Rigol DS1000Z series (ds1054z, ds1074z, ds1104z and -s models) Bugs/Wish List

[MrWolf]

This thread originally started as a pretty concise collection of bug reports and improvement suggestions for the DS1000Z scope family.

Good point. So heres a bug:

Signal:
32768Hz square wave, 1Vpp, <10ns rise, low jitter.
Timebase: 5us
Acquire mode: Average
Averages: 1024
Mem depth: Auto
Enable "Rise" auto-measurement with stats.
Reset stats, take note of Min, Max, Avg.
Enable zoom.
Zoom to 5ns timebase.

Due to specific inner workings of device Max value stays on "unzoomed" 200ns, Min gets precise "zoomed" value of 8.9 and Avg gets spot in the middle...

[Howardlong]

Did you try resetting your stats after zooming?

This is an artefact of the previous problem you described, although I agree it would make sense for the stats to auto-reset after any change in measurement sampling rate. Whether averaging is on or not makes no difference.

Keysight x3000 series won't even collect stats if a measurement is undersampled or out of range. The Tek MDO3000 still collects stats but displays "Low Resolution". It resets stats on a change of sample rate.

Interestingly, on the vertical, the Rigol does reset stats automatically when it clips.

[Fungus]

This thread originally started as a pretty concise collection of bug reports and improvement suggestions for the DS1000Z scope family.

Good point. So heres a bug:

Due to specific inner workings of device Max value stays on "unzoomed" 200ns, Min gets precise "zoomed" value of 8.9 and Avg gets spot in the middle...

'Minimum' and 'maximum' are supposed to remember the minimum and maximum values they've seen. It would be a bug if they reset when you zoom in, not the other way around.

This is an artefact of the previous problem you described, although I agree it would make sense for the stats to auto-reset after any change in measurement sampling rate.

I think the opposite - it makes more sense not to reset them.

Resetting them potentially destroys information. Pressing 'reset stats' to manually restart them is the way to go.

[Howardlong]

This thread originally started as a pretty concise collection of bug reports and improvement suggestions for the DS1000Z scope family.

Good point. So heres a bug:

Due to specific inner workings of device Max value stays on "unzoomed" 200ns, Min gets precise "zoomed" value of 8.9 and Avg gets spot in the middle...

'Minimum' and 'maximum' are supposed to remember the minimum and maximum values they've seen. It would be a bug if they reset when you zoom in, not the other way around.

This is an artefact of the previous problem you described, although I agree it would make sense for the stats to auto-reset after any change in measurement sampling rate.

I think the opposite - it makes more sense not to reset them.

Resetting them potentially destroys information. Pressing 'reset stats' to manually restart them is the way to go.

The problem to my mind is that it's contaminating two differently extracted sets of data, so it's throwing apples and oranges into the same basket, making the result even more nonsensical.

Irrespective, whether it's a bug or a feature, it's never been something that has been of any particularly conscious concern for me. If I see, say, a nice round number like 200.0ns in a measurement, I am naturally and automatically going to be suspicious, and know it's time to question that number and make sure I understand why.

Once you've experienced, learned and understood how the instrument works and its limitations, something like this is not really a problem in practical terms, or even something to get irritated about. Whether it's a $400 or a $400,000 machine, you always need to be able to know when something doesn't look right, if you don't question your results and measurements then you're almost certainly not doing it right!

I'd say that at some point, say beyond the $50,000 price point, you are _more_ likely to find bugs because the gene pool of users and experience is so much smaller, and those units will not have been characterised to anywhere near the same extent as a $1,000 instrument.

[Fungus]

If I see, say, a nice round number like 200.0ns in a measurement, I am naturally and automatically going to be suspicious, and know it's time to question that number and make sure I understand why.

Yep.

Once you've experienced, learned and understood how the instrument works and its limitations, something like this is not really a problem in practical terms, or even something to get irritated about.

Hence my earlier comment (which was mocked by some people), "If it's going to be wrong, I prefer it to be very wrong."

50% wrong is far more dangerous than 5000% wrong.

Whether it's a $400 or a $400,000 machine, you always need to be able to know when something doesn't look right

With experience you learn to look at what's on screen as well as what the number says.

You should only think "OK" when they both match (and the number makes sense).

Maybe this is a benefit of learning with analog 'scopes. With analog 'scopes you learn to look at the screen and count the squares.

[MrWolf]

Irrespective, whether it's a bug or a feature, it's never been something that has been of any particularly conscious concern for me. If I see, say, a nice round number like 200.0ns in a measurement, I am naturally and automatically going to be suspicious, and know it's time to question that number and make sure I understand why.

Well depending on your background. Z box price is clearly in "complete noob" class. When I bought my first (very cheap and "nooby"!) DSO I had interest only in project on my desk and could not care less about inner workings of DSO. I had multimeters costing more than that DSO, but 25MHz was plenty enough for the task so why pay more? It took maybe a year before I paid any attention to "sampling rate" concept and many other "DSO specific" things. My project was highly experimental and it would have failed totally if my DSO wouldnt be programmed in extremely honest way and displayed "no reading" on wrong settings for the task etc.
Bottom line: In its current state DS1000Z requires very skilled operator to deliver. It can be considered more a "poor pro DSO" than "noob DSO". I thinks thats a problem for young players. What if someone is also doing something totally insane but very interesting and will think concept failed, while actually DSO played some little tricks on him like some gypsy on the flea market?
Noobs need DSOs programmed in extremely dry and "metrological" manner. I suspect Digilents Analog Discovery 2 might be fine example. I cannot see why theres such acute reaction by some against getting DS1000Z into same "dry and metrological" condition. Cannot see any limitation hardware-wise 

[David Hess]

Offset controls inject a signal earlier in the signal chain before some or all of the attenuation and gain stages.

You can see the Rigol doing this when you run the calibration. The traces move up and down the screen looking for 'zero'.

Dave's reversed engineered schematics of the Rigol DS1000Z input circuits do not extend to the integrated switched gain amplifier and digitizer.  What he shows as a vertical position control may actually be for balancing the offset into the switched gain amplifier.

It is possible that they used that circuit also for the position control but we cannot know without seeing more of the schematic.  That would mean that part of the calibration is to allow an offset adjustment to operate as a position control because gain changes would change the position control gain and hence range.

[Fungus]

Bottom line: In its current state DS1000Z requires very skilled operator to deliver.

Complete and utter rubbish.

All it takes somebody who does the perfectly natural thing of setting the timebase to actually look at the rising edge they're measuring. The 'scope has been on the market for two years and nobody else has even noticed this. That's how anti-intuitive you're being.


b) You're judging the whole 'scope based on one particular thing?? 

c) You're still pointing your finger squarely at the DS1054Z even though we've seen that most other 'scopes do it, too. 

[David Hess]

b) You're judging the whole 'scope based on one particular thing?? 

I have a whole list of complaints.  The thing I like about finding problems with the Rigol DS1000Z series is that it trains me for what to look for in modern DSO.  I used to just check for peak detection capability and transient response.  Now I include RMS measurements of noise and weird display record measurement problems.

c) You're still pointing your finger squarely at the DS1054Z even though we've seen that most other 'scopes do it, too. 

I do not care if other DSOs do it or not; if they do, then they are broken as well.  Them being broken does not fix the Rigol and my criteria for working properly does not include everybody else's product is broken also.

If I had to pick two things wrong with the Rigol right now, they would be the broken RMS measurement capability which would require making measurements on the acquisition record to fix and the transient response.  The later seems to indicate that the oscilloscope does not support a 100 MHz bandwidth do to slew rate limiting or something; maybe a stage is saturating or going into cutoff.  This issue might not be present in the 50 and 70 MHz models because of their more limited bandwidth.

[metrologist]

If I had to pick two things wrong with the Rigol right now, they would be the broken RMS measurement capability which would require making measurements on the acquisition record to fix and the transient response.  The later seems to indicate that the oscilloscope does not support a 100 MHz bandwidth do to slew rate limiting or something; maybe a stage is saturating or going into cutoff.  This issue might not be present in the 50 and 70 MHz models because of their more limited bandwidth.

I thought the RMS measurement issue was fixed in last (pulled) FW. Also, transient response, I thought you made the comment about an MSO1074Z measurement we saw on youtube. I did not see a problem when trying to test mine.

[nctnico]

Irrespective, whether it's a bug or a feature, it's never been something that has been of any particularly conscious concern for me. If I see, say, a nice round number like 200.0ns in a measurement, I am naturally and automatically going to be suspicious, and know it's time to question that number and make sure I understand why.

Well depending on your background. Z box price is clearly in "complete noob" class. When I bought my first (very cheap and "nooby"!) DSO I had interest only in project on my desk and could not care less about inner workings of DSO. I had multimeters costing more than that DSO, but 25MHz was plenty enough for the task so why pay more? It took maybe a year before I paid any attention to "sampling rate" concept and many other "DSO specific" things. My project was highly experimental and it would have failed totally if my DSO wouldnt be programmed in extremely honest way and displayed "no reading" on wrong settings for the task etc.
Bottom line: In its current state DS1000Z requires very skilled operator to deliver. It can be considered more a "poor pro DSO" than "noob DSO". I thinks thats a problem for young players. What if someone is also doing something totally insane but very interesting and will think concept failed, while actually DSO played some little tricks on him like some gypsy on the flea market?

You are totally over reacting here! My $25k Agilent DSO7104A also measures using screen data and if you dig even deeper into an oscilloscope you'll find it can show the weirdest signals. For example: when I evaluated my GW Instek GDS2204E I could make it to show very weird signals using some setting and a certain input signal. When I told GW Instek how I produced the result they send me a screendump of their Tektronix scope (IIRC 3000 or 4000 series so not low end) which showed a similar effect. I then tested on my DSO7104A and I was able to reproduce the effect on that scope as well. My test was stretching the limits too far.

The bottom line is: You have to know what the outcome is before you make the measurement. Otherwise you have no idea whether the result is correct or not. You can't go around sticking probes into a circuit and take what is being displayed at face value.

[Fungus]

I do not care if other DSOs do it or not; if they do, then they are broken as well.  Them being broken does not fix the Rigol and my criteria for working properly does not include everybody else's product is broken also.

That's like saying that a Ford Fiesta is broken if it can't do 200mph and the fact that none of its competitors can do 200mph either isn't an excuse.

If I had to pick two things wrong with the Rigol right now,

Out of curiosity, how many things can you pick that are right with it?

[David Hess]

I thought the RMS measurement issue was fixed in last (pulled) FW. Also, transient response, I thought you made the comment about an MSO1074Z measurement we saw on youtube. I did not see a problem when trying to test mine.

The RMS thing is a different problem we ran across in a recent discussion where we tried to test the analog versus DSO noise difference.  I think what is going on is that the processing to produce the display record alters the standard deviation of the signal, which is not surprising, but it was a confirmation to me that measurements are made on the display record which I did not want to believe before.  If you use RMS to measure noise on this DSO, it returns the wrong value; it is not even close.

For the transient response, I said the test was ambiguous in the Youtube video because we do not know the state of the feedthrough termination which was not used on any other oscilloscope.  Later I was doing some transient response measurements and it occurred to me that transient response anomalies are limited by the vertical bandwidth; you cannot create nonlinearities which would add high frequency content with transient response misadjustments.  Ok, that makes sense.  But the results from the video do not seem to show that.  If the feedthrough attenuator was bad, it should not produce that weird slant in the signal edge.  What that waveform *does* look like is nonlinear distortion caused by an amplifier driven into cutoff or saturation or distortion during overload recovery.

Now Dave identified the transistors in the first differential stage of the DS1000Z as MMBT3904s and Rigol had to use emitter networks to boost the high frequency gain.  That makes sense in comparison to much older 100 MHz oscilloscope vertical amplifier designs which had the 2N3904 available but did not use them.  They either used faster transistors or they graded the 2N3904s for low base transit time (how is this done?) which suggests to me that those transistors are just not fast enough.  That weird cross coupled differential cascode looks suspicious to me but maybe the differential tail current is low.  Why is the differential tail current even adjustable?

So why didn't this show up in your test?  Maybe Rigol grades these oscilloscopes like 2N3904s were graded in vertical amplifiers; if this is the case, the behavior would not show up in stock 100 MHz units.  Or maybe your test signal was not fast enough to produce the behavior.

So the transient response thing might not be a flaw, but it sure looks like one based on the performance and the design unless it only shows up in Rigol's hacked for higher bandwidth.

I do not care if other DSOs do it or not; if they do, then they are broken as well.  Them being broken does not fix the Rigol and my criteria for working properly does not include everybody else's product is broken also.

That's like saying that a Ford Fiesta is broken if it can't do 200mph and the fact that none of its competitors can do 200mph either isn't an excuse.

I do not *expect* automobiles to do 200mph unless I specifically bought them for that application.  200mph automobiles are rare.

I also do not expect the Rigol or any analog oscilloscope or DSO to have the fast overload recovery of a digital storage sampling oscilloscope or any sampling oscilloscope so that fact that its overload recovery is horrible is irrelevant.  Analog and digital sampling oscilloscopes are rare.

If I had to pick two things wrong with the Rigol right now,

Out of curiosity, how many things can you pick that are right with it?

I can simplify the answer to it is better than an even worse DSO.  I do not see it as a general purpose DSO because of its limitations.  I came to the same conclusion about Rigol's earlier DSOs.

I have got a question though.  Does making measurements on the display record mean that the Rigol actually has a record length of (checks manual, 600x800 display) 12 horizontal divisions * 50 points/division (estimated) = 600?  Is the long record length only available when acquisitions are stopped?  Wouldn't that make it like the early Tektronix DPOs?

[Karel]

Does making measurements on the display record mean that the Rigol actually has a record length of (checks manual, 600x800 display) 12 horizontal divisions * 50 points/division (estimated) = 600?  Is the long record length only available when acquisitions are stopped? 

The display memory contains 12 div. x 100 = 1200 pts. These 1200 pts can be aquired via usb or lan at any moment without
stopping the acquisition.
Stopping the acquisition is only necessary for downloading the deep memory waveform data e.g 24Mpts.

[Fungus]

I have got a question though.  Does making measurements on the display record mean that the Rigol actually has a record length of (checks manual, 600x800 display) 12 horizontal divisions * 50 points/division (estimated) = 600?  Is the long record length only available when acquisitions are stopped?  Wouldn't that make it like the early Tektronix DPOs?

The programming guide uses numbers in the range 0 to 1199 for setting start/end points. This suggests it works with 2 samples per pixel (600x2 = 1200 points).

[MrWolf]

The bottom line is: You have to know what the outcome is before you make the measurement.

Would that make you... god? Or policmean that spotted red 3-series BMW with big wing that just must to be speeding, and if not now then in future for sure, so better to ticket now 

c) You're still pointing your finger squarely at the DS1054Z even though we've seen that most other 'scopes do it, too. 

Yes but they seem to either indicate reduced accuracy or at least use some reasonable sized secondary buffer. Siglent SDS1102X+ seems to have 10kpts horizontal resolution for end-user. If to be compared to Rigol, around 40kpts buffer then instead of 1.2kpts? Edit: In fact buffer is even bigger according to rf-loop here  :
https://www.eevblog.com/forum/testgear/siglent-sds1000x-series-oscilloscopes/msg1092595/#msg1092595

And since you seem to suggest that I have sort of "attitude problem"... Did you failed to notice that I filed bug reports also on Siglent SDG2000X? After short and very constructive discussion it was all settled.
If we would have similar 5+ reps here from Rigol it would be all fine and civilized, but instead we have who... you? And you seem not to be a rep because they know that customer is always upwind when you try to take a piss on him.
In fact I personally always highly regarded customers who give feedback (even negative) on my product/system. Can hire fewer testers then 

The programming guide uses numbers in the range 0 to 1199 for setting start/end points. This suggests it works with 2 samples per pixel (600x2 = 1200 points).

Yes but actual data suggests 300pts end-user resolution. Aliasing effect on further processing on these points?

[nctnico]

The bottom line is: You have to know what the outcome is before you make the measurement.

Would that make you... god?

No, a good engineer! A good police officer usually knows from experience if a car is speeding or not (and a ball park figure on the speed) in order to pick the ones he/she measures with the lidar. In this case the measurement is also a confirmation of something which is already known.

[tautech]

Siglent SDS1102X+ seems to have 10kpts horizontal resolution for end-user. If to be compared to Rigol, around 40kpts buffer then instead of 1.2kpts? Edit: In fact buffer is even bigger according to rf-loop here  :
https://www.eevblog.com/forum/testgear/siglent-sds1000x-series-oscilloscopes/msg1092595/#msg1092595

Just to be sure you're clear that there's 3 different models from 3 series of Siglents in risetime measurement tests ATM.
SDS1102CML+, SDS1202X and SDS2304X, all with significantly different specs.
Just be sure you're quoting the right one. 

[rf-loop]

Just to be sure you're clear that there's 3 different models from 3 series of Siglents in risetime measurement tests ATM.
SDS1102CML+, SDS1202X and SDS2304X, all with significantly different specs.
Just be sure you're quoting the right one. 

We know what is this test but I want here note for some  random readers.
These some tests shown are  not at all for evaluate Siglent scope risetimes.




Generally:

These particular tests are for get bit more knowledge about automatic measurements system technical "features" and this is generally important and interesting thing, independent of scope manufacturer and brand and model. 

And for example this purpose I have used externally controlled rise and fall times what are far slower than oscilloscope limits because I need data for estimate amount of data used for automatic measurements and for also proof that even with deep memory, every sample is mapped to display (I can count sample points when I look edges when I use  slow wfm/s acquisition mode where every displayed TFT refresh have only single shot wfm, so I can proof all samples are mapped to display but not all is used for measurements).  And some of these features can not read from data sheet. Example about automatic measurements resolution with different memory and timebase settings. There is  nothing in data sheets. So, user need evaluate these (if want do also more serious things than just looking strange and fun images about electrical phenomenons.)    Just for "know your equipments" -- for better confidence and better understanding results and for avoid mistakes.
This is why this kind of tests are important and useful.

[David Hess]

I have got a question though.  Does making measurements on the display record mean that the Rigol actually has a record length of (checks manual, 600x800 display) 12 horizontal divisions * 50 points/division (estimated) = 600?  Is the long record length only available when acquisitions are stopped?  Wouldn't that make it like the early Tektronix DPOs?

The programming guide uses numbers in the range 0 to 1199 for setting start/end points. This suggests it works with 2 samples per pixel (600x2 = 1200 points).

Is it returning 16 bit values?

Alternatively it could be setup to return minimum and maximum values when peak detection is used.  On my DSOs, the record length is halved when peak detection is used which is annoying but acceptable since they have such high resolution anyway.

Otherwise it seems odd that the display would be 600 wide while the returned data is 1200 wide.

[Karel]

Is it returning 16 bit values?

No, 8 bit.

[Fungus]

Otherwise it seems odd that the display would be 600 wide while the returned data is 1200 wide.

Why? They might have done it to get a teeny bit more accuracy in the on screen calculations (rise times, etc). It makes perfect sense to do that if you've got a bit of CPU time left over.

[David Hess]

Is it returning 16 bit values?

No, 8 bit.

Otherwise it seems odd that the display would be 600 wide while the returned data is 1200 wide.

Why? They might have done it to get a teeny bit more accuracy in the on screen calculations (rise times, etc). It makes perfect sense to do that if you've got a bit of CPU time left over.

My guess then is that it either has something to do with peak detection which to make full use of the 600 horizontal point display would require a 1200 point record or something do to with rendering the index graded display like antialiasing.

If it has to do with peak detection which I think is the most likely option, then it will be easy to determine; check to see if the number of horizontal display points is halved when peak detection is used.  In order for that to work with a 600 point display without halving the horizontal resolution, the digitizer has to return 1200 point records and it probably always does that.  The design is just simpler that way.

[Karel]

If it has to do with peak detection which I think is the most likely option, then it will be easy to determine; check to see if the number of horizontal display points is halved when peak detection is used.

It always returns 1200 pts, also in peak det. mode.

Offtopic, what I personally find strange is, why it uses a vertical resolution of only 25 lsb's per division.
For comparison, the DS6000 uses 32 lsb's per vertical division and thus uses the full dynamic range of the ADC for the display.

Edit: Found some info here:

https://rigol.desk.com/customer/en/portal/articles/2282294-why-does-my-data-not-appear-to-have-8-bits-of-resolution-

[ebastler]

Offtopic, what I personally find strange is, why it uses a vertical resolution of only 25 lsb's per division.
For comparison, the DS6000 uses 32 lsb's per vertical division and thus uses the full dynamic range of the ADC for the display.

Not further off topic than the last 3 pages of this thread... 

I think the DS1000Z's vertical resolution has been discussed on this forum before. Rigol seems to have mapped 200 ADC counts to 400 pixels vertically. The integer conversion factor probably helps in driving the display with limited FPGA resources.