DS2072 HW FAIL SW BUG or What the hell? SOLVED

[Carrington]

More images:

[Teneyes]

Please look carefully these images (If anyone wants try to replicate it. The sigal is a pulse of 3.3V, wide 50ns and rise time 2ns):

@Carrington
  You may wish to try using math function to zoom and have cursors operate.
First it is good to adjust scale to full vertical display ( plus a bit more for max resolution.
Next Use Math function to remove DC offset , and scale up the remainder
see pics
Add cursors using Math, and remember to scale down the cursor reading

HiRres may Help also

[Teneyes]

@Carrington
Be Aware of the SinX/X displaying of sample points
here are sample points and the resulting Rigol display
1 sample off a line
2 samples off a Line

[Carrington]

@ Teneyes
Math function to zoom? Does not seem very convenient.
With what equipment generated this samples? -> 1 sample off a line and 2 samples off a Line.
Thanks.


DS2072s  input stage (it is not my oscilloscope):
I don't think that this effect occurs in the rebranded LMH6518. So only remains the first stage. Although I'm not sure, because it is the only one who varies the gain.
Damn, I would love to measure with another oscilloscope at several points of that input stage. But it violate the warranty.   

I tried with another 3.3V pulse (tr1.5ns) generated in a CYCLONE II and exactly the same.



Schematic of a input stage (does not correspond to the DS2072):

[Carrington]

Oscilloscope probe with spring, that's the setup that I used to measure.
Probably marmad is right, but I'm still not totally convinced.
  - Can someone with a DS2072, DS2102 or DS2202 try to replicate my results?

[echen1024]

And what do you want mean, that the internal DC offset in conjunction with the input stage distort the signal.

Yes.

Why this does not happen with a cheap SDS8102V?

As mentioned, it can vary from scope to scope - some analog scopes I've used do it easily - while other's don't.

I don't have the equipment to do.

I'm not sure I understand you - I just mean can you cause the Rigol to give different measurements at those scale settings using any other signal source.

My crap siglent used to rarely overdrive, but now my DS1000Z does. I know that some Tek scopes do as well.

[Carrington]

You have increased the BW in your DS100Z?


Normally, the input divider with compensation (resistors + capacitance) is factory adjusted, but:
And if it is not adjusted properly on my device due to a BW increased? But this explanation is not very convincing to me.

[Carrington]

Let us suppose that the first stage (trts) has unity gain. And that a lower internal DC offset applied to this first stage implies less distortion by saturation.

Now I use the SDS8102V to measure the previous square pulse. Then I choose the AC coupling mode in the DS2072 and I measure the signal in the input stage of SD8102V (yes internally, just at the output off SDS8102V's first input stage). In this way I can shift the signal with the SDS8102V's internal DC offset. Whereupon I get this signal (see attached). To frame it properly I just had to apply a few mV of offset in the DS2072. But despite everything I get the same results. Yes, I know the signal is not entirely within of the reticle. But for all levels no matter how small it always going to saturation? 

[Carrington]

The SDS8102V's input stage is very similar to this, but is based on the BF862 and BFT93 but it work completely with +-5V.

[Teneyes]

@ Teneyes
With what equipment generated this samples? -> 1 sample off a line and 2 samples off a Line.
Thanks.

 A Sonic Screwdriver, or a Neutrino gun on the ADC  

NOT ! just hand patched the waveform File and re-loaded it into the DSO

[Carrington]

A Sonic Screwdriver, or a Neutrino gun on the ADC  

NOT ! just hand patched the waveform File and re-loaded it into the DSO

Ok, it seemed too good to be true...

[marmad]

But despite everything I get the same results. Yes, I know the signal is not entirely within of the reticle. But for all levels no matter how small it always going to saturation? 

I appreciate your frustration, but you're trying to do something that the instrument isn't really designed to do.

And you know, it wasn't that long ago that the only way you could measure waveform voltage levels, etc, was to manually count graticule divisions. In fact that's still how I do it on my old Tektronix analog.

[Carrington]

I appreciate your frustration, but you're trying to do something that the instrument isn't really designed to do.

I understand your position, and I appreciate your help, but what bothers me is that if something is known then it can be avoided.

[Carrington]

Which part cause this problem in the DS2072?
I wonder if changing some part (trt or rebranded LMH6518) by other in the input stage would help to solve the problem.
Although, perhaps, is due to the preamplifiers configuration.

By now we know that this also happens in the DS1000Z.
In the old DS1000 series. Also happens?

[Carrington]

Signal Acquisition Considerations

Some signals are just more difficult to work with than others. For example, looking at the ripple or overshoot on the high-voltage switching waveform of a power supply can be a problem. If you attenuate the signal to fit within the graticule, the small details may be lost, as shown in the 8-bit simulation in the figure. A high-resolution scope with sufficiently low noise may allow you to zoom the captured waveform while retaining enough resolution to make your measurements.

This kind of application is difficult to deal with because several factors can catch the unwary engineer. AC coupling isn’t going to help because the noise/ripple frequencies aren’t much different from the higher frequencies that define the switching edge.

You could try turning up the sensitivity and triggering on one of the waveform edges. That may produce a stable image, but you probably would be overdriving the scope channel preamp, which means that you may be looking at the preamp recovery from saturation rather than the shape of the switching waveform. For this reason, an analog scope also would have trouble with this application.

Your scope specifications may include the recovery characteristics, so this setup can still work if you only measure the waveform after allowing for the preamp’s recovery time. Adding input offset could be useful to the degree that it might allow you to minimize the amount of preamp overdrive.

Sometimes, engineers build special signal-conditioning circuits to apply ahead of a scope. In this case, a fast diode clipping circuit in combination with an amount of offset could preserve the overshoot and ripple but significantly reduce the overall signal amplitude. All these problems are avoided by using a high-resolution scope and keeping the entire switching signal within the preamp and ADC dynamic range.

Source: http://www.evaluationengineering.com/special-reports/201307/instruments/Resolving-Finer-Detail.html

[tinhead]

You overdriving the input stage, on some DSOs this might looks better, on others worse, but overdrive is an overdrive and when you overdrive the measured result might be everything but ok.

Attached results of similar setup on my Tekway, made with 3 probes :
 - original Hantek PP150 probe
 - Testec HF512 (passive 500MHz) probe
 - Texas 250/II (passive 250MHz) probe

Note, the shape of the overdrive depends as well on the probe BW and the probe itself. With proper HF probes (the ones having compensation for LF and HF, like the Texas and Testec) the shape looks better - but of course it still overdrives. It depends as well on the DSO BW, so when you do such measurments don't enable any BW filters.

As you can see Hantek probe is not good enough to handle fast edges, not only due missing HF compensation (only LF) but as well due the probe BW. The Texas probe looks on 500mV/DIV and the 2.5x overdrive good, but with 5x overdrive it does overshoot too much (it can not be compensated out, so it is like it is). The best seems to be the Testec probe, it does have some ringing but the signal shape is constant (of course more overdrive = more ringing)

It is hard to say if the signal shape on your Rigol looks like it looks due wrong probes, wrong probe compensation, not matching probes, broken DSO or the design itself - or combination of one or more. Maybe someone else (but please, not a Rigol fanboy) could try similar setup and report the results. For me especially the 200mV/DIV looks too bad to be pure overdrive.

-----

Rigol probably even know that there is an bit more overdrive than expected, but maybe decided to not do anything about. Or they expected to get more, but then recognized that it is within their expectation. Or maybe they simply copied part of DS1000CA series schematics and even routed the one diode but then recognized that "this is not necessary" - choose one or more

Seriously, hard to say what idea was behind, but when you look exactly there is not populated sot23 between the two opamps in input stage - the diode used here (on Rigol CA) does have T2H marking (actually T2 is marking, H is date code) which is HSMS-286C (Microwave detector diode). So when you really whish to "play" with overdrive, the best place is between the two opamps, i would not touch the first part of the input stage, there is nothing wrong (as it is exact the same on my Tekway and the shape looks better even with the crap Hantek probe). But still, wait first until someone else confirm the signal shape during overdrive. These diodes are not really expensive, but who knows if the firmware has been made to work with them together.

[Carrington]

Honestly, I never had this problem before. So I asked for help in the forum. Thank to all who have contributed with their comments (especially to marmad and tinhead).

And leaving aside the brand, I appreciate your sincerity. Also I must admit, that I am a little distrustful.
I hope you understand. I apologize if I offended someone.

[tinhead]

but please, not a Rigol fanboy

Yes, probably much better to have it done by someone who sells a competing brand of DSO.

i would, generally said, not recommend to trust an distributor or fanboy or any other persons affected by any other kind of "affiliation" (by manufacturer or distibutor or any other revenue source). Sure, there are some trustful people even if i know they work for X or Y, best example is 'w2aew', but there are as well others (good know here in forum) who are only revenue biased.

On the other side this is a simple test which can be done by every owner of X or Y brand, so only an idiot would try to fake results.

There might/will be difference between same probes/DSO models, but they have to be there due compensation errors and parts tolerance. So compensate LF first, then try to change HF compensation (if available) to get best shape results and amplitude accuracy of a test signal first, and then the overdrive result will be the best you can get. The signal source must be something between 500ps and 5ns, square, and 3-4Vpp. That's all, so in principle any 5-20MHz quarz oscillator (with matching series resistor to get nice signal shape) is good enought. And yeah, same probes as used by Carrington would be the best.

[Hydrawerk]

Have you seen this Overcoming Overdrive Recovery on High-Speed Digital Storage Oscilloscopes?
http://cp.literature.agilent.com/litweb/pdf/5989-0068EN.pdf

[tinhead]

Overcoming Overdrive Recovery on High-Speed Digital Storage Oscilloscopes[/b]?

right, this is exactly what i remembered having read somewhere. One have however know that there is a small problem with "digital magnification", you will never get the full DSO resolution in the magnified signal. So it make sense, when possible, to enable high-res mode to at least increase resolution of the part of the waveform which will be magnified.

[Carrington]

Have you seen this Overcoming Overdrive Recovery on High-Speed Digital Storage Oscilloscopes?
http://cp.literature.agilent.com/litweb/pdf/5989-0068EN.pdf

@ Hydrawerk, thanks.

I suppose that this was what Teneyes meant in his previous post.

One have however know that there is a small problem with "digital magnification", you will never get the full DSO resolution in the magnified signal. So it make sense, when possible, to enable high-res mode to at least increase resolution of the part of the waveform which will be magnified.

Would be very interesting make a PC application to implement the technique recommended by Agilent.


I think that the following information was published by tinhead somewhere here.

Code: [Select]

TESTEC TT-HF512 500MHz x10 Probe

  1MHz    0dBm
 50MHz   -0.1dBm
100MHz  -0.1dBm
150MHz  -0.2dBm
200MHz  +0.2dBm
250MHz  +0.2dBm
300MHz  +0.4dBm
350MHz  -0.3dBm
400MHz  -0.7dBm
450MHz  -1.7dBm
500MHz  -2.2dBm
550MHz  -2.8dBm

TEXAS 250/II 250MHz x10 Probe

  1MHz     0dBm
 50MHz  -0.3dBm
100MHz  -0.5dBm
150MHz  -0.1dBm
200MHz  -0.4dBm
250MHz  -0.1dBm
300MHz   0.0dBm
350MHz  +0.3dBm
400MHz   0.0dBm
450MHz  -0.5dBm
500MHz  -1.1dBm
550MHz  -2.4dBm

Hantek P6200 200MHz x1/x10 Probe

  1MHz     0dBm
 25MHz     0dBm
 50MHz  +1.7dBm
 75MHz  +2.9dBm
100MHz  +3.8dBm
125MHz  +4.4dBm
150MHz  +4.5dBm
175MHz  +4.7dBm
200MHz  +4.9dBm

[tinhead]

i would, generally said, not recommend to trust an distributor or fanboy or any other persons affected by any other kind of "affiliation" (by manufacturer or distibutor or any other revenue source).

I'm not sure what "fanboy" is doing in this list - unless you're confusing some German word with the English one you wrote.

The word was coined in English to apply to someone who was a passionate fan of something (normally from geek culture - e.g. sci-fi, comics, Star Wars, video games, anime, etc), often with an uncritical eye - but certainly not someone with an "affiliation" or "revenue source" - and not someone who would be affected by whether brand X or Y sold more units.

So, no, in English, fanboys do not fit into the same category as manufacturers, distributors, and sellers of a particular product - in this case, DSOs. But it is interesting to note the continued increasing usage of the word online as an argumentum ad hominem.

http://en.wiktionary.org/wiki/fanboy

fanboy - Someone who is utterly devoted to a single subject or hobby, often to the point where it is considered an obsession

So yes, the definition of fanboy i know (nerd) is not a problem, but when fanboy starts to be too obsessed, and this is probably one of the reason why usage of this word increasing in online discussions, it start to be a problem. Of course there will be fanboys capable of seeing the reality, but due human nature they will be minority in the fanboy group. Another reason why the usage on fanboy increasing if the fact that today it is enought to say only few times something possitive about single "thing" to get a fanboy sticker, no matter if one was right or not. Due this, i'm using the word "fanboy" only to describe people who simply too obsessed and blind, and "nerd" for all the others (the not blind "fanboys" and all the geeks).

[Orange]

Another pissing contest between marmad and thinhead 

[Orange]

Well thank you for this compliment !

[Carrington]

@ tinhead, marmad, please don't follow this path. As a hippie would say, peace.