Making a differential scope probe, how tough?

[Chris Carlen]

Hi:

I've been wondering about the feasibility of making a differential scope
probe that can work as well as or better than the low-end ones on the
market. Since it would be a home-made thing, it wouldn't need to have
any certs.

I'd like to achieve these specs:

25MHz bandwidth.

Less than 2% error introduced in a 10V gate drive signal when swinging
+/-100V at moderate switching speeds of up to 100kHz. So if the rise of
the switching is 100ns or so, that's spectral content at about 3.5MHz.
Let's say CMRR of

20*log[200/(0.02*10)] = 60dB at 3.5MHz.

Let's shoot for 65dB at 1MHz and 80dB at 100kHz and below.

Input: 1 or more Megohm differential input R and =>1 Megohm common mode
input R. Less than about 13pF input C.

This CMRR spec is moderately better than the 50-60dB at 200kHz-1MHz
typical of the stuff I'm finding on the market.

No range switching, just a hardwired unit for each input division ratio.

I'd like to have an input division of 1/20 for a 100V common mode and
100V differential voltage readable range option (with tolerance of 200V
max at either input without breaking), as well as possibly a unity gain
input for low level stuff.

A second stage of amplification could offer 1x, 10x, or 100x to make it
possible to get up to about 10uV/div sensitivity on a typical DSO.

The package would be something like a Pomona box with the differential
input being perhaps just a two pin header sticking out of a hole at one
end to which short test leads could be attached. To avoid noise and
keep things simple, a pair of 9V batteries for power or perhaps a cabled
source of +/- DC voltage.

Can this be done without using a discrete differential input stage (ie.,
regular op-amps) ?

Is the bulk of the difficulty focussed on matching input parasitic
capacitances to get the CMRR? Careful layout and a few trimmer
components should make this doable, no?


Thanks for input.



--
Good day!

________________________________________
Christopher R. Carlen
Principal Laser&Electronics Technologist
Sandia National Laboratories CA USA
[email protected]
NOTE, delete texts: "RemoveThis" and
"BOGUS" from email address to reply.

 

[John Larkin]

Hi:

I've been wondering about the feasibility of making a differential scope
probe that can work as well as or better than the low-end ones on the
market. Since it would be a home-made thing, it wouldn't need to have
any certs.

I'd like to achieve these specs:

25MHz bandwidth.

Less than 2% error introduced in a 10V gate drive signal when swinging
+/-100V at moderate switching speeds of up to 100kHz. So if the rise of
the switching is 100ns or so, that's spectral content at about 3.5MHz.
Let's say CMRR of

20*log[200/(0.02*10)] = 60dB at 3.5MHz.

Let's shoot for 65dB at 1MHz and 80dB at 100kHz and below.

Input: 1 or more Megohm differential input R and =>1 Megohm common mode
input R. Less than about 13pF input C.

This CMRR spec is moderately better than the 50-60dB at 200kHz-1MHz
typical of the stuff I'm finding on the market.

No range switching, just a hardwired unit for each input division ratio.

I'd like to have an input division of 1/20 for a 100V common mode and
100V differential voltage readable range option (with tolerance of 200V
max at either input without breaking), as well as possibly a unity gain
input for low level stuff.

A second stage of amplification could offer 1x, 10x, or 100x to make it
possible to get up to about 10uV/div sensitivity on a typical DSO.

The package would be something like a Pomona box with the differential
input being perhaps just a two pin header sticking out of a hole at one
end to which short test leads could be attached. To avoid noise and
keep things simple, a pair of 9V batteries for power or perhaps a cabled
source of +/- DC voltage.

Can this be done without using a discrete differential input stage (ie.,
regular op-amps) ?

Is the bulk of the difficulty focussed on matching input parasitic
capacitances to get the CMRR? Careful layout and a few trimmer
components should make this doable, no?


Thanks for input.

Why not battery power, fully floating, and a fiberoptic link to the
receiver box on the scope? Hmmm, maybe I should manufacture those!


John

 

[Stanislaw Flatto]

Hi: Hi!

Can this be done without using a discrete differential input stage (ie.,
regular op-amps) ?

Is the bulk of the difficulty focussed on matching input parasitic
capacitances to get the CMRR? Careful layout and a few trimmer
components should make this doable, no?

Thanks for input.

NO!
[Scratching the old memory cells]
"Medical electronics" developing an input stage for ECG recorder. The
signal is in range of 0.5-5 mV on a human body noisy with about 1V line
frequency voltage. You need about 8 hands to trim all the trimmers each
time you blink your eye. So the "expensive?" solution was to go into
microelectonic solution of first amplifier stage with the input and
feedback resistors monolitic with the division ratios certified by
Vishay to x/ppm. Only then the rest of the circuit became trivial.
So if you can get hold of such a stage it will save you a lot of
frustration.
But it is your "craziness", so good luck!

"We always learn!" - a Hassidic proverb.

Stanislaw
Slack user from Ulladulla.

 

[Chris Carlen]

On Tue, 20 Jun 2006 15:22:30 -0700, Chris Carlen
Why not battery power, fully floating, and a fiberoptic link to the
receiver box on the scope? Hmmm, maybe I should manufacture those!

Your probably capable of just such a thing, which would be pretty cool.
Of course it would cost a bundle.

One question, does having an ill defined relationship to ground make the
input behavior hard to tune?

What proportion of the typical gadget's cost is composed of regulatory
compliance? Ie., UL, FCC, EMC, CE, RoHS, IEC, etc. certification costs,
and/or the costs in time or consultant fees to engineer compliance?




--
Good day!

________________________________________
Christopher R. Carlen
Principal Laser&Electronics Technologist
Sandia National Laboratories CA USA
[email protected]
NOTE, delete texts: "RemoveThis" and
"BOGUS" from email address to reply.

 

[David L. Jones]

Hi:

I've been wondering about the feasibility of making a differential scope
probe that can work as well as or better than the low-end ones on the
market. Since it would be a home-made thing, it wouldn't need to have
any certs.

I'd like to achieve these specs:

25MHz bandwidth.

That's the killer.

If you were happy with about 100KHz, Analog Devices and Burr Brown (TI)
do some nice isolation amps like this:
http://www.analog.com/en/prod/0,,759_783_AD215,00.html

Dave

 

[John Larkin]

Your probably capable of just such a thing, which would be pretty cool.
Of course it would cost a bundle.

Not really. 850 nm VCSEL lasers and silicon photodiodes are fairly
cheap.

One question, does having an ill defined relationship to ground make the
input behavior hard to tune?

The circuit would still have a local ground/Farady cage, probably a
metal box or equivalent, tucked inside a plastic case.

What proportion of the typical gadget's cost is composed of regulatory
compliance? Ie., UL, FCC, EMC, CE, RoHS, IEC, etc. certification costs,
and/or the costs in time or consultant fees to engineer compliance?

I wouldn't do any of that, and none are needed for the US market
anyhow.


John

 

[Chris Carlen]

That's the killer.

If you were happy with about 100KHz, Analog Devices and Burr Brown (TI)
do some nice isolation amps like this:
http://www.analog.com/en/prod/0,,759_783_AD215,00.html

Dave

Sometimes 100kHz is pretty useful. I've still rarely been upset by only
having a 100kHz current probe at my disposal.

It isn't sounding very encouraging about the 25MHz diff voltage probe
though.


--
Good day!

________________________________________
Christopher R. Carlen
Principal Laser&Electronics Technologist
Sandia National Laboratories CA USA
[email protected]
NOTE, delete texts: "RemoveThis" and
"BOGUS" from email address to reply.

 

[Chris Carlen]

Not really. 850 nm VCSEL lasers and silicon photodiodes are fairly
cheap.

I wasn't talking about the parts cost. I suppose if you are correct
about not needing to address compliance issues, then it might not be so bad.

The circuit would still have a local ground/Farady cage, probably a
metal box or equivalent, tucked inside a plastic case.

It is a very cool idea though. Now tell me, would it make more sense to
digitize each input separately, then subtract digitally, or actually
implement an analog differential amp, then digitize a single ended result?

What rate of oversampling would you use? I'd like to see at least 10x
if not 20x. At these speeds, it wouldn't be easy to get more than 8-10
bits. Hmm, maybe that means the analog diff front end is still needed.

Heck, even I can do an optical data link. Done one before with free
space transmission off a running engine piston!

But discrete analog and high-end stuff isn't my thing.

I wouldn't do any of that, and none are needed for the US market
anyhow.

Interesting. None are needed for the "instrument" market, or any
market, such as consumer products? Are these certs all just voluntary?

Wait a minute, I thought at least FCC part 15 or something was required
for anything with a clock running at more than some dismal speed?


Thanks for comments.




--
Good day!

________________________________________
Christopher R. Carlen
Principal Laser&Electronics Technologist
Sandia National Laboratories CA USA
[email protected]
NOTE, delete texts: "RemoveThis" and
"BOGUS" from email address to reply.

 

[Paul Mathews]

Are you wanting high voltage differential probe or logic level diff
probe...2 very different animals. Assuming the former, it's not too
difficult to do something like Tek 5200. You attenuate both inputs with
matched resistors to get the voltages down to reasonable values,
connect to an instrumentation amplifier and a line driver. Tek 5200
uses readily available Intersil ICs, if I recall correctly.
Paul Mathews

 

[Ancient_Hacker]

Less than 2% error introduced in a 10V gate drive

If you're interested in measuring the source to gate voltage, you may
be in luck.

If the source is a relatively low-impedance point, you can just use a
single-ended amplifier, that eliminates HALF of the amplifiers and
balance adjustments.

-----

Another way is to just buy an old Tek 465B with battery pack, hang it
from the ceiling with fishing-line, and just use it with a regular
scope probe, hooking the ground lead to the source. Should be less
than 10pf to ground.

Regards,

George

 

[John Larkin]

I wasn't talking about the parts cost. I suppose if you are correct
about not needing to address compliance issues, then it might not be so bad.



It is a very cool idea though. Now tell me, would it make more sense to
digitize each input separately, then subtract digitally, or actually
implement an analog differential amp, then digitize a single ended result?

If it's fully floating, there's no "differential" input; there's the
input, and there's the return, the "ground" clip, the circuit common
of the probe. And I wouldn't digitize... probably go FM, which is how
the Tek scopes do it, I think.

A differential input needs *three* leads, but a scope probe only needs
two.

What rate of oversampling would you use? I'd like to see at least 10x
if not 20x. At these speeds, it wouldn't be easy to get more than 8-10
bits. Hmm, maybe that means the analog diff front end is still needed.

Heck, even I can do an optical data link. Done one before with free
space transmission off a running engine piston!

But discrete analog and high-end stuff isn't my thing.




Interesting. None are needed for the "instrument" market, or any
market, such as consumer products? Are these certs all just voluntary?

Wait a minute, I thought at least FCC part 15 or something was required
for anything with a clock running at more than some dismal speed?

Last time I looked, test equipment is exempt, and you'd have to get
caught anyhow.

John

 

[John Larkin]

Another way is to just buy an old Tek 465B with battery pack, hang it
from the ceiling with fishing-line, and just use it with a regular
scope probe, hooking the ground lead to the source. Should be less
than 10pf to ground.

Just the capacitance of the enclosure to the universe will be a lot
more than that. And the ground side will squirt massive ammounts of
hum and noise into the circuit under test.

John

 

[Joerg]

Hello Stanislaw,

Can this be done without using a discrete differential input stage
(ie., regular op-amps) ?

Is the bulk of the difficulty focussed on matching input parasitic
capacitances to get the CMRR? Careful layout and a few trimmer
components should make this doable, no?

Thanks for input.

NO!
[Scratching the old memory cells]
"Medical electronics" developing an input stage for ECG recorder. The
signal is in range of 0.5-5 mV on a human body noisy with about 1V line
frequency voltage. You need about 8 hands to trim all the trimmers each
time you blink your eye. So the "expensive?" solution was to go into
microelectonic solution of first amplifier stage with the input and
feedback resistors monolitic with the division ratios certified by
Vishay to x/ppm. Only then the rest of the circuit became trivial.
So if you can get hold of such a stage it will save you a lot of
frustration.
But it is your "craziness", so good luck!

ECG inputs are usually fully isolated. Have to be.

 

[Joerg]

Hello John,

If it's fully floating, there's no "differential" input; there's the
input, and there's the return, the "ground" clip, the circuit common
of the probe. And I wouldn't digitize... probably go FM, which is how
the Tek scopes do it, I think.

A differential input needs *three* leads, but a scope probe only needs
two.

You could also use a couple of tiny iso transformers. One wideband for
the signal, the other to drive the DC restore.

 

[Ancient_Hacker]

Just the capacitance of the enclosure to the universe will be a lot
more than that. And the ground side will squirt massive ammounts of
hum and noise into the circuit under test.

John

oops, you're right, a one-meter sphere is around 60pf, the Earth is
around 700uF, so a Tek scope is going to be 20-40pF I'd guess.

So jsut add another kludge-- sense the source voltage, buffer it, and
feed it to a screen cage around the scope. That will take care of the
capacitance and the noise. Watch the fingers though.

 

[John Larkin]

oops, you're right, a one-meter sphere is around 60pf, the Earth is
around 700uF, so a Tek scope is going to be 20-40pF I'd guess.

So jsut add another kludge-- sense the source voltage, buffer it, and
feed it to a screen cage around the scope. That will take care of the
capacitance and the noise. Watch the fingers though.

A common-mode choke/balun can help a lot, or wrap the probe lead
around a ferrite toroid a few times. That at least improves the hf
cmrr a bit.

John

 

[Jim Yanik]

If it's fully floating, there's no "differential" input; there's the
input, and there's the return, the "ground" clip, the circuit common
of the probe. And I wouldn't digitize... probably go FM, which is how
the Tek scopes do it, I think.

A differential input needs *three* leads, but a scope probe only needs
two.


Last time I looked, test equipment is exempt, and you'd have to get
caught anyhow.

John

If you just want a floating scope input,get a TEK A6902.
Or order the service manual for the P52XX and build one yourself the way
you want.

 

[Stanislaw Flatto]

Hello Stanislaw,


ECG inputs are usually fully isolated. Have to be.

Nothing to do with measuring the signal and CMRR. Safety = 50 uA (
that's my keyboard spelling micro) at line frequency can (does NOT have
to ) throw the heart into pachanga dance (medically called fibrillation).

This part of medical electronics many times IS the major time consuming
part of designing an intensive care room equipment.
There are NO excuses.

Have fun

Stanislaw
Slack user from Ulladulla.

 

[Rich Grise]

On 21 Jun 2006 13:21:23 -0700, "Ancient_Hacker" <[email protected]>

^^^^^^
This rhymes with "sludge". It's KLOOGE! >:-[

A common-mode choke/balun can help a lot, or wrap the probe lead around
a ferrite toroid a few times. That at least improves the hf cmrr a bit.

The times I've needed differential probing, I just used the 454, with two
of the same model of probes, and put the vertical on A-B.

Cheers!
Rich

 

[Keith]

On 21 Jun 2006 13:21:23 -0700, "Ancient_Hacker" <[email protected]>

^^^^^^
This rhymes with "sludge". It's KLOOGE! >:-[

Nope: http://www.m-w.com/dictionary/kludge

The times I've needed differential probing, I just used the 454, with two
of the same model of probes, and put the vertical on A-B.

That works for *really* slow signals where CMRR isn't an issue.