Author Topic: Anomalies and/or Bugs in/between the Rigol MSO/DS1000Z, MSO/DS2000A, MSO/DS4000  (Read 18450 times)

0 Members and 1 Guest are viewing this topic.

Offline marmad

  • Super Contributor
  • ***
  • Posts: 2986
  • Country: aq
    • DaysAlive
I'm starting this thread to discuss any and all anomalies and/or bugs in or between the Rigol UltraVision DSOs (MSO/DS1000Z, MSO/DS2000A, MSO/DS4000).

I'm currently working on a video examining the differences between High Res mode on the DS1000Z and DS2000, but meanwhile I'm discovering other anomalies as well so I wanted a place to post them. I figured this might be a handy way to highlight some differences between the models that is not readily apparent from the specs.

Personally, I'm mostly interested in comparisons between 2 or more of the series, but feel free to discuss any bugs or problems you've discovered in any of the UltraVision models - anything is open for discussion - if you want to be an off-topic policeman, find another thread  ;)

I noticed when I was setting up some tests today that the MSO1000Z displays a Ground-coupled channel input in an unexpected way (to me anyway). It looks more or less identical to the DC-coupled setting when there is no input to the channel (i.e. with some noise), while the DS2000 'simulates' a perfectly noise-free grounded input. So I figured this weird little anomaly was a point with which to begin this thread:




« Last Edit: November 23, 2014, 06:53:46 AM by marmad »
 

Offline motocoder

  • Supporter
  • ****
  • Posts: 768
  • Country: us
  • Electrical Engineer
Here's a link to my thread, in which we are discussing a lock-up triggered by certain events on the network interface.

http://www.eevblog.com/forum/testgear/rigol-ds2072-rma-ed-to-rigol-for-repair/
 

Offline marmad

  • Super Contributor
  • ***
  • Posts: 2986
  • Country: aq
    • DaysAlive
Here's a link to my thread, in which we are discussing a lock-up triggered by certain events on the network interface.

http://www.eevblog.com/forum/testgear/rigol-ds2072-rma-ed-to-rigol-for-repair/
Hey motocoder - welcome!  :)  I've been following your thread with interest.

One quick question for you: there were some seemingly drastic changes with the LAN code in the DS2000 FW, beginning with, I believe, v.3 (although maybe it was v.2). Have you downgraded to FW 02.01.00.03 or FW 01.01.00.02 to see if the lock-ups persist?
« Last Edit: November 23, 2014, 06:50:34 AM by marmad »
 

Offline i4004

  • Regular Contributor
  • *
  • Posts: 70
  • Country: hr
perhaps "grievances" would be a best choice of words?   :-//
(as this is not a bug or anomaly, it's just way to make more money....lol!)

like mentioned here
http://www.eevblog.com/forum/blog/eevblog-683-rigol-ds1000z-ds2000-oscilloscope-jitter-problems/msg555785/#msg555785
1054z has 700x200 resolution, so nobody should scratch their heads when they see jaggy screenshots....  :-//
and i think that goes for whole z series...all models have 8bit adcs?

if i read the specs pdf, even 4k series mentions 12bit in "only if" terms, simillar to 1000z's "12bits max".

so it's 8bit everywhere.
so it's low vertical res everywhere. out of curiosity, is this the norm today, are "more vertical bits" reserved for much more expensive scopes?

as for this reply from marmad
Quote
Well, forgetting Hi Res for a moment, the true resolution of the ADC is 256 bits - so the question becomes: how do you map 256 voltage levels to 400 vertical pixels? You have to so some upsampling. I haven't examined the DS1000Z closely, but on the DS2000, Rigol maps each level to 2 pixels, so the full ADC = ~10.2 vertical divs.
on "old" thread,
i would say they just "crop" those extra 56 levels (i would like it if they were to put "256 bits" adc like you've mistaken above!), because how do you decide which pixel will be "2 pixel height" and which "just one", and if "just one" doesn't even exist (in any screenshot), then you surely have only 200 pixels.

edit/ they can also map 256 levels to 200 pixels by making each "double" pixel have more than one color tone, which they probably do, but did anybody check the vertical resolution by providing input signal with 256 (and more) voltage levels?

in my screenshot from that thread, i just see "2 pixel height".



offcourse resampler you use to blow the image counts, must be "point resize" or so...irfan view should do it if you just hit the "+" (zoom in) button...see the image..any vertical samples of just 1 pixel height?
inspect the irfan-view blown image below.

« Last Edit: November 23, 2014, 07:10:43 AM by i4004 »
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 5514
  • Country: us
  • DavidH
I noticed when I was setting up some tests today that the DS1000Z displays a Ground-coupled channel input in, to me anyway, an unexpected way. It looks more or less identical to the DC-coupled setting when there is no input to the channel (i.e. with some noise), while the DS2000 'simulates' a perfectly noise-free grounded input.

The old Tektronix 2232 series of DSOs have a settable "smooth" function which zeros out noise below a threshold in the same way that an audio noise gate operates.  The later 2440 series as "smoothing" which is completely different.

Smooth

Uses a digital process to smooth the waveform display, yet retain the glitch catching capabilities of Peak Detect or Accumulate Peak acquisition modes.  (Smooth applies only to the Peak Detect or Accumulate Peak modes.)


The result is that peak detection which would normally show peak-to-peak noise from the digitizer instead produces a clean trace.  I find it annoying because the extra processing slows down the acquisition rate so I usually leave it disabled but it does produce a nice looking display.

Maybe the DS2000A series is doing something similar.  Peak detect or envelope mode might change this behavior.
 

Offline marmad

  • Super Contributor
  • ***
  • Posts: 2986
  • Country: aq
    • DaysAlive
perhaps "grievances" would be a best choice of words?   :-//
(as this is not a bug or anomaly, it's just way to make more money....lol!)

 ;D  Perhaps - but one man's grievance might be another man's expected behavior  ;)

Quote
so it's low vertical res everywhere. out of curiosity, is this the norm today, are "more vertical bits" reserved for much more expensive scopes?

Yep - pretty much. Attached is a zoomed and cropped image from Agilent 3000 X-Series, showing the same upscaling for their 400 pixel vertical display.
 

Offline marmad

  • Super Contributor
  • ***
  • Posts: 2986
  • Country: aq
    • DaysAlive
The result is that peak detection which would normally show peak-to-peak noise from the digitizer instead produces a clean trace.  I find it annoying because the extra processing slows down the acquisition rate so I usually leave it disabled but it does produce a nice looking display.

Maybe the DS2000A series is doing something similar.  Peak detect or envelope mode might change this behavior.

Interesting. Peak detect made no difference (and there is no Envelope mode on the DS2000), but High Res (at >=5us/div) changed the results slightly.

EDIT: Although I'm now wondering if that slight change is just due to the drop in waveform captures from 104 to 17 wfrm/s (when High Res is switched on while in AUTO-trigger mode @200us/div).
« Last Edit: November 23, 2014, 07:32:22 AM by marmad »
 

Offline i4004

  • Regular Contributor
  • *
  • Posts: 70
  • Country: hr
your agilent screenshot shows both 1 and 2pixels vertical samples, you saw same thing on any rigol screenshot?
(or is it 2 and 4 vertical pixels i'm seeing there on agilent?)

ie 1 discrete 'vertical' pixel?

as we already saw 2pixels samples in abundance on rigol.

although they're both 8bit scopes, it's just that they use different means to display it...i would prefer the agilent way because it's closer to the voltage waveform....


(for me it's just silly i bumped to 8bit adc so quickly in 2014, i mean it's 90s technology..and in the scope....!!??  :scared: )
« Last Edit: November 23, 2014, 07:39:30 AM by i4004 »
 

Offline marmad

  • Super Contributor
  • ***
  • Posts: 2986
  • Country: aq
    • DaysAlive
your agilent screenshot shows both 1 and 2pixels vertical samples, you saw same thing on any rigol screenshot?
(or is it 2 and 4 vertical pixels i'm seeing there on agilent?)

No, it's 2 and 4 vertical pixels - look at Photoshop ruler.

Quote
although they're both 8bit scopes, it's just that they use different means to display it...i would prefer the agilent way because it's closer to the voltage waveform....

The Agilent image just happens to be a stopped DSO - in other words, not an intensity-graded image. The Rigol looks exactly the same when stopped.
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 5514
  • Country: us
  • DavidH
The result is that peak detection which would normally show peak-to-peak noise from the digitizer instead produces a clean trace.  I find it annoying because the extra processing slows down the acquisition rate so I usually leave it disabled but it does produce a nice looking display.

Maybe the DS2000A series is doing something similar.  Peak detect or envelope mode might change this behavior.

Interesting. Peak detect made no difference (and there is no Envelope mode on the DS2000), but High Res (at >=5us/div) changed the results slightly.

EDIT: Although I'm now wondering if that slight change is just due to the drop in waveform captures from 104 to 17 wfrm/s (when High Res is switched on while in AUTO-trigger mode @200us/div).

The displayed level of noise in the original DS2000A image does not look consistent with what I would expect from a buffer with a 15 picofarad input capacitance and excluding quantization noise.  What is the input bandwidth?  At 200 MHz with a shorted input, I would expect a lot more than even 1 millivolt of peak-to-peak noise.

High resolution or averaging will reduce the displayed noise considerably of course.
 

Offline i4004

  • Regular Contributor
  • *
  • Posts: 70
  • Country: hr
here is 1054z's

1-gnd-ed input
2-gnd-ed on probe
(while grounding the probe to the 'calibrator' ground produces more noise than it can fit the screen on 10mV/div)

i compared this scope without a connected probe to that what connor wolf mentions on his ds4k test yt video, and i actually got a bit lower front-end noise than he, but not that much lower...
(even though my 240v mains are not grounded...it's just L and N in the wall socket..connecting ground brings this lower yet...by some 80uV (0.08mV) to about 1.36mV front-end noise...)

i dunno why would 2k be perfectly still in that situation, some noise must exist there....
 

Offline marmad

  • Super Contributor
  • ***
  • Posts: 2986
  • Country: aq
    • DaysAlive
The displayed level of noise in the original DS2000A image does not look consistent with what I would expect from a buffer with a 15 picofarad input capacitance and excluding quantization noise.  What is the input bandwidth?  At 200 MHz with a shorted input, I would expect a lot more than even 1 millivolt of peak-to-peak noise.

The display does not change based on the vertical scale - it seems as if the input to the ADC is shorted (or something similar). I've seen this same method used on other DSOs.

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

BTW, continuing our previous conversation from the other thread (Hi Res mode and Tektronix - deemed OT by OCs), I found some specifications from Tektronix as to how they map waveform values to the display (at least on the TDS XXXA series):

According to the specs, Normal mode maps 6.8 bits to 460px (i.e. ~4.128 pixels per value - some rather serious upscaling) - while Hi Res mode maps 9.7 bits to 460px (i.e. ~0.55px per value). From these numbers it appears as if you would only see a 1:1 mapping when Hi Res was equivalent to 8.85 bits of resolution, otherwise it would almost always be downscaling (unless I'm completely missing some salient point - which is totally possible  ;D ).

Other than that, I have found no indication that the added resolution in the waveform record would be usable in real time. It would only be useful when the DSO was stopped - or for viewing/processing on an external device (which, granted, would still be a nice feature to have).
« Last Edit: November 23, 2014, 10:24:43 AM by marmad »
 

Offline marmad

  • Super Contributor
  • ***
  • Posts: 2986
  • Country: aq
    • DaysAlive
here is 1054z's

1-gnd-ed input

@i4004: Would you mind checking your GND-coupled input @ the lowest timebase and vertical scale? Here is the MSO1074Z that I have @ 5ns/1mv/div:
 

Offline i4004

  • Regular Contributor
  • *
  • Posts: 70
  • Country: hr


No, it's 2 and 4 vertical pixels - look at Photoshop ruler.


so both agilent and rigol have a usefull vertical display resolution of 200pixels...


yep, lo-res it is!

but not as low as 1052.

and dave just made a new thread on hires mode...heh...hold on, can't find it now(probaby saw this one and erased it), so it's ok...i'll say it here:
this scope doesn't have 12bits and doesn't have any hires mode, and neither does tek with such low display resolutions!  ;)
12bit adc would be 4096 levels which would be best translated to 4096 pixels of vertical resolution on the display. in that case horizontal res. should be notched up a bit, to do true 8k display for the scope.....yiiiihhhhaaaaa!  :scared:


no, i wouldn't mind that at all, it's running anyway....fastest timebase and least v/div here for you
(default and max sample memory)
0v trigger.

and turn on that vpp counter if you're interested in noise...  ;)


« Last Edit: November 23, 2014, 10:45:16 AM by i4004 »
 

Offline eV1Te

  • Regular Contributor
  • *
  • Posts: 153
  • Country: se
  • Your trusted friend in science!
    • richardandersson.net
In order to really see the 12 bit of information when using a Hires mode one would need to export the data or stop the scope and "zoom" into a portion of the captured signal (vertically). Also the FFT of the signal would get a much lower noise level.

I took these screenshots on my DS1104z-S after stopping the scope at 100 mV/div vertical, and then changed the vertical scale 10 times to 10 mV. It is obvious that the data is displayed in 8 bits no matter which timebase or setting you use (at least I have not succeeded).

Both the high res and the averaging 256 times should be able to generate 12 bit effective resolution.

First picture  is in Normal mode:


This is in Hires mode (no difference in resolution still 8 bits):



This is Averaging 256 times (still 8 bits):
« Last Edit: November 23, 2014, 12:26:26 PM by eV1Te »
 

Offline marmad

  • Super Contributor
  • ***
  • Posts: 2986
  • Country: aq
    • DaysAlive
In order to really see the 12 bit of information when using a Hires mode one would need to export the data or stop the scope and "zoom" into a portion of the captured signal (vertically). Also the FFT of the signal would get a much lower noise level.

I took these screenshots on my DS1104z-S after stopping the scope at 100 mV/div vertical, and then changed the vertical scale 10 times to 10 mV. It is obvious that the data is displayed in 8 bits no matter which timebase or setting you use (at least I have not succeeded).

Yes, as I've mentioned before, the Rigols (although I'm not sure about the DS4000) don't save the High Res averaged data in sample memory - meaning they only work on the displayed image, which is downscaled while running. If it wasn't, then when you, for example, switched on High Res using the full 12-bits (at least on the DS2000), the scale of the display would immediately change by a factor of 10.

The problem is, when you STOP the DS1000Z and then increase the vertical scale, you're only changing the scale of the display memory - which has already been downscaled from the averaged data. On the DS2000, when STOPPED, it recalculates the averaging from the original samples every time you change the horizontal or vertical scale.
« Last Edit: November 23, 2014, 02:11:43 PM by marmad »
 

Offline marmad

  • Super Contributor
  • ***
  • Posts: 2986
  • Country: aq
    • DaysAlive
Here are two series of images demonstrating the High Res recalculations being done by the Rigol DS2000 while stopped.

Both series of images are of the exact same zoomed-in portion (vertical scale) of a sawtooth waveform (a ramped edge) - and are the same series of horizontal timebases, which start at the timebase when High Res first becomes active (500ns) and continue to the first timebase when it averages to a full 12-bits.

The first series is done with the DSO in Normal mode - and the second series is done with High Res active.
 

Offline marmad

  • Super Contributor
  • ***
  • Posts: 2986
  • Country: aq
    • DaysAlive
Here are two more images that perhaps show the High Res recalculations more succinctly.

DS2000 stopped while in Normal mode, then vertical scale decreased from 200mv/div to 20mv/div:




Then High Res mode turned on:

« Last Edit: November 23, 2014, 02:39:27 PM by marmad »
 

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 2485
  • Country: fi
  • Starting with DLL21

I noticed when I was setting up some tests today that the MSO1000Z displays a Ground-coupled channel input in an unexpected way (to me anyway). It looks more or less identical to the DC-coupled setting when there is no input to the channel (i.e. with some noise), while the DS2000 'simulates' a perfectly noise-free grounded input. So I figured this weird little anomaly was a point with which to begin this thread:

My opinion is that finally they have hit how it need do.  If GND is just as writing straigh line to display what information it give. Nothing, exactly nothing. But well done GND + front end to display, it tell something about front end working example in case of fault.  Why "simulate" GND coupling, it is better to do real GND coupling as have done last 50 years.

(it is even perfect if there is also real GND switch before first stage in analog front end what means that front end input is connected to GND and BNC input pin is leaved very Hi-Z floating with minimal capacitance)

When set input GND, if you change input V/Div do this noise change. How about if change horizontal t/div.
I like  if it works so that real input is tied to GND and INput BNC center is floating. If it is done so that input amplifiers and ADC is still producing normal noise and input is like "no signal" it is still acceptaable and well better than "simulating" GND writing constant zero to display. (what hell this straigh constant line in display memory tells to user  = do not simulate if you can do it real)
If practice and theory is not equal it tells that used application of theory  is wrong or the theory itself is wrong.
It is much easier to think an apple fall to the ground than to think that the earth and the apple will begin to move toward each other and collide.
 

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 2485
  • Country: fi
  • Starting with DLL21


Yes, as I've mentioned before, the Rigols (although I'm not sure about the DS4000) don't save the High Res averaged data in sample memory.

Do it mean that you can not get out 12-bit  waveform data from scope if it is captured in Hi-Res mode?  Do measurements/cursors  use this 12bit data or 8bit data?
If practice and theory is not equal it tells that used application of theory  is wrong or the theory itself is wrong.
It is much easier to think an apple fall to the ground than to think that the earth and the apple will begin to move toward each other and collide.
 

Offline i4004

  • Regular Contributor
  • *
  • Posts: 70
  • Country: hr
Quote
Here are two more images that perhaps show the High Res recalculations more succinctly.

it's just calculating how to make that thick line thinner, but nothing is really gained with regard to sampling, you could also make that line thinner in your head with same results, you didn't really gain any more adc vertical resolution.

so it's really cheating, and the waveform shown is just a guess....

if you had a scope with more bits in adc, it would show something else, something more genuine...this is a play with 8bit data....
 

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 2485
  • Country: fi
  • Starting with DLL21


if you had a scope with more bits in adc, it would show something else, something more genuine...this is a play with 8bit data....


With 8bit ADC we can get more than 8 bit.  But it need averaging with noise. Noise can be input stage normal noise or if this noise is very small then there need do noise injection.
But, price is speed.  It really reduce speed lot of.
It can do with waveform averaging if average enough long time. But this make whole waveform all changes slow.
With low horizontal speed we can use other method.
Example if scope have 1GSa/s speed.  When you turn horizontal so that you see scope have now example 1MSa/s. What is ADC itself speed, it is still 1GSa/s.  System only take every 1000's  sample.  999 samples are flushed out.
But if we use these 1000 samples for averaging and then produce this one bit we get more accurate sample. (if there is enough noise and noise quality is good for this). 
What we loose. We do not know in horizontal axis what is right time position for this averaged result. We can do it even better if take all these 1000 pixels and note every pixel time and then some more math but no reason to do it in normal lab oscilloscope.

What is real gain of absolute accuracy perhaps not need even tell but rised relative accuracy and resolution  inside some captured waveform may be "lot of".



« Last Edit: November 23, 2014, 08:41:12 PM by rf-loop »
If practice and theory is not equal it tells that used application of theory  is wrong or the theory itself is wrong.
It is much easier to think an apple fall to the ground than to think that the earth and the apple will begin to move toward each other and collide.
 

Offline TMM

  • Frequent Contributor
  • **
  • Posts: 413
  • Country: au
Here are two more images that perhaps show the High Res recalculations more succinctly.

DS2000 stopped while in Normal mode, then vertical scale decreased from 200mv/div to 20mv/div:

http://www.eevblog.com/forum/testgear/anomalies-andor-bugs-inbetween-the-rigol-msods1000z-msods2000a-msods4000/?action=dlattach;attach=120108


Then High Res mode turned on:

http://www.eevblog.com/forum/testgear/anomalies-andor-bugs-inbetween-the-rigol-msods1000z-msods2000a-msods4000/?action=dlattach;attach=120110
I'm trying to replicate this on the DS1000Z, let me know if i'm doing anything wrong.

Using 1Vp-p 1.5KHz ramp from DG4062,
200mV/10uS per div, normal acquisition:


Then adjusted to 20mV/div after stopping the above:


Same thing again, but high-res acquisition instead of normal:
 

Offline TMM

  • Frequent Contributor
  • **
  • Posts: 413
  • Country: au
And again with shorter memory depth (i set the memory depth to auto).

Normal:


High Res:
 

Offline TMM

  • Frequent Contributor
  • **
  • Posts: 413
  • Country: au
And slowed down by 1 and 2 orders of magnitude:





Last test run again with forced short memory.


« Last Edit: November 23, 2014, 08:53:29 PM by TMM »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf