14 Mar
2023
14 Mar
'23
6:38 p.m.
On Mar 14, 2023, at 2:00 PM, Jonathan Chapman via
cctalk <cctalk(a)classiccmp.org> wrote:
Often that's true. The issue I'm fighting right now is on a baseband data
transfer which occurs at 11k per second, and some of them seem to be getting messed up.
So there are long gaps in between. I only see one or two of them before memory runs out.
It may be what I need to do is unplug the serial link, which is running with a continuous
clock (6 MHz or so), that's probably sucking up buffer entries.
The analyzer I have is a Philips PM3585 -- 96 channels (though I don't have a full set
of cables and pods), 2k samples. So in theory, with some care and enough wires unplugged,
I should be able to see 100 or 200 cycles of the problematic transfers. That might just
be good enough.
paul
On HP: yes, perhaps. I used one of those back at
DEC, in the mid 1980s. Nice machine, but my suspicion is that I'd run into the small
memory problem again that plagues me with the Philips/Fluke analyzer I use right now.
We use several HP analyzers around here, all from the 80s. Even the "small"
ones are 72 channel and more than sufficient for all your old computer needs. The big ones
have Ethernet and will talk X11 across the network, which is super handy. The older models
are serial-only, but can print to a ThinkJet emulator running on a PC. That's how I
made e.g. this capture image:
http://users.glitchwrks.com/~glitch/images/xtide/xt_ide_rev4/read_delay.gif
Before getting my first HP 1650, I had an older Sony/Tektronix that suffered from the
"not deep enough capture memory" issue. Haven't run into that with the HPs
yet, especially the big HP 16500 series. I've found that if I'm running out of
memory depth, I'm probably not hooked up right. 