Author Topic: Square signal level > 20ms not looking good - Siglent SDS800X HD  (Read 3425 times)

0 Members and 1 Guest are viewing this topic.

Offline shapirus

  • Super Contributor
  • ***
  • Posts: 1575
  • Country: ua
It would be interesting to know (if the probe is the cause), why a signal takes 20ms to become stable. To me it seems it comes from the probe, because touching a battery (with additional cap) looks the same. Also on my testec, i can not get a perfect square, since there always will be a short overshoot at the beginning, or a tiny and longer flat drop after it.
You also got pretty much the same result using a coaxial cable with a 50Ω terminator, didn't you? So the probe can be ruled out.

And since you got the same result with a 12V battery... Do you have any other scope to compare with the one you've been using? :)

Well, strictly speaking, you haven't fully ruled out the probe and the source of the signal (the one that's not the battery). To do that, you need to use the battery with something that is not a 10x probe.
Since the battery is not a proper 50Ω source and simply adding a 50Ω resistor in series with it is not really an option because of the power dissipation ((12V)²/100Ω = 1.44W), you can use a higher value resistor to create what's known as a resistive probe to test that. It's basically a piece of 50Ω coaxial cable terminated with a 50Ω load (a feed-through terminator in your case) that has a resistor in series with the center conductor on the DUT side. A 2400Ω resistor, for example, will create a 49:1 attenuation ratio and such a probe will draw only about 50 mW from the battery, so there will be no overheating issues.

Alternatively, use a smaller resistor on the battery side and use a lower voltage battery.

If even in this case you will be seeing the slow settling time... well, then it's the scope. At least in the scenario when you are overloading the input to zee the zoomed in portion of the waveform. You will need to repeat the same tests that you did in post #8 to make sure it's not a consequence of the input overload.
 

Offline shapirus

  • Super Contributor
  • ***
  • Posts: 1575
  • Country: ua
Speaking of range switching, here's a demonstration.

Same 10x probe, same scope, same signal source (50Ω output impedance), same input.

100 Hz 3V p-p square wave.


1. 1 V/div (actually 100 mV/div). This is the waveform used to adjust the probe compensation.

2253706-0


2. 500 mV/div (actually 50 mV/div). An input relay clicks when the vertical scale is changed into this position. Same probe, same compensation setting!

2253712-1


It's quite clear that the input capacitance is quite different in these two cases (and it sucks!)


3. Coaxial cable, feed-through terminator, 300 mV p-p across the terminator. 100 mV/div, same input path as in case 1. A tiny bit of overshoot/undershoot (as in an overcompensated probe) can be barely recognized.

2253718-2


4. Same, but 50 mV/div, same input path as in case 2. A little bit of what would be undercompensation in case of a 10x probe can be barely seen.

2253724-3


Someone on this forum has a signature about lies, big lies and what you see on the scope screen. This is a perfect illustration. One always has to take this into account and understand that there is always some uncertainty at fast transitions. Better built scopes will likely be better at this.

p.s. The above is reproducible very consistently across all the 4 channels, so it doesn't look like a simple deviation of component values within a tolerance range. It must have something to do with the actual implementation of the analog frontend.

p.p.s. Easy: as the vertical scale is switched and the relay clicks, the actual measured input capacitance changes from ~15.8 pF to ~16.3 pF. My bet is on a PCB parasitics difference coming from different lengths of the traces. PCB designer's failure!

p.p.p.s. Now I've got an itch to add something for an extra 0.5 pF on the lower-capacitance input paths in each channel. I will try to resist it and simply remember to re-adjust the probe compensation for each of the two possible input signal paths whenever it may matter.
« Last Edit: May 28, 2024, 09:27:53 pm by shapirus »
 
The following users thanked this post: thm_w, eTobey

Offline eTobeyTopic starter

  • Frequent Contributor
  • **
  • Posts: 797
  • Country: de
If it helps, see below for a screenshot from my old Infiniium 54825A scope

What a nice signal. Would it look the same, if you put the probe on a battery (maybe with a cap)?

"Sometimes, after talking with a person, you want to pet a dog, wave at a monkey, and take off your hat to an elephant." (Maxim Gorki)
 

Offline eTobeyTopic starter

  • Frequent Contributor
  • **
  • Posts: 797
  • Country: de
I  had some intuition, and changed the duty to 99.9%, and there you go, flat as Australia. Except a little spike on the probe (green), but that looks the same as before. The coax (blue) looks perfect. Its a 40us gap, but the coax looks pretty good until about a gap of 300us.

I had this intuition, but cant really tell whats up with this. My guess is, that scope has some issues?

p.p.p.s. Now I've got an itch to add something for an extra 0.5 pF on the lower-capacitance input paths in each channel. I will try to resist it and simply remember to re-adjust the probe compensation for each of the two possible input signal paths whenever it may matter.

Maybe a blob of hot snot on the traces would add 0.5pF?
« Last Edit: May 28, 2024, 09:54:10 pm by eTobey »
"Sometimes, after talking with a person, you want to pet a dog, wave at a monkey, and take off your hat to an elephant." (Maxim Gorki)
 

Offline shapirus

  • Super Contributor
  • ***
  • Posts: 1575
  • Country: ua
p.p.p.s. Now I've got an itch to add something for an extra 0.5 pF on the lower-capacitance input paths in each channel. I will try to resist it and simply remember to re-adjust the probe compensation for each of the two possible input signal paths whenever it may matter.
Maybe a blob of hot snot on the traces would add 0.5pF?
That's an interesting thought. I wonder if there are any solderless ways of slightly increasing capacitance of a trace. Need to add just 0.5 pF, it's not that much.
A piece of copper tape connected to ground and put over the trace? Hmm...
 

Online thm_w

  • Super Contributor
  • ***
  • Posts: 6573
  • Country: ca
  • Non-expert
I  had some intuition, and changed the duty to 99.9%, and there you go, flat as Australia. Except a little spike on the probe (green), but that looks the same as before. The coax (blue) looks perfect. Its a 40us gap, but the coax looks pretty good until about a gap of 300us.

You also changed the voltage from 10V to 5V, if that matters.
Profile -> Modify profile -> Look and Layout ->  Don't show users' signatures
 

Offline eTobeyTopic starter

  • Frequent Contributor
  • **
  • Posts: 797
  • Country: de
That's an interesting thought. I wonder if there are any solderless ways of slightly increasing capacitance of a trace. Need to add just 0.5 pF, it's not that much.
A piece of copper tape connected to ground and put over the trace? Hmm...

You may not even need to connect it to GND, because of displacement currents?
"Sometimes, after talking with a person, you want to pet a dog, wave at a monkey, and take off your hat to an elephant." (Maxim Gorki)
 

Offline eTobeyTopic starter

  • Frequent Contributor
  • **
  • Posts: 797
  • Country: de
I have now made a testsignal with 400us rising edge. Still looks ugly.
"Sometimes, after talking with a person, you want to pet a dog, wave at a monkey, and take off your hat to an elephant." (Maxim Gorki)
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 3068
  • Country: gb
If it helps, see below for a screenshot from my old Infiniium 54825A scope

What a nice signal. Would it look the same, if you put the probe on a battery (maybe with a cap)?
It looks the same on a battery although it took quite a few attempts to avoid bounce effects when connecting it. A cap would help here.

I also tried changing the V/div on the Infiniium scope when set to 1ms/div (to match shapirus' settings) and the x10 probe compensation remained flat on every setting.

This 500MHz Infiniium scope only has about 7.5pF input capacitance and the input attenuator is designed on a film PCB. See the image below. This is actually an image of someone else's scope attenuator (damaged?) but I think mine is the same design. You can see the designer used quite complex printed shapes to try and maintain good performance for each attenuator.

A lot of modern scopes (even from HPAK and Tek) save on cost by attempting to do this stuff with ordinary SMD and ordinary PCB technology and they do it all on the main PCB. These scopes generally have about 13-18pF input capacitance and I guess the performance will be less consistent across all V/div settings.


 

Offline shapirus

  • Super Contributor
  • ***
  • Posts: 1575
  • Country: ua
I also tried changing the V/div on the Infiniium scope when set to 1ms/div (to match shapirus' settings) and the x10 probe compensation remained flat on every setting.
Does your scope have a relay that clicks when the vertical scale is changed from a certain value to another? If the response remains flat in both ranges, then the scope has a well matched input capacitance for both paths, or maybe it compensates for the difference in software (not sure if it's even possible though). Mine unfortunately doesn't.
 

Offline G0HZU

  • Super Contributor
  • ***
  • Posts: 3068
  • Country: gb
I also tried changing the V/div on the Infiniium scope when set to 1ms/div (to match shapirus' settings) and the x10 probe compensation remained flat on every setting.
Does your scope have a relay that clicks when the vertical scale is changed from a certain value to another? If the response remains flat in both ranges, then the scope has a well matched input capacitance for both paths, or maybe it compensates for the difference in software (not sure if it's even possible though). Mine unfortunately doesn't.
Yes, it has relays. If you look closely, you can see the PCB pads for the relays in the previous image. There are six relay pads for each attenuator section. I think that having relay switched attenuators also helps with the noise performance.

I think that some later Infiniium scope models (as used at my place of work) don't use relays. These scopes (eg 600MHz 8064A) are much noisier than my old 54825A and they have about double the input capacitance.
 

Offline eTobeyTopic starter

  • Frequent Contributor
  • **
  • Posts: 797
  • Country: de
That's an interesting thought. I wonder if there are any solderless ways of slightly increasing capacitance of a trace. Need to add just 0.5 pF, it's not that much.
A piece of copper tape connected to ground and put over the trace? Hmm...

You may not even need to connect it to GND, because of displacement currents?

@shapirus : What do you think about this? I should note, that there has to be GND, that the copper tape overlaps.
"Sometimes, after talking with a person, you want to pet a dog, wave at a monkey, and take off your hat to an elephant." (Maxim Gorki)
 

Offline shapirus

  • Super Contributor
  • ***
  • Posts: 1575
  • Country: ua
That's an interesting thought. I wonder if there are any solderless ways of slightly increasing capacitance of a trace. Need to add just 0.5 pF, it's not that much.
A piece of copper tape connected to ground and put over the trace? Hmm...

You may not even need to connect it to GND, because of displacement currents?

@shapirus : What do you think about this? I should note, that there has to be GND, that the copper tape overlaps.
I don't know. Might work. This has to be tested, before that there's little point in speculating. I'm not too enthusiastic about taking the scope apart now. Maybe some day later.

p.s. but the idea itself may be easily verified on a piece of a random PCB.
 

Offline eTobeyTopic starter

  • Frequent Contributor
  • **
  • Posts: 797
  • Country: de
I have made more measurements:
The change of the rising edge does not seem the alter the shape of the trace at all, just cutting it off.

It looks, like the magnitude of the fluctuation depends on how the signal was before the rising edge! That would match with the observation, that with only a gap in between, there are no fluctuations.


"Sometimes, after talking with a person, you want to pet a dog, wave at a monkey, and take off your hat to an elephant." (Maxim Gorki)
 

Offline eTobeyTopic starter

  • Frequent Contributor
  • **
  • Posts: 797
  • Country: de
Measurments on a shorter gap:
"Sometimes, after talking with a person, you want to pet a dog, wave at a monkey, and take off your hat to an elephant." (Maxim Gorki)
 

Offline eTobeyTopic starter

  • Frequent Contributor
  • **
  • Posts: 797
  • Country: de
Measurment on a different level rise (0-5V):

"Sometimes, after talking with a person, you want to pet a dog, wave at a monkey, and take off your hat to an elephant." (Maxim Gorki)
 

Offline shapirus

  • Super Contributor
  • ***
  • Posts: 1575
  • Country: ua
@shapirus : What do you think about this? I should note, that there has to be GND, that the copper tape overlaps.
I don't know. Might work. This has to be tested, before that there's little point in speculating. I'm not too enthusiastic about taking the scope apart now. Maybe some day later.

p.s. but the idea itself may be easily verified on a piece of a random PCB.
The effect is surprisingly strong!

I did a quick test with a piece of PCB.

1. First I measured the capacitance between the two marked points and got ~12 pF:




2. I applied a patch of some copper tape (~11 mm x 15 mm) that overlaps the trace and the ground plane, and now it measured 19 pF (!):




So it should not be difficult at all to tune the required track to have an extra 0.5 pF capacitance, as long as it's not routed between two other non-ground tracks. That'll be an interesting experiment, if I ever have enough inspiration to do it :).
 
The following users thanked this post: thm_w

Offline eTobeyTopic starter

  • Frequent Contributor
  • **
  • Posts: 797
  • Country: de
Have you got it still stuck on the PCB? Rub it with some hard round object to make it really flush. How much does it read then?

BTW: Its a bit offtopic, and i think its worth a dedicated topic for it.
"Sometimes, after talking with a person, you want to pet a dog, wave at a monkey, and take off your hat to an elephant." (Maxim Gorki)
 

Offline EEVblog

  • Administrator
  • *****
  • Posts: 37933
  • Country: au
    • EEVblog
FYI, some probes have both low frequency and high frequency compensation adjustments.
 

Offline BillyO

  • Super Contributor
  • ***
  • Posts: 1536
  • Country: ca
FYI, some probes have both low frequency and high frequency compensation adjustments.
Not these $20 70MHz probes.  Which, in my testing, do quite well.

The OP is not using any kind of controlled procedure here.  Everything is in doubt on this one.
« Last Edit: May 30, 2024, 07:05:37 am by BillyO »
Bill  (Currently a Siglent fanboy)
--------------------------------------------------
Want to see an old guy fumble around re-learning a career left 40 years ago?  Well, look no further .. https://www.youtube.com/@uni-byte
 
The following users thanked this post: tautech, shapirus

Offline shapirus

  • Super Contributor
  • ***
  • Posts: 1575
  • Country: ua
Have you got it still stuck on the PCB? Rub it with some hard round object to make it really flush. How much does it read then?
27 pF!

It'll need to be a really tiny piece (or far enough from the nearest ground) to add just 0.5 pF :).

BTW: Its a bit offtopic, and i think its worth a dedicated topic for it.
Not a big deal at this point, I think. It'll be a separate topic if/when I get to tuning my scope's input capacitance using this method.
 

Offline eTobeyTopic starter

  • Frequent Contributor
  • **
  • Posts: 797
  • Country: de
Not these $20 70MHz probes.  Which, in my testing, do quite well.

The OP is not using any kind of controlled procedure here.  Everything is in doubt on this one.

I would love to know how you tested it.

It doesnt look much different when using a probe or a coax with a resistor as 50Ohm terminator. But i guess a terminator would make no difference at all, when the rising edge is 4ms?
"Sometimes, after talking with a person, you want to pet a dog, wave at a monkey, and take off your hat to an elephant." (Maxim Gorki)
 

Offline eTobeyTopic starter

  • Frequent Contributor
  • **
  • Posts: 797
  • Country: de
FYI, some probes have both low frequency and high frequency compensation adjustments.

I found here https://cdn-reichelt.de/documents/datenblatt/D100/TT_HF_612.pdf, that compensation of the low frequency is done with a 1kHz square. That has 0.5ms of high level, but the maximum of the fluctuation is somewhere around 5ms. That compensation would not really help here, would it?
« Last Edit: May 30, 2024, 11:09:00 am by eTobey »
"Sometimes, after talking with a person, you want to pet a dog, wave at a monkey, and take off your hat to an elephant." (Maxim Gorki)
 

Offline BillyO

  • Super Contributor
  • ***
  • Posts: 1536
  • Country: ca
I would love to know how you tested it.

I used a 600MHz scope and a frequency sweep.  The probes themselves were -3 dB @ 214MHz.  More than 3 times the rated bandwidth.  So I imagine the problem you are having has nothing to do with the probes.  As to the scope, the bandwidth was "improved" using a certain python script to generate license keys and was tested using a Bodnar fast risetime oscillator and found to be approximately 255MHz.

I did a short video.



The conditions you are looking at with a "microscope" seem rather pointless and out of bounds of any realistic use case.  I'm just wondering what it is you are looking to find?

Bill  (Currently a Siglent fanboy)
--------------------------------------------------
Want to see an old guy fumble around re-learning a career left 40 years ago?  Well, look no further .. https://www.youtube.com/@uni-byte
 

Offline eTobeyTopic starter

  • Frequent Contributor
  • **
  • Posts: 797
  • Country: de
The conditions you are looking at with a "microscope" seem rather pointless and out of bounds of any realistic use case.  I'm just wondering what it is you are looking to find?

At the moment, i would like to know, why there is a fluctuation of that signal, that is quite long. Also why it is depending on how the signal was milliseconds before. One interesting thing is, that with different risingtimes, the curve does not change at all, apart that it is "cutoff" or somehow "hidden" (https://www.eevblog.com/forum/beginners/probe-compensated-but-level-of-square-signal-gt-20ms-not-looking-good/msg5521063/#msg5521063).

I do not care so much about how high the deviation is (except that spike in the 100's mV), but rather how much later it happens, and why that happens. It would be also good to know if there are other things that can make it worse or better.
« Last Edit: May 30, 2024, 02:19:44 pm by eTobey »
"Sometimes, after talking with a person, you want to pet a dog, wave at a monkey, and take off your hat to an elephant." (Maxim Gorki)
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf