48K Issue 3, noisy display.

edited March 2012 in Hardware
Has anyone here experienced and been able to resolve a problem that looks like this:

Photograph of display.

It is in fact quite a bit worse than it appears in that photo. The picture generally waves and jiggles about (but with no loss of synchronisation) and the pattern in the centre of the display (including the border) slowly and smoothly scrolls upwards.

The video feed was taken as a composite signal with the modulator disconnected. This Spectrum has all new electrolytic capacitors and a new 7805. Apart from the poor display, it seems to work correctly. The pattern rather makes me think the noise is digital in origin.

I've used an oscilloscope to look at the 12V line at the LM1889 and found what looks like a rough square wave with an amplitude of about 100mV p-p with about 100mV p-p of noise superimposed on it. The same measurement on an issue 2 Spectrum that gives a good display finds almost exactly the same.

I've got some further tests in mind, but I thought I'd ask here in case anyone has encountered this before.
Post edited by Zorn on

Comments

  • edited February 2012
    Ive encountered this on issue 2 boards, but a quick turn of one of the pots usually resolves it, not on an issue 3 though, no pots!
  • edited February 2012
    Is that an LCD display? My issue 3 48k looks exactly the same if I hook it up to the little LCD portable I have.

    I suspect the noise on the signal confuses the analog to digital conversion in the display as it has to quantise the signal. On an analogue display it's just grey and a bit fuzzy.
    My rubbish website including the redrawn Amstrad schematics and the new home of the Sinclair FAQ wiki.
  • edited February 2012
    As for how to fix it, perhaps replacing the capacitors might keep some of that noise out of the video circuits.
    My rubbish website including the redrawn Amstrad schematics and the new home of the Sinclair FAQ wiki.
  • edited February 2012
    It's an analogue monitor. It produces a good display with other Spectrums. The problem is much worse than the usual bit of fuzziness.

    As I pointed out, I have replaced all of the electrolytic capacitors. I suppose I may check some of the others in the video circuitry.
  • edited February 2012
    Try to find a different GND point for video out first. If not working try to add a big wired GND connection between GND pin of ULA and GND pin of LM1889. If even not good add big capacitors (220 ?F + 0.47?F parallel) between VCC and GND at LM1889 and probably the ULA.

    Maybe the ULA itself produces bad signal quality. In than case it is very difficuilt to improove the signal. But it maybe also improoved by compensation - needs time and patience ...

    Regards Ingo.
  • edited February 2012
    I will have to experiment further. I have tried swapping ULAs about and they all have a similar problem in this Spectrum, but not in another one. Another issue 3 Spectrum doesn't show any hint of the same problem, so there must be something not quite right with this one.
  • edited February 2012
    Zorn wrote: »
    I will have to experiment further. I have tried swapping ULAs about and they all have a similar problem in this Spectrum, but not in another one. Another issue 3 Spectrum doesn't show any hint of the same problem, so there must be something not quite right with this one.

    My old Amiga 600 RF out started to show similar noise then I found some of it's capacitors leaked all over the board.

    So I guess, guesser's recommendation is a good one.

    By the way, Eightyone (the emulator) is emulating this exact effect :)
  • edited February 2012
    I have seen this on issue3. IIRC it had to do with decoupling the +12V line that feeds LM1186. Check the new capacitors?
  • edited February 2012
    Replace the two capacitors on the far left of the board (but prefferably replace all electrolytic caps).
    If that doesn't help, replace the two transistors right below the RF metal box.
    Finally if that doesn;t completely solve it, replace the 7805.
  • edited February 2012
    roko wrote: »
    I have seen this on issue3. IIRC it had to do with decoupling the +12V line that feeds LM1186. Check the new capacitors?
    It kept me puzzling. I strongly believe now that I saw it when I (for some silly experiment) had removed one (or maybe both), electrolytic capacitor from the powerbrick.
  • edited March 2012
    Thanks for all the suggestions, but you were all wrong! :-)

    The problem turned out to have a very simple cause, but it took a while to find. After experimenting with extra power decoupling and testing several components I decided to attack it more methodically with an oscilloscope probe.

    Below is an example of the composite video output of the problematic Spectrum (top) along with the same from one that gives a good display:
    ZX106-composite.jpg

    Some significant interference of a digital nature is very obvious during the horizontal sync. pulse. Moving one step back through the circuit, I found the video signal to be perfectly clean at the base of TR2. This meant that either TR2 was behaving in an exceptionally strange way or the noise was somehow being added to the final output signal.

    I followed the traces from TR2's emitter around the board and over to the edge connector, looking for anything that could be bridging it to another trace. Nothing was particularly obvious. So, I started looking at the signals on traces that run next to it, looking for one with pulses approx. 500ns wide and found one!

    Here is the composite signal at the output along with the Z80's A13 line (pin 3):
    ZX106-composite2.jpg

    I think that makes it rather obvious where the interference must have come from. I measured 870 ohms resistance between A13 and the video trace, but could see no obvious sign of anything bridging them.

    I noticed that both traces pass underneath C26, so I removed this capacitor to have a look underneath. The board looked a bit grimy, but there was nothing obviously wrong. I cleaned the board and the body of C26 (after finding it to be conductive!) and then my meter indicated an open circuit between the two affected traces.

    With C26 back in place, this Spectrum now produces an excellent display.

    I wouldn't be surprised if dirt under C26 is a common cause of poor displays on Spectrums because it is located just inside from the edge connector.
  • edited March 2012
    Zorn wrote: »
    So, I started looking at the signals on traces that run next to it, looking for one with pulses approx. 500ns wide and found one!
    No address line is safe when you are on the move mr. Zorn! I pity the poor Spectrums that fall in your diabolical hands :-)
    Well done, and thanks for sharing!
Sign In or Register to comment.