Differences in Measurements 50Hz Keysight 1000X scope.

[Twistx77]

Hi,

Recently a friend of mine have discovered that his 1000X series scope is measuring wrongly a 50hz sine wave. The measurement changes when changing from DC to AC coupling, measuring higher values when using AC coupling.

When he uses a Rigol scope he doesn't have that problem and the measurement is closer to the one shown by a DMM.

At first glance, one would think that the mismeasurement is caused by the value of the dc blocking capacitor of the scope, but the weird thing is that the value increases when using AC coupling.

We think that in theory should decrease since in the best case you are blocking some DC that is causing the misreading or the impedance of the capacitor is not low enough for the 50 hz signal.

Another reason we think it might be reading a higher value, is because of the deformation suffered by the signal when switching between DC and AC. coupling

I have tried the same thing with my scope which is exactly the same model and I get the same thing.

Here is a little video showing the a simple test: https://www.youtube.com/watch?v=Qqv0gDAt9oI&feature=youtu.be

How can this be explained?

Thank you.

[genghisnico13]

interesting.... perhaps repeating the test while displaying the FFT will give you more information. you could sweep the internal generator from 5-100Hz with DC and AC coupling to see the frequency response difference, maybe there is a small peak around 50Hz with AC coupling.

[David Hess]

That should not be happening.  Do the same test with a low frequency square wave source or even the calibrator output to see how it affects the transient response.  Also try it with a different input channel.

[rf-loop]

This can happen depending waveform. Of course (more or less) -  who experienced scope user have never meet this "feature" when  signal have (and as we can see in video this 50Hz waveform have)  some harmonics. Signal is far away from pure 50Hz sine wave. Keysight data sheet tell that AC corner freq is around 10Hz -3dB.

This effect can see well if use very extremely bad "sinewave" with extremely lot of harmonics... all scopes do this (but amount depending AC coupling freq response and used test freq.)

If it helps thinking lets find limes cases what is some times good for thinking: Think that other limes for this case  is extremely pure sine wave and other (bad case) limes is "bad sine" aka  square wave and both have 50Hz 1st harmonic. More we go out from pure sinewave example with odd harmonics (example with series of harmonics what sum is more and more like square then more dramatic this effect is (because our 1. harmonic 50Hz is so close this AC response corner). Try with more low than 50Hz square, say example 10Hz and it can see how big difference is p-p value when switch  between AC and DC coupling. With 50Hz this happen also but less.

When AC coupled and corner is 10Hz it can well see that 50Hz square wave is not at all square wave. example 10Vpp square p-p value is now really lot of more than 10V - naturally.  Then go slowly more and more close to pure sine and you see  less and less p-p error... until error fall  outside of detection threshold.

Of course this OP's video example is not pure sine and it is not worst case but just enough that this effect come visible and perhaps added by some bit weird thing what happend to waveform shape but perhaps even this is just this same "math" added with oscilloscope errors. (lets hope AC/DC coupling do not change input reactrance significantly -  if this happen then it goes bit more complex.)

I can not show this because I do not have now scope available in hand here what have 10Hz AC coupling corner f.
But just demonstrated this with scope what have lower (1.3Hz) AC coupling corner freq and  then used much lower "bad sine" 1st harmonic freq (6.5Hz) and effect can see - of course.

This is not from this case but look figure and think... it may give wink... of course situation is not at all so dramatic. But this is possible principle what is root of this case. Then change thus blue square with poor sine... less dramatic but still can detect small rise as in OP video show.

[David Hess]

Of course the AC coupling acts as a high pass filter however what we see shows both a loss of amplitude of the fundamental and a greater loss of amplitude of the harmonics when inserting a high pass filter should instead cause the opposite.

I have lots of oscilloscopes with an AC coupling cutoff frequency of 10 Hz and they have never displayed this behavior when looking at a power line frequency sine wave.

[sibeen]

OK, just tried the same thing with my 3024T, and there is a difference. Not as pronounced, and it is lower when being AC coupled, which seems to make more sense. I went back and forth a few times and the difference between the two measurement was in the range of 600 to 700 mV with teh DC coupled reading being the higher.

[Twistx77]

Hi,

Thank you all for commenting.

rf-loop I can see what you propose as a possible cause but the problem is that even if the signal is not a perfect sine, I don't think that the little imperfections it might have is not enough to cause such a big difference in the measurement.

OK, just tried the same thing with my 3024T, and there is a difference. Not as pronounced, and it is lower when being AC coupled, which seems to make more sense. I went back and forth a few times and the difference between the two measurement was in the range of 600 to 700 mV with teh DC coupled reading being the higher.

600mV for a 240V signal is not even close to the error measured with our oscilloscope.

In the next video you can see that even with the internal function generator the difference is petty big almost 1V for a 12Vpp wave which is almost 10% error...

https://www.youtube.com/watch?v=mXpQm62uARQ&feature=youtu.be

[rf-loop]

Hi,

Thank you all for commenting.

rf-loop I can see what you propose as a possible cause but the problem is that even if the signal is not a perfect sine, I don't think that the little imperfections it might have is not enough to cause such a big difference in the measurement.

OK, just tried the same thing with my 3024T, and there is a difference. Not as pronounced, and it is lower when being AC coupled, which seems to make more sense. I went back and forth a few times and the difference between the two measurement was in the range of 600 to 700 mV with teh DC coupled reading being the higher.

600mV for a 240V signal is not even close to the error measured with our oscilloscope.

In the next video you can see that even with the internal function generator the difference is petty big almost 1V for a 12Vpp wave which is almost 10% error...

https://www.youtube.com/watch?v=mXpQm62uARQ&feature=youtu.be

Yes this amount of error is weird. If it exist also with pure sine wave then more weird.

Can you test it also without probe using direct cable.

Because now I look more carefully your first video... it really looks like that also fundamental 50Hz rise as D.H. explain.
And now with new video with quite pure sinewave this is clear. My explanation do not anymore match for this.

How it look if you show good 50Hz square wave with AC coupled and DC using probe (is it 1:10 probe?) and then direct cablel for exclude possible probe related things.

How your probe show 1kHz square (probe cal signal) using AC and DC coupling.


Here is one tiny example fore some random readers who may wonder what I'm talking in my previous msg:
5Hz sine together with some individual levels of added 3. 5. 7 and 9. harmonics observed with Siglent. Used 5Hz fundamental because Siglent AC coupling -3dB corner is very low (with new better measurement near 1Hz).

Both channels input signal equal. (meas: Cstd (Cycle standard deviation) is same as Cycle RMS without possible small DC offset/bias). Using AC/DC coupling just be careful what and how measure for avoid this "trap".  Of course as told previously with some freq and waveforms effect may be huge. Also in some scopes AC/DC coupling may change scope input reactance more or less or negligible.

ETA: For clarify: As can see in images, when all works right, including measurement.  "amplitude" may rise depending signal shape (harmonics) but  RMS drop when AC coupled! (here RMS measured using cycle standard deviation to overcome the possible small offset effect )   Because AC coupling low pass filter attenuate fundamental. As can see in image mesurements.

[Svuppe]

In the next video you can see that even with the internal function generator the difference is petty big almost 1V for a 12Vpp wave which is almost 10% error...

I have no explanation for this weird behaviour, but I can confirm it. I tried the same test (50 Hz pure sine) on a 1000X series scope at work, and it does the same thing. Not quite as much as your scope, but noticeable none the less. AC measures higher than DC (unfortunately I didn't write down the exact numbers).

At home I tried it on my DSOX2024A, but that behaves as normally expected. AC is a little lower than DC at 50 Hz.

[Keysight DanielBogdanoff]

Our support team is looking into it!

[David Hess]

OK, just tried the same thing with my 3024T, and there is a difference. Not as pronounced, and it is lower when being AC coupled, which seems to make more sense. I went back and forth a few times and the difference between the two measurement was in the range of 600 to 700 mV with teh DC coupled reading being the higher.

More importantly, the shape did not change.

Run the same test using a x1 probe instead of a x10 probe.  Obviously this will require a transformer.  The high pass cutoff frequency will be significantly higher, I measured 20 times but this is not a trivial measurement, with a x1 probe.  Exactly where the high pass cutoff is depends on the internal coupling capacitance but I learned long ago to always use a x10 probe for measurements at power line frequencies where AC coupling is used if I want the best accuracy.

[Keysight DanielBogdanoff]

Ok, here's what R&D is telling me.

AC coupling is useful for view waveforms with large DC offsets.  AC coupling places a 10 Hz high-pass filter in series with the input waveform and removes  DC offset voltage from the waveform.  Keep in mind that 10 Hz is the filter’s -3dB point (and still has some attenuation). AC coupling should only be used on sine waves greater than 100 Hz and square waves greater than 200 Hz. DC coupling should be used on frequencies lower than that.

So, it's basically a filtering thing when using AC coupling.

[Keysight DanielBogdanoff]

Also, updating to 1.10 firmware will mention this in the built-in help. It will also give you more Bode plot control.

https://www.keysight.com/main/software.jspx?cc=US&lc=eng&ckey=2958151&nid=-32976.1203266.02&id=2958151

[2N3055]

I apologize if I'm out of line but, did anybody read what OP wrote or I'm crazy...

OP says in first post that he/she tried both AC and DC coupling, and that DC coupling is actually showing SMALLER RMS value than when AC coupled..

He/she says that he/she expected that AC coupled should show less than DC but it didn't. DC coupled is showing less than AC coupled... There is almost 2V difference (DC showing circa 8%  less than AC coupling)

Svuppe also tested it on his scope and confirmed odd behaviour...

In the next video you can see that even with the internal function generator the difference is petty big almost 1V for a 12Vpp wave which is almost 10% error...

I have no explanation for this weird behaviour, but I can confirm it. I tried the same test (50 Hz pure sine) on a 1000X series scope at work, and it does the same thing. Not quite as much as your scope, but noticeable none the less. AC measures higher than DC (unfortunately I didn't write down the exact numbers).

At home I tried it on my DSOX2024A, but that behaves as normally expected. AC is a little lower than DC at 50 Hz.

So no, he/she is not having problems with AC coupling low pass filtering at 50 Hz...
It seems only David Hess understood the question...

Those two 1000X series scopes (two confirmed) have something funny in how they calculate/measure RMS value on DC coupling...
It might be interesting to check software version on those two. And for more people to check if they can replicate behaviour.

Regards,

Sinisa

[Svuppe]

I am back with more information from the 1000X scope at work.
It is a DSOX1102G with firmware version 01.01.2016092800

I made a test setup, splitting the generator output up in two, and using equal length cables to connect both channel 1 and 2.
All the following screendumps are taken with 16 times averaging.

First image: Both channels are DC coupled and at 50 Hz.
Second image: Channel 2 (green) changed to AC coupled.
Third image: At 500 Hz there is no real difference.
Fourth image: The peak amplitude on the AC coupled channel is at 77 Hz.
Fifth image: Below 77 Hz the amplitude dropped again, and were back at the DC channel at 36 Hz (but with phase shift).

[Performa01]

Some seem to think AC coupling is just a simple first order high-pass (RC) filter at the scope input, which might be true for simple low-bandwidth designs only. The unspecific statement from Keysight R&D is not helpful either.

The actual reason for any weird effects including wiggly frequency responses and implausible differences between AC/DC coupling is the complex interaction of various time constant (including the crossover networks) in the split path input buffer. It is particularly difficult to get flawless performance over a wide range of source impedances at low frequencies, which can be 50 ohms for a direct coax connection but also 9 Mohms for a 10:1 passive high impedance probe.

I have quickly designed and simulated a split path input buffer with 10Hz corner frequency for AC coupling. The difference in frequency response for AC/DC can be seen below. Draw your conclusions…


DSO Input Buffer FR

[2N3055]

Some seem to think AC coupling is just a simple first order high-pass (RC) filter at the scope input, which might be true for simple low-bandwidth designs only. ......

Hello Performa01!
First of all, let me use this opportunity to thank you for your kind sharing of your knowledge. You have a talent of clarity in expression, explain things well and are very professional in both method of approaching problem and in presentation. Your work on testing Siglent scope is superb.. It serves as both a excellent source of information about scope in question, and also about scope theory and use in general and shows method how to do things right...

That being out of the way, I would like to also thank you for this excellent post. It stands in full, and explains some things well..

But I, for one, am aware that scope input path (together with probe and circuit under test) is anything but simple...

It might as well be that something very similar to what you explain is at work here...

My problem with it is that it is almost 10% error, at 50 Hz.. Right there where you would measure if you are looking into linear power supplies or audio......
And it is on a scope that costs 3x more than Rigol Z-box or 2x than Siglent ... That  don't have such issues as far as we know...

Which would be OK if 1000x manufacturer's condescending slogan wasn't, and I quote : "Scrap the toys, get a real oscilloscope. Have confidence in your measurements with ............."

I have enormous respect for HP/Agilent/Keysight. I grew up with them being standard of highest quality and corporate ethics all should aspire to.

But at his moment, I honestly believe that not buying Rigol 2000A or Siglent scope for the money they ask for 1000X would be just bad  choice.

Only thing that would work for 1000X series would be that their slogan was true.. Which at this point is not so clear..

I hope Keysight will  live up to their reputation, and investigate this. If there is an issue, for all affected it should be fixed in warranty..
They paid more money than they would to other manufacturers, because of reputation and expecting better support ..

And if I had to take a wager, i would think the Keysight will sort this out...

Thank you and best regards,

Sinisa

[Performa01]

Hello 2N3055 (the transistor that became famous in the 70s of last century for its power dissipation of 115 watts)!

Thank you for your kind words and my apologies if I sounded like defending Keysight or their product, let alone their advertising. I just noticed that all attempts to find an explanation for this flaw were based on the assumption of a simple first order high-pass filter right at the scope input. Even the response from Keysight R&D did not clear this up, so I felt the urge to point you to a different direction.

While the AC coupling actually is just an RC high-pass, it sits in the DC/LF path that has several additional time constants and also local as well as global feedback, the latter including the AC/HF path of the input buffer. This makes it an active filter and a bad choice of the time constants could even turn the entire cicuit into a notch filter...

So yes, there obviously is a flaw, but it cannot be explained using a simple first order low-pass filter model.

It is certainly not easy to get a reasonably flat frequency response at low frequencies, and other scopes might have problems too, but at much lower frequencies, simply because most will have a significantly lower AC corner frequency than 10Hz. So the issue might just go unnoticed.

In my review of the Siglent 1000X-E, I have included some tests for checking the proper design of the split path input buffer, but I was only concentrating on what I thought would be relevant in practical use, i.e. a 100Hz square wave and a linear sweep from 10Hz to 1kHz, all of this with DC coupling only. I did not check the difference between DC and AC coupling and since this scope has an AC corner frequency of ~1.2Hz, there might be a problem around 6Hz for instance. I always come across new test scenarios that would be useful to be included in a comprehensive review and I will certainly take a closer look at that and publish my findings. But after experimenting a bit with my circuit simulation, I would be surprised if there was not some problematic frequency range as well, since it really doesn't seem to be trivial to avoid that.

Well, the obvious cure might actually be the use of a lower AC corner frequency. Even without that issue, 10Hz is a rather unlucky choice for my taste.

[2N3055]

Performa01,

115W was a bit of marketing for 2N3055... 50W was more realistic, and that was with good cooling... When I was a kid I liked making PSU-s and have many fond memories of good ol' trusty 2N3055...
In former country they made them on old single diffusion process, those were slow but had massive SOA... And they were cheap as chips... You put few in parallel, LM723, and you had very nice PSU, quiet good regulation and so robust you could weld with it....

I also apologize if I made an impression to accuse you of protectionism... Quite the opposite, I respect that you simply say how it is... Bad and good. Let the numbers speak.

I agree, that cutoff frequency is a little high...  I took a look and it seems that Rigols are specced at <= 5Hz/-3dB..

If I catch some time, I can try your test to see if DS1000Z has some peaks in low frequency range..

Anyways,  it's a pleasure to talk to you.

Good night!

Sinisa

[Performa01]

I have checked the split path input buffer performance of the SDS1104X-E once again with emphasis on the differences between AC/DC-coupling and the results can be found here:

https://www.eevblog.com/forum/testgear/siglent-sds1104x-e-in-depth-review/msg1454076/#msg1454076

[2N3055]

Just a quick report..

My Rigol DS1074Z has cutoff frequency at cca 4,5 Hz.. Difference in amplitude between AC and DC is less than 1 % from 5 Hz to 50 Hz. No visible overshoots.
My Picoscope 3406D goes down to less than 2Hz, with no detectable amplitude difference above cutoff between AC and DC coupling.

Regards,

Sinisa

[rf-loop]

Here Siglent SDS1104X-E  Signal from SDG1032X using RG223 coax (without 50ohm termination).
Sweep times 100s

AC coupling 0.01 - 10Hz   (1Hz  1div)
Y Marker ~-3dB  X markers ~0Hz and  ~1.3Hz (13s)

AC coupling 0.1 - 500.1Hz (50Hz 1div)

AC coupling (yellow) DC coupling (green) 0.1 - 500.1Hz (50Hz 1div)
Both traces ~6Vpp and display 200mV/div so signal total is 30div p-p   
In this image cursors and also 2 channels are just as "cursors" +/-  ~0.3div related to signal pos p only for more easy visaualize amount of "error".  Siglent is ot "state of art" but good if look this part of level flatness. What is optimal AC coupling high pass filter corner freq - I do not know any "one truth" for it.  It depends... Of course I like if it is adjustable (as many other things also - as example also adjustable (analog side) low pass brick wall filter)... all can wish.

[TK]

I tested on my EDUX1002G and I am getting the same result as OP