Siglent SDS1202X-E compensation issue (and other manufacturers ?)

[nctnico]

There is a test you can do: use a direct connection between the calibrator output and the input (or use a function generator with a square wave output). If the effect is gone then the problem is likely that the input capacitance of the scope isn't constant and throwing off the adjustment for the 1:10 probe. If the effect persists then it is likely a factory calibration error or out of spec component.

In this case my opinion is that it is possible that square wave is disturbed due to small change in oscilloscope 1Mohm input (capasitive) reactance change.  V bands 2 and 3 have not so much difference between each others but V band 1 change capasitance more. Change is perhaps even some 1pF's and it affect to this probe Low Freq compensation.  {Siglent development team need take this into consideration}

Every V band (I, II and III) have separate pathway what are selected by relays. There is not adjustment for null these capasitive input reactance differences between vertical bands.
I do not want speculate if different probes are less sensitive for this small reactance change.

If want, it can also say that this error do not violate given specifications for frequency response error limits or level error limits, not even close.

It is way out of specification. At low frequencies (say up to 1MHz) an oscilloscope should be as accurate as it's DC specifications but this error is way over! Besides that I have never seen this kind of behaviour on any oscilloscope I have used.

[rf-loop]

At low frequencies (say up to 1MHz) an oscilloscope should be as accurate as it's DC specifications...

This is your opinion. This is not truth - it is trumpth.
 
Specifications tell frequency flatness is + 1dB from DC to 20MHz.
For draw acceptable 1kHz square what frequencies of sinewaves you need so that sum is what you see is "square wave" on this case.
Think that every these single sinewaves may have level error something between + or - 1dB.  (including that also this DC may have 3% error. Now draw specifications acceptable error window example for this top part of this square wave. Is it inside this window.
You or who ever can define what is nice to have or how it should looks like what you like is ok but if it violate only this opinion it do not mean it violate given specifications.

And I do not mean that it should not be better. It should, of course.

When designing front end layout and topology it need look with signal eyes from dc to rf. It must not look with human eyes as we look if things on table are in good order.  It need care more about how signal see it.

Besides that I have never seen this kind of behaviour on any oscilloscope I have used.

I have seen same situation also example in Owon XDS3000. And in my lifetime experience I have seen this situation many times.



Here Owon. Error is not so big but also there can see separate vertical bands have different input reactanse. (mostly capacitive)  But owon input nominal capacitance is much lower and also changes are lower between bands.
Attacments just for example. (I do not have enough high voltage level gen for drive system with probe set 10x ) )

In images used 200MHz probes connected to Gen output where is also 50ohm extra feed through directly attached to gen output before probe adapter. 20M BW is on for reduce Owon noise and it do not affect this shape. (gen out imp is 50ohm and after then 50ohm feed thru and probe connected to this with probe bnc adapter)
Between images probe is untouched. You can see which image probe is compensated and then changed level and scope Vertical level band (Level band relay switch just between these levels)  It can see with other level band probe is not anymore perfectly LF compensated.

[TK]

I have seen same situation also example in Owon XDS3000. And in my lifetime experience I have seen this situation many times.

Can you name other scopes where you experienced the same "feature"?

[nctnico]

I have seen same situation also example in Owon XDS3000. And in my lifetime experience I have seen this situation many times.

Can you name other scopes where you experienced the same "feature"?

I agree. I see the 'Siglent bug sugar coating team' is hard at work to justify why a faulty product isn't faulty. Before we know it they try to sell the emperor some new clothes as well!

If a scope from a brand like Keysight, Tektronix, Yokogawa, Iwatsu, R&S or Lecroy shows this behaviour their customers will come over with pitchforks, boiling tar and feathers.

[MrW0lf]

Think this needs more scientific approach. Problem is not something most people would notice unless specifically looking for it. So would suggest to actually measure it on various super-duper-quality scopes and then talk. For measure zoom in vertically so signal would have same general size (unlike with Rigol example provided). Preferred near-full screen. Possibly apply some kind of averaging since 8-bit ones may get quite fuzzy. Measure with cursors.
In fact good video material for Dave, much like catastrophic problems with microphonics in super-duper R&S.
Overall mr rf-loop is right, everything turns to s**t if look too close. Here science comes in. Is s**t big, small, how does it smell and is it natural form for given species or perhaps somehow twisted.

[StillTrying]

Some CML+ results. On the CML+ the relay click occurs between 200mV <click> 206mV at the scope.

With the probes on X1 there's very little visible change to the square-ness of the trace as the relay clicks between 200mV <-> 206mV.

With the probes on X10, using the tkonekt test where the probes are re-compensated each side of the relay click, my results are nearly as bad as the OP's, although I don't think it's a massive problem.

So it's a X10 compensation thing ?

EDIT.
After a rough measurement, it seems to be the scope's input capacitance changing when the relay clicks, ~17pF to ~20pF.

[MrW0lf]

PicoScope 2408B
Supplied TA132 150MHz X1/X10 probe at X10 setting
3.2Vpp 1kHz from signal gen, BNC adaptor, directly on output
4Vpp, 10Vpp, 100Vpp ranges, 12bit high res mode



Some effect is there but suppose no issue with typical usage. But good to learn have to re-cal if doing something ultra accurate.

[nctnico]

For a good comparison you should calculate the relative error. The Siglent from the OP is in the 8% ballpark where the Picoscope's error is less than 2%.

[MrW0lf]

Also tried with Testec TT-HV 250 x100 probe and interestingly effect was smaller. This is 300MHz probe with 4pF capacitance:
http://www.testec.de/en/products/product-categories/standard-probes/
So interesting if try some higher bandwidth or high impedance probes with Siglent, will effect also be smaller?

[Paul Moir]

Mine displays this a bit but not nearly as bad as the OPs.  I measure about a 4pF change when the attenuator switches out on both channels which I think jives with how much I would have to twiddle the compensation by.
EDIT:  Pretty sure my old Tek 465 has holes on the side through which you can match up the attenuators.  My Heathkit IO-12 certainly did so that tells you what high quality gear is supposed to be like.   

EDIT2:  I finally got a chance to double check the measurements and it's spot on 4.0pF on one channel (maybe 4.02pF but you know how that goes) and 4.1pF on the other.  The input measures 13.1-13.4pF + 4.0pF so it's "within spec" of 18pF.   It almost seems like a calibration error it's so exact.  I wonder what other's measure?

[egonotto]

Hello,

with the DSO 2074G it is more worse

https://www.eevblog.com/forum/testgear/experience-with-dso-2074g/msg1320502/#msg1320502

egonotto

[borjam]

I noticed a very small effect on my SDS1202X-E, both using the built in calibration generator and the one from a Rigol DS1054Z.

I have just tried the Siglent provided probes but I have four nice LeCroy PP005 (I found a really nice deal on eBay for brand new probes).

I'll repeat it using a LeCroy.

[Serge Petrov]

Hi, borjam!

So, what is the difference between Siglent probes and LeCroy probes in the context of this problem?

[StillTrying]

Micsig, Keysight, Tektronix, Yokogawa, Iwatsu, R&S or Lecroy.

Anyone else done the relay click with X10 probes test.

[Chris-G]

I too have the same problem as the original post.  It looks worst when using the fine control between 1.20v and 1.22v (Relay Click).

I'm not sure of the capacitance change at the input when the relay clicks, but measuring the resistance, the input is 1.0023Mohm (1.20v/Div) and 0.9990Mohm (1.22v/Div).  A change of 3.3Kohm.

The problem is on both channels and absolutely identical.

[StillTrying]

I'm not sure of the capacitance change at the input when the relay clicks, but measuring the resistance, the input is 1.0023Mohm (1.20v/Div) and 0.9990Mohm (1.22v/Div).  A change of 3.3Kohm.

I don't think the 3k3 change would do it.
Assuming a SDS1202X-E, connecting a 3M3 resistor directly between the scopes BNC input and its 1kHz should give an on screen measurable change in the rise time or RC time of the scopes input capacitance as the relay clicks between 120mV and 122mV/div.

[borjam]

Micsig, Keysight, Tektronix, Yokogawa, Iwatsu, R&S or Lecroy.

Anyone else done the relay click with X10 probes test.

Interestingly, LeCroy mentions some differences between V/Div settings in their P007 probe manualhttp://cdn.teledynelecroy.com/files/manuals/pp007_om-e.pdf.

Although they mention overshot and ringing when mentioning the HF compensation.

(page 10)

"Some overshoot and ring will be present at some settings of V/Div.
Adjust both trimmers for the overall best response on all ranges."

(The trimmers are two hidden trimmers in some LeCroy probes, at least my PP005's have them)

[Chris-G]

I don't think the 3k3 change would do it.
Assuming a SDS1202X-E, connecting a 3M3 resistor directly between the scopes BNC input and its 1kHz should give an on screen measurable change in the rise time or RC time of the scopes input capacitance as the relay clicks between 120mV and 122mV/div.

I didn't think the 3K3 would be a problem, after all it's only a change of 0.33%, and yes its an SDS1202X-E.

I've done the test you suggest using a 3M0 resistor (I couldn't find a 3M3), and attached the results.

[StillTrying]

"I've done the test you suggest"

Your % change in rise times is twice that I see on the CML+. When viewing a 3Vpp to 8Vpp square wave on 1.2 to 1.22V/div I'd expect your HF compensation error to look much worse than mine or the OP's in the first post. 

I can't get the above link, page 12 in this one but it doesn't say much.
http://ppmtest.com/wp-content/uploads/pp007_om-e.pdf

[Chris-G]

I'd expect your HF compensation error to look much worse than mine or the OP's in the first post.

I've replicated the first posts screenshots (1-5) for Ch1 input.  They look the same to me!

[StillTrying]

"I've replicated the first posts screenshots (1-5) for Ch1 input.  They look the same to me!"

And me, there's an updated test in reply 8 that keeps the height of the waveforms very similar.

https://www.eevblog.com/forum/testgear/siglent-sds1202x-e-compensation-issue/msg1324788/#msg1324788

I think mine are just as bad as the OP's in the above post, I'll be adding this to my CML+ list at this rate!

[tautech]

Siglent are in the last stages of validating a remedy, they say they'll have something for us next week.

[Chris-G]

Siglent are in the last stages of validating a remedy, they say they'll have something for us next week.

Wonderful...  I'd love to know how they are going to do it though.  If it is an hardware problem (which I assume it is), I can only think they are applying a digital filter to compensate for the error, dependent on the voltage range and probe type.

[StillTrying]

"I'd love to know how they are going to do it though."

So would I !

[Serge Petrov]

Siglent are in the last stages of validating a remedy, they say they'll have something for us next week.

May be it will be compensated with SW patch