Can anyone explain what's going on here?

[rhb]

These are screenshots from a GW Instek MSO-2204EA  (V1.32 FW) with input from one of Leo Bodnar's 40pS pulsers using a 50 ohm thru terminator.

REF1 is a capture of CH1 moments before.  I'm using the math function to display REF1 - CH1 with CH1 in two positions relative to the step on REF1.

I'd expect DS0010.png to show very little difference.  I'd expect DS0011.png to show a large difference between the steps.  Under no circumstances can I imagine a reason for an alternating polarity 750 mV spike train as the difference.

[newbrain]

I have no knowledge of that GW Instek scope, but that's what I would expect to see if the time scale of the red difference track was compressed WRT the reference and channel 1 tracks:

• short spikes when they are almost in phase,
• wider ones when their phase differs some more.

Is this what that 0.08 red number right of the V/div indicates?

[amspire]

Does the scope have an auto-calibration procedure?

[tautech]

Reg
There's 2 issues I see possibly affecting your results.
Horizontal trigger position change between Ref waveform and Ch 1; that will affect the Math result.
AC triggering ? What's the need for it ?

[SWR]

I would also guess that it is some sort of small phase delay.
0.08Div could indicate a small phase delay of 0.08 divisions???
I haven't received my scope or the LB generator yet, so I can't check it for myself (but it's in transit now ).

One way to find out would be to make small trigger level adjystments to see if you can minimize the difference and then read the phase shift between Ref and Ch1.

There is a CAL output on the back, but I haven't been able to find any detailed info on that.
I'm looking forward to investigating it further when I get the scope.

[tautech]

There are *no* controls over the horizontal scale for either "MATH"  or "ADVANCED MATH"  that I can find.  The only controls are vertical position and scale.

The "

Then why is H= 6.880 ns for the first and -2.700 ns for the second are displayed ?
Did you adjust it ?

Hit default to reset Holdoff and Trigger coupling to DC and give us the same again please.

[rhb]

There are *no* controls over the horizontal scale for either "MATH"  or "ADVANCED MATH"  that I can find.  The only controls are vertical position and scale.  I think the assessment that the horizontal sweep is much slower for the difference trace is the explanation. 

The "ADVANCED MATH" shows approximately what I expect.  I got similar results from "MATH"  at slower sweeps, but the jitter error seemed excessive so I went to a faster sweep.

The last two figures are after setting "Default" and setting a 256 sample average trace, storing a new reference trace and configuring the math operations again.  Note that the difference is not the difference of the averaged trace and the reference, but the sample trace and the reference trace which is not helpful.

In all the figures I am comparing a trace to a stored copy of itself.  I adjusted the horizontal position as a means of creating a known difference from the reference trace.

I removed the accidental post tautech responded to.  The 15 & 16  figures did not capture the rapidly changing math result.  I've added a figure with the pulser off which makes clear that the horizontal scale of the math trace is 100 nS/Div  despite the 0.00Div label.

[tautech]

For interests sake can you try a different approach by using trigger holdoff to get the signal jitter under control instead of using averaging ?

[rhb]

I shall.  I've got a pretty full day ahead of me doing plumbing.  I've got to make a major modification to a shower fixture, unsolder a fitting, cutoff the stainless steel tube that carries the soap dish and hand shower brackets, resolder the fitting and then turn the joint for a tight fit and cleanup the heating marks.  And this requires making wooden jaws to hold everything so I don't make marks on the tube or the O ring assembly.  I'm very much hoping they used soft solder rather than silver solder as I don't have any hard solder and would have to make it and roll it.  Or else wait for commercial solder to arrive in the post.

[nctnico]

There are *no* controls over the horizontal scale for either "MATH"  or "ADVANCED MATH"  that I can find.

The horizontal math scale always follows the scale of the trace.

I think what you are seeing is the time variance between the pulses produced by the pulse generator.

[rhb]

There are *no* controls over the horizontal scale for either "MATH"  or "ADVANCED MATH"  that I can find.

The horizontal math scale always follows the scale of the trace.

I think what you are seeing is the time variance between the pulses produced by the pulse generator.

It *should*, but DS0018.png proves that it is not.  Leo's pulser is a 10 MHz square wave.  According to Leo, the pulser jitter is a few picoseconds.  So any jitter is the DSO timebase.

[nctnico]

I did some testing and managed to see similar effects. It seems the oscilloscope is using the original timebase (time/div) and V/div from the reference trace no matter how the reference trace is zoomed in/out. If you scroll the trace a bit to left or right then you get the narrow peaks.

[rhb]

Here are a series of displays.  I erased memory and set defaults before adjusting the scope to display the waveform;

input  - the input in dot mode with infinite persistence
ref1  - the REF1  settings
math - the difference using "math"
adv_math - the difference using "advanced math"
adv_math_2 - rescaled from 50 mV/Div to 200 mV/Div

Repeating the process resulted in adv_math using 100 mV/Div by default.  It appears that erasing memory and setting defaults does not set positions to zero nor does it set vertical ranges consistently.

No zoom was used and no horizontal settings were changed.  The "math" option is clearly setting the trace display to 100 nS/Div  independent of the other trace display settings. 

[rhb]

All of this started because I tried to compare a thru terminator to a tee and terminator.  It just gets worse the farther I look.

To get better contrast for a 1 nS/Div plot, I decided to try using CH3. The "math" function gives the same 100 nS/Div display, but "advanced math" gives nothing at all at any vertical scale up to 500 TV/Div!  It defaulted to 2 pV!   Pretty fancy math for an 8 bit scope with a minimum of 1 mV/Div and  a maximum of 10 V/Div.  I wonder what would happen if you connected a  probe to a 100 TeraVolt signal?

So I went back to CH1.   I've aligned REF1 as close to  CH1at the zero crossing as I can get it.  We see the "interesting" property that there are discontinuities in the difference plot at every time grid line. At 1 nS after the zero crossing, the advanced math plot claims that CH1 is 250 mV greater than REF1.  In fact, REF1 is greater than the CH1 points.  So both the sign and the magnitude are wrong.

Edit:  It's actually worse.  The advanced math only works for CH1 at 10 Kpts.  It does not work at any other length and does not work at all on any other channel!  But the CSV file has what looks to be decent data, though I've not analyzed it in full.  I just plotted it with gnuplot.  The forum does not allow CSV files so I can't post it.

[tautech]

The forum does not allow CSV files so I can't post it.

You can simply by changing the file extension but it still must be smaller than the forum max file size limit of 1 MB.
It's a good idea to add into the filename some indication of the actual file type.
Filename CSV.txt as an example. Add a mention of the real file name into the body of the post and most won't stuff it up.

[Proto]

Ok, here is what I see. 

1. The (M = R1 - 1) difference signal goes low and high due to the phase difference in the acquisition of the live signal copy stored as the reference signal vs the live signal realtime capture.

2. The phase difference arises from the fact the the digital snapshots that are occurring at a 1GS per second rate are asynchronous to the 10MHz input signal.  The capture drifts within the time span of at least a 1LSB x 1nS window.

3. When the input signal stored in reference memory was captured, the trigger point (where the input signal exceeds the trigger level) was at one position within the 1nS span between sample points.

4. When the LIVE input signal was captured, the trigger point (where the input signal exceeds the trigger level) was at another position within the 1nS span between sample points.

5. (IMAGE DS0010.PNG) When the first live capture of the input with the reference and difference signal was taken, the live trigger point was very close to the reference copy but not at the same instant resulting in a spike in the difference signal

6. (IMAGE DS0011.PNG) When the second live capture of the input with the reference and difference signal was taken, the live trigger point was much further from the reference copy resulting in wider pulse (~4.8div x 2nS/div = ~9.6nS) in the difference signal

7.  Referring the image (IMAGE DS0010.PNG), the interval between the rising edge of R1 (WHITE) and the rising edge of the live signal (YELLOW) causes positive difference signal M (800mV) since R1 is at +400mV amd the live signal is at -400mV.  (+400 - (-400)) = +800

8.  Again, referring the image (IMAGE DS0010.PNG), the interval between the unseen falling edge of R1 (WHITE) and the unseen falling edge of the live signal (YELLOW) causes negative difference signal M (-800mV) since R1 is at -400mV amd the live signal is at +400mV.  (-400 - (+400)) = -800

9.  You should try the same experiment with longer acquisition memory up to 10Mpts.  The pulse width span from test to test will occupy a narrower range.

10.  Who can figure out the upper bound on M 's pulse width since it is around 10 times greater than the 1nS sample interval in the second figure?



Make any sense - it's late.

Regards,