Step Response on Rigol scopes

[Jwalling]

Came across this on S.E.D. (That's a Usenet group called sci.electronics.design for you young-uns) posted by the Engineer/owner of Highland Technology in SF California. Posting it verbatim; I'd be curious to see the responses to this.

"The Rigol 4054 (500 MHz) and 6104 (1 GHz) scopes were tested for step response.

The step generator is a Tek SD24 TDR sampling head, TDR pulse out. Very similar results
were seen using a Highland P400 pulse/delay generator.

Both scopes have reasonable, somewhat ringy step response when working in 50 ohm mode,
and an ugly step response in 1M mode.

These measurements are consistant when the two different step generators are used, when cable lengths are varied, and when a 10 dB 50 ohm attenuator is added directly at the scope input. Other scopes do not show the funny steps when presented with the same signals.

The 6104 shows a similar stepped rise when used with its Rigol passive 10x probe.

The 4054 is also strange when used at 50 ohms and 250 MHz bandwidth limit.

We have seen a similar ugly step response on the 200 MHz Rigol 1204B, which is always hi-Z.

It's strange that all the Rigol scopes have this weird step response in hiZ mode."

John Larkin  Highland Technology Inc   July 13, 2016

Any thoughts?

EDIT: Can anybody reproduce this?

Jay

[hendorog]

This is a bit of a guess, but I think this might be due to the frequency response - specifically the wide input bandwidth.
Someone here did a chart of the roll off above 500MHz and other scopes roll off more rapidly than the DS4054 does.

IIRC I've managed to get a trace up to around 1.5GHz on my DS4014.

I did some quick tests and flicking on the 200MHz bandwidth limiter removes the 'hump' at the top of the cliff.
This is using the scopes aux output which can generate a fast rise time signal.

This theory could be tested using a 500MHz low pass filter.

[David Hess]

The 1st photo looks like a 2 nanosecond reflection (two ways though the cable) into the 1 megohm input from a roughly 10 inch cable between the pulse generator and oscilloscope.  I do not see any feedthrough attenuator at the input to the oscilloscope which is backed up by the doubled pulse voltage.

The 2nd photo looks reasonable but is not consistent with the 3rd photo; see below. 

The 3rd and 4th photos show vertical sensitivity of 1 volt/div which is not consistent with the Tektronix SD24 250 millivolt maximum output pulse into 50 ohms.  Maybe the DS6104 oscilloscope thinks a x10 low-z probe is connected?  This would be consistent with the text which mentions a x10 probe being used earlier.

If that is the case, then the 4th photograph is just broken and the 3rd photograph is not much better.  The internal 250 MHz bandwidth limit should produce a nearly perfect 1.4 nanosecond edge on a 1 GHz oscilloscope.  Maybe the patch cable which is different from the one shown in the other photographs is bad.

[Jwalling]

John did some more testing and posted the results to SED again.

>Any thoughts? Have you tried it with the Tek?

Here's a direct comparison: same setup, only the scopes changed.

https://dl.dropboxusercontent.com/u/53724080/Gear/Rigol/Tek_X.JPG

https://dl.dropboxusercontent.com/u/53724080/Gear/Rigol/Rigol_X.JPG

Both scopes have that little plateau on the rising edge in 1M mode,
but the Rigol is much worse, bigger and longer.

The cable is a semi-hardline that TDRs nicely, 51 ohms and 480 ps
one-way. I trust the step pulse.

All three Rigol scopes have that step. My Tek DPO2024 doesn't.

https://dl.dropboxusercontent.com/u/53724080/Gear/Rigol/Tek_D.JPG


We may be able to use the Rigol in 50 ohm mode to make the critical
measurements in our test sets, but we'd have to be careful about
probing. Fet probes should be OK.

Jay

[David Hess]

Both scopes have that little plateau on the rising edge in 1M mode,
but the Rigol is much worse, bigger and longer.

The plateau is the reflection from the 480 picosecond cable into the 1 megohm inputs.  The 200 MHz oscilloscope is too slow to show it.

All three Rigol scopes have that step. My Tek DPO2024 doesn't.

Looks like the lower bandwidth did not see the reflection and my guess is that the overshoot is caused by a resonance in the high impedance input attenuator.  Odd things can happen if you drive a 200 MHz oscilloscope with an 8 GHz edge.

Is he expecting these oscilloscopes to all show the same response when operated under abnormal conditions?

[Jwalling]

Both scopes have that little plateau on the rising edge in 1M mode,
but the Rigol is much worse, bigger and longer.

The plateau is the reflection from the 480 picosecond cable into the 1 megohm inputs.  The 200 MHz oscilloscope is too slow to show it.

All three Rigol scopes have that step. My Tek DPO2024 doesn't.

Looks like the lower bandwidth did not see the reflection and my guess is that the overshoot is caused by a resonance in the high impedance input attenuator.  Odd things can happen if you drive a 200 MHz oscilloscope with an 8 GHz edge.

Is he expecting these oscilloscopes to all show the same response when operated under abnormal conditions?

I think he expects the two 500MHz Tek and Rigol scopes to show the same response. Wouldn't you?

Jay

[David Hess]

I think he expects the two 500MHz Tek and Rigol scopes to show the same response. Wouldn't you?

With 50 ohm inputs or 50 ohm feedthrough attenuators, I would expect close to identical results at the same bandwidth but he did not show the later even using his 10dB attenuator which can sort of act as a poor feedthrough attenuator.  The slightly peaked response on the Rigol 4054 with 50 ohm input is typical.  The ringing shown in the Rigol 6104 with 50 ohm input is very poor and the 6104 with 50 ohm input and 250 MHz bandwidth limit is completely broken.

I would not trust any of the 1 megohm input results or even expect them to be consistent.  Oscilloscope and probe inputs are specified with a 25 ohm source impedance which results from either a 50 ohm loop-through or a 50 ohm transmission line terminated into 50 ohms which is what the oscilloscope does internally or the user does externally with a 50 ohm feedthrough terminator.

[MarkL]

The screen shot of the Rigol DS6104 with 50R termination looks like the step response you would get with a flat (non-gaussian) filter design on the scope's input.

Rigol may have chosen that type of filter because of the proximity of the Nyquist limit when they're sampling at 2.5GS/s (2 adjacent channels on).  I couldn't find any info on what they've actually done.

An Agilent/Keysight appnote discusses the various input filter strategies and their effect on displayed waveform:

  http://literature.cdn.keysight.com/litweb/pdf/5988-8008EN.pdf

[Bud]

Do not seek logic in rigol's designs. It is a bogus company that makes bogus stuff, as i showed in my Project Yaigol post. The best that could happen they copied someone else's design and none of them can explain how it works.

[David Hess]

With no preshoot, it does look like a butterworth step response.  That does not explain the completely broken 250 MHz bandwidth filter though.

As a practical manner, I would still consider a butterworth step response broken.  The Tektronix 1 GHz+ oscilloscopes I have evaluated had no such problem.