Square signal level > 20ms not looking good - Siglent SDS800X HD

[tautech]

I would like to know that equipment you are using.

This question explains a whole lot, the man just cannot read or take note of any fine detail. 

I couldn't find what you used as well in your four posts.

Note the attachment filenames. 

[joeqsmith]

I would like to know that equipment you are using.

This question explains a whole lot, the man just cannot read or take note of any fine detail. 

I couldn't find what you used as well in your four posts.

Note the attachment filenames. 

So that's using the probes that came with the scope and internal calibrator? 

[eTobey]

I just came across this issue again, while measuring my BLDC driver:
I looks exact like the deviation in that post: https://www.eevblog.com/forum/beginners/probe-compensated-but-level-of-square-signal-gt-20ms-not-looking-good/msg5524375/#msg5524375

Its the same shape, and it also levels out after 25ms. (compare with the timebase)

picture 73 to 75 are actually the same capture, but the scope decides to mess up the averaging trace after changing the timebase. 

I was again confused about my circuit, but its actually the scope.

Edit:
I predicted it:

The fluctuations are now "hidden", but i guess they will confuse me one day, and lead me to look for problems in my circuit that are not there.

[eTobey]

2 screenshots from a perfectly compensated 10x probe.

Compares a 20ms timebase measurement with a 50us timebase measurement and even a 50ms signal with a 500us...
Does not answer my question about this wrong comparison... 
That is ... (cant find any words for this)


Now i am looking at scopes from the last century, that do better in this regard.

(Yes i know, this topic is old, but that limitation of this scope came up again.  )

[shapirus]

Let me repeat a question that, as far as I recall, you never answered: what if you try the dot mode display? I believe your scope can do it. It will reveal if that's an interpolation artifact or something actually present on the ADC input.

[eTobey]

Let me repeat a question that, as far as I recall, you never answered: what if you try the dot mode display? I believe your scope can do it. It will reveal if that's an interpolation artifact or something actually present on the ADC input.

There is no such question in this topic...

But i probably ignored that "suggestion", because it seemed not plausible at all to me. As you say "display": its just another way of displaying. And there are no repeating waveforms, that there would be any artefacts right?

[eTobey]

Here is my current use case:
I am measuring the current of the lowside mosfet when switching. Using the 10x probe (red trace) is yielding quite some deviations. With 1x probe (green) it looks much better.

The traces are the low side mosfet drain to source. There is a current flowing of about 77A before A. The mosfet:
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&cad=rja&uact=8&ved=2ahUKEwjKrdHLqK-KAxWCgv0HHfnAKD8QFnoECBYQAQ&url=https%3A%2F%2Fwww.vishay.com%2Fdocs%2F76233%2Fsir180dp.pdf&usg=AOvVaw17UiD23QCaDR2MAh1WU18F&opi=89978449

A: The highside mosfet turned off.
B: The lowside mosfet starts conducting
C: The lowside mosfet is fully conducting

Just before B is the forward drop of the diode in the mosfet.

[shapirus]

Let me repeat a question that, as far as I recall, you never answered: what if you try the dot mode display? I believe your scope can do it. It will reveal if that's an interpolation artifact or something actually present on the ADC input.

There is no such question in this topic...

But i probably ignored that "suggestion", because it seemed not plausible at all to me. As you say "display": its just another way of displaying. And there are no repeating waveforms, that there would be any artefacts right?

It's in post #13. Not a question, techically, but come on, you understand what I mean.

Dot display is meant to show raw data points coming from the ADC before any interpolation takes place. This will allow to rule out one possible culprit, namely, the interpolation algorithm. That's how you troubleshoot things, generally: you rule out each factor one by one, thus narrowing down the set of possible reasons causing the issue.

[eTobey]

I have now made many more tests, and i came to the following results:
- The probes seem to have little to no effect on this matter. Except switching 1x to 10X (each input has its own "fingerprint")
- The fluctuations seems to be somewhat proportional to the voltage change (see images)
- 7.5V: Having the signal vertically fill out the screen has less noise, but is generally a bit to high (see yellow trace of "7V square.png"
- 2V/DIV seems to be the best setting for measuring after big voltage drops (if it was a substantial time high).

The signal shown, is right after it went from xV to 0V. It was 70ms high(!). The measurements were averaged. The stock probe (1x) has been used with a croc clips at the wavegen (UTG2025A).
20M BW limit.

measurements.xml is the setup file.

[eTobey]

Another interesting measurement:
Tried 2 different probes with different settings:
- g = green probe (stock) / m = probe with metal BNC
- 10x and 1x attenuation
- probe g changed compensation (comp0 done by standart compensation procedure / comp2 tried to match other probe)

The probes are compensated quite good (see pictures - comp2 of green probe - Y1 shows level after settling). With 1x att, the probes have practically identical proberties here (dark red and red overlay each other).

Its very good to see, where the compensation actually comes into effect!

I wonder, if there is a scope, that could show traces on a logarithmical scale??

[eTobey]

I have now compensated my probe with the 10V square (70ms high) at the point of 1us, which looks much better than my compensation in the conventional way.
Also i have switched to another channel. You can clearly see, that there are fluctuations, that depend on the channel, and that are rather big. Here i also compensated on 1us, but that does not affect the "bad" part on the right.

[eTobey]

After one day and many dozends of measurements...

I have made the final shootout of measurements:
The best channel of the scope, with the best probe i have, vs. my cheap tiny DSO quad. I am quite dissapointed of the SDS800. The Quad has settled after 20us, but the SDS800 only after a time that is 1000 times longer... orders of magnitude!

[ Specified attachment is not available ]

These measurements were done via a probe coax adapter and coax cable directly to the signal gen. The waveform of the DSO Quad was done in 2 steps, because it has limited memory.

[iMo]

So your "settling time" (let us call it that way) is around 20ms? That is too long duration for having something with probe's compensation, sure. Even when considering the resistance of the probe's inner wire (the central wire in the coax) be typically around 200ohm.
That is a rather long settling time or transient coming either from the generator's output circuits (ie. its power source response to the sharp rising/falling edge sent into the probe), or in the scope's input circuitry - like a low frequency feedback somewhere in the analog front end (as people described here in past), or a kind of "AGC response" there in the input circuitry (like attack/release time), if any..

[tautech]

A probe is not the correct tool for precise step response measurements.

[iMo]

TEK on scope probes (from 1969, 9MB pdf)

https://w140.com/tekwiki/images/6/62/062-1146-00.pdf

A sim..

[eTobey]

I believe its coming from the analog front end. Because the part after 200us only changes, if i change the channel on the scope. I wonder if it could be calibrated bit better?

Another question:
If i get a probe, that can be adjusted in LF too, could i compensate for these fluctuations?

https://youtu.be/XLC1CPWznTE?si=uyq5GsawfXIFtcpS&t=591

[iMo]

You would need the front end schematics, or provide a reverse engineering as Dave did ~10y back with 1054Z afaik, or Siglent will do it for you.. If it is in the front end you must see your long "settling time" with direct coax (terminated by 50ohm) fed into the input as well (more less the same effect)..

[tautech]

I believe its coming from the analog front end. Because the part after 200us only changes, if i change the channel on the scope. I wonder if it could be calibrated bit better?

That also is my view.

You have already demonstrated some channels are better than others.
Should I get an adjustment procedure maybe I can share it .....IDK.
It will require my own tests.....

That's all I'll say on this....at this time.

[eTobey]

Should I get an adjustment procedure maybe I can share it .....IDK.

That would be great!

I made some more measurements:
Compared all the channels, and measured the effect of the level of the signal on the screen.

I hope all the labels explain the measurement sufficient.

I found, that i am best of using CH1 2/div for measureing levels at < 200us, and for levels at > 200us CH4 1/div.

One could go further, and subtract the errors, but that would go a bit to far here. 

[tautech]

eTobey
Could you please show screenshots of each channel with vertical Zoom.
Direct BNC connect, no probes !

Source = your SDG set to 1M output for voltage levels to be correct.
Signals:
Square wave with amplitude of 1.2 Vpp and frequency of 10 kHz, and set the vertical scale of the scope channel to 200 mV/div

Change the amplitude of the square wave to 12 Vpp and vertical scale of the scope channel to 2 V/div

Compare zoomed and unzoomed waveforms for clean Step response.

[eTobey]

Yeah nah, 10kHz is far away from my measurements.

What would you expect to see here? I was already about to use direct BNC, but i am already a bit tired with this issue.

I would expect the signal to look much better, but i would not change the signal to make it look better on the scope  . But indeed, my testsignal is not exactly like my use case.

[eTobey]

Made another test run:
With a MCP1407, and gnd spring of course.
7V/div and a 3V - 18V step.

+-80mV of fluctuation on a 56V scale comes out as beeing as good (700 values), as a 10 bit scope , that does not have such fluctuations. But it still gets worse, as the step increases.

[eTobey]

Forgot the picture of the setup to my last post.

I have also now a measurement of my project: i am not sure much i can trust this signal (pink, coming from about 38V), with that accuracy mentioned above.

[eTobey]

After a test with shorting 40V, that the probe was on, i came to the conclusion, that this signal is distorted quite heavily. The actual measurment and test was done with gnd spring and alike.

The arrows show how i suspect the actual signal to be.