Probe into probes. What's up?

[bdunham7]

I don't know what terminating load the Bodnar pulser requires but the conditions for verifying probe performance traditionally calls for a 25 source (ref. Tektronix Circuit Concept series, Oscilloscope Probe Circuits, p27). This source impedance is typically made up of the pulse generator source impedance of 50 paralleled by a 50 through termination. So even if you are not using a interposing cable the output of the pulse generator still requires a terminator.

If you were characterizing the probe and wanted to fairly assess whether it met its specifications for BW, etc, this would be correct.  However, if you are just probing some circuit, you have to take it as it comes.  My Tek P6139A looks entirely different then BillyO's, so I can't demonstrate with that.  However, if you try it with a P6156 500R 10X, it looks good either way and just the amplitude changes.  The reason 25R matters is that with 8pF and 500MHz, you get a calculated Z of only 40R, and of course the actual impedance of the probe is fairly complex at that point.  I doubt that this signal is going to magically clean itself up with an extra 50R terminator, but I'll wait and see the result.

[BillyO]

Results edited into reply #24.

[bdunham7]

Results edited into reply #24.

That certainly made a lot of difference, and if you were looking to measure the rise time (characterizing re a spec) it should improve.  But is it really better or more accurate or do you just dislike the aberrations less?

Here is a P6156 500R 10X and a P6139A (doesn't look like yours, has a thread-on attenuator and can't use a BNC adapter at all, so all probing done by hand)  without an extra terminator and then with one.

[tautech]

Here is a P6156 500R 10X and a P6139A (doesn't look like yours, has a thread-on attenuator and can't use a BNC adapter at all, so all probing done by hand)  without an extra terminator and then with one.

Nicely captured SMPS ripple on the max and min. 

LED lighting ?

[bdunham7]

Nicely captured SMPS ripple on the max and min. 

LED lighting ?

No, my LED lighting is now completely quiet.  Besides, if that's noise it has got to be up around a GHz or so given that I'm at 20ns/div.  Those are all single-shots, so you see stuff you might not otherwise.  Perhaps it is internal, maybe ADC noise?

[BillyO]

That certainly made a lot of difference, and if you were looking to measure the rise time (characterizing re a spec) it should improve.  But is it really better or more accurate or do you just dislike the aberrations less?

I think both to some extent.  That document on probe circuits mentioned above states several times that Tek probes are spec'd and calibrated using a 25 ohm source and they make a point of mentioning finding low impedance probe points for HF signals.

It's not really fair to compare a P6139 to the P6156 for rise time and susceptibility to source impedance, however I am scratching my head as to why your 6139 shows practically no degradation in rise time with the 50 ohm source.  What source are you using (not that that would matter)?  Also, what other modifications were done to you 6139?

Since the scope itself would not be much effected by the source impedance through a 9M ohm probe, and it is responding properly to the pulsers when directly connected, then any aberrations must be in the probes.  But they do behave fairly well when applied as suggested by Tek.

In any case, the document discusses the circuit used in the 6139 on page 23 and gives an outline of the adjustment procedure so I now have a fairly good place to start with and see if I can improve things.  However, I feel fairly certain I'm not going to get what you are seeing with you 6139 using a 50 ohm source.

[bdunham7]

It's not really fair to compare a P6139 to the P6156 for rise time and susceptibility to source impedance, however I am scratching my head as to why your 6139 shows practically no degradation in rise time with the 50 ohm source.  What source are you using (not that that would matter)?  Also, what other modifications were done to you 6139?

In any case, the document discusses the circuit used in the 6139 on page 23 and gives an outline of the adjustment procedure so I now have a fairly good place to start with and see if I can improve things.  However, I feel fairly certain I'm not going to get what you are seeing with you 6139 using a 50 ohm source.

I'm using the same Bodnar pulser.  My 6139A is older than yours, has the threaded attenuator and is not made in China.  It has not been modified that I know of.  My guess is that the degradation doesn't get much worse because there isn't as much in the first place.

I don't know if you can get the same results I do.  David Hess suggests that Tek is juicing the response of these to improve HF response of particular oscilloscopes they were used with, which in your case would be the TDS3000 series if I have that right.  By 'juicing the response' I mean deliberately loading the circuit more at HF in order to make up for losses elsewhere.   So in that case, the source impedance would have an even larger effect than might normally be expected. 

[jonpaul]

Rebonjour a tous...

very Fine thread, probes are one of my favorite tools..

At >100 MHz the probe C dominates the impedance and thus the 10X high Z  probes are not so good. 

1/ For 95% of our   P6137 TEK are best. We found some NIB in orig case with all tips/clips/adapters.

2/ At >100 MHz the Zo probes like P6136 are excellent, despite the 500 Ohm..5K load, only  use on 50 Ohm input scopes.

3/ My Easy and cheap wideband DIY:

Thin 50 Ohm  coax Rg175/U or similar , .5..1.5 M
Term to BNC one side
Use  with short shield direct 
OR insert series 450 or 4950 R at cable shield/tip to make a DIY Zo probe.

4/ The TEK calibrate with the comp trim with a 1 kHz cal signal. Beware some scopes change CAL freq with time base (485?)

5/ Active probes (diff/FET) are very rarely needed and can be easy to damage physically or electrically

6/ In SMPS/EBU work,  current is often more useful than voltage info.
We provide small bus loops with Teflon tube insulation on a proto or SMPS PCB.

7/Using  P6021, P6022, and the DC Hall effect probes/amps P6302, 6303, AM502B.
A fine passive current transformer  is the very wideband Pearson CT, 410, 411, The  very best and widest freq resp and current capability.
Use a wire passed thru the toroid...

Hope this is interesting!

Happy probing!

 
Jon

[BillyO]

Could it be that my Bodnar is faulty, or at least borderline?

Even if I feed it directly into the scope, the signal into 1M is very much different than into 50.  See the shots below - 1M first, then 50.  You can see the 1M rise time is many times longer, and there is that shift 66% of the way up the rising edge.

@bdunham7, what does yours do in the same situations?

[jonpaul]

used Leo Bodnar 40ps pulser for years.

Rated only with 50 ohm load

No spec for open circuit load

Ask Leo,

Jon

[bdunham7]

Here is 1M and 50R superimposed at the same vertical setting.  The slew rate and rise times are similar until the 1M rise slows down.  This is a function of the scope, the Bodnar pulser is fine.  The pulser itself does not need 50R termination to put out a clean square wave, as I have demonstrated with the P6156 or could be demonstrated by using an active-FET probe. 

This is a good discussion to demonstrate the difficulty, or even futility, of trying to use passive Hi-Z probes above 200MHz or so.  The Hi-Z probe will put a complex load on the source, causing the actual signal at the probe tip to be altered--which also can be demonstrated by using a P6156 or active-FET in parallel.  HF probes specific to scopes will attempt to compensate for this to some extent so as to give the appearance of an enhanced system reponse.  The dilemma is then to decide which lie you would rather be told--do you want the probe/scope system to reflect the actual waveform including the alterations caused by the probe load itself or would you like it to attempt to 'compensate' and attempt to show you what would be there if the probe were not in the system?

[David Hess]

The dilemma is then to decide which lie you would rather be told--do you want the probe/scope system to reflect the actual waveform including the alterations caused by the probe load itself or would you like it to attempt to 'compensate' and attempt to show you what would be there if the probe were not in the system?

Manufacturers take that into account, but Tektronix and HP used different methods.  I forget which was which, but one calibrated their probes to show the signal at the probe tip with the probe loading, and the other calibrated their probes to show what the signal would look like without the probe.

[jonpaul]

bdunham7, David H,  very fine notes.

No pulse gen, Leo B 40ps, Tek PG506, PG502 or the many HP high speed pulse gens  will supply the specified output without a 50 Ohm load.


The problem  of high Z probes and FET probes in wideband work is 

a/ Ground return lead inductance (must use a Bayonet probe tip or adapter)

b/ Xc of probe capacitance at the desired frequency.

Those two factors are slight at low F but enormous above 100 MHz.

Here is a  calculation of Xc  impedance of the probe tip C

1/   X10 probe, 15 pF probe tip
Xc = 1/2pi FC
 200 MHz
Xc = 1/(15 exp -9 * 6.28 * 200 exp 6) = 50 Ohms

2/  FET probe, C= 3pF
Xc = 280 Ohms.

Thus the entire raison d'etre, for the X10 or high Z probe is in question due to  capacitance Xc.

For accurate wideband  scope use,  we use  either a Zo probe (P6136) into a 50 Ohm scope,

OR make a 50 Ohm coax RG174/U direct to a 50 Ohm ckt or with 450/4950 series R for Zo probe.

Just the thoughts of an old retired EE, using TEK scopes and probes since 1967!

Have an ABSOLUTELY FANTASTIC day!

Jon

[bdunham7]

No pulse gen, Leo B 40ps, Tek PG506, PG502 or the many HP high speed pulse gens  will supply the specified output without a 50 Ohm load.

But I showed it doing just that in my earlier reply #27....

2/  FET probe, C= 3pF
Xc = 280 Ohms.

3pF is better than 8pF, but a high-BW FET probe will be even lower.  <1pF should give you similar results to  the P6156 even without termination.

[BillyO]

While certainly no P6156 I have always made my own Z₀ probes.  Cheap and cheerful.  I usually solder them directly to the DUT, but in this case I used a BNC connector.

This one is 1m long, 500R.  It looks like it's suffering from a little bit of reflection.

[gf]

This one is 1m long, 500R.  It looks like it's suffering from a little bit of reflection.

You can try a 500R/56R divider instead of the 500R series resistor. If there is a load mismatch at the scope end of the cable, then an additional source termination of the cable helps to dampen repeated back and forth reflections. You can also check the scope's 50 Ohm input with a VNA to verify that it is really 50 Ohm and ohmic up to say 1GHz.

[bdunham7]

It looks like it's suffering from a little bit of reflection.

Even the P6156 has a bit of ruffling in the first part of the signal which is related to the length of the cable and so on.

Here is another comparison, the scope set for 50R vs the scope set for 1M with a 50R pass-thru terminator on the pulser.

[Njk]

Sometimes, it's not only the pulse edge fidelity that matters. For floating measurements in an LV applications, I'd usually powered the scope from the battery (TDS3BAT), because using the math function for two channels would be less accurate (all the errors doubled). What is less expensive now, a battery or a super-duper floating probe?

[jonpaul]

Njk:

Wideband fast measurements (as in the posts above) are seldom done on battery scopes.

In power/mains/HV work, indeed a battery operated scope is useful

The differential probes are costly and seldom needed.
In Over 50 years of work we used a simple setup:

mains>>variac>>1:1 isolation transformer

Usually a Staco or GenRad 3..20A variac and   Signal 2..5 kVa 120/240:120/240 transformer.

Then a normal grounded scope and NO diff probe.

Jon

[Njk]

mains>>variac>>1:1 isolation transformer

Usually a Staco or GenRad 3..20A variac and   Signal 2..5 kVa 120/240:120/240 transformer.

Sure, but all that stuff can form a bulky setup.

I recall in a POTS line connected devices, the isolation strength requirement (between the line side and the SELV side) is of kV units, because of possible lightning strikes, etc. So in telephone modems, the common approach was to use a transformer. Later, the chips were introduced (pioneered by Silicon Labs, IIRC) that were using a capacitive digital link over an isolation barrier to communicate between the line part and the SELV part within the chip. And the last modem designs were transformer-less. I was thinking the same approach could be used to design a floating probe. Not of GHz BW, of course. A diff probe is totally different thing as it requires ground connection.

[BillyO]

I had some idle time today and decided to do some bode plots on the tek and no-name probes.  A bit pointless, I know, but ..  I only have an "improved" SDG1032X so could only go to 60MHz.  As expected they did just fine.  The behavior at the end is due to the sig. gen.

Top picture is the Tek, next the no-name and the last is the SDG1032X by itself.

[David Hess]

Wideband fast measurements (as in the posts above) are seldom done on battery scopes.

But battery powered oscilloscopes commonly have galvanically isolated input channels, making them particularly useful for wideband isolated measurements on the primary side of off-line designs.  The switching waveforms on more recent off-line switching power supplies have transition times in the nanosecond range requiring 100+ MHz measurement performance.

The big advantage they have over using a high voltage differential probe is an essentially infinite common mode rejection ratio.

So in telephone modems, the common approach was to use a transformer. Later, the chips were introduced (pioneered by Silicon Labs, IIRC) that were using a capacitive digital link over an isolation barrier to communicate between the line part and the SELV part within the chip. And the last modem designs were transformer-less. I was thinking the same approach could be used to design a floating probe. Not of GHz BW, of course. A diff probe is totally different thing as it requires ground connection.

It has been done at various times.  The Tektronix A6902 is a good example which is fully documented, and there are modern faster probes.

[Bud]

mains>>variac>>1:1 isolation transformer

Folks in the other thread must be getting a heart attack caused by your post

https://www.eevblog.com/forum/chat/do-you-guys-use-an-isolation-transformer-for-your-labworkbench/msg4675405/#msg4675405

[bdunham7]

But battery powered oscilloscopes commonly have galvanically isolated input channels, making them particularly useful for wideband isolated measurements on the primary side of off-line designs.  The switching waveforms on more recent off-line switching power supplies have transition times in the nanosecond range requiring 100+ MHz measurement performance.

Yes, something like my TPS2024 has better real performance than even my A6909 isolator, let alone modern discount differential probes.  Here is a shot of me measuring the output of the Bodnar pulser using a P6139 (not properly compensated, but it matters very little) with just my fumbly hands and a regular ground lead--I didn't bother putting the ground collar thingy on it. 

[David Hess]

Yes, something like my TPS2024 has better real performance than even my A6909 isolator, let alone modern discount differential probes.

Tektronix took advantage of their CCD based sampling technology to make a small and low cost sampling front end for that series of DSOs.  Once the high speed sampling is complete, the analog samples are convert at low speed and the digital data is transferred across the isolation barrier.

An all digital design would have required a high speed digitizer and pile of fast SRAMs for each channel making it much more expensive.  Tektronix had DSOs like that, but they were not portable and never had isolated channels.