Lo-Z Probe

[Mark_O]

I picked up some new old stock of LeCroy PP061 resistive probes that I have no documentation for.  These were very inexpensive.

Thanks for posting the VNA tests (a year ago!).  Interesting stuff, especially the "bumpy ride" in the amplitude domain.  This response tends to be more typical of homemade devices, but considering how old these are, may have been the norm back then.

(The first time you mentioned the probes in the video, you referred to them as PP066:  "I also purchased some PP066 probes."  Later on you did identify them as PP061 though.)

Re: the PP061, all I can tell you is they were phased out by Feb'99, and replaced by the PP062 probes.  Their characteristics don't seem to vary much from the story your VNA told you (500-ohm, 1.5pF, 1 GHz BW rating).  See the attached chart...

I noticed earlier in 1999, they still had the PP063 in their passive probe lineup, but it had been dropped by 2000.  That one was also a 500-ohm unit, but had an 8 GHz BW, made possible by a <0.5 pF capacitive loading.

[Archaeological digging courtesy of the Wayback Machine.]

[mich41]

I wonder how such setup deals with the capacitance of the scope input? With 1k resistor, even few pF translates to time constant of few ns, which doesn't look good. Am I missing something?

[rfeecs]

High speed probes of all types:
http://www.ggb.com/

And of course:
https://www.cmicro.com/products/probes/rf-microwave

I know, not cheap and not home made.  But the 50 ohm ones are pretty simple.  Just basically a tiny length of coax.

[joeqsmith]

I picked up some new old stock of LeCroy PP061 resistive probes that I have no documentation for.  These were very inexpensive.

Thanks for posting the VNA tests (a year ago!).  Interesting stuff, especially the "bumpy ride" in the amplitude domain.  This response tends to be more typical of homemade devices, but considering how old these are, may have been the norm back then.

(The first time you mentioned the probes in the video, you referred to them as PP066:  "I also purchased some PP066 probes."  Later on you did identify them as PP061 though.)

Re: the PP061, all I can tell you is they were phased out by Feb'99, and replaced by the PP062 probes.  Their characteristics don't seem to vary much from the story your VNA told you (500-ohm, 1.5pF, 1 GHz BW rating).  See the attached chart...

I noticed earlier in 1999, they still had the PP063 in their passive probe lineup, but it had been dropped by 2000.  That one was also a 500-ohm unit, but had an 8 GHz BW, made possible by a <0.5 pF capacitive loading.

[Archaeological digging courtesy of the Wayback Machine.]

  If I only make one mistake per video, I'm doing pretty good.  In most cases, there are enough breadcrumbs to figure it out like the title in this case.   

The oldest LeCroy catalog I have is from 1990 and they are not listed there.   They were brand new in the package so I couldn't pass them up for the price. 

[Gyro]

I wonder how such setup deals with the capacitance of the scope input? With 1k resistor, even few pF translates to time constant of few ns, which doesn't look good. Am I missing something?

The 1k resistor is feeding directly into a 50 ohm cable, terminated at the scope by a 50 ohm terminator. The whole thing is a transmission line - edges propagate along the line at some fraction of the speed of light and are absorbed by the terminator at the scope end (google a reference on transmission line theory). As a result the signal at the 1k resistor never 'sees' the input capacitance of the scope (or the much higher lumped capacitance of the cable itself). It will of course see the capacitance at DC or slow moving signals, but the RC time constant has no consequence at those speeds.

To get flattest HF response it is sometimes necessary to ADD a little parallel capacitance at the input end of the cable to compensate for the (tiny) parasitic capacitance of the 1k resistor itself. The resistor can be any value, within reason, limited only to zero at low end (50 ohm probe) to so high that the attenuation ratio becomes silly, or its parasitic capacitances becomes the dominant factor.

[mich41]

Oh, thank you for pointing me at the resistor "not seeing" this capacitance. I completely forgot that RC filter relies on R "knowing" the voltage across C in real time 

This being said, I know this capacitance simply must have some bad effect on performance Flattened edges in one-shot transitions? Inverted reflections causing FR ripple? I think somebody reported the latter, blaming it on bad terminating resistor.

edit:

To get flattest HF response it is sometimes necessary to ADD a little parallel capacitance at the input end of the cable to compensate for the (tiny) parasitic capacitance of the 1k resistor itself.

So same thing as compensation of the garden variety probes, basically.

[tggzzz]

This being said, I know this capacitance simply must have some bad effect on performance Flattened edges in one-shot transitions? Inverted reflections causing FR ripple? I think somebody reported the latter, blaming it on bad terminating resistor.

Yes. The effects can be easily seen using a simple spice simulation (an ideal transmission line model will give a slightly pessimistic result), or via a Smith Chart, or via numerous online VSWR calculators. If the scope's input capacitance is 15pF, then model the termination as 50ohm//15pF

[Gyro]

Oh, thank you for pointing me at the resistor "not seeing" this capacitance. I completely forgot that RC filter relies on R "knowing" the voltage across C in real time 

This being said, I know this capacitance simply must have some bad effect on performance Flattened edges in one-shot transitions? Inverted reflections causing FR ripple? I think somebody reported the latter, blaming it on bad terminating resistor.

Theoretically there 'should' be no reflection to worry about. In practice nothing is perfect, I use a through-terminator, which is much better than a T connector and BNC terminator, but hopefully worse than a scope with internal 50R termination (not sure if this is actually the case on modern low cost ones). There will always be some slight discontinuity in practice, but there is on a normal passive probe too. Remember at this point, the time constant you're talking about with the scope input capacitance is against 50R, not 1k (<1ns). There is no time constant associated with the cable distributed capacitance as this, coupled with distributed inductance  is what makes up its characteristic impedance. 

The other imperfect area of course is the probe tip, where the resistor must couple into the end of the 50R coax. Parasitics both capacitive  and inductive of the resistor and coupling. It is these factors that are the reason that joeqsmith and others (see the linked articles early in the tread) have measured ripple at high frequency, and tend to limit its use to 1-2GHz (where everything struggles anyway apart from a direct 50R connection).

All that said, it still tends to work better than a passive probe at high frequencies. True, it has lower input impedance at low frequency, but at 100MHz the 8pF capacitance of the Tek P6139A (a 'good' 500MHz probe) brings its input impedance down to 200R, while the Zo probe is still up at nearly 1k.

I tried both probes (the Zo and the Tek) at the same time on the output of a Williams type avalanche pulse generator, both give similar traces but if I disconnect one or the other, the P6129A tip is clearly distorting the pulse generator output a lot more than the Zo one, the same applies to logic outputs. As I say, I make no claims about the flatness of the ones that I built (others have certainly built and characterized better ones).


Time for a question from me... In Douglas Smith's implementation, http://www.emcesd.com/1ghzprob.htm why does he include 50R termination (the 4x200R smd resistors) at the probe end of the cable? I realize it's a double parallel termination of the cable (presumably there's 50R termination at the scope end too), but what benefit does it give in this application, given that it seems to work pretty well without? One downside is obviously the increased attenuation for any given tip resistor. Confused.

[macboy]

Time for a question from me... In Douglas Smith's implementation, http://www.emcesd.com/1ghzprob.htm why does he include 50R termination (the 4x200R smd resistors) at the probe end of the cable? I realize it's a double parallel termination of the cable (presumably there's 50R termination at the scope end too), but what benefit does it give in this application, given that it seems to work pretty well without? One downside is obviously the increased attenuation for any given tip resistor. Confused.

I'm just thinking out loud here, but here is my take.
You see that rippled pass-band frequency response in Ivan's post: https://www.eevblog.com/forum/projects/lo-z-probe/msg801477/#msg801477 ?
That is caused (I assume) by a slight mismatch between the 50 ohm termination R_T and the characteristic impedance of the cable Z₀. Here is why: Any mismatch results in a bit of reflection. Let's say 5% of the energy is not terminated but is reflected (it is reflected either at the same polarity or opposite depending on whether R_T is higher or lower than Z₀). That reflected wave travels back to the tip, where it encounters almost no termination (the 1 kOhm or so is similar enough to an open cable end), so most (~95%) of the energy reflects again back to the scope side. This adds to the signal seen by the scope. For a pulse, you will see a time-delayed glitch (small step up or down in the top of the pulse depending on over- or under-termination), but for a repetitive signal, especially a sine wave, this adds at some phase that will either increase or decrease the amplitude. This causes the bumpy frequency response, as this reflected signal slides through in-phase to out-of-phase to in-phase again as frequency changes. The spacing of the bumps is determined by the length of the probe cable. Now consider the effect of adding that extra terminator at the probe end of the cable. Now the probe tip will absorb most of the reflected wave instead of reflecting most of it back again toward the scope. So at the scope end you might see 5% of 5% of the original wave as a total reflection, or .25% total. Without the extra terminator you might see 95% of 5%, or basically 5%. The double terminated case will have much less passband frequency response ripple.

[Gyro]

Your aloud thinking seems entirely plausible, thanks. It does sound like the only reason to parallel terminate the source end. It's a shame that Douglas doesn't have any plots easily accessible on his site. I'm a bit surprised that he only 'specifies' his probe up to 1GHz given the improved termination, though he does talk about component limitations.

A  quandary now, I'm actually in a position to be able to do this on my probes, using 0603 resistors, It's a question of whether it's worth the trade-off of higher attenuation. A no-brainer I suppose for a SA, not as clear cut for a scope.

[Ivan7enych]

Time for a question from me... In Douglas Smith's implementation, http://www.emcesd.com/1ghzprob.htm why does he include 50R termination (the 4x200R smd resistors) at the probe end of the cable? I realize it's a double parallel termination of the cable (presumably there's 50R termination at the scope end too), but what benefit does it give in this application, given that it seems to work pretty well without? One downside is obviously the increased attenuation for any given tip resistor. Confused.

Idealy, both ends of coax cable should be properly terminated to minimize reflections on both ends of the cable.
The scope end should have 50om input resistance.
The probe end should be behave like an "ideal voltage source" + 50om resistance in series.

If 50om termination of the scope is ideal in all freq range (no reflection at all), you don't need proper termination on probe end. But in my case it isn't true, my analyzer input resistance is not perfect and reflects some part of signal back to the cable.

I've added 56om resistance on probe end of coax, so my probe tip is actualy a resistance divider, 560om + 56om, this divider output resistance is close to 50om. As a result, it almost removes any ripple in 0..1GHz range. (blue trace) Compare this trace with my previous one, with much higher ripple on lower range -
https://www.eevblog.com/forum/projects/lo-z-probe/msg801477/#msg801477

Adding a piece of metal foil around resistors partialy decreases sensitivity on >2GHz range to normal value. (pink trace)

[Gyro]

Thanks Ivan, that's an impressive ripple reduction! I'm surprised that the ripple extends right down to the lowest frequencies in your previous plot (I should have looked more closely), I was assuming it was only a high frequency problem for some reason.

It's a shame there's no way of assessing the quality of the scope end termination without an analyser - I would assume that the termination in your analyser is actually pretty good in comparison, so some level of reflection is inevitable.

It looks as if I may have to mess up my gold plating a bit then (though the added attenuation is still a concern).

Chris

P.S. I'm impressed with the SMD construction, that must have hurt a bit!

[Gyro]

Really scraping the bottom of the barrel now in terms of keeping the attenuation down...

I'm wondering if AC termination (ie. 50R + C) at the probe end would be sufficient to minimize the ripple. The reflection time should be directly related to the cable length, so calculating the minimum time constant isn't difficult. With careful tuning it ought to be possible to clean up the edges without reducing the overall attenuation ratio....maybe 

Just a thought anyway.

[Mosaic]

Shooting the breeze perhaps but:
Why not build in that SMT attenuator onto the BNC connector crimp at the scope end. The result is a Pi pad configuration with  a 56? SMT 'short' then a 560? series for attenuation and a 50? 'short' by the scope termination. Thus the probe source sees about a 51? end termination as does the TDR from the scope termination. Net 20X signal to sample ratio.

Then just use the 50 ohm coax as the probe, thus as the probe end needs refreshing from soldering etc, you just trim.
Now, since the ripple in play has to do with the leakage TDR beating on the signal, it's going to have less effect as the TDR induced ripple is gonna be at a high freq. and naturally attenuated (becomes not visible) by the scope or SA analyzer bandwidth up to perhaps 3Ghz.
So instead of about a 205Mhz TDR beat freq over a 1M cable you'll have a 6.8Ghz TDR beat between the scope and the smt attenuator (30mm away). Thus the reflected attenuation of the beat interference is 30x more over the same time constant. (30dBmV down).

[Zero999]

I've recently made a low-Z probe for my Chinese 'scopes (OWON and Rigol). Unfortunately 100MHz 'scopes, they don't have a 50R input impedance option, so I connected two 100R resistors in parallel, right next to the BNC connector on the 'scope's input. I didn't have a 450R resistor so used 270R & 180R in series at the probe end.

It works quite well, although as it's just a 100MHz 'scope, I have no way of testing it at higher frequencies. Assuming the 'scope input is just 10pF and the 450R 50R  divider has an output impedance of 45R, the -3dB point should be around 350MHz, which is more than good enough.

I wonder if 75R co-axial cable and a 75R termination resistor near the 'scope would be any better? A 750R input impedance won't load the device under test so much but the -3dB point will be lower: probably around 236MHz, still good enough for a 100MHz 'scope.

[Mosaic]

Well I think the purpose of a Lo Z probe IS to load the DUT with a matched impedance for HF work.

Why would u need a Lo Z probe otherwise?

[c4757p]

Low capacitance.

[Zero999]

Well I think the purpose of a Lo Z probe IS to load the DUT with a matched impedance for HF work.

Why would u need a Lo Z probe otherwise?

No, the purpose of a lo-Z probe is to load the device under test as little as possible: it may have a lower impedance at DC but it has a lower capacitance than a hi-Z probe, so will have a much lower impedance at higher frequencies.

[T3sl4co1l]

A "Lo-Z" probe is used when it has a higher Z than a conventional probe.  A 10pF probe has less than 500 ohms reactance at only 30MHz.

It's also used on naturally low impedance circuits, for example RF amplifiers with 50 ohm system impedances: a 500 ohm load diverts only 10% of the voltage, which is only -0.82dB, hardly noticeable.

Tim

[Gyro]

No, the purpose of a lo-Z probe is to load the device under test as little as possible: it may have a lower impedance at DC but it has a lower capacitance than a hi-Z probe, so will have a much lower impedance at higher frequencies.

I think you have a typo there, I think you meant ".... , so will have a much higher impedance at higher frequencies".

[nctnico]

A "Lo-Z" probe is used when it has a higher Z than a conventional probe.  A 10pF probe has less than 500 ohms reactance at only 30MHz.

It's also used on naturally low impedance circuits, for example RF amplifiers with 50 ohm system impedances: a 500 ohm load diverts only 10% of the voltage, which is only -0.82dB, hardly noticeable.

Still in some cases 500 Ohms is too much of a load. In one of my current RF projects I added several 1:50 passive divider test points in the prototype in order to keep the load on the signal low. When measuring LVDS signals a 500 Ohm load is definitely too much so I also have 1:100 Lo-z probes but the attenuation is becoming troublesome. At some point you'll need active FET probes.

[Mosaic]

For RF work ,perhaps directional couplers (with attenuation for power RF) would be the better way to look at signals.

[Mosaic]

A "Lo-Z" probe is used when it has a higher Z than a conventional probe.  A 10pF probe has less than 500 ohms reactance at only 30MHz.

It's also used on naturally low impedance circuits, for example RF amplifiers with 50 ohm system impedances: a 500 ohm load diverts only 10% of the voltage, which is only -0.82dB, hardly noticeable.

Tim

20log.9 = -0.91dB , not so?

[Howardlong]

A "Lo-Z" probe is used when it has a higher Z than a conventional probe.  A 10pF probe has less than 500 ohms reactance at only 30MHz.

It's also used on naturally low impedance circuits, for example RF amplifiers with 50 ohm system impedances: a 500 ohm load diverts only 10% of the voltage, which is only -0.82dB, hardly noticeable.

Still in some cases 500 Ohms is too much of a load. In one of my current RF projects I added several 1:50 passive divider test points in the prototype in order to keep the load on the signal low. When measuring LVDS signals a 500 Ohm load is definitely too much so I also have 1:100 Lo-z probes but the attenuation is becoming troublesome. At some point you'll need active FET probes.

I am wondering, without wishing to tell grandma how to suck eggs ;-) did you try AC coupling? Certainly a 500 ohm dc coupling referenced to ground rather than the common mode offset voltage will be fairly nasty on LVDS.

But I agree, a fet probe, even better a differential fet probe, is probably the instrument of choice in this situation.

[T3sl4co1l]

A "Lo-Z" probe is used when it has a higher Z than a conventional probe.  A 10pF probe has less than 500 ohms reactance at only 30MHz.

It's also used on naturally low impedance circuits, for example RF amplifiers with 50 ohm system impedances: a 500 ohm load diverts only 10% of the voltage, which is only -0.82dB, hardly noticeable.

Tim

20log.9 = -0.91dB , not so?

500/550 = 0.9090..
Although on a doubly matched 50 ohm line, the Thevenin is 25 ohms, so the ratio is 0.95ish.

Still in some cases 500 Ohms is too much of a load. In one of my current RF projects I added several 1:50 passive divider test points in the prototype in order to keep the load on the signal low. When measuring LVDS signals a 500 Ohm load is definitely too much so I also have 1:100 Lo-z probes but the attenuation is becoming troublesome. At some point you'll need active FET probes.

Well, I did specify 50 ohms...

Inner networks of RF circuits frequently go into the kohms (e.g., many crystal, ceramic and SAW filters are around 1.5kohms), which of course aren't good points to probe.  Often you'll need to add a buffer (which might be just an emitter follower) to get useful probing.

Buffering in the circuit is even better than FET probes.

Tim