Products > Test Equipment
Let’s Talk About LeCroy Scopes, AKA… the “Wuerstchenhund Holds Court” Thread
Wuerstchenhund:
--- Quote from: AutomationGuy on September 14, 2016, 07:18:25 pm ---I would like to have WaveScan on FFT which would include math functions on a FFT trace.
--- End quote ---
WaveScan on FFT is unlikely as FFT is processed data (and WaveScan works with the original sampled data).
I haven't tried on the WS3000 but as to math on FFT, you should be able to use the FFT trace (which is just a math trace like any other math function) as an input to another math function.
--- Quote ---Another nice feature on the WS3000 would be a SENT decoder.
--- End quote ---
LeCroy has stated they'll expand the options for the WS3k depending on customer feedback (they already implemented CAN and FlexRay which were both missing when the scope came out) so they might as well implement SENT if enough people show interest.
--- Quote ---One thing about the WS3000 which annoys me is the autosetup which probbaly breaks sampling for a few seconds each time its triggered. I could not yet find out when the autosetup is triggered. There is a button for autosetup on the scope but it has no use for me since the scope starts it automaticly.
The autosetup is realy helpfull and gives real good results but it interrupts sampling.
--- End quote ---
the scope should not start auto setup by itself, only if you press the button (if it does then there's something wrong with your unit).
The WS3000 does a self-cal cycle after a certain amount of time running but that is once only.
--- Quote ---Direct replay from a trace to the build in function generator would be nice.
--- End quote ---
As with SENT, ask LeCroy, they might just implement it in a next software release.
Wuerstchenhund:
--- Quote from: nctnico on September 15, 2016, 12:48:21 pm ---The thing is that peak-detect combined with roll mode will show all peaks. If Wavescan is an acquire-process-acquire-process- system then it will have a considerable blind time. IMHO you are too much focussed on finding glitches in known signals but you have to take a few steps back to a signal you know nothing about and want to get a feel for.
--- End quote ---
I do, and that's what WaveScan can be used for. It can find problems you didn't know where in the signal. You seem to think that you have to know the problem for it to work, but you don't.
Yes, the blind time will be higher than with PD in roll mode. But then all roll mode tell you if there are any drop-outs or if there aren't (and that only if you didn't miss anything), you still haven't analyzed the problem.
--- Quote ---Without peak detect this is nearly impossible and it sounds to me Wavescan isn't solving that.
--- End quote ---
It's difficult to do remotely (and I don't have a WS3000 at hand anyways) but I'm still confident I could find with WaveScan anything you find with PD.
Just out of interest, what frequencies/pulse widths/PRFs are we talking here? You mentioned H-Sync/V-Sync which suggests some kind of video signal?
--- Quote ---Compared to the Lecroy Wavesurfer 3000 it seems the R&S RTM2000 series is in the same price range and it does have peak detect.
--- End quote ---
The RTM2000 is a very nice scope (and yes, it does have PD), however it has a max update rate of some 11k wfms/s, a much smaller screen, and you only get 128k FFT, plus options and probes are pretty expensive (and probe selection is very limited). However it does have some nice features (like dimmable LED indicators), it's really very silent, and even comes with a Z input (brightness modulation for true X/Y/Z mode) which is a rarity on DSOs. It also has a memory search tool that helps to find certain glitches and runts post-acquisition.
Some time ago I bought a RTM1054 (500MHz 4ch) 2nd hand which for the price (<$2k) was an absolute bargain. The RTM2000 is pretty much the same plus a few improvements (128k instead 64k FFT, 10M instead of 8M, MSO option, more serial decode standards) but the platform is the same (BlackFin based). The scope is now with someone else who talked me into selling it to him and he's quite happy. I'm not sure I'd have paid the $6k or so R&S wanted for it when new.
I know we talked about this and you had some problems finding representation in your country but if you have the chance then I'd recommend to get loaners and spend some time exploring the various tools like WaveScan in your test environment. I might well be that I'm wrong and WaveScan can't help in your situation (although I doubt that), but I'd really recommend to give it a try.
The worst thing with WaveScan is that it is so poorly documented by LeCroy (like a lot of things, unfortunately).
AutomationGuy:
I don't know the insightes of the WaveSurfer3K but technicaly it should be possible to implement WaveScan at least partly in FPGA and process all data in real time. How many math functions are implemented in VHDL could depend on the scope price. There might still be some functions implemented on an "offline" CPU. That would make WaveScan run "offline" in just some configurations. Many might run in real time.
I know there is a limit on high speed signals but in many usecases WaveScan could run in real time. Especialy on video or audio signals.
LeCroy will not publish its scope insights. Especialy not for marketing reasons.
I just need occasionaly peak position and peak width in numbers on an FFT to quickly read out without fiddling with cursors. A math function could implement that. If FFT could be implemented in VHDL plus math that could even run in real time in WaveScan.
WaveScan might not be fast enough to always catch all anomalies but its so quickly configuered and can run for days which saves so much time for me to find problems.
Someone:
--- Quote from: nctnico on September 15, 2016, 12:48:21 pm ---
--- Quote from: Wuerstchenhund on September 15, 2016, 12:34:59 pm ---WTF are you talking about? WaveScan directly uses life acquisition data, there's nothing inherently "offline" to it as you claim. The update rate goes down (naturally because of the processing required), by how much depends on the circumstances (i.e. what WaveScan settings, scope generation and CPU cache size and speed). Nevertheless it scans all acquired data.
As to my "narrative", all I said was that the tool you clearly know jack shit about should be a sufficient replacement for situations where people tend to use Peak Detect.
--- End quote ---
The thing is that peak-detect combined with roll mode will show all peaks. If Wavescan is an acquire-process-acquire-process- system then it will have a considerable blind time. IMHO you are too much focussed on finding glitches in known signals but you have to take a few steps back to a signal you know nothing about and want to get a feel for. Without peak detect this is nearly impossible and it sounds to me Wavescan isn't solving that. Compared to the Lecroy Wavesurfer 3000 it seems the R&S RTM2000 series is in the same price range and it does have peak detect.
--- End quote ---
This is the important point for everyone to understand, searching the acquired data after the fact looking for anomalies has a much higher blind time than looking at realtime or your extreme example of peak detect in roll mode which theoretically has no blind time if you check the screen at least once per horizontal period. Different ways to look for different characteristics, and all useful techniques.
--- Quote from: AutomationGuy on September 15, 2016, 06:45:49 pm ---I don't know the insightes of the WaveSurfer3K but technicaly it should be possible to implement WaveScan at least partly in FPGA and process all data in real time. How many math functions are implemented in VHDL could depend on the scope price. There might still be some functions implemented on an "offline" CPU. That would make WaveScan run "offline" in just some configurations. Many might run in real time.
--- End quote ---
Yes, this is the approach taken by some other manufacturers with hardware mask testing or advanced triggers. Lecroy goes further and offers a very unique triggerScan mode where it will step through a list of triggers to test against but sadly doesn't apply them in parallel so the blind time becomes proportional to the number of features you are looking for exceptions on (similar scaling for wavescan). So its about evaluating the possibility for different tools to a) see the anomaly, and b) the rate at which it can check for that specific anomaly amongst all the other possible anomalies it is looking for.
More advanced searches can be conduced offline, so they increase the ability to detect anomalies, but they capture slower and with higher blind times than simpler tools such as mask testing or visual assessment of an eye diagram. None of the tools are better for all situations no matter how much Mr W says so.
David Hess:
--- Quote from: Wuerstchenhund on September 15, 2016, 12:34:59 pm ---*Nothing* runs in the acquisition memory in a scope, it's a store for sampled data :palm:
--- End quote ---
While literally true, this statement is misleading; operations like digital triggering, peak detection, timebase decimation, and DPO functions work on the non-trivial real time subset of samples in parallel with or between the digitizer and acquisition memory limited only by logic resources. While I expect designers attempt to make this subset of samples as small as possible because logic costs more than memory (memory is incredibly cheap in all respects), it always exists unless none of these operations are present like on very early or very cheap DSOs. What gets stored in the acquisition record has been processed except in trivial cases or apparently if you are LeCroy.
--- Quote from: Wuerstchenhund on September 15, 2016, 12:34:59 pm ---As to my "narrative", all I said was that the tool you clearly know jack shit about should be a sufficient replacement for situations where people tend to use Peak Detect.
--- End quote ---
I am not interesting in "sufficient" replacements; I am interested in superior replacements which if they do not cover all previous applications, at least do not preclude them using the older method. DPO is usually but not always a superior replacement for peak detection so the later is still needed. I am not sanguine that WaveScan is a superior replacement for either.
Marketing said the same thing about large acquisition memories replacing the need for delayed acquisition and peak detect yet the result has been lower performance or outright uselessness in a minority of applications which older DSOs without large acquisition memories have no trouble with; if a DSO misses the long tail of applications, then that is not an improvement no matter how much marketing tries to convince otherwise. For instance if a current DSO cannot capture the jitter on a GPS pulse per second output at a sampling rate commensurate with its bandwidth, then I am not interested in it no matter how good its automatic measurement capability, touch screen, "one true knob" user interface (HP), and dual aspect ratio display (LeCroy) is.
Dig up Bob Colwell's Stanford lecture where he mentions "Blue Crystals" for an example of marketing driving development to its detriment. If you cannot find it, I can provide it as a bittorrent link. Marketing reminds me of the law aphorism about pounding the table but suitably modified:
If you have the record length on your side, pound the record length. If you have the WaveScan on your side, pound the WaveScan. If you have neither on your side, pound the table.
Your comment about LeCroy designs originating with physics applications where all original data is preserved and then analyzed fits with how WaveScan is described as working however I do not think that model is necessary or even suitable for design, development, and troubleshooting.
It would be interesting to consider how a modern DSO could work to provide superior performance to older designs. Processing after acquisition strikes me as insufficient by itself unless either deadtime is acceptable or processing occurs in real time in which case why is there an acquisition record at all?
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version