EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: wetbox on February 07, 2021, 06:59:43 pm
-
I need to continuously monitor 10-30 channels, each receiving 0 to occasionally up to around 100 logical pulses within a second. On average let's say 1-5 pulses per second per channel. Pulses can arrive at any time on any or all channel, but no two pulses on the same channel will arrive closer than ca. 100 ns of each other (and if they ever do, it's ok to only catch the first one). It is important that pulses arriving at exactly the same time on many or all channels are all captured. Ideally the measurement would be a stream of 'timestamp, channel #' where timestamp has a 1 ns accuracy or better. What matters is the accuracy of relative timing between pulses and across channels. Stable skew between channels is ok, I can correct for it. Absolute accuracy of the timestamp is not important because one of the channels will always receive a stable pps from a time reference. Within reason, I can adapt pulse shape, levels, polarity, line impedance and connectors to suit the instrument.
Some logic analyzer may be usable for this, but it's not their typical application and looking only at specs may miss some key consideration. Budget is up to 1-2 kEUR max (incl. shipping and 25% DK customs) and I'm happy with old-ish instruments if cheaper and with decent docs. The Zeroplus LAP-C Pro 32064M may almost do what I need, but I'm concerned that the only place telling about the sampling rate dropping with number of active channels (500 MHz above 16 channels) is the back of the package as seen in an online review. Kind of predictable, but the spec. in the manual and website don't say. Not sure I want to pay more than 1700 eur for a half-mystery gadget, and with a reportedly clunky software if I one day I need to use it as an actual LA.
For now it's only for use in the lab, so boat anchor class is fine, with a reasonably common connection to a 0-20 years old pc (GPIB, ethernet, rs232, even LPT) unless the thing can itself log at least a month of data internally. Of course something smaller and USB-powered would be handy. Thanks in advance for any comment and suggestion!
-
What you haven't said is what the acceptable latency of the readings would be. There a big difference between delivering a result with ns scale precision within a few tens or hundreds of ns, and delivering the same data after several 10s, 100s, or 1000s of milliseconds. The former requires clocking counters and the like in the GHz region, the latter can be done with a relatively slow clock (say 10MHz) and some analogue interpolation.
You're also constrained by how often a timing event will happen on an individual channel - you can't use an analogue interpolator that takes 1000-2000 slow clocks to deliver a result if another pulse/edge to be measured arrives while the interpolator is busy with a previous pulse. Your ~100ns inter-pulse interval would seem to rule out this kind of analogue interpolation.
You might be able to repurpose one of the 'time of flight' ICs used for distance measurement to do the interpolation between pulse events and a slower counter clock. Some of those can have quite low latencies, but without researching it I couldn't hazard a guess if any are fast enough to deal with your potentially short pulse intervals.
-
What you haven't said is what the acceptable latency of the readings would be.
Good point. Latency can be pretty much anything, like seconds or even an hour would still be fine!
Your ~100ns inter-pulse interval would seem to rule out this kind of analogue interpolation.
The 100 ns figure comes from the performance of the devices producing the pulses. A longer dead time may still be acceptable but it will then become one of the limiting factors in the system.
You might be able to repurpose one of the 'time of flight' ICs used for distance measurement
Thanks, that's a great idea if nothing is in readily available in my price range. Of course then it becomes a project in itself.
-
I think a TLA700 series logic analyser from Tektronix could be a good fit. These have timestamped recording capability which means it only captures data when there is a change in any of the channels. You can make extremely long recordings this way. I recommend to get at least a TLA715 mainframe.
This looks like a complete kit to get you started:
https://www.ebay.com/itm/Tektronix-TLA715-TLA-715-Logic-Analyzer-w-probes-2x-TLA7N2-modules/362591303674 (https://www.ebay.com/itm/Tektronix-TLA715-TLA-715-Logic-Analyzer-w-probes-2x-TLA7N2-modules/362591303674)
IIRC the modules in this mainframe (TLA7Nx) have 2ns timestamp resolution and timestamp capability 100 days with 2ns resolution.
The TLA7AAx modules have 500ps timestamp resolution but the probes are harder to work with; you'll need an adapter PCB (which is somewhere on this forum) to get to header pins.
-
IIRC the modules in this mainframe (TLA7Nx) have 2ns timestamp resolution and timestamp capability 100 days with 2ns resolution.
I see they go down to 125 ps over 3 days (overkill but nice). Can they stream as they go, or would I need to wait either for the memory to fill up or the 3 days to lapse?
When the datasheet specifies the TLA7N1, TLA7N2, TLA7N3, TLA7N4 acquisition memory depth as '64 K or 256 K or 1 M or 4 M samples', is that per channel or shared among all active channels?
-
IIRC the modules in this mainframe (TLA7Nx) have 2ns timestamp resolution and timestamp capability 100 days with 2ns resolution.
I see they go down to 125 ps over 3 days (overkill but nice). Can they stream as they go, or would I need to wait either for the memory to fill up or the 3 days to lapse?
You have to wait. A logic analyser captures first and then you can analyse. Though you can stop the capture process prematurely and analyse the data captured so far.
When the datasheet specifies the TLA7N1, TLA7N2, TLA7N3, TLA7N4 acquisition memory depth as '64 K or 256 K or 1 M or 4 M samples', is that per channel or shared among all active channels?
For Tektronix that is the samples per channel. Using timestamped mode or not doesn't not impact the sample depth. The modules are easy to hack to get maximum possible sample depth.