Passive probe compensation - reason for subtle abnormal waveform

[trobbins]

I am trying to appreciate the reason for a subtle abnormal squarewave response when compensating a motley selection of x10 probes to a simple 20MHz analog scope (GW GOS-622B).

The 1kHz calibration signal is fine for x1 probes, but all my motley probe collection show the following generic range of response when compensating for x10 input (on either channel).  With the scope end trimcap compensated to achieve a flat latter part of the waveform, the initial portion shows a generic overcompensated characteristic (image #2).  With the scope end trim cap re-adjusted so that the start of the waveform rises to the nominal level, there is then a slight undershoot before returning to a flat latter part of the waveform (image #1).

Can anyone confirm that this characteristic response is mainly related to the level of shunt capacitance across the x10 probe tip 9Meg resistor when working with the combination of probe trimcap circuitry and scope input capacitance/impedance (my guess), or is it perhaps related to just the input circuit/impedance of the scope?

The scope input is 1Meg 30pF nominal.  6 other motley 10x probes, and a 100x, show the same generic slight undershoot characteristic.  A low capacitance Tek P6139A (8pf) shows an expected under-compensated waveform, but still shows the undershoot characteristic.

I don't have the exact scope schematic, but the closest schematic I can find (Kikusui COS5021) shows some HF compensation on the 1/10 and 1/100 input attenuators that may be an influence, but I have checked the scope input performance/waveform using two different 1MHz squarewave generators, and neither show up any abnormal scope waveform response issue no matter what the scope input attenuator settings, so I am loathe to play with any of the scope internal trimming capabilities (especially as access to that circuitry is very poor).

Ciao, Tim

[tggzzz]

Thoughts for consideration...

Probes's compensation range should encompass the scope's input capacitance.

Some probes have internal compensation networks inside the plug at the scope end of the lead. These are mainly higher performance probes, i.e. >>20MHz. What are the probes' nominal bandwidths?

I don't think your 1MHz test demonstrates anything useful, since we are concerned about LF effects.

What happens with *10 probes on your signal generator operating at 1kHz and with a correct termination (probably 50ohms or 600ohms).

What happens at different signal amplitudes and on different scope input ranges?

[David Hess]

Is that happening at the same volt/div setting when using 1x and 10x probes?  It should be apparent even though the x10 probe will produce a much smaller displayed amplitude.

If you have a variable 50 ohm RF attenuator, use it to test every volts/div setting for proper response when connected directly to a square wave calibration source like a function generator.

A x1 probe and direct 50 ohm connection should produce the same low and medium frequency display.

Also test your x1 and x10 probes against each other with the square wave output from a function generator instead of the oscilloscope's calibration signal.

[trobbins]

Thanks for responses - much appreciated.

The scope displays textbook squarewaves from separate sig gen on any scope vertical V/div, and any frequency, when using direct coax connection, and with a 50 or 600 ohm term.  Also textbook is the scope compensation signal when connected via a 1x probe (a few probes have 1x switched settings, and I also have TEK P6028). 

Any 10x probe shows a similar response when connected to either the scope 1kHz cal signal, or an external gen 1kHz squarewave - if I was to split hairs, the scope cal signal shows a slightly more pronounced level of undershoot than the external sig gen, but it is still identifiable.  This response also persists with smaller/larger signal levels from external sig gen, using smaller/larger scope volts/div ranges.  Using a decade lower frequency still shows the same initial undershoot, but obviously just for the very initial portion after each step. The effect is effectively lost above about 5kHz.

Many of the probes I have I can't find datasheets for (Konig OT 100/3, K P060-3, Hitashi AT-10AF, unbranded), but I get the same response when using a TEK P6139A 500MHz, and chinese 6100 and P4100 knockoffs.  All but one probe has its main compensation trimmer in the BNC scope end, and all probes but the P6139A achieve a trim range that gives nominal square wave response (except for the initial undershoot) and so are effectively compatible with the GW scope I have.

I am just about to go carefully through the probes to see if any tip ends allow easy access to the 9M resistor.

[CDaniel]

Probably the scope input is somewhat faulty ... at 1KHz you cold easilly compensate the probes in any scope and this is pretty standard

[tggzzz]

The scope displays textbook squarewaves from separate sig gen on any scope vertical V/div, and any frequency, when using direct coax connection, and with a 50 or 600 ohm term.  Also textbook is the scope compensation signal when connected via a 1x probe (a few probes have 1x switched settings, and I also have TEK P6028). 

Any 10x probe shows a similar response when connected to either the scope 1kHz cal signal, or an external gen 1kHz squarewave - if I was to split hairs, the scope cal signal shows a slightly more pronounced level of undershoot than the external sig gen, but it is still identifiable.  This response also persists with smaller/larger signal levels from external sig gen, using smaller/larger scope volts/div ranges.  Using a decade lower frequency still shows the same initial undershoot, but obviously just for the very initial portion after each step. The effect is effectively lost above about 5kHz.

So, the *1 and *10 probes are the effectively the same with 1kHz square wave at any signal level, and that response is textbook. (I'm assuming "shows a similar response" refers to the description in the previous paragraph).

Apart from that, it sounds as if you should describe responses in terms of time, not a signal's first harmonic frequency.

I am just about to go carefully through the probes to see if any tip ends allow easy access to the 9M resistor.

If all the probes have the same response, what makes you think they all have the same fault?

[trobbins]

tggzz, sorry if my earlier post was a bit ambiguous.  The textbook squarewave response is with direct coax connection of sig gen to scope input (ie. no probe used), and when any 1x probe is used.  The last sentances that refer to frequency are referring to the squarewave frequency, ie. when viewing a 100Hz squarewave then the initial undershoot caused by the 10x probe is still observable during the first few hundred microsecs of each 5ms flat portion.

I appreciate that the probes are not 'at fault'.  What I am trying to get a better appreciation of is whether a set of probes could be slightly tweaked to give a nominal squareware response when using this particular scope.  I have a few sets of probes that I could just assign to the scope, if that ended up the simplest solution (although on review of my probe collection it looks like an ugly task to 'break in' to any of my probe ends to gain access to the 9Meg resistor). 

This is just a mundane 20MHz analog scope, and will be quite onerous to try and check through its front-end to see if or where it may be contributing to this subtle abnormal waveform response, as I don't have an exact schematic or service manual to work with, and practical access to the part side of the pcb with power on and all connections made would be ugly.

So I am coming to a conclusion that I have no other way to progress this, and will live with the subtlety for now, and just do a sinewave amplitude response sweep to confirm there is no significant frequency response deviation when using a 10x probe.  Still, it is always worth exploring the options along the way.

[David Hess]

So there is a subtle problem with the high impedance input and not the high impedance x10 probes, and a low impedance source produces the correct results.

I have looked at the schematics and do not see anything obvious to check.  The only two adjustments are input capacitance and attenuator compensation, as usual.  Attenuator compensation is adjusted using a low impedance source like a function generator and input capacitance is adjusted using a "normalizer" as described below.  But I do not think either of these things will correct the problem.

What I might do is make a "normalizer" to test the high impedance input.  This acts as a x2 probe and is composed of a 1 megohm series resistor in parallel with a variable capacitor which is adjusted for proper compensation.  It goes between a low impedance source like a function generator and the oscilloscope input.

High impedance wideband circuits are tricky.  What occurs to me is that the circuit board or wiring between the input BNC and the switches for the high impedance attenuators might be contaminated or dirty.  I would try cleaning them with a toothbrush and some alcohol.

[trobbins]

Thanx David. 

I have attached the input circuit of the Kikusui COS5021, which appears to be correct for my scope except for the part designators.

I had cleaned/flushed (as well as I could gain access) the AC/GND/DC switch and the Volts/div switch contacts, and have visually inspected the track side of the pcb and connections from the BNC's.

I ran a set of sinewave frequency responses with various probes and setups in to just channel 1 of the scope, using a 20MHz HP3325A for signal source.  All probe combinations (direct, probe x1, probe x10 and using a few different probes) do not show any discernible signal magnitude change from 10Hz through to circa 300k to 1MHz, which indicates the subtle abnormal compensation waveform does not impart any noticeable variation in frequency response. 

This set of frequency response tests does show that the scope itself has a -2.5dB gradual rolloff to around 10MHz, with a rapid decline after 10MHz.  That is somewhat under-par for the claimed -3dB at 20MHz.  I also checked the rise time which showed about 50ns, given a <20ns spec for the HP3325A, and a claimed 18ns rating, although this observation is prone to some tolerance.  There could be some degradation of circuitry, which might show from also testing the other channel, but I think this line of testing has allowed me to characterise how well I can utilise the scope, and for now I will happily proceed on using it 'as is'.  All my probes do not show any additive bandwidth reduction when used with this scope, so they are happily blameless for now (and this process has helped recover two of my P6139A probes which had shown shorting between coax and signal wire, which was apparently a result of some swarf or metal whisker in the enclosed region where the 9Meg tip capsule screws in to the probe cable, and the short was displaced with a squirt of contact cleaner and some tapping).

[tggzzz]

All instruments are imperfect; knowing what is "good enough" and compensating for the imperfections is what is important.

The waveform's shape is a combination of the frequency and phase response.

Your measured 10MHz vs 20MHz is indeed suspicious. Added to the observation that it is the same for both channels, it might be worth looking at and tweaking the performance of the single Y-amplifier between the CRT and where the two channels are multiplexed into one signal. If there are tweaks there, then mark the original control position before tweaking, and they might benefit from turning/cleaning.

I concur with David's assessment of possible problems at the input.

[Nanoman]

Note that FR4 pcb materials have a capacitance that changes at low frequencies (sub 300kHz). This can affect attenuating probes because the input capacitance is a substantial part of the load.  It's generally referred to as hook.

Not that the same effect can sometimes be seen on the higher ranges of scopes even with x1 probes, the same nonlinear cap affects the internal attenuators.

There was an Electronics world mini article on the subject taken from the book Wideband Amplifiers by Staric and Margan.

It could be a broken protection diode, that can cause odd effects.

[macboy]

Try switching between DC and AC coupling. I don't expect it to make much difference in this case but you never know.

When adjusting the basic compensation trimmer, you look only at the first 1/3 or so of the square wave. You want this to be as flat as possible, with the sharpest corner possible without causing peaking.

Better probes will have additional high frequency compensation adjustments to perform similar adjustment on an increasingly shorter portion of the square wave top. With 20 MHz scope, you don't need to be concerned about those adjustments at all, and they wouldn't help with this case anyway.

If you were to test with a leveled sine generator, you would likely find that the frequency response rolls off gently toward the high end. Obviously, roll-off is expected, but on a 20 MHz scope, not noticeably until a few Mhz or so. Yours would be rolling off starting in the kHz range, not MHz, based on the slope of the square wave top.

Scopes are usually designed with two distinct signal paths: the DC/LF path, and a HF path. The transition is typically in the kHz range. This might lead one to believe that there is a simple gain mismatch in the two signal paths.