green signal derived from them.

Note the frequent use of the words "mostly" and "usually", there were/are many exceptions and variations.

Steve A.

An overview of component (RGB,YUV) video is here...it's a shame that the words "composite" and "component" are so similar...

https://en.wikipedia.org/wiki/Component ... nent_video

[Steve Anderson]

The part I intend to tackle first in this is the post-demodulation filter as this for me is somewhat of a bug-bear. Previously I used conventional analogue filters, Butterworth, Chebychev, Bessel and the like[1], and although most 'sort of' worked OK in terms of frequency separation passing up to 1kHz and rejecting all above 2.4kHz (the second harmonic of sync at 1.2kHz), the phase response is not linear resulting in large overshoots on rising/falling edges within the video waveform. Though this possibly had the beneficial side-effect of edge-sharpening on lesser amplitude signals.

One method to reduce/eliminate this may be the use of an FIR filter instead. The signal needs digitizing anyway, so why not try it? The rest of the system is straight-forward digital signal processing and D-A output as previous. There is a chance that this FIR method may fail, but at least it was tried. As it's mainly a software exercise there should be few hardware implications if it goes nowhere. Though a filter of some sort is still needed.

Steve A.

[1] I did also try some supposedly 'phase-linear' analogue filters, they were only marginally better. i.e. still not good enough.

[Klaas Robers]

From my past experience on TV systems I know that Butterworth is the best filter characteristic for video signals. The cut-off or fall-off is moderate and the overshoot is not too large for video signals. An all-pass in succession for delay compensation will divide the pre-shoot and after-shoot to be equal. This is comparable to a FIR-filter, which also should follow the Butterworth character. So the choise between analogue or FIR is independent to the cut-off character.

I will follow your design. It was too difficult for me.

[Steve Anderson]

Klaas, I agree, the use of an FIR filter is a new area to me, but the potential advantages are worth a try. Initially I'll just try it as a 1kHz low-pass audio filter to ascertain its worth, if it seems OK I'll add the rest of the demodulation circuitry and see how it all performs.

FIR filters are supposedly linear-phase, but I doubt that's the end all and be all of it. I'm sure there will be a 'gotcha' or 'but' or two...I expect there may be some over/undershoot still evident, but hopefully less.

Other potential advantages are far fewer op-amps, few/no close tolerance and sometimes awkward value components (inductors), as well as less board space required.

Steve A.

Here's an example of the overshoot that really annoyed me on the extreme left-hand side. A screen-shot of an old JVC colour TV where the SSTV image has been up-converted to 625. If it were not for that overshoot I would leave it well alone, it's good enough otherwise for me!

I just checked, the filter is/was a 1kHz, -24db/octave Butterworth, two TL082 op-amps (one 8-pin DIL package).

YOKH Epsom UK 3.jpg (42.08 KiB) Viewed 555 times

[Steve Anderson]

Looking through my notes for when this MkI up-convertor was built (2015) I was reminded that there's also a filter at the baseband input of the SSTV modulator, a separate unit. Could the pair of them be compounding the problem? Worth looking into!

Steve A.

The next day...I've got a feeling that I checked this out some years ago and the modulator input filter was not really to blame. It contributed a small amount, but significantly less than the demod filter, say 20% or less. it may have been (and perhaps is) a Bessel version. I need to find the data for the modulator which is here...somewhere!

Found the circuits and data for the modulator...indeed, the input filter is a 4-pole Bessel at 1.3kHz with a gain of 4.0 (see attached), so that confirms its lesser contribution...in fact the simulator showed an almost zero overshoot, the figure of 20% was from memory (unreliable) and may be simply capacitor tolerances (5%), resistors are all 1%, no inductors used.

So the plan of attack is to first confirm the modulator filter behaves as below, then confirm the demodulator filter is really as bad as thought. If so draw up a design for an FIR filter that has less over/undershoot. Though I guess there will still be some, but hopefully noticeably less. An FIR filter is linear-phase, not zero-phase.

Four-Pole Bssal 1-graph.gif (10.09 KiB) Viewed 521 times

[Steve Anderson]

Looking back over the screen-shots from 2015 of this MkI up-converter the overshoot that used to really annoy me then isn't so bad after all. If it can be improved, all well and good. For some reason I don't find it as irritating as I used to! Looking at other versions most display some evidence of this too.

Still working on simplification of the hardware, two PIC micros as before, but possibly 1x14-pin and 1x20 devices instead of 1x28 and 1x40-pin devices. RAM changed from 1x32-pin device to 1x8-pin device, possibly two.

Initially with 625/50 output with a view to a 800x600 VGA output later. The SSTV image would occupy 512x512 of the full screen.

I may have a go at the PD120 mode in the future, though I did say I'm not really interested in colour or hi-res SSTV modes. It would be an interesting challenge, but with the impending demise of the ISS its usage may decrease.
There is PC software that can decode it anyway, making a hardware method probably somewhat redundant.

Steve A.

[Steve Anderson]

A slight change to the processors here, the 14-pin device I was hoping to use simply doesn't have enough I/O. So it might become two identical 20-pin DIL devices, though loaded with different software for the read and write functions. The chips are looking like they'll be PIC18F13K22 or PIC18F14K22 devices at the moment.

The RAM to be either two 23LC512 or 23LC1024 8-pin DIL serial devices.

The 'glue logic' devices required (not that many) are generally a few 74ACxxx series for VGA output, though 74HCxxx should be OK for a 625-only version.

Steve A.