Here's a trace taken on an EXDOS card plugged into the expansion port of a standard EP128:


[ Guests cannot view attachments ]

The purple trace is /RD and the blue trace is the same signal after it has passed through an OR gate (other input tied low) on the EXDOS card. The dotted horizontal line is at 0.8v - the "0" to "1" threshold for LS TTL.


Here's the same signal, /RD, on my EPNET card (different timebase on 'scope though):

[ Guests cannot view attachments ]


Just wondered if anyone else has seen this before?

B.

Thanks Pear.

The OR gate I mentioned is on the EXDOS disk card - nothing to do with me! My guess is that Dave Woodfield who designed the EXDOS card (and Dave chip) knew about the glitch, which is why it is buffered by the OR gate and it comes out smaller. It then goes through a NAND gate and comes out even smaller (and inverted). It does make it as far as the WD1772 floppy controller chip but (as far as I have seen) only when the chip is not enabled, but my scope can't trigger on that condition.

I see exactly the same glitch with my EPNET card but it happens when the network module is enabled so I get a double read, which matters particularly because it has an auto-incrementing register. I've played around a lot with decoupling and improving the power supply tracks but with no change.

I don't see it with nothing plugged into the side connector though. I don't have anything else to plug in and try, which is why I wondered if anyone else had seen it with different hardware. I can work around it, just need to be sure I'm not causing it!

What you see in pure EP128? For example on the internal memory expansion card.

It is there, but smaller, 0.6v. Needs to be 0.8v to count as a "1".

[ Guests cannot view attachments ]


Some pretty nasty noise on "high" too, but doesn't quite go low enough to actually matter:

[ Guests cannot view attachments ]


These measured on the edge connector fingers but with nothing plugged in.

Zozo has found that some Z-80's on other EP's act the same as mine. This can be your problem.

Bruce! Your another machine (EP64) also producing same?

Unmodified EP64 with BASIC cartridge (but the -12v has broken again, I'll mend and re-check):

[ Guests cannot view attachments ]



It's not in quite the same place but there's definitely a spike there! And the downwards one is only 0.1v away from hitting the "low" threshold.

Pear: it's only occasionally so you might not see it if you just look at /RD. I am lucky, my 50th birthday present to myself last year was a new scope, and it can trigger on upwards or downwards glitches, that is how I first saw it. I have a borrowed expansion bus with LS245s etc, and the /RD glitch gets through those.

On the EXDOS card the spike is reduced by passing it through an LS gate; on my card using HCT that seemed to make it worse, turning a little spike into a full low-high-low glitch.

I only noticed it because the network module has an auto-incrementing register and I was missing bytes sometimes. I sat in a loop reading the fixed ID register and it read back the correct value every time for over an hour, even with the glitches. So depending on your h/w you might not notice any symptoms.

My "fix" at the moment is to pass /RD through a schmitt trigger.

And no better backward connected Schottky diode with pull-up after it ?
It's reduce the levels of approx. 0.2 V to next gate.

Will try that, thanks for the suggestion, would save a chip!

I can now confirm if I just let the scope free run and press "stop", every few stops I can see a glitch, 1uS timebase

Other signals (WR,IORQ,MREQ) also have the problem?
Did you replaced the C9 at the latest repair?

Ah em...it turned out I still had the monitor brightness turned down to 0 after I was taking photos a few weeks ago and for some reason my EP64's ouput is much darker than my 128's !!!

I can see the glitch on /WR, /IORQ and /MREQ but they are much less...too small to cause any harm. But the /RD spike is so big on the EP64 that it triggers my schmitt trigger gate "fix", so turning a spike into a big glitch with nice square corners