EEVblog #1367 - 5 Types of Oscilloscope Passive Probes COMPARED

[nctnico]

No, you are just cherry picking based on a single measurement with no control experiment to back it up (which is exactly why I initially also tested with a Z0 probe with the same attenuation so I have two different tests showing the same result).

If you look closely at the signal I recorded with the Wavepro 7300A you'll see there is a step in there as well. This is missing from your simulation because the model for an ideal transmission line you are using isn't good enough. If you copy my circuit and start adding the parasitics you'll see you get closer to what the 7300A is showing.

Again: try the 1k resistor at the end of a piece of coax!

[gf]

@nctnico, why did you do the simulation with a 75 Ohm cable? [ sqrt(0.378e-6/67.3e-12) = 74.944 ]
Of course there will be reflections when terminated with 50 Ohm.

[nctnico]

@nctnico, why did you do the simulation with a 75 Ohm cable? [ sqrt(0.378e-6/67.3e-12) = 74.944 ]
Of course there will be reflections when terminated with 50 Ohm.

I thought I'd got the parameters for a 50 Ohm cable but I see that is not the case and with this mismatch there will be reflections back into the 1k resistor.

[tggzzz]

No, you are just cherry picking based on a single measurement with no control experiment to back it up (which is exactly why I initially also tested with a Z0 probe with the same attenuation so I have two different tests showing the same result).

I have a much better control experiment than yours: I measure a pure 50ohm input and also a 50ohm//20pF input.

If you look closely at the signal I recorded with the Wavepro 7300A you'll see there is a step in there as well. This is missing from your simulation because the model for an ideal transmission line you are using isn't good enough. If you copy my circuit and start adding the parasitics you'll see you get closer to what the 7300A is showing.

Again: try the 1k resistor at the end of a piece of coax!

Let's keep our eye on the ball, not irrelevancies. The key bone of contention is whether the WavePro's input is 50ohm or 50ohm//15pF.

My model (the 50ohm//15pF) clearly demonstrates the notch seen in your waveform (and my 50ohm//20pF waveform).
Your model clearly does not demonstrate the notch.

My measurements clearly demonstrate the notch seen in your waveform - and the absence of a notch with a pure 50ohm input.

So, should we trust models or measurements? I prefer models backed up with measurements.

Over to you to produce a model that demonstrates that notch you see in your measurements....

[tggzzz]

@nctnico, why did you do the simulation with a 75 Ohm cable? [ sqrt(0.378e-6/67.3e-12) = 74.944 ]
Of course there will be reflections when terminated with 50 Ohm.

Snort!

It is noteworthy that he claims his model matches measurements when there are such gross errors!

[gf]

@nctnico, why did you do the simulation with a 75 Ohm cable? [ sqrt(0.378e-6/67.3e-12) = 74.944 ]
Of course there will be reflections when terminated with 50 Ohm.

I thought I'd got the parameters for a 50 Ohm cable. But nevertheless there is a clear difference between having a massive mismatch at the beginning of the cable versus a much smaller mismatch. So either way it doesn't really matter for the conclusion.

If the cable is perfectly terminated with its Z0 at the end, then there won't be any reflections at the end, and no Z0 termination at the beginning were necessary. If the end is mis-terminated, though, then an (additional) Z0 termination at the beginning helps of course to prevent the reflected signal being reflected back from the beginning to the end again (or at least to reduce the amplitude of the 2nd reflection if the termination at the beginning is not perfect either).

[nctnico]

@nctnico, why did you do the simulation with a 75 Ohm cable? [ sqrt(0.378e-6/67.3e-12) = 74.944 ]
Of course there will be reflections when terminated with 50 Ohm.

I thought I'd got the parameters for a 50 Ohm cable. But nevertheless there is a clear difference between having a massive mismatch at the beginning of the cable versus a much smaller mismatch. So either way it doesn't really matter for the conclusion.

If the cable is perfectly terminated with its Z0 at the end, then there won't be any reflections at the end, and no Z0 termination at the beginning were necessary. If the end is mis-terminated, though, then an (additional) Z0 termination at the beginning helps of course to prevent the reflected signal being reflected back from the beginning to the end again (or at least to reduce the amplitude of the 2nd reflection if the termination at the beginning is not perfect either).

You are right about this. Still it doesn't explain the results I'm seeing.

[gf]

You are right about this. Still it doesn't explain the results I'm seeing.

Do you have a VNA? Would you consider measuring S11 (or the impedance) of the 50 Ohm scope input? Is it purely resistive and independent of frequency?

A direct coax connection between signal generator and scope is also source-terminated by the generator, thus a mis-termination at the scope input may not matter too much. For a Z0 probe w/o source termination of the cable it certainly does matter.

Is your cable 50 Ohm (or accidentally 75 Ohm too, like in the simulation)?

[nctnico]

You are right about this. Still it doesn't explain the results I'm seeing.

Do you have a VNA? Would you consider measuring S11 (or the impedance) of the 50 Ohm scope input? Is it purely resistive and independent of frequency?

A direct coax connection between signal generator and scope is also source-terminated by the generator, thus a mis-termination at the scope input may not matter too much. For a Z0 probe w/o source termination of the cable it certainly does matter.

Is your cable 50 Ohm (or accidentally 75 Ohm too, like in the simulation)?

I could use a network analyser but that only goes to 300MHz. Cable should be 50 Ohms but since I got the same results with a Tektronix probe it is reasonable to assume the cable I used is 50 Ohms. Some further investigation is needed as I would be highly surprised (and dissappointed -not for being wrong but due to equipment not being as good as I expect it to be-) if it turns out the oscilloscope is the cullprit after all.

[gf]

I could use a network analyser but that only goes to 300MHz.

300 MHz is better than nothing. A parallel capacitance of a few pF should already show up at this frequency.

Another question: Does an external 20dB attenuator (a good one, with proper frequency rating) in front of the scope input make a difference?

[StillTrying]

I usually find that LT's lossless transmission line has bugs, even after a new OS and latest version of LT it still has them.
Changing the value of a completely unrelated V or I source changes the TL's output.

[tggzzz]

I could use a network analyser but that only goes to 300MHz.

Here are some very quck and dirty results from my NanoVna and Tek 485. They are sufficient to show that you can measure the difference at 300MHz.

Sweep is from 100MHz to 900MHz, marker at 340MHz.

The first is connected to the scope's proper 50ohm attenuator.
The second is connected to an inline 50ohm attenuator.
The third is connected to the inline attenuator load connected to the scope's 1Mohm//20pF attenuator. [EDIT: inline 50-ohm load]

It will be interesting to see your results with the LeCroy.

[2N3055]

I guess that MSOX3104T has true 50 Ohm path....





Note: This was with loaded master calibration. If I were to calibrate it would have been more accurate, but for this purpose it is enough..

[tggzzz]

I guess that MSOX3104T has true 50 Ohm path....

Yup

[gf]

The third is connected to the inline attenuator connected to the scope's 1Mohm//20pF attenuator.

Inline attenuator (if yes, how many dB?), or just inline terminator?

[radioing]

Hi all, brief theory about the Z0 probe:

                   Z0, gamma (R, L, C, G)
1k tip    Zin -> ================================ Zl (termination, oscilloscope side, 50 ohms)
                 |<-------------l ------------->|


The input impedance Zin of the lossy transmission line (nonzero R & G line parameters) is given by
Zin = Z0 * [Zl + Z0*tanh(gamma*l)]/[Z0 + Zl*tanh(gamma*l)],
where
characteristic impedance of cable Z0 = sqrt[(R + j*omega*L)/(G + j*omega*C)],
propagation constant of cable gamma = sqrt[(R + j*omega*L)*(G + j*omega*C)],
load (terminating) impedance is Zl,
l is cable length,
j is imaginary unit,
tanh is tangent hyperbolic,
and
L, C, R, G are cable parameters (per one meter),
omega = 2*pi*frequency.

And now:

1. When the cable is lossless (R = G = 0), the Zin formula can be rewritten to
Zin = Z0 * [Zl + j*Z0*tan(beta*l)]/[Z0 + j*Zl*tan(beta*l)],
where
characteristic impedance of cable Z0 = sqrt(L/C) and it is pure real number,
beta = omega*sqrt(LC) and it is also pure real number.

From this Zin formula, it results that when we terminate the lossless line with load Zl = Z0,
the impedance seen at the cable input (Zin) is exactly Z0.
As there are no reflections at the Zl (reflection coefficient is 0 if Zl = Z0), there can't be also
reflections at unmatched cable input returned toward Zl. Due this fact, any voltage measured at
the cable input will be also measured at the cable load Zl, of course delayed due to nonzero cable length.
The situation changes when the Zl is not eq. Z0 (but hopefully close). In this case, there are reflections at Zl (osci side) and there are also
reflections at the cable input (1k tip side), because the reflected signal going back through cable sees 1k + some parasitic C and L.
Note that at Zin side, there are nearly total reflection due to such impedance discontinuity.
How to improve this situation:
A. Of course, try to match Zl to Z0 (better osci etc.;-).
B. Terminate cable at Zin side so that reflections spreading from Zl see impedance close to Z0 and are not reflected back.
The solution is to plug 50 ohms resistor at the cable input. In this case, the Zin changes to 25 ohms approx., so the division ration doubles.
To increase sensitivity back, the 1k can be replaced by 470 ohms resistor (but the load of measured circuit increases).

2. When the cable is lossy (nonzero R and/or G) and loaded (terminated) by pure real impedance Zl (osci 50 ohms)
reflecting only L and C parameters of cable.
There are the reflections at the the Zl and there are also reflections at the cable input if it is also not matched (like in prev. example).
Thus, the signal measured at the Zl is jammed by multiple reflections from Zl mismatch and cable input impedance mismatch.
The situation is even worse as the parameters R and G are significantly frequency dependent and the transmission line becomes dispersive,
i.e., the signal will spread out in time.
There are too limit cases:
a. The cable is not too lossy (up to some desired frequency, e.g. cable with good dielectric, short cable,...).
Then the matching of complex Z0 and pure real Zl is good enough so that the reflections coming to Zl measurement point are small
in amplitude and does not disturb the original signal too much.
b. The cable is too much lossy.
In this case, the reflections are attenuated by cable itself (imagine cable with 10 dB att -> first reflection is attenuated by 20 dB (supposed total refl. at Zin point) -> 0.1 amplitude)
and they are also hopefully marginal at the measurement point Zl.
What can be done to dump reflections coming to Zl measurement point?
A. Match Zl to complex Z0 -> hard to do as Z0 varies with frequency.
B. Use better cable.
C. Use long cable to increase reflections attenuation -> not good idea, we lost the sensitivity.
C. Match cable at the cable input to reduce reflections of reflections from Zl.

[tggzzz]

The third is connected to the inline attenuator connected to the scope's 1Mohm//20pF attenuator.

Inline attenuator (if yes, how many dB?), or just inline terminator?

0dB Yes, I'm an idiot: it is an inline load.

Original edited to avoid more readers wasting their time. But thanks for reading it!

[gf]

The third is connected to the inline attenuator connected to the scope's 1Mohm//20pF attenuator.

Inline attenuator (if yes, how many dB?), or just inline terminator?

it is an inline load.

I just asked, because if it were an attenuator (say 20dB), then I would have expected better decoupling (i.e. that the capacitive load attached to it would not affect the input impedance so much).

EDIT: Btw, could you please also also sweep S21 of the Z0 probe (from the tip to the end of the cable - thus including all the parasitics), in order to get its frequency response, and also let the Nano calculate the corresponding time domain impulse/step response [I think the Nano can do that, can't it]? Just curious - particularly what effect the parasitics have.

[nctnico]

I did some testing with the Lecroy 7300A but didn't got much wiser. First of all it turns out my Mini-circuits directional coupler is broken and the other one I have is a cheap one from Ebay with only goes to 200MHz. With the latter the SWR line (using Anritsu MS4630B network analyser) isn't flat (from 1.00 @30MHz to 1.05 @200MHz) when connected to the 7300A's input. But I get the same result with the R&S RTB3004.

I get exactly the same trace on both the Lecroy 7300A and the R&S RTB3004 in 50 Ohm mode using two different generators. A 12dB inline attenuator doesn't change the signal shape at all on the Lecroy 7300A.

All in all nothing conclusive yet.

BTW: the NanoVNA seems like a very useful tool instead of having to pull out a spectrum analyser and RF generator to make more in-depth measurements. I'm going to order one.

[tggzzz]

The third is connected to the inline attenuator connected to the scope's 1Mohm//20pF attenuator.

Inline attenuator (if yes, how many dB?), or just inline terminator?

it is an inline load.

I just asked, because if it were an attenuator (say 20dB), then I would have expected better decoupling (i.e. that the capacitive load attached to it would not affect the input impedance so much).

Yes, exactly, and your question was very sensible.

EDIT: Btw, could you please also also sweep S21 of the Z0 probe (from the tip to the end of the cable - thus including all the parasitics), in order to get its frequency response, and also let the Nano calculate the corresponding time domain impulse/step response [I think the Nano can do that, can't it]? Just curious - particularly what effect the parasitics have.

Yes, I could - but not today.

The hp10200a spec is 1.5GHz with <0.7pF tip capacitance.

[tggzzz]

BTW: the NanoVNA seems like a very useful tool instead of having to pull out a spectrum analyser and RF generator to make more in-depth measurements. I'm going to order one.

It is worth getting one, but you have to choose which variant. The differences can be found by googling.

Mine has the annoying feature that the calibration dara is not saved across power downs. It has to be recalibrated every time it is turned on.

[gf]

For some time I'm after the V2 Plus4 model but is seems to be out of stock at the moment in all stores.

[nctnico]

I did some testing with the Lecroy 7300A but didn't got much wiser. First of all it turns out my Mini-circuits directional coupler is broken and the other one I have is a cheap one from Ebay with only goes to 200MHz. With the latter the SWR line (using Anritsu MS4630B network analyser) isn't flat (from 1.00 @30MHz to 1.05 @200MHz) when connected to the 7300A's input. But I get the same result with the R&S RTB3004.

I get exactly the same trace on both the Lecroy 7300A and the R&S RTB3004 in 50 Ohm mode using two different generators. A 12dB inline attenuator doesn't change the signal shape at all on the Lecroy 7300A.

All in all nothing conclusive yet.

I did some further investigation. It seems that the problem is not the oscilloscope input but the interaction between the signal source and the probe itself. If I put an attenuator or 50 Ohm feed-through between the Lecroy 7300A calibrator output and the probe the signal improves. Something else to keep in mind is that the Lecroy 7300A has more than 8 times the bandwidth compared to tggzzz's Tek 485 so it will show narrow spikes much larger.

[tggzzz]

I did some testing with the Lecroy 7300A but didn't got much wiser. First of all it turns out my Mini-circuits directional coupler is broken and the other one I have is a cheap one from Ebay with only goes to 200MHz. With the latter the SWR line (using Anritsu MS4630B network analyser) isn't flat (from 1.00 @30MHz to 1.05 @200MHz) when connected to the 7300A's input. But I get the same result with the R&S RTB3004.

I get exactly the same trace on both the Lecroy 7300A and the R&S RTB3004 in 50 Ohm mode using two different generators. A 12dB inline attenuator doesn't change the signal shape at all on the Lecroy 7300A.

All in all nothing conclusive yet.

I did some further investigation. It seems that the problem is not the oscilloscope input but the interaction between the signal source and the probe itself. If I put an attenuator or 50 Ohm feed-through between the Lecroy 7300A calibrator output and the probe the signal improves.

More likely "In that case the effect you see without the source 50ohm might be due to the cal out behaving properly when correctly loaded."
https://www.eevblog.com/forum/blog/eevblog-1367-5-types-of-oscilloscope-passive-probes-compared/msg3438928/#msg3438928

The Tek 485's cal out signal is most entertaining in that respect: it looks like an exponential risetime quantised into discrete levels

Something else to keep in mind is that the Lecroy 7300A has more than 8 times the bandwidth compared to tggzzz's Tek 485 so it will show narrow spikes much larger.

Any energy reflected by an improper termination will be independent of the scope's bandwidth. Thus there is an argument that the area under the spike will be the same.

[nctnico]

I did some testing with the Lecroy 7300A but didn't got much wiser. First of all it turns out my Mini-circuits directional coupler is broken and the other one I have is a cheap one from Ebay with only goes to 200MHz. With the latter the SWR line (using Anritsu MS4630B network analyser) isn't flat (from 1.00 @30MHz to 1.05 @200MHz) when connected to the 7300A's input. But I get the same result with the R&S RTB3004.

I get exactly the same trace on both the Lecroy 7300A and the R&S RTB3004 in 50 Ohm mode using two different generators. A 12dB inline attenuator doesn't change the signal shape at all on the Lecroy 7300A.

All in all nothing conclusive yet.

I did some further investigation. It seems that the problem is not the oscilloscope input but the interaction between the signal source and the probe itself. If I put an attenuator or 50 Ohm feed-through between the Lecroy 7300A calibrator output and the probe the signal improves.

More likely "In that case the effect you see without the source 50ohm might be due to the cal out behaving properly when correctly loaded."

But isn't the whole point of probing to see the signal as it is as much as possible? Requiring a signal to be 'correctly loaded' is not always possible or even desireable (think about the good old PCI bus for example as an unterminated high speed bus).