Page 2 of 4

Re: Update to SSTV demodulator

PostPosted: Thu Sep 19, 2019 6:09 pm
by Steve Anderson
Harry Dalek wrote:I feel looking at it again and your advice i think i will stick to the 2 diode idea ..


As I mentioned, it's fine for moderately light loads, and those where the drain on the two DC supplies are relatively equal. But if you have an large imbalance on the positive and negative supply load you end up with effectively DC passing through the transformer windings. This can lead to saturation of the transformer core leading to more heat - not good.

I repeat, "Get the right transformer for the job." Never skimp on power supplies, without them meeting the challenge your whole project is doomed.

I could quote Hans Summers from his website (http://www.qrp-labs.com), except I cant find the exact passage right now, but it's basically this...many failures or poor performance of our kits are simply down to poor power supplies...(he goes on to elaborate).

Steve A.

Re: Update to SSTV demodulator

PostPosted: Fri Sep 20, 2019 8:15 pm
by Klaas Robers
Steve, can you not incorporate the FM-detector in the PIC? I would measure the time lapse between the ajacent zero-crossings. This will give a number proportional to 1 / frequency, so with a look-up table that number can be converted into a number proportional to frequency. Then this running number can be put into the FIR-filter.

I tried to work with a counting detector, but that gave problems with the sampling frequency of the FIR-filter. But if you use a controller with a hardware multiplier, this may not be a problem any more.

Then you need only the limiting amplifier to convert from analogue audio to a square wave. No A-D converter needed.

Re: Update to SSTV demodulator

PostPosted: Sat Sep 21, 2019 3:44 pm
by Steve Anderson
I did try a zero-crossing detector some time ago, it actually wasn't that bad. I'm sure I documented it somewhere here...or at least the results..

The major problem with SSTV is that the modulating frequencies (0-1000Hz approx.) are so close to the sub-carrier (1200-2300Hz). So you really need to capture those FM sidebands and process them correctly.

There's a trade-off (as always) in what you decide is 'good enough' or perfection. There's nothing wrong with aiming for perfection, but at some stage we all have to say, "That's good enough for me." Or whatever other constraints that are placed upon you.(cost, complexity, time).

So, I'm going to stick with the demodulator pretty much as it is with the exception of the output filter...until something else better comes along then I'l embrace it.

I'm always open to ideas...

However, I will take a second look at the time-dependent alternative.

Steve A.

Re: Update to SSTV demodulator

PostPosted: Sat Sep 21, 2019 8:28 pm
by Klaas Robers
I think it could go this way:

1. Have the limited input audio as a binary input,
2. have a first interrupt at the upgoing edge of the input signal:
- that resets a first running timer,
- and reads a second running timer,
3. have a second interrupt at the downgoing edge of the input signal:
- that resets the second running timer,
- and reads the first running timer.
4. The latest read timer value is a value for the lapsed time between two edges in the input signal.
5. Convert that value to a frequency value by a look-up table (that is the fastest way).
6. That value becomes the input of the FIR-filter until it changes.

These interrupts should react very fast, because each jittering delay is a phase modulation = frequency modulation. That is why my previous experiments failed.

Re: Update to SSTV demodulator

PostPosted: Mon Sep 23, 2019 6:20 am
by Klaas Robers
Steve, I think the FM-detector described above CAN be used without a (FIR) low pass filter, because it has already a low pass action with a null at twice the audio frequency, so at 3 kHz for black, to 4.6 kHz for white. But when the samples are in the controller, why not let it be followed by a low pass filter. The higher frequencies only add noise.

Do you have already a sample frequency in mind for the FIR filter? I am thinking of 8 KHz for a low pass with a cut off of 1000 Hz. I don't know how many stages you can allow the FIR-filter to be with a sample frequency of 8 kHz. I never used a PIC with a hardware multiplier. How many T-cycles does a multiplication need? Is it more than an addition (1T)?

Re: Update to SSTV demodulator

PostPosted: Tue Sep 24, 2019 3:04 pm
by Steve Anderson
I did quite a long reply to the two posts above...but when I pushed the 'submit' button I got a reminder to be logged in - so my message simply vanished...I WAS logged in - that I know for sure. This used to happen some years ago, a signed-in user got 'dropped' part-way through composing a reply...

I'll have another go some other time...

Steve A.

Re: Update to SSTV demodulator

PostPosted: Tue Sep 24, 2019 5:45 pm
by Klaas Robers
I see Steve, I have had the same experience. Then it took too long to finish the post because something happened in the mean time, it interrupted the writing process. Now when I fear this to happen, I select the whole text and type Ctrl-C (copy) before I hit "Submit". Then I can restore the text any way after relogging in.

What I also do now and then is to ask for a preview half way. By doing that the bulletin board knows at least that I am still doing something and a new time-out is started. But I know, writing this brings not back all your work. Sorry.

Re: Update to SSTV demodulator

PostPosted: Tue Sep 24, 2019 6:16 pm
by Dave Moll
Yes, it's probably a good discipline when typing a long post to copy it periodically to the "scratchpad" in case of problems. The difficulty is remembering to do it before a problem arises!

Re: Update to SSTV demodulator

PostPosted: Wed Sep 25, 2019 1:29 pm
by Andrew Davie
Steve Anderson wrote:I did quite a long reply to the two posts above...but when I pushed the 'submit' button I got a reminder to be logged in - so my message simply vanished...I WAS logged in - that I know for sure. This used to happen some years ago, a signed-in user got 'dropped' part-way through composing a reply...

I'll have another go some other time...

Steve A.



The forum's "maximum time to submit forms" was set to 7200 seconds. Two Hours. I have just disabled the timeout, but be warned there are other factors that can affect this - sessions can be dropped, ISPs can change/drop connections... or something like that. If you're leaving a form partially filled for over two hours, you're asking for trouble! But, having said that, I have now removed the limit. Have at it!

Re: Update to SSTV demodulator

PostPosted: Wed Sep 25, 2019 1:56 pm
by Steve Anderson
I used to do the Ctrl-C (copy) thing every so often years ago, but stopped when it appeared to not be required anymore - it seemed that all was getting more reliable...until yesterday! So I'll reinstate that habit...I suggest others do too...

I don't think my compose/edit time went beyond the old two hour limit though...I can type at a moderately fast rate...who knows? Probably a dropped connection, that still happens occasionally.

Steve A.

Re: Update to SSTV demodulator

PostPosted: Thu Sep 26, 2019 6:18 pm
by Steve Anderson
Klaas Robers wrote:Do you have already a sample frequency in mind for the FIR filter? I am thinking of 8 KHz for a low pass with a cut off of 1000 Hz. I don't know how many stages you can allow the FIR-filter to be with a sample frequency of 8 kHz. I never used a PIC with a hardware multiplier. How many T-cycles does a multiplication need? Is it more than an addition (1T)?


Klaas, we are thinking along the same lines...yes 8kHz with a cut-off at 1kHz, also (just for the heck of it) 16kHz with a cut-off at 1kHz. After a D-A it's wise to filter the output, if you shift the sampling frequency upward the analogue filter has a a better phase characteristic (say a Bessel). Anyway, I'll give it a go and see if it makes any practical difference. There's no practical difference in cost or complexity...just factors...

Steve A.

Re: Update to SSTV demodulator

PostPosted: Fri Sep 27, 2019 2:09 pm
by Steve Anderson
Here's my initial stab at a phase-linear post-D-A filter. I think all the info is in the GIF...

I've adjusted the -3db point to 1500Hz, but it can be lowered back to 1kHz if required...in fact I'll do it anyway..

In both cases they need to be fed from a comparatively low-impedance source, if not (and it's resistive) adjust the first R1 to suit...

Steve A.

The software is a freebie downloadable from TI...I've shrunk the gifs to fit the board display page size, hence the 'de-focused' appearance, but I think it's still legible.

Re: Update to SSTV demodulator

PostPosted: Fri Sep 27, 2019 2:12 pm
by Steve Anderson
Here's the 1.00kHz version...

Steve A.

Re: Update to SSTV demodulator

PostPosted: Fri Sep 27, 2019 2:39 pm
by Steve Anderson
Here's the 1kHz Bessel version...

Steve A.

Re: Update to SSTV demodulator

PostPosted: Sat Sep 28, 2019 2:18 pm
by Steve Anderson
I noticed a 'silly' (an error) in one of the postings above, now corrected...

Steve A.