Hello world!
I'm seeing an occasional error when reading from my peripheral board
circuit, about once every 50 million reads, and I'm trying to narrow
down the issue. I decided to pull out my card (from a IIgs) and put a
blank one in and attach probes to just the ground and phase 0 (pin 40)
and see a lot of ghosting/jitter or whatever that is and I wanted to see
if you guys think this is normal because it doesn't look kosher to me!
Probes at 1x: >https://drive.google.com/file/d/1xaisSJ6kjPNAKoPKuPd1VkV0HrSmoYhj/view?usp=sharing
Probes at 10x: >https://drive.google.com/file/d/1pO1FS7R1Fo8lGp97Dbfhuf4htH0n2f1-/view?usp=sharing
I looked briefly at the 1x. You have a lot of ringing. First, some ringing is fine, if that's the way the circuit is designed. Old NMOS/CMOS stuff
is not trying to balance driver and receiver impedence, so I would expect things to not look "perfect". If there is real ringing, it's only a problem if the amplitude is so much that it causes the receiver to misread the signal (so, a clock ringing so much that it could be intereted as a new
edge is a problem). As I explain to many people--some ringing is usually fine, it needs to be severe to be a problem.
But: you showed the image of your probe--that's very useful. And you're probing incorrectly. You need the tip of the probe to be within say 1cm
of the signal to be probed. You can not attach a long wire, and then
attach the probe to that wire. If you want to see a signal on the motherboard, the probe tip must go right down there. You can solder on
a VERY SHORT stub of less than 1cm (I often solder on a resistor, and
then clip the lead down to <1cm since I can easily hold a resistor with pliers, it's thinner than usual solid wire, and it makes the soldering
easy. You clip off the resistance blob itself). Then attach the probe
to that SMALL stub.
It looks like in slot 2 you have a prototyping card with a breakout connector, and you've stuck some wires in the connector for the signal
and ground. These wires look to be about 6 inches long, and you've
attached the probe to these wires. If you put a resistor in that
connector instead, clip it to be short (and get rid of the resistance
part), and probe that. No soldering needed. Not as good, since you're
using an effective 3cm stub (the connector counts against you, sorry).
The best is to take a real card, like the card in slot 5, and find the
back side of a through hole chip with the signal you want. Solder on a resistor lead, then snip it to be very short (5mm is very easy to do,
3mm is great). Note: I don't use new solder usually, just melt the
solder already there and it's enough to hold the stub. I find the probe won't really clip on to the through-hole leads as they are since there's
not enough to grab onto beyond the solder ball. Then put the probe
directly on that stub, and it should grab on easily. This is about the
best I can do.
One trick is to test out ringing by manually holding the probe tip on the signal as well, and see if the ringing gets better. If the 1cm stub shows ringing, try manually holding the probe and see if it gets better. If it still does, then there's just ringing on the signal. It happens.
It's ok to probe "badly" if it's easy, or that it's not worth the
trouble to probe properly. Try to look at the signals as "digital", and ignore kinks, ringing, etc.
Kent
I looked briefly at the 1x. You have a lot of ringing. First, some ringing >> is fine, if that's the way the circuit is designed. Old NMOS/CMOS stuff
is not trying to balance driver and receiver impedence, so I would expect
things to not look "perfect". If there is real ringing, it's only a problem >> if the amplitude is so much that it causes the receiver to misread the
signal (so, a clock ringing so much that it could be intereted as a new
edge is a problem). As I explain to many people--some ringing is usually
fine, it needs to be severe to be a problem.
It's ok to probe "badly" if it's easy, or that it's not worth the
trouble to probe properly. Try to look at the signals as "digital", and
ignore kinks, ringing, etc.
Kent
Hi Kent! Thank you so much for responding, this is a great explanation
of what I've been doing wrong. I'll do what you suggested and see what
it looks like. I figured the problem was likely to be me, and thanks to
you I'll be able to make a more informed test from now on.
Sysop: | Keyop |
---|---|
Location: | Huddersfield, West Yorkshire, UK |
Users: | 491 |
Nodes: | 16 (2 / 14) |
Uptime: | 107:14:44 |
Calls: | 9,684 |
Calls today: | 5 |
Files: | 13,725 |
Messages: | 6,175,499 |
Posted today: | 1 |