check this out:https://docs.google.com/spreadsheets/d/1e0dgVVgFcP2gGTa8Z8X4Z88gdzFuHal3OJlJbwn8spE/edit#gid=777902241
This is a simulation of a stuck second-order bit on one of the ADC's two
outputs. Bears a likeness to your images - right?
So, given this, I'm pretty sure now that you have a stuck second-order bit,
the question is where.
The options are:
1. An open on one of the ADC's output channels.
This would manifest as a problem on every other sample, when you're
going at 250MS/s (the max sampling rate for one ADC).
2. A bad RAM chip.
This would manifest as a problem on every eight sample, probably
irrespective the sample rate, but for sure at 250MS/s.
3. A bad MUX.
Let's hope this is not the case, as I don't think these are easy to come
4. A stuck bit somewhere else.
I can't think where there could be a stuck bit manifesting precisely
like this, so let's ignore that possibility :).
Note that varying the input offset and magnitude of the input will hange
the manifestation of the stuck bit, and hopefully you can repro the trouble
I forget (or never knew) whether this scope will inject a DC offset to the
input channel when you adjust the trace position, but for sure it'll inject
a DC offset when you adjust the input offset in the vertical menu.
On Thu, 5 Nov 2015 at 22:08 Sigurður Ásgeirsson <firstname.lastname@example.org> wrote:
The other possibility that'd fit the facts so far, is if there's a stuck
bit on the ADC outputs, due e.g. to a broken trace or such. Did you do any
repair in the vicinity of U700?
Again, playing with DC inputs will help, as that'll narrow the options. A
stuck line from the ADC should manifest in every other sample, and at lower
speeds, it's possible that only half of the output lines are used
(explaining why this doesn't hit at all settings, all the time on CH2).
So, ET off, both channels on (to force interleaving off), sin(x)/x off,
dots mode if available, 1M input impedance.
Input DC on CH2 and vary voltage or offset to full scale positive and
negative. If this reproduces the issue, take note of whether it affects the
entire sample record, or a fraction. What's the distance between anomalous
Take note of whether and how the trace follows the input voltage.
On Thu, Nov 5, 2015 at 18:48 Sigurður Ásgeirsson <email@example.com> wrote:
On Thu, 5 Nov 2015 at 15:11 'adri' firstname.lastname@example.org [TekScopes] <
So you think it might be a defective memory ICI don't see anything inconsistent with that hypothesis in your pictures.
Looking at the A10 schematic, it looks like there are eight distinct CS
lines, and the RAM is written four-byte (64 bit) wide.
If you look closely at your captures, I think you'll see that on CH2,
every fourth sample is bad, for half of the total capture record.
When this hits on CH1, you should see every eight sample is bad, again
for a total of half the waveform record.
To look at this closer, turn ET off, turn sin(x)/x off, if this scope has
a dots mode, turn that on (instead of vectors), and you should be able to
... time passes ...
Actually, looking closer at your pictures, I'm starting to wonder,
because it looks like the "noise" always hits at the top and bottom of the
signal. If this is a RAM chip, then that'd mean the same chip is getting
hit for the same portion of the signal each time, and I can't think why
that would happen.
I think we should be able to distinguish RAM vs DAC/MUX failure by
playing with DC inputs to the scope - see below.
Do you know if the ICs are currently available ?I'd be very surprised if they aren't - these are just 2K (or 8K) static
And what about raising one (or more) pin to re-create the problem andYou could do that, but there are 64 total pins to try :/.
validate the hypotesis ?
Can you reproduce this problem with a DC signal, and do you have a second
(storage) scope to diagnose this?
Try putting your signal gen on AUX (or CH1) and trigger on it there, then
look whether you get a messed-up waveform on screen with DC on the input.
From your pics it looks like you might have a stuck bit, so play with the
DC input. You can also ground the input and play with the input offset for
the same effect as inputting DC.
Assuming you can provoke the problem with a DC signal (or offset) you
could narrow this down to a chip by looking at the data lines (there's 64)
and looking for the the bit that looks bad.
[Non-text portions of this message have been removed]