Author Topic: Rigol DS1000Z signal level bug - false Sin(x)/x and "fake" 1 - 2mV/div.  (Read 5240 times)

0 Members and 1 Guest are viewing this topic.

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 2363
  • Country: fi
  • Starting with DLL21
I did some tiny work with DS1074Z (same as DS1054Z and DS1104Z). Signals what I was looking was very different what now here in test images. First I did not note signal level due to continuously variable live signal but then when I do some measurement I find that part of signal level was too low.

After short check I find that really, if I change ONLY timebase and all other things unchanged.  Slower than 100ns/div level is constant. With 100ns/div or more fast level change. There is severe level step depending of signal details. Alone this design error drop this equipment to "toy" class. More bad is that user can not mostly even turn off this function. Why?  Is it because entry level hobbyists want "nice image" instead of trusted measurements. This kind of "smooth" filter is then named Sin(x)/x. Bad is that if flush to toilet all true sample points and the reality is replaced by an invented data.  If some analog scope in old times do this kind of error then responsible people set adhesive sticker over display "out of order" or just "do not use" so that peoples in lab do not accidentally use it before it is repaired.


After then I take other scope (because after this finding I did not know what I can trust with this Rigol) and continue my work  and later I did some tiny tests about what is going on.

Pick-up some tiny examples from tests: part I

Here tested with steady and enough good quality sinewave for this kind of test.
Signal source HP8657B. Signal 70MHz sinewave. Level set so that signal is around 6 div high peak to peak with scope setting 50mV/div. Scope end terminated with Tek 50ohm feed thru. From HP to Rigol 50ohm cable.

Oscilloscope display mode: Dots
Sampling mode normal, Input DC, Trig C. DC, Trig Rising edge.
All 4 channels on. Samplerate 250MSa/s

First 3 images:
200ns/div ; Sin(x)/x  OFF  scope show normal level (as also with more low speeds)
100ns/div ; Sin(x)/x  OFF  level drops down and then stay same level from this speed down to 5ns/s
005ns/div ; Sin(x)/x  OFF level drops down

Next 3 images:
200ns/div ; Sin(x)/x  ON   level ok (as also with more low speeds)
100ns/div ; Sin(x)/x  ON   level ok
005ns/div ; Sin(x)/x  ON   level ok

Also if think typical criteria for sin(x)/x interpolation there is 250MSa/s and input is 70MHz sinewave. Samplerate/input frequency = 3.57 what is ok. Even with 100MHz it is still in acceptable range (2.5)

But what is this problem?  Note that it happends when Sin(x)/x interpolation is selected OFF! and even more, there is not even vectors, display mode is Dots not Vectors!
(yes I have tested also same so that only display mode changed to Vectors. Result exactly same!

So, what this machine is doing?

But I have also question if someone know real true and not only quess.

If I want save  sampled data as .CSV file,  are these sample points true native raw data from ADC (only scaled with known vertical settings and/or other exactly known parameters.) or are these saved points after some undefined "Rigol make-up" process. Are there any solution to get real ADC RAW for later external analysis. If analyze data what is manipulated with unknown way it is nearly waste of time analyze these. Garbage in - analyzing - garbage out.


Later, when I have time to extract out more my test notes and arrange data and images then more in part II what  rise more questions related to this and display mode Vectors/Dots and with also Sin(x)/x ON/OFF.

Here more readings also:
http://cdn.teledynelecroy.com/files/whitepapers/wp_interpolation_102203.pdf
http://m.eet.com/media/1051226/Sin(x)x_Agilent.pdf

« Last Edit: September 16, 2016, 03:15:00 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 rf-loop

  • Super Contributor
  • ***
  • Posts: 2363
  • Country: fi
  • Starting with DLL21
Re: Rigol DS1074Z signal level fault - bug.
« Reply #1 on: December 06, 2014, 11:53:06 PM »
Part II

In all images test signal is exactly same. 70MHz sinewave as also previously.

1 Sin(x)/x ON. Display Vectors. Signal level on screen 300mV p-p

2 Turned Sin(x)/x OFF.  Signal level on screen drops to around 240mV p-p. Around 20% drop!

3 Sin(x)/x ON. Display Dots.

4a Sin(x)/x OFF. Display Dots. Signal drops around 20%.

4b as 4a but scope stopped. Zoomed out and in so that one waveform capture dots only are visible. These looks like real unmanipulated samples but then rise question what happends between timebase 200us and 100us/div in part I. What are they really doing after ADC and before display?

4c as 4b but now Display Vectors. This is how Rigol think "vectors". In my eyes it looks like Sin(x)/x for these sample points.

Then scope run with  same signal. This is because Sin(x)/x do not affect anything when scope is stopped. Display Dots. Sin(x)/x ON.
 

5a  Scope stopped,  Zoomed out and in so that single acquire dots are visible. But I suspect these dots are not real ADC samples. It looks like these are manipulated using some undefined function so that levels are rised. If now save .CSV there is these manipulated sample points, not true ADC sample points. How to analyze externally this data what is manipulated with some undefined method. 

5b  same as 5a, scope still stopped but now display mode Vector. Now it looks like it do Sin(x)/x for these dots what can see in image 5a.


Also need note that these are made using quite pure sinewave signal. Normally we do not so much inspect this kind of signals. Normally oscilloscope is used for inspect unknown  signals or check condition of known komplex signals what include lot of sinewaves with different frequencies and levels, in simple case example some kind of square waves or pulses. This kind of design fault (or kind of bug)  in oscilloscope can not accept and it may lead  errors in signal analysis.

Later part III.
« Last Edit: December 07, 2014, 06:49:18 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 rf-loop

  • Super Contributor
  • ***
  • Posts: 2363
  • Country: fi
  • Starting with DLL21
Re: Rigol DS1074Z signal level fault - bug.
« Reply #2 on: December 08, 2014, 03:08:47 AM »
Part III

Here one example about how it works right with some no named (and much more expensive) oscilloscope.
This may give better vision what is wrong with Rigol DS1000Z design but just  this detail is not question about price. It need only that designer take his work seriously. There can easy add thousends of dingle dongles and rise number of features but these most basic fundamentals need first be in hand.

With this example:
Sin(x)/x  ON or OFF do not destroy/change original real sample points.
Horizontal speed setting do not alter signal level.


Samplerate and input sinewave frequency is selected so that sampling frequency and signal frequency ratio is same.
Because other scope is much faster I have used 280MHz sinewave so this ratio is same 3.57 as used with Rigol DS1074Z.


Signal: 280MHz quite pure sinewave and other way same as used with Rigol but not 50ohm Tek feed thru due to this have 50ohmj inputs. Level adjusted for around six div peak to peak.
Selected two channel for use so that samplerate is 1GSa/s (lowest native ADC samplerate with this scope, just as 250MSa/s is lowest native samplerate with Rigol DS1000Z) Display mode vectors exept if told that mode is dots.


Images a1 and a2:
Sin(x)/x ON or OFF do not affect signal level with 200ns/div

Continuing with same signal. No change in level when step to 20ns/div

Images b3 and b4
Sin(x)/x ON or OFF do not affect signal level with 20ns/div

Continuing with same signal. No change in level when step to 1ns/div.
Images c5 and c6
Sin(x)/x ON or OFF do not affect signal level with 1ns/div

Note for c5:  In vector mode with running scope there can see natural things. Rising slope near trigger point is very narrow. But in same image signal other rising and all fallinf slopes are wide what looks somehow loke "timing jitter".  This is natural. Think sampling points period is 1ns and vectors between these points need fine adjust so it cross perfectly over trigger point.  Every new waveform capture have sampling points in different position related to input signal.
There are  lines drawed between these points. But every line what is between nearest points before and after trigger point need adjust so that this line cross exactly over trigger point. So there need be position fine adjust. This moves samples backward or forward related to trigger point.
But this is fun geometry, after Sin(x)/x trace is in right place, without any trick.



Image c7
Continuing with same signal scope is now stopped and display mode Dots.

Keep stopped

Image c8
Continuing with same signal and scope stopped and now display mode Vectors.

Keep stopped and keep display mode Vectors

Image c9
Still continuing with same signal and skope stopped and now display mode Sin(x)/x ON
(note, if now select display mode Dots, there is still original sample dots just as image c7)

Later Part IV  how this affect with other than pure sine wave.
« Last Edit: December 08, 2014, 03:36:28 AM 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 marmad

  • Super Contributor
  • ***
  • Posts: 2986
  • Country: aq
    • DaysAlive
Re: Rigol DS1074Z signal level fault - bug.
« Reply #3 on: December 08, 2014, 12:56:56 PM »
Part II

In all images test signal is exactly same. 70MHz sinewave as also previously.

1 Sin(x)/x ON. Display Vectors. Signal level on screen 300mV p-p

2 Turned Sin(x)/x OFF.  Signal level on screen drops to around 240mV p-p. Around 20% drop!

3 Sin(x)/x ON. Display Dots.

4a Sin(x)/x OFF. Display Dots. Signal drops around 20%.

4b as 4a but scope stopped. Zoomed out and in so that one waveform capture dots only are visible. These looks like real unmanipulated samples but then rise question what happends between timebase 200us and 100us/div in part I. What are they really doing after ADC and before display?

4c as 4b but now Display Vectors. This is how Rigol think "vectors". In my eyes it looks like Sin(x)/x for these sample points.

Nice find. I just tested this, and I can confirm it happens also on the MSO1074Z that I have. The Sin(x)/x ON/OFF switch does absolutely nothing whatsoever to change the interpolation - if the DSO is using Sin(x)/x, it stays Sin(x)/x - and if it's using Linear, it stays Linear.

But I think Rigol has completely screwed-up with the labeling of this switch - I think, instead of interpolation, it's doing some kind of BW-limiting (maybe 50MHz?) on all of the input channels simultaneously - to try to prevent aliasing when 3/4 channels are active with the reduced 250MSa/s rate. But I'm not sure why it only seems to be active between 100 - 5ns/div.

Image #1 shows 100MHz sine to CH 1 & 4: "Sin(x)/x ON"
Image #2 shows 100MHz sine to CH 1 & 4: 20MHz BW limit on CH1 - "Sin(x)/x ON"
Image #3 shows 100MHz sine to CH 1 & 4: 20MHz BW limit on CH1 - "Sin(x)/x OFF"
« Last Edit: December 08, 2014, 01:33:18 PM by marmad »
 

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 2363
  • Country: fi
  • Starting with DLL21
Re: Rigol DS1074Z signal level fault - bug.
« Reply #4 on: January 16, 2015, 06:09:54 PM »
I will rise this up.

Scope is full of many kinds of dingle dongle features but Basic Fundamentals are "out of order".

I hope Rigol note this design error and take this seriously. This need repair.

I have not seen this kind of error in any basic level oscilloscope exept in Rigol. 

This (some amount different) joke/crap was also implemented in old DS1000E series.

Rigol do not even listen there is problem? Do they think perhaps that kids need toys and thousend dingle dongle features for amazing and play with.

Professonals need oscilloscope where basic fundamantals are working right so that user can trust what scope show.

I know many (specially noobs) peoples want make-up images from signals so they do not need wonder what is this cratch and what is this pixel up and down in my ideal signal what come from my 5$ crap "wave generator". 

Please Rigol.

Be honest and show what scope is sampling.  Show real dots (and only sampled real dots if user have selected dipsplay mode "dots") then show straight lines between dots if user have selected "lines" (without cosmetics image make-up) and finally also real Sin(x)/x function if user have selected it. And do not if user have selected it off.   

(If  Rigol EE even know what it means. If you Rigol EE can read (I suspect it difficult and it can suspect afteer we know how you have read components data sheets, example PLL. (look PLL kidding)).

It is very very simple paper available if school book did not tell it, read example Chris Rehorn's (Hewlett-Packard  -- Agilent (--Keyshit today) simple explanation what  is Sin(x)/x (Sinc) and think also what it is not. Or ask Siglent how to do it.

Please, Rigol, give some explanation what this DS1000Z is doing when it do some "magic" functions between ADC and Display.
I need know signal in DUT - not your image cosmetics.
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 Solder_Junkie

  • Regular Contributor
  • *
  • Posts: 66
  • Country: gb
Re: Rigol DS1074Z signal level fault - bug.
« Reply #5 on: January 16, 2015, 10:23:41 PM »
If you leave Sin(x)/x ON, it seems to behave itself...

However, try this anomaly:
Input a sine wave at 50 MHz, 100 mV RMS (actual level perhaps irrelevant).
Single channel on.
Set timebase to 10 mS...

On my DS1054Z, it shows a sinewave around 500 Hz. Change the time base to 10 nS and it shows a nice sinewave.

Anyone else see this? I can't find a setting to stop it.
 

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 2363
  • Country: fi
  • Starting with DLL21
Re: Rigol DS1074Z signal level fault - bug.
« Reply #6 on: January 17, 2015, 01:06:26 AM »
If you leave Sin(x)/x ON, it seems to behave itself...

However, try this anomaly:
Input a sine wave at 50 MHz, 100 mV RMS (actual level perhaps irrelevant).
Single channel on.
Set timebase to 10 mS...

On my DS1054Z, it shows a sinewave around 500 Hz. Change the time base to 10 nS and it shows a nice sinewave.

Anyone else see this? I can't find a setting to stop it.

Basic fundamentals of digital oscilloscopes.
What is sampling speed with timebase 10ms/div.
Is it enough for 50MHz signal? 100MSa/s with 50MHz sinewave is just Nyquist freq of 100MSa/s.

If take theoretically exactly 100MSa/s and if input is exactly 50MHz (and they are phase locked) it show  DC level.  Level of DC  what is on display is related only with sampling and input signals phase.  (they need be phase locked.)

If you take 50000000Hz signal and sampling frequency is 100000001Hz you can see 1Hz sinewave then add more difference and some practical jitter, phase noise etc and we can see  something what we then can just wonder.

We have one famous name for it: Aliasing.

(what I told is theory and in this theory also practice is same. But what I do not remember, DS1000Z sampling rate with 10ms/div)

And then, this have nothing to do with these things what are in this topic.
« Last Edit: January 17, 2015, 01:20:04 AM 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 rf-loop

  • Super Contributor
  • ***
  • Posts: 2363
  • Country: fi
  • Starting with DLL21
Re: Rigol DS1074Z signal level fault - bug.
« Reply #7 on: April 11, 2015, 04:54:45 PM »
I will rise this up.

Scope is full of many kinds of dingle dongle features but Basic Fundamentals are "out of order".

I hope Rigol note this design error and take this seriously. This need repair.

I have not seen this kind of error in any basic level oscilloscope exept in Rigol. 

This (some amount different) joke/crap was also implemented in old DS1000E series.

Rigol do not even listen there is problem? Do they think perhaps that kids need toys and thousend dingle dongle features for amazing and play with.

Professonals need oscilloscope where basic fundamantals are working right so that user can trust what scope show.

I know many (specially noobs) peoples want make-up images from signals so they do not need wonder what is this cratch and what is this pixel up and down in my ideal signal what come from my 5$ crap "wave generator". 

Please Rigol.

Be honest and show what scope is sampling.  Show real dots (and only sampled real dots if user have selected dipsplay mode "dots") then show straight lines between dots if user have selected "lines" (without cosmetics image make-up) and finally also real Sin(x)/x function if user have selected it. And do not if user have selected it off.   

(If  Rigol EE even know what it means. If you Rigol EE can read (I suspect it difficult and it can suspect afteer we know how you have read components data sheets, example PLL. (look PLL kidding)).

It is very very simple paper available if school book did not tell it, read example Chris Rehorn's (Hewlett-Packard  -- Agilent (--Keyshit today) simple explanation what  is Sin(x)/x (Sinc) and think also what it is not. Or ask Siglent how to do it.

Please, Rigol, give some explanation what this DS1000Z is doing when it do some "magic" functions between ADC and Display.
I need know signal in DUT - not your image cosmetics.

Rising up this question and hope that some day Rigol do something for this design error.  This is some kind of wrong application of sinc function.  Do not name it Sin(x)/x. Use "smooth" or someth9ing like it.  And then repair this really unacceptable level error. 
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: 2363
  • Country: fi
  • Starting with DLL21
Re: Rigol DS1074Z signal level fault - bug and and "fake" 1 and 2mV/div
« Reply #8 on: August 10, 2015, 08:40:40 PM »
I was using this scope for long time collecting some data and I hit problem where data is not as expected. Normally if and when there is (random)noise, distribution is quite nice and without "dead bands" what I hit in collected data...
After then I look some amount more deep (and throw to garbage whole long time collected data because data was garbage due to this scope "feature")

What I find or what I think I find.

5mV/div front end wiorks normally and whole 8bit ADC is in use.
2mV/div looks like it is only vertical zoom somehow.
1mV/div looks also it is only more vertical zoom somehow.

Images tell more than words. first 6 images (triangle and rectangle): Signal level changed so that vertical height of signal is around same for 1, 2 and 5mV/div
(in all cases there is 20dB attenuator + 50ohm feed thru  in scope input. (signal itself noise level is quite small)
Last 3 images (rectangle with very slow fall edge) all have same signal level and only scope changed 1mV- 2mV and 5mV/div


Of course settings have selected so that this situation can detect more easy.

So it looks like 2mV and 1mV/div do not give any more true sensitivity  than 5mV/div.

« Last Edit: August 10, 2015, 08:47:47 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 sync

  • Frequent Contributor
  • **
  • Posts: 799
  • Country: de
Re: Rigol DS1074Z signal level fault - bug and and "fake" 1 and 2mV/div
« Reply #9 on: August 10, 2015, 09:35:24 PM »
5mV/div front end wiorks normally and whole 8bit ADC is in use.
2mV/div looks like it is only vertical zoom somehow.
1mV/div looks also it is only more vertical zoom somehow.

Images tell more than words. first 6 images (triangle and rectangle): Signal level changed so that vertical height of signal is around same for 1, 2 and 5mV/div
(in all cases there is 20dB attenuator + 50ohm feed thru  in scope input. (signal itself noise level is quite small)
Last 3 images (rectangle with very slow fall edge) all have same signal level and only scope changed 1mV- 2mV and 5mV/div


Of course settings have selected so that this situation can detect more easy.

So it looks like 2mV and 1mV/div do not give any more true sensitivity  than 5mV/div.
It's more complicated. See http://www.eevblog.com/forum/testgear/new-rigol-ds1000z-and-mso4000/180/
 

Offline rf-loop

  • Super Contributor
  • ***
  • Posts: 2363
  • Country: fi
  • Starting with DLL21
Re: Rigol DS1074Z signal level fault - bug and and "fake" 1 and 2mV/div
« Reply #10 on: August 10, 2015, 10:10:30 PM »
5mV/div front end wiorks normally and whole 8bit ADC is in use.
2mV/div looks like it is only vertical zoom somehow.
1mV/div looks also it is only more vertical zoom somehow.

Images tell more than words. first 6 images (triangle and rectangle): Signal level changed so that vertical height of signal is around same for 1, 2 and 5mV/div
(in all cases there is 20dB attenuator + 50ohm feed thru  in scope input. (signal itself noise level is quite small)
Last 3 images (rectangle with very slow fall edge) all have same signal level and only scope changed 1mV- 2mV and 5mV/div


Of course settings have selected so that this situation can detect more easy.

So it looks like 2mV and 1mV/div do not give any more true sensitivity  than 5mV/div.
It's more complicated. See http://www.eevblog.com/forum/testgear/new-rigol-ds1000z-and-mso4000/180/

All is more complicated. ;)

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 smgvbest

  • Supporter
  • ****
  • Posts: 344
  • Country: us
    • Kilbourne Astronomics
Re: Rigol DS1074Z signal level fault - bug and and "fake" 1 and 2mV/div
« Reply #11 on: August 11, 2015, 12:18:38 AM »
Have you reported this directly to rigol?   talking to Rigol here I doubt will do any good.  if you've not reported it I would do so also it is help full if more that one person reports an problem.

Rigol like any company will note a single report,   they will start to take action when there are many reports.

Sandra
(Yes, I am a Woman :p )
 
The following users thanked this post: videobruce


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf

 

http://opalkelly.com/