REVIEW - Rigol DS2072 - First Impressions of the DS2000 series from Rigol

[marmad]

Are you sure?
I can also clearly see on the screen, that the signal is scaled. The steps are clearly visible and larger as compared to e.g. 1mV/DIV.

I'm not sure what that proves. Depending on the signal, the steps can be clearly visible - and look similar to what you posted - at any vertical setting. Rigol made a big deal advertising the 'true' 500uV setting of this DSO - and while there's many things I dislike about the company - I've never experienced them lying about their specifications.

As I mentioned, you're correct when you say the LSB is 0 when you read DISPLAY memory when set to the 500uV/div setting - but when you read SAMPLE memory, the LSB is 0 or 1 - I just tested it.

[Teneyes]

Another thing I noticed - maybe already known and not a bug.
In 500uV/DIV mode, the scope is internally set to 1mV/DIV and the data scaled by a factor of 2. This can be noticed when reading out the waveform data. The LSB is always 0. This effectively reduces the ADC resolution by 1 bit / dynamic range.

@Xyphro
Did you test this on your DS2000?
What version of Firmware was this bug?
as a bug dealing with 500uV scale was fix in the latest firmware.

But with  full scale being -2.5 mV to +2.5mV
and a noise floor of  about 400uv un-filtered and Not Averaged  I think 7 bit resolution is good

Below is a Sweep of 4-40MHZ at 5mVpp on the 500uV/div scale , the traces are noisy but Good

[Xyphro]

The signal was an analog one and the slope is slow at some positions. I doubt it is the signal source being responsible for this discrete steps in this case.

So it looks they have a bug showing the trace correctly in 500uv/div setting in case the sample memory shows the correct lsb.

If the spec says true 500uv/div support it is actually wrong due to this bug, because the main purpose should be in most cases to look at the trace on the screen.

[Xyphro]

It was the very latest firmware version which appeared just some days ago. Currently I'm not close to my scope, so I cannot check it now.

[Xyphro]

I agree, it is not a serious issue :-)
Just a funny fact.

[marmad]

If the spec says true 500uv/div support it is actually wrong due to this bug, because the main purpose should be in most cases to look at the trace on the screen.

Sorry, this is NOT a bug. It just has to do with sample -> display scaling. As I mentioned before, the data in display memory is heavily manipulated - via decimation, interpolation (if on) and display scaling (the display is 400 pixels high - so ALL sample data is scaled for the display height). If you set the vertical scale to 550uV/div, you again have LSBs with 0 and 1 in display memory.

I repeat: THIS IS NOT A BUG.

[Teneyes]

I agree, it is not a serious issue :-)
Just a funny fact.

Just a specific setting on the Scope


Check out this display in DOTS , Lots of resolution here
A Capture of the change of frequency from 2.2Mhz up to 52.2 MHz (manually )
with persistance set to 10 sec ,then freeze with a "Stop"

[zibadun]

How to do it? It's not rocket science - over-sample, then randomly decimate for the 'required' sample speed = stochastic sampling. Done.

If it's so simple why don't you implement it in RUU marmad?

[Chalky]

I found a new bug while programming this here: http://www.xyphro.de/blog/index.php?entry=entry130705-223439

The DS2072 does not accept an INITIATE_CLEAR class command over the USB interface. This command is a required command according to the USBTMC specification. The Scope does not react at all to it, and the USB transfer times out.

Yeah also doesn't implement some of the LXI 1.3 stuff over TCP/IP, such as 10.2 rule get XML ID document via '<hostname>/lxi/identification'

[marmad]

If it's so simple why don't you implement it in RUU marmad?

There's no way I could get sample data fast enough from the DSO to do it in real time. I could do it slowly - but what good would that be?

Anti-alias is ideally something you want to have while probing unknown signals at slower horizontal speeds - but once you figure out what you're looking at and adjust your settings accordingly - it becomes a bit irrelevant. But it's something that needs to take place between the sampling and the display of the data - and so even though the basic implementation (in terms of math) is not complex - it DOES have it be done quickly.

[zibadun]

If it's so simple why don't you implement it in RUU marmad?

There's no way I could get sample data fast enough from the DSO to do it in real time. I could do it slowly - but what good would that be?

Anti-alias is ideally something you want to have while probing unknown signals at slower horizontal speeds - but once you figure out what you're looking at and adjust your settings accordingly - it becomes a bit irrelevant. But it's something that needs to take place between the sampling and the display of the data - and so even thought the basic implementation (in terms of math) is not complex - it DOES have it be done quickly.

If RUU is fast enough to render 3d it should be fast enough to draw "anti-alias". You can do it on a stored waveform, as a proof of concept.   A picture worth a thousand words in this case

[darrylp]

Have noticed from videos of people ds2xx2 that all earlier shipments come with switchable x1 x10 probes. Some people mention 350MHz bandwidth. Well my scope came with x10 non selectable , rigol info sheet says 300MHz bandwidth. Now I've got other probes for accessing the 500uV setting, but who else has the x10 only probes ?

The RP3300 x1/x10 switchable probes have been the standard included probes for the DS2000 series since they first started selling them. Where did you buy your DSO?

This is from the DS2000 User Manual:

I'm away from my scope at the moment, but the sheet tucked into each probe matched the actual probe. Ie no mention of a x1 position or switch. I'm 99% sure they are listed as RP3300.
Will post the pic tonight / tomorrow. Its a new scope bought from rigol-uk. Otherwise known as telonic instruments.

--
 Darryl

[jc101]

I'm away from my scope at the moment, but the sheet tucked into each probe matched the actual probe. Ie no mention of a x1 position or switch. I'm 99% sure they are listed as RP3300.
Will post the pic tonight / tomorrow. Its a new scope bought from rigol-uk. Otherwise known as telonic instruments.

--
 Darryl

My DS2072 also came from Telonic, and did come with the RP3300 probes which are 1/10 switchable.  Although, on one of the two supplied the switch would not budge at all, it was stuck on x1.  After an e-mail to telonic they simply swapped it.  I got mine early April after an extended delay from order to delivery of 3 working days 

[Xyphro]

The Firmware version is 00.01.01.00.02.

I made some tests.

I'm very confident, that the 500uV mode is a scaled mode and the gain in the analog path is the same as for the 1mV/DIV setting.

What I did:
I injected a 10 MHz signal from a signal generator with -50dBm (@50Ohm Termination) , set the scope to 200ns/DIV, max. sampledepth.

Then I made 2 measurements:
1.) 500uV/DIV => Single shot => saved CSV Data as 'Newfile1__500uv.csv'
2.) 100uV/DIV => Single shot => saved CSV Data as 'Newfile1__1000uv.csv'

Then I used this sample Matlab script to calculate a Histogram and check how many discrete values were actually used:

Code: [Select]

sig500uv  = dlmread('Newfile1__500uv.csv', ',', 2, 0);
sig1000uv = dlmread('Newfile1__1000uv.csv', ',', 2, 0);

sig500uv  = sig500uv(:,2);
sig1000uv = sig1000uv(:,2);

hist500uv = hist(sig500uv, 10000);
hist1000uv = hist(sig1000uv, 10000);

c500uv = find(hist500uv > 0);
c1000uv = find(hist1000uv > 0);

c500uv = length(c500uv)
c1000uv = length(c1000uv)

c500uv was in my case 72 and
c1000uv was 73

With a real *2 Gain in the analog path you c500uV would be twice as high as c1000uV.

[marmad]

If RUU is fast enough to render 3d it should be fast enough to draw "anti-alias". You can do it on a stored waveform, as a proof of concept.   A picture worth a thousand words in this case

The math data on how to do it is readily available online - and used (at the very least) in the Agilent X-Series. Exactly what am I trying to prove and to whom? 

[Xyphro]

Teneyes: i tried the same, but get to see discrete steps (blank lines) on my scope.
Which fw version do you use? Mine is listed in the previous posting.
Maybe this is a new "feature" of my version?

[marmad]

I'm very confident, that the 500uV mode is a scaled mode and the gain in the analog path is the same as for the 1mV/DIV setting.

So send an email to Rigol and get their response.

[Wim13]

Teneyes: i tried the same, but get to see discrete steps (blank lines) on my scope.
Which fw version do you use? Mine is listed in the previous posting.
Maybe this is a new "feature" of my version?

I have also a kind of interlaced screen only on 500 uV see pictures below, 500 uV and 1 mV
Both had same input on both pictures. on 1 mV the screen is not interlaced.

i dont get this picture of Tenyes on 500 uV. stays interlaced on dots.

[zibadun]

If RUU is fast enough to render 3d it should be fast enough to draw "anti-alias". You can do it on a stored waveform, as a proof of concept.   A picture worth a thousand words in this case

The math data on how to do it is readily available online - and used (at the very least) in the Agilent X-Series. Exactly what am I trying to prove and to whom? 

Agilent does this by first estimating the highest frequency component in the measured signal and then selecting a sampling rate which is fast enough to prevent aliasing.  The captured samples are displayed using a technique you describe to avoid moire patterns on the screen.  Rigol does the same thing more or less.  I uploaded an example of this which you simply dismissed  

If you override scope settings and force 200 ksps on a 1 Mhz signal no algorithm will be able display noise instead of the aliased signal.  In fact sampling 1 MHz @200 ksps would give you a 0 Hz aliased signal, i.e. a DC offset      

I think you are confusing two different issues, interference patterns on the screen and RF aliasing. The latter cannot be fixed by algorithms.   If you believe otherwise write a function that shows you are correct. 

[marmad]

Agilent does this by first estimating the highest frequency component in the measured signal and then selecting a sampling rate which is fast enough to prevent aliasing.  The captured samples are displayed using a technique you describe to avoid moire patterns on the screen.  Rigol does the same thing more or less.  I uploaded an example of this which you simply dismissed

If you override scope settings and force 200 ksps on a 1 Mhz signal no algorithm will be able display noise instead of the aliased signal.  In fact sampling 1 MHz @200 ksps would give you a 0 Hz aliased signal, i.e. a DC offset      

I think you are confusing two different issues, interference patterns on the screen and RF aliasing. The latter cannot be fixed by algorithms.   If you believe otherwise write a function that shows you are correct.

Sorry, man, but I think you're the one who is confused.    Both about what Agilent is doing - and about what I've been talking about. And NO, the Rigol is NOT doing the same thing as the Agilent - the problem is that the Rigol has user-selectable sample lengths - the Agilent does not. And it seems Rigol didn't figure out an elegant way to deal with this (e.g. forcing the sample length to a specific size).

The Agilent X-Series doesn't 'select a sampling rate' at all - it always samples at the same speed: e.g. 2GSa/s (in the case of the 2000X) - and just varies how many samples it decimates based on the horizontal scale. I NEVER said that the Rigol could anti-alias a 1MHz signal by just sampling at 200kSa/s - I only used the image as an example of the Rigol not performing anti-aliasing. But you can show the exact same failure of the Rigol using any signal (like the attached image of an aliasing 100kHz sine @ 100kSa/s with ANTI-ALIASING turned on).

What should happen when you turn on anti-aliasing? The Rigol should be over-sampling (e.g. 2GSa/s) and randomly decimate to simulate 100kSa/s (or whatever your fixed rate should be) - just like the Agilent. Again - I don't need to write any function to see how it would work because I understand the math - if you don't, sorry, but it's not my problem. 


If you don't think it's possible, you should be getting an Agilent owner to prove me wrong. For example:
@Hydrawerk - can you please demonstrate? For example, send a 100kHz sine into the DSO - and adjust the horizontal scale until the sampling rate shows 100kSa/s - then grab a screen shot and post it here?

[Xyphro]

I'm very confident, that the 500uV mode is a scaled mode and the gain in the analog path is the same as for the 1mV/DIV setting.

So send an email to Rigol and get their response.

I do, but do you really expect any response from them?

[marmad]

I do, but do you really expect any response from them?

Sure, why not? I've had them fix bugs I've reported to them - so I would imagine they could clarify this issue for you.

Edit: Although, I'd probably ask my dealer to ask Rigol - rather than emailing them directly: the dealers can often get questions answered more quickly.

[Xyphro]

Here the histograms of a -50dBM / 50Ohm sinusodial signal with 500uV and 1000uV setting just for reference => nearly same amount of discrete values, while it should be twice as much at 500uV, if it would be "TRUE 500uV" with scaling applied in analog domain.

FW Version 00.01.01.00.02.

WIM13 / Teneyes, can you post your FW versions?

[darrylp]

Have noticed from videos of people ds2xx2 that all earlier shipments come with switchable x1 x10 probes. Some people mention 350MHz bandwidth. Well my scope came with x10 non selectable , rigol info sheet says 300MHz bandwidth. Now I've got other probes for accessing the 500uV setting, but who else has the x10 only probes ?

The RP3300 x1/x10 switchable probes have been the standard included probes for the DS2000 series since they first started selling them. Where did you buy your DSO?

This is from the DS2000 User Manual:

I'm away from my scope at the moment, but the sheet tucked into each probe matched the actual probe. Ie no mention of a x1 position or switch. I'm 99% sure they are listed as RP3300.
Will post the pic tonight / tomorrow. Its a new scope bought from rigol-uk. Otherwise known as telonic instruments.

--
 Darryl

Well they are listed as RP3300A.


--
 Darryl

[darrylp]

So the manual is dated mar 2012 and says RP3300A. The spec sheet says fixed x10.

So hmm, not much can do. The rigol packing sheet says rp3300a x2 .
I wonder if rigol will stop advertising the 500uV unless they put an asterisk, and say with optional probe. !


--
 Darryl