High voltage, large bandwidth instrumentation amplifier

[Bob]

I did a project several years ago where I needed an instrumentation
amplifier that had these basic characteristics:

- DC to 10MHz (-3dB) bandwidth
- Common-mode input max of +-200V with respect to earth
- Differential input max of +-100V
- Differential and common-mode input impedance >10Kohms
- Common-mode rejection > 60dB over its bandwidth for signals with up to 1k
ohm source impedance
- Output will drive an ADC whose supplies will be referenced to earth but
will be noisy with respect to earth (5v peak max from DC to 10MHz)


I had looked at several techniques including some discrete solutions and
some canned solutions from Apex Microtech and Burr Brown (now TI). These
involved using expensive laser-trimmed input networks.

However, with these solutions, I could not get the bandwidth, common-mode
rejection, and/or input impedance that I needed.

I ended up using a tranformer-coupled design, so of course I lost my low
frequency capability. This was okay but not desirable.

Any ideas on a solution that would meet or come close to the listed specs?

Thanks much,
Bob

 

[Fred Bartoli]

Le Thu, 16 Aug 2007 15:47:08 -0700, Bob a écrit:

I did a project several years ago where I needed an instrumentation
amplifier that had these basic characteristics:

- DC to 10MHz (-3dB) bandwidth
- Common-mode input max of +-200V with respect to earth - Differential
input max of +-100V
- Differential and common-mode input impedance >10Kohms - Common-mode
rejection > 60dB over its bandwidth for signals with up to 1k ohm source
impedance
- Output will drive an ADC whose supplies will be referenced to earth
but will be noisy with respect to earth (5v peak max from DC to 10MHz)


I had looked at several techniques including some discrete solutions and
some canned solutions from Apex Microtech and Burr Brown (now TI). These
involved using expensive laser-trimmed input networks.

However, with these solutions, I could not get the bandwidth,
common-mode rejection, and/or input impedance that I needed.

I ended up using a tranformer-coupled design, so of course I lost my low
frequency capability. This was okay but not desirable.

Any ideas on a solution that would meet or come close to the listed
specs?

Thanks much,
Bob

Bob,

I'm currently working on such a thing (but up to 4kV and with a 1:2000
ratio) and it's not for the fainted heart (lots of minute details
will get in the way). 200V will ease things a bit, but still you've got
to be very, very, cautious.
There are two ways of doing this: xformers and the 'classic' dual
dividers way.
I proposed the best one, my client choosed the worse (which will not
necessarily be the same for you, given your lower voltage level).
A few 'nice' things among others: dividers compensation isn't as simple
as it seems because of caps dielectric behavior vs frequency (hey JL!)
varying enough from one cap to another to compromise not only CMRR, but
frequency response too. And of course, the caps from the high and low
branches have pretty different values, (hence construction) which
magnifies the pb.
Varying PCB parasitics with temp, humidity and frequency is another.
Tight geometry control is yet another (that was the last catch). I hope
that's really the last one.

In our case, we have to 'manually' tune each board (a handful of poles
and zeros) to trim the frequency response and CMRR.
I'm currently working on the automatic values extraction because there's
too much interaction bewteen the adjustments to make 'scratch head
tuning' really possible.

You can contact me off list if you're interested. My email address have
to be unmangled.

 

[John Larkin]

Le Thu, 16 Aug 2007 15:47:08 -0700, Bob a écrit:


Bob,

I'm currently working on such a thing (but up to 4kV and with a 1:2000
ratio) and it's not for the fainted heart (lots of minute details
will get in the way). 200V will ease things a bit, but still you've got
to be very, very, cautious.
There are two ways of doing this: xformers and the 'classic' dual
dividers way.
I proposed the best one, my client choosed the worse (which will not
necessarily be the same for you, given your lower voltage level).
A few 'nice' things among others: dividers compensation isn't as simple
as it seems because of caps dielectric behavior vs frequency (hey JL!)

Tektronix did things like this in their differential-input plugins,
all the way from the tube types (Z, W plugins) through the 7A13 for
the 7000 series scopes, which had phenomenal common-mode and overload
recovery to 100 MHz bandwidth. They were basically diff pairs of tubes
or fets, heavily cascode bootstrapped.

I've seen nice bootstrapped opamp circuits, where the opamp power
supplies are driven to track the opamp output or common-mode voltages.

The Tek isolated-input scopes (like TPS2024) have insane specs. I
think they do transformer-coupled gigahertz FM coupling or something.

Air has pretty low dielectric absorption.

John

 

[Rich Grise]

ratio) and it's not for the fainted heart (lots of minute details

"faint of heart".

Cheers!
Rich Grise, Insufferable Pedant ;-)

 

[Fred Bartoli]

Le Fri, 17 Aug 2007 22:11:37 +0000, Rich Grise a écrit:

"faint of heart".

Cheers!
Rich Grise, Insufferable Pedant ;-)

At least I didn't get the apostrophe wrong

 

[Rich Grise]

Le Fri, 17 Aug 2007 22:11:37 +0000, Rich Grise a écrit:


At least I didn't get the apostrophe wrong

It also occurred to me, I might have been thinking of "faint _at_ heart";
either way, it's a pretty easy mistake to make.

I once saw a guy writing up a dorm-room inventory form, and one of his
items was "Chester drawers." ;-)

Cheers!
Rich

 

[Rich Grise]

Le Fri, 17 Aug 2007 22:11:37 +0000, Rich Grise a écrit:



At least I didn't get the apostrophe wrong

Ah, that's also true! I guess when stuff is right, I just let it be
right and hardly notice it.

Cheers!
Rich

 

[JosephKK]

Bob [email protected] posted to sci.electronics.design:

I did a project several years ago where I needed an
instrumentation amplifier that had these basic characteristics:

- DC to 10MHz (-3dB) bandwidth
- Common-mode input max of +-200V with respect to earth
- Differential input max of +-100V
- Differential and common-mode input impedance >10Kohms
- Common-mode rejection > 60dB over its bandwidth for signals with
up to 1k ohm source impedance
- Output will drive an ADC whose supplies will be referenced to
earth but will be noisy with respect to earth (5v peak max from DC
to 10MHz)


I had looked at several techniques including some discrete
solutions and some canned solutions from Apex Microtech and Burr
Brown (now TI). These involved using expensive laser-trimmed input
networks.

However, with these solutions, I could not get the bandwidth,
common-mode rejection, and/or input impedance that I needed.

I ended up using a tranformer-coupled design, so of course I lost
my low frequency capability. This was okay but not desirable.

Any ideas on a solution that would meet or come close to the
listed specs?

Thanks much,
Bob

I do not speak for anyone but myself, but the application
description screams out for a tube solution.