An unhappy oscilloscope, Lecroy 454 saga part 4

[vaualbus]

Someone here can sell me some of the hybrid ADC for a lecroy waversurfer 454 (or also the entire acq board)?  I can find less to zero info on the web
I really point ot think of adc failure for my scope. I can get a good sinewave if I'm at 50Gs/s  but as soon as I go down to less than 2Gs/s (so I'm using only one ADC) I get crap data. this happen for channel 2. channel 1 is always stuck with noise)
If I put a signal on channel 4, I can see what it seems to be a sinewave on channel 3 but Ican't control the attenuation of that so I think I'm saturating the ADC, the same happen for channel 3.
So  take a looking at the schematics, channel 1 adc is used also for channel 2 and the same think for channel 3/4. So I bet 90% the ADC are bad.
But I don't understand why I can get channel 2 to work. If channel 2 work it mean that also channel 1 adc have to be right, but If I'm only looking at channel 2 with channel 1 disable I get garbage data (this happen below 2Gs/s) but as soon as I turn on channel 1 I get my sinewave and I can control the amplitude and time division.

Here a link to a little video explaning better the situation:
 

[alm]

What is the point of splitting the same problem up across multiple threads? Do you expect readers to go back and find all history to get a complete picture? Can you at least link to the previous parts?

Channel 3/4 clearly looks like dead ADC to me. Channel 1/2 I am not sure about. Could also be in the analog part. Maybe there is noise on power rails or coupled into the input circuit / attenuator? Does input impedance or attenuation ratio (including the highest and lowest) make any difference? The behavior above 5 GHz could be something like noise cancelling out between two ADCs.

Are the ADC hybrids socketed? Then I would try swapping ADCs between channels (swap 1 with 2, swap 2 with 3) to see if the symptoms move with it. Of course this is not feasible if they are soldered down. Any other parts (like input or attenuator assemblies) that could be swapped between channels?

[vaualbus]

What is the point of splitting the same problem up across multiple threads? Do you expect readers to go back and find all history to get a complete picture? Can you at least link to the previous parts?

Channel 3/4 clearly looks like dead ADC to me. Channel 1/2 I am not sure about. Could also be in the analog part. Maybe there is noise on power rails or coupled into the input circuit / attenuator? Does input impedance or attenuation ratio (including the highest and lowest) make any difference? The behavior above 5 GHz could be something like noise cancelling out between two ADCs.

Are the ADC hybrids socketed? Then I would try swapping ADCs between channels (swap 1 with 2, swap 2 with 3) to see if the symptoms move with it. Of course this is not feasible if they are soldered down. Any other parts (like input or attenuator assemblies) that could be swapped between channels?

I just made a new blog because this is a completely different problem. Before I had a power supply and pfc failing but I menage ti repair them without not much trouble.

By the way the ADC are not in a socket.

[vaualbus]

No one have any suggestion?

Also is there any self test/diagnostic menu I can access?

[abraxa]

alm gave very good suggestions already and I could only repeat what he wrote because that's the easiest way to troubleshoot such a hardware problem (without the complete circuit diagram, that is). Did you try as he suggested? If not, why?

[vaualbus]

By the way the schematics are avaible, I link it down here, and I have arleady check that the power supply have not noise and are in specs.
Here the link to the schematics: https://www.dropbox.com/s/szrury12gm2ltjl/WaveSurfer%20454%20Schematics.pdf?dl=0

[lukier]

Hmm, there might be multiple things wrong here.

First of all, are your calibration files (D:\Calibration) intact and correct? Any error messages on the XStream app power-up? Any errors/warnings in the logs (enable developer mode, but be careful there)?

On the ADCs, I don't know the architecture of this family, but in my WaveRunner 6100A there are multiple components that could produce bad sample data.

On the WaveRunner the ADCs (HAD639) run at full speed all the time (5 or 10 GSPS) and are controlled by the triggering subsystem, but the decimation (i.e. lower sample rates) is done by the MAM439 DRAM memory chips, 3 per ADC (24 MB total).

Failures in any of these could cause bad data.

However, there is another subsystem that ties everything together, all the ADCs and memory chips - triggering. In WaveRunner 6100A these are chips HTB645, MST429A and ATC FPGA. These things are responsible for controlling the ADCs, decimation in the memory chips (i.e. sampling) and obviously the classical and smart triggering itself. I would bet this part is also responsible for RIS (i.e. equivalent sampling), in your case > 2 GSPS.

As you have one or another type of failure on all the channels I would start looking there.

Of course there is also the interface to the PC, I don't know how it is done on your WaveSurfer, but there might be something there as well. For example a faulty S(D)RAM chip, on the WaveRunner there is one for the PC link related FPGA and one in the triggering-sampling subsystem. Here, for example, one WavePro user had to replace the SRAMs (on the PCI card itself in this case) to get rid of strange decimation problems:
https://www.eevblog.com/forum/testgear/upgrading-mainboard-in-lecroy-dda-3000-(aka-wavepro-7300a)/msg1000237/#msg1000237

Edit: I see you have the schematics, which is rare (haven't found one for the WR6100A, just the service manual) and nice, because if the failure is not in the custom ASICs then there is a high chance of repair as I guess the other parts are common off the shelf.

[vaualbus]

So ya I don't have the cal data or at least I have the one from another 454 scope that someone here in the blog have provide me, but the scope don't accept it and saying that he can't find the cal data. (But I think the scope should run fine without it).

How I enable developer mode??? I can acess service mode but here no logs are shown.

[lukier]

So ya I don't have the cal data or at least I have the one from another 454 scope that someone here in the blog have provide me, but the scope don't accept it and saying that he can't find the cal data. (But I think the scope should run fine without it).

Huh, that's not good. I don't know to what extent these calibration files matter from one series to another, but in my case (WR6100A) in addition to the files responsible for front-end response corrections there are files for trigger calibration (TDC lines, thresholds, offsets etc).

I'm not surprised now that you have, for example, 100V offsets. Some of the files seem to have the scope serial number inside - check that. If there is no CRC or similar measures then the software might accept them.

How I enable developer mode??? I can acess service mode but here no logs are shown.

I don't know if I should post the number here. LeCroy seems to be very litigious, see this thread that had the post removed:
https://www.eevblog.com/forum/testgear/upgrading-mainboard-in-lecroy-dda-3000-(aka-wavepro-7300a)/msg983565/#msg983565

You'll have to do it yourself and look into the servicesvr.dll file. And be careful, a lot of options there could de-calibrate your scope even worse I guess.

[vaualbus]

There is a check on serail number, i try modified the files by changing serail number but not luck.

[vaualbus]

So looking at the log after getting into developer mode really tell that my worst fear are true....
If a go into all logs and into the HwInfo, I get the following Info: "No Acquisition Memory chips (MAMs) on channel 1", but strange enough this message is just a warning not an error. Any way how I can perform the self test??

On the other end I also fail cal with error Trigger threashold cal failed.

[lukier]

So looking at the log after getting into developer mode really tell that my worst fear are true....
If a go into all logs and into the HwInfo, I get the following Info: "No Acquisition Memory chips (MAMs) on channel 1", but strangle enough this message is just a warning not an error. Any way how I can perform the self test??

No idea. I've been told that the self test from the production/developer menu is an internal LeCroy testing that requires some sort of a production test jig. Normal testing is done at the XStream app startup.

At least you have some hints where to look now (and the schematics) so I would start with measuring the voltages

On the other end I also fail cal with error Trigger threashold cal failed.

I don't know the exact message you have and there are huge differences between WaveSurfer and WaveRunner 6kA, but I sometimes get SWTriggerThresholdSearch errors at startup since I've upgraded to a Haswell i7-4770 motherboard with an external cheapo PCIe to PCI bridge (for the LeCroy aquisition card). Usually I need to turn the trigger levels and/or wait for the channel auto-calibration cycle and things are fine and the scope works.

If I move back to my previous PC hardware (Q8300 CPU, G45 Intel motherboard) that has a native PCI slot this error never happens.

[vaualbus]

Checking voltages.. would be easy but I can't acess the acquisition board when the scope is running....
Anyway at least in this scopo you really can't do much with this developer mode.... 
Maybye there are more hidden, for calibration for example?
Anyway I basicly sure ADC1 is fried at this point.
I really have to get the cal data working first I think.

[lukier]

Checking voltages.. would be easy but I can't acess the acquisition board when the scope is running....

It is similar with the WaveRunner. Fortunately, I didn't have to debug anything, but if I had then I would have to make PSU cable and LVDS cable extenders.

Anyway at least in this scopo you really can't do much with this developer mode.... 
Maybye there are more hidden, for calibration for example?

No idea, large part of this menu depends on your particular HW (i.e. WaveSurfer). On my WaveRunner there are options there to edit/tune all the internal calibration constants.

Anyway I basicly sure ADC1 is fried at this point.
I really have to get the cal data working first I think.

ADC is fried or memory (MAM)?

I would definitely try to load the calibration files. Maybe try to do some reverse engineering what the XStream software is checking when loading these files. I guess you must be 100% correct & certain on the software side before blaming the HW.

[vaualbus]

Ya I should really reverse engineering what the software does when it load the call data. Probably it does some check withe the hardware. I have not that much experience but I will try.surly if somebody here can help me it would be good 

[vaualbus]

But what file I should check? Is that all that you point out before menage the dso app??

[lukier]

LeCroy's XStream app is tricky to analyze, as it is a distributed system made out of ~100 dlls that are in fact DCOM OLE services. Therefore, even with IDA, analyzing the call flows requires a lot of actual debugging with plenty of various breakpoints and, of course, luck.

One can get some initial hints by tracing the WinAPI calls, as at some point one of the services will look for the calibration files and try to open them with the system call. I can recommend this tool:
http://www.rohitab.com/apimonitor

There are separate services that communicate with the HW (the PCI card). In WaveRunner 6100A case the serial number (along with other parameters, like back-up calibration constants I guess) is on a TSSOP48 FLASH on the acquisition board, but the options are stored on a 1-Wire EEPROM on the PCI interface card.

The module responsible for options, loading the scope serial number and computing the scope ID seems to be ConfigMgrSvr.dll, an encrypted/packed DLL that also contains the option key generation/verification routines. It calls the scope specific HW access service to get the serial number and the options stored in the 1-Wire EEPROM etc. Then all the other services query the ConfigMgrSvr and ask if this or that option is enabled, which model it runs on. This happens very early in the app startup process. It is also the ConfigMgrSvr module that tells the app that no HW is detected, therefore the message 'Not authorized to run on this system' is displayed (if you install XStream on a PC, without the scope bits).

As a side note, interestingly, there seems to be an option to run the software alone in a "simulation" mode and simulate various LeCroy scopes HW. I still haven't figured it out how to enable that, so one could run the app without the 'Not authorized to run on this system' message.

In a recent Tektronix video on MSO58 design efforts they've boasted that they had to build a SW-simulator, as the hardware was not ready (btw this is common practice in the embedded industry). Funny that LeCroy had this for over a decade I guess (along, among other things, with labelled axes, separate display grids and even Tek's UI looks similar to XStream, menu on top, channel, timebase, trigger buttons on the bottom etc).

To the subject. I didn't look how the app loads the calibration files in detail, as I didn't need that. If you manage to get some info on that area please share it here, as I bet it will be useful for other users of second-hand LeCroys.

[vaualbus]

Ok if I menage ti find something I will tell you. Any way I tried change the serial number and the board version to the one in the cal file I received but still I get the error cannot find the factory calibration file. Interesting.

[vaualbus]

I think maybe there is some checksum to the file, but my really question how the hell you create a new factory calibration file? Because probably there is  a way to do that.

[vaualbus]

So I was looking back at the logs, and the only real error I have happen is:
Source: TrigVref, Error: U2L:8700 L2U:16700 AVG:6000 DIFF:8000 accurancy testing NG
and a similar error: 
Source: TrigVref, Error: U2L:20279 L2U:20770 AVG:6000 DIFF:491 accurancy testing NG
so what the hell this mean? Prblem with trigger Vref? Look at the scematica I can't see a trigger vref...

[gkrusi]

vaualbus, have you succeeded in fixing it?

I came to a similar problem on my Wafesurfer 422 after motherboard's bad capacitors replacement and have no idea what to to do...

[vaualbus]

Yes come up with the sure of bad ADCs, well I could try cange.
If the ADC is not the problem could be something bad in the clock generation circuitary.
As my hope are almost null, the scope still stay in my home is still nice to have for free ahah

[vaualbus]

Well Well I have some news....
I finally decide to take serious the repair and so I order some PCI extender cable to be able to probe the acq board out of the instrument.
Well the set up I got was messy as hell I think when some chips arrive I will make a photo of it. This scope was not meant for repair for sure!
Any way I found some interesting things.
Ok so channel 1 hybrid is dead but that is something I knew already.
Now a part from that this is what I found:
          - Channel one VGA is dead aka signal enter no signal out
          - Channel two seem ok but given the fact it use both ADC (channel 1 & 2) after some sample rate, If have only channel 2 wnable I see bad signal on the screen.
          - Now the interesting part come with channel 3 and 4 where there is absolutely a fault. I tried to set up the those two channel as channel two and measure differences using the "good" channel 2 as a reference. What I found is amazing,
            It seem that they not receive proper signaling because for example the relay the route the signal to the ADC are not working because no command signal arrive at them. (Bot are controlled from 21IC10 and 21IC4)
            Also the DAC that control the channel 3&4 stuff show zero volts on all the output but have good ref voltage. So I think that something is bad before this chip, looking at the schematics both IC are controlled from a demux
            27IC5 that is selected from the previous IC, 27IC2.
So I have ordered those chips, as well as a new multimeter ahah, and I will try to replaced them maybe that allow me to get at least a trace on the screen for channel 3 & 4.