EEVblog #683 - Rigol DS1000Z & DS2000 Oscilloscope Jitter Problems

[pa3bca]

http://bugmenot.com/view/ni.com

The first login works

Thank you Sir!

[leppie]

I was also getting out of memory errors when capturing anything above 2Mpts.

Have you set the memory depth options correctly on your scope? (don't leave it on auto). Each of the memory depth settings works fine for me on a 1074.

Maybe a screenshot of the error? I think I've encountered it before, its driver/usb related and a reboot fixed it.

1: Scope maxed out on memory, click capture, get this.
2: Click OK, get this.
3: Click OK, get this, the click capture, UI stuck.

BTW on my 2560x1440 screen the UI (graphs rather) is rather 'tiny'.

[AlessandroAU]

I was also getting out of memory errors when capturing anything above 2Mpts.

Have you set the memory depth options correctly on your scope? (don't leave it on auto). Each of the memory depth settings works fine for me on a 1074.

Maybe a screenshot of the error? I think I've encountered it before, its driver/usb related and a reboot fixed it.

1: Scope maxed out on memory, click capture, get this.
2: Click OK, get this.
3: Click OK, get this, the click capture, UI stuck.

BTW on my 2560x1440 screen the UI (graphs rather) is rather 'tiny'.

Looks like it is physically running out of memory, I've removed some of the more computationally hard parts of the code (the part it got stuck at) and added more memory options. How many points can you download now?

[leppie]

I was also getting out of memory errors when capturing anything above 2Mpts.

Have you set the memory depth options correctly on your scope? (don't leave it on auto). Each of the memory depth settings works fine for me on a 1074.

Maybe a screenshot of the error? I think I've encountered it before, its driver/usb related and a reboot fixed it.

1: Scope maxed out on memory, click capture, get this.
2: Click OK, get this.
3: Click OK, get this, the click capture, UI stuck.

BTW on my 2560x1440 screen the UI (graphs rather) is rather 'tiny'.

Looks like it is physically running out of memory, I've removed some of the more computationally hard parts of the code (the part it got stuck at) and added more memory options. How many points can you download now?

This one works all the way to 4M. At 5M, I get out of memory. I run Win7 64-bit and have 16GB RAM.

It is not the reading bit that seems to be a problem but the FFT bit. I will DL the 64-bit runtime and see if it works better.

Update: 64-bit driver is a no go, dont bother... app wants 32-bit driver.

[Teneyes]

I would like to ask owners of Rigol DSO to review my Methodology of testing .
I am using the Passfail function to detect when there is an abnormal triggering.
I am inputting into Channel 1 a 200Khz Square wave
settings:
      Input coupling to AC
      Trigger coupling to AC
      Trigger level  to 300mV
      Time base to 10nSec/Div
      Persistence to infinite
DO a single trigger scan
Create a Mask from single scan
Run PassFail Test
The Pictures below show the Mask and 2 tests.
Is there anything more I should Set??
     

[Teneyes]

Well I see a few people seen the previous post and NO comments.
Here is a test of a 15MHz sinwave for more than 0.5 Giga Waveforms and there is No Jitter Error

[Teneyes]

Ok time to make a more interesting post
Here are 2 displays of the tests for Jitter
 1. No   Error  testing 20 Mega Waveforms
 2  827 Errors testing 10 Mega Waveforms
Only 1 DSO setting was changed and by only a relative small amount.!

I hope this will help Rigol narrow down the problem, and Address it if possible

Is it Jitter or just a fixed trigger position error?

[Bud]

Can you try the highest frequency that you can generate, close to the scope bandwidth limit? 15MHz is rather on the low side. The scope must perform up to the specs frequency limit, or it would not be a fair test.

[Teneyes]

Can you try the highest frequency that you can generate,

Good point Bud, how is this

[Bud]

Can you crank up the offset to 50...100uS and retest.
thks.

[Teneyes]

Can you crank up the offset to 50...100uS and retest.

I assume you are stating Trigger Delay.
I also show 55uS Delay
Just 1,000,000 waveforms tested

[AlessandroAU]

Ok the issues with the program should be fixed, it was running out of memory calculating waveform statistics, now it only uses the first 50k points to calculate waveform stats. I was also storing the waveform array with unnecessary precision, anyway it should be ok now.

Let me know of any bugs

[leppie]

Ok the issues with the program should be fixed, it was running out of memory calculating waveform statistics, now it only uses the first 50k points to calculate waveform stats. I was also storing the waveform array with unnecessary precision, anyway it should be ok now.

Let me know of any bugs

Thanks

My main issue is with the Labview Runtime actually uninstalling Ultra Sigma completely.

What should be selected in the install screen to not have this happen?

[AlessandroAU]

Ok the issues with the program should be fixed, it was running out of memory calculating waveform statistics, now it only uses the first 50k points to calculate waveform stats. I was also storing the waveform array with unnecessary precision, anyway it should be ok now.

Let me know of any bugs

Thanks

My main issue is with the Labview Runtime actually uninstalling Ultra Sigma completely.

What should be selected in the install screen to not have this happen?

I just installed the Labview RTE and then ultra sigma and don't seem to have it a problem, what happens for you?

[leppie]

Ok the issues with the program should be fixed, it was running out of memory calculating waveform statistics, now it only uses the first 50k points to calculate waveform stats. I was also storing the waveform array with unnecessary precision, anyway it should be ok now.

Let me know of any bugs

Thanks

My main issue is with the Labview Runtime actually uninstalling Ultra Sigma completely.

What should be selected in the install screen to not have this happen?

I just installed the Labview RTE and then ultra sigma and don't seem to have it a problem, what happens for you?

My UltraSigma got uninstalled Not sure if this happened during install or uninstall of the Labview runtime as I was trying to see if 64-bit runtime would work.

I will trying again today to see where it happened.

Update:

Installed UltraSigma again to make it work, everything else uninstalled.

This time I only used the 32-bit runtime, and only selected that in the installer.

I was prompted to reboot, but did not. (Will check if this causes the uninstall).

The good news is that I can capture all the way to 24Mpts, no crashes. 

Here is the FFT of a 24Mpts capture of an 8Mhz crystal, using 1024 averaging. Takes about 15 seconds to read the data from the scope.



Looks very sexy, still wish I can maximize the window

[Bud]

You do not want that kind of sex, the scope is really sick.

[leppie]

You do not want that kind of sex, the scope is really sick.

All these spikes to tease you with no dominant moment ;p

I just hope Rigol can sort this out in firmware, but I am not feeling very lucky now.

Even it just gets back to the original FW's behavior, I would be a little happier.

[poida_pie]

You do not want that kind of sex, the scope is really sick.

All these spikes to tease you with no dominant moment ;p

I just hope Rigol can sort this out in firmware, but I am not feeling very lucky now.

Even it just gets back to the original FW's behavior, I would be a little happier.

Yes, Leppie, I can live with a poorly designed PLL filter. That can be fixed via the application
of Bud's brain and a soldering iron once we know the values sent to the PLL chip by the new firmware to come.
The keyboard lockups however are something I do not want to have anything to do with.

just for fun I vomited up some python to display the ugliness of the PLL phase noise.
It produces an FFT very similar to MarkL. I could not get Alessandro's program to work.

Just strip the first 3200 bytes from a waveform file and rename to "data.raw"
Only the first 512K samples are used, I did not feel ready for the full impact
of the chaos clock in it's 1M sample glory.

Code: [Select]

import numpy as np
import matplotlib.pyplot as plt

global n, data,time

n=1024*512
dt = np.dtype('uint8')
data = np.fromfile("data.raw",dtype=dt,count=n,sep="")
time = np.linspace(1e-9,n * 1e-9,num=n)
fftpow = np.linspace(0,1,n+1)

wblackman = np.blackman(n)
wdata = wblackman * data[0:n]
fftres = np.fft.rfft(wdata,n)
fftpow = np.log10(np.absolute(fftres)/n)
fscale = np.linspace(0,1e9/2,1+n/2)

plt.subplot(311)
plt.plot(time,data)

plt.subplot(312)
plt.plot(fscale,fftpow)

plt.subplot(313)
plt.plot(fscale,fftpow)
fc = 20e6
plt.xlim(fc*0.9,fc*1.1)

plt.show()


[AlessandroAU]

You do not want that kind of sex, the scope is really sick.

All these spikes to tease you with no dominant moment ;p

I just hope Rigol can sort this out in firmware, but I am not feeling very lucky now.

Even it just gets back to the original FW's behavior, I would be a little happier.

just for fun I vomited up some python to display the ugliness of the PLL phase noise.
It produces an FFT very similar to MarkL. I could not get Alessandro's program to work.

Curiously, what was the problem with my program?


Edit:

Also here is my contribution:

top image 100khz sine measured with rigol at 24M points.

Bottom image: same source measured with ad7760 (24bit 2.5MSPS adc)

[poida_pie]

Curiously, what was the problem with my program?

I don't have a PC, only an imac. So I'm using VMware fusion, and a Windows XP VM and things don't quite work.
I can install the runtime from NI, connect the USB from the DS1054Z to the XP VM and then run your application. It shows via the changed button that it's downloading data but just does not.
I know the issue is the fact I'm running this inside a VM.

I run python on this VM fine, it connects realiably to the scope via USB - good for trying various options codes as well as transferring captured data.

[fusebit]

I tried the LabViev application, but I can't get useful results.
With an 5 MHz signal my FFT looks like this. If I change the X-axis scaling there is nothing above 0.5 Hz 
But the found freq field shows 5m!?

[AlessandroAU]

I tried the LabViev application, but I can't get useful results.
With an 5 MHz signal my FFT looks like this. If I change the X-axis scaling there is nothing above 0.5 Hz 
But the found freq field shows 5m!?

That's really strange, I can't recreate the fault, what are your capture parameters? Scope Model etc?

[fusebit]

I'll try it again in the evening.
It's a 1074z, I'll post the parameter details later...

The captured waveform is shown in the upper window of your VI, but the FFT doesn't make sense.
Is there a particular setting at the scope I should be careful with to get good results?

I'm just a hobbyist and this is my first useful scope (beside a 10 MHz single channel Hameg  ), so it's easily possible that the problem is located between my ears...
 

[Teneyes]

Well , 94  Errors even before  2 Giga Waveforms were sampled   He he 

[pa3bca]

Off topic (nothing to do with scope jitter) but just to show how useful AlessandroAU's Labview program is, for those who do not have an SA:

Just a 200 KHz ARB from my DG1032Z.
Waveform + FFT on the DS1074Z, FFT from 0-6MHz:


FFT from 1 Megasamples with AlessandoAU's program:


And what it "really"  looks like if analyzed with a DSA815 SA:


Pretty neat huh?