Falcon sound problems (DSP)

[kohli79]

Thanks

Yes, many many thanks to Exxos!! After the whole odyssey, a happy ending and a richer experience!

[dhedberg]

Yes, many many thanks to Exxos!! After the whole odyssey, a happy ending and a richer experience!

...and another Falcon saved! These machines are so rare now that every single one counts! Great work Chris!

[exxos]

I have been thinking about this and the resistor is on the wrong end, it should be SDMA end.. So will try that when I get back home later.

Note that the resistor is 100K and nothing to do with the "Termination resistor" as described in other mods.

[kohli79]

I have been thinking about this and the resistor is on the wrong end, it should be SDMA end.. So will try that when I get back home later.

Note that the resistor is 100K and nothing to do with the "Termination resistor" as described in other mods.

OK.
Would have been interesting to know whether it would have worked with the "old" SDMA.
The only difference was that the new one no longer responded to the cold spray. The rest of the problems were the same ...

[exxos]

I need to clean it up a little bit, I soldered a 0805 100K onto the clock line to gnd instead of the resistor mod on the clock board. Falcon on test again now. So far it seems to be fine. So will leave it running for the next hour or so.

IMG_6517.JPG (376.73 KiB) Viewed 3945 times

[Steve]

Chris,
Again for educational purposes, you mention this is different from the sdma termination patch which used a 47ohm resistor to GND with the original board resistors intact. So of course this is a 100k resistor, plus your clock patch, could you explain the difference a little? I know resistors are for voltage of course, I understand that electricity follows the easiest route. So if the current is ignoring this 100k resistor to GND and still following its intended route to the clock patch, how is it affecting the signal electrically? Thx

[exxos]

The 47R is basically a crowbar for the clock. Its pretty bad as if you take 5 / 47 = 106mA peak current draw where normally outputs can't drive anywhere near that. As the voltage rises the current will fall. But I really do fear people who do that 47R mod it's a wonder it's not blowing the clock drivers up by now. My falcons won't boot with it as it clamps the clock to about half a volt! I think I've already written about all that before.

Anyway, it's not termination which "solved" the problem. It's the fact it's overloading the clock and all but killing it. What it actually does is short the clock out to the point its half the voltage and because it's crowbarred down to 0v, it alters its DC offset which forces it towards 0v more and it then has a lower logic low level which actually works, albeit pending blowing stuff up.

Problem is as always noise and grounding problems. You actually end up with something like 2v DC offset between the clock source and destination. So if "logic low" is 2v, you're not being at a logic low at all, hence the whole clock issue epic. In fact the logic low only happens because of undershoot. So when its crowbarred with 47R it's literally forcing the DC offset down to more like 1V and "mostly" the voltage swing is high enough to still push a logic high over 2.5v. Only on my 2 Falcons it clamps the clock down to more like 0.5v and that's not enough by far. Hence then my falcon won't boot as its clock is basically low all the time.

I solve this differently. I come from an audio background so know all about DC offsets and logic levels with MOSFET digital amps. Anyway, I don't crowbar the voltage to get it lower, I just remove the DC offset instead. I then have something like 33R to help stop ringing. But as the clock is going right across the board, all the inductance and capacitance and noise pickup all adds up. It's still there at the SDMA end,BUT the ringing / under /overshoots all happen outside of TTL logic level thresholds. So it basically doesn't matter.

So for some reason the DC offset drifts on this falcon over time which it shouldn't really be doing. This voltage very slowly increases until the logic low level gets to the point it was before all this modding was done. Basically audio crackles etc. It further drifts out then causes video issues like in the image I posted. Then eventually it just crashes and half the time won't even reset again.

Now we get back to the 100K which just pulls the clock down to 0V enough to keep the proper clock biasing. Basically you end up with like 0v to 4v on the buffer board end for example, then without the buffer board to correct the voltages, you end up with more like 2-6v on the SDMA as there is bad grounding between the 2 areas of the motherboard. With my buffer board, it auto corrects to mirror the voltage on the buffer board. So you end up with a nice clean 0-4v clock on the SDMA.

In fact I've documented these problems many times over the years. I noted it's possible to get a 1v drop between gnd pins on the plcc 68000. It threw me off in my early booster work. Its why I use solid gnd layers now in all my PCBs as it's a huge problem.. And well.. Might as well throw in the bad DMA BS again just for good measure

[Steve]

@exxos thx for the explanation!! Maybe @mikro might be interested in quoting some of you for his clock patch website. (By the way mikro you might want to add the latest Exxos clock patch as it is different from the original you have listed and works with boosters)

The reason I've been asking these questions is because I'm actually working on a falcon with the same issues you discovered in this thread. It came with the 47ohm crowbar and has stability issues, I tried adding the clock patch which then gave me the display issues going haywire. I'll try the 100k and see how I go. By the way does it need to be an SMD resistor? I only have regular resistors in stock.

[exxos]

By the way does it need to be an SMD resistor? I only have regular resistors in stock.

I used a SMD as I don't really fancy putting a through-hole resistor on the track as it is easily damaged.

[exxos]

It has been on all morning (4+ hours) I have been running all the test programs and listening to a tracker music and checking for crackling etc and always good. So will look at sending it back next week @kohli79