> > I have looked at this problem a little more. I have two motherboards, > neither of which work, but one at least produces a corrupted video pattern. > The one that works best appears not to be writing to the video memory. When > I look at the EMEM pin on the Paradise PVGA1A chip I can see a signal but > the scope shows a trace that is very faint. When I look at the same pin on > the other motherboard, I get a nice clear bright trace on the scope, using > all the same settings on the scope. This pin is driven by a non-inverting > buffer (74LS126). The input side of the buffer is tied to 0V, the enable > signal comes from a custom gate array. Comparing the buffer's enable signal > on the two boards I see the same dimming effect on the board with the > corrupted video pattern, and no dimming on the other board. >
A dim trace suggests that the trace is changing too rapidly to see it properly. Try increasing the scope brightness and sweep rate and adjusting the various triggering options to see if you can get a better trace that reveals what is really happening at that pin. It might be useful to try triggering the scope from whatever clock is used locally as the signal might be synchronised to that. Once you can see the trace properly, it should be easier to figure out where it is coming from. > > I have checked the other pins on both the buffer and the gate array and I > don't see anything suspicious. > It may be worth looking very closely at the power supply pins for difficult to see spikes that might be caused by decoupling failures. It would also be good to make sure ground is really ground at the chips in question. > > I am thinking of speculatively replacing the 74LS126 because I can go and > buy replacement parts for it, I can't replace the gate array (although I > could conceivably swap the part on the two boards). > If the 74LS126 has some fault at it's input which is affecting the signal coming from the gate array, it seems to me that it would be more likely to load it down or up rather than cause it to change rapidly (if that is in fact what it is doing), unless it has somehow managed to turn itself into an oscillator. On the other hand, if the gate array output is open collector, it could be relying on the 74LS126 to provide a collector load and not getting it if the 74LS126 is faulty. This seems unlikely though. I suppose another possibility is that the gate array output could be tristate and not enabled leading to noise pickup from nearby traces and components and perhaps across the PCB surface, given there was a battery leak at some point. Getting back to the oscillator theory, I wonder if it is possible that the non-inverting buffer could be oscillating at a much higher frequency than normally found on the board due to feedback from it's output to it's input as result of the battery leakage? I don't normally like suggesting cutting tracks but if the board already has damaged and repaired tracks, you might feel ok about cutting the track between the gate array and the 74LS126 to determine which of them (or both together) is responsible for the unusual signal there. I would be very wary of replacing the unobtainable gate array with your only replacement until all other possiblities are eliminated in case the gate array was damaged by a fault elsewhere. Looks like I've provided more questions than answers :-( Regards, Peter Coghlan.
