Maker Pro
Maker Pro

low-cost high-voltage electrometer amplifier

R

Rene Tschaggelar

John said:
John Larkin wrote:
[snip]

The depletion Fets are single sourced ? I only
heard about one manufacturer once, and even
forgot about it. I'd prefer enhancement MosFets.

You can do some cool stuff with depletion fets, like very nice current
limiters. We've never had a problem getting the Supertex parts.

I'm aware of the simple current sources and limiters,
the supertex parts are just somewhat exotic over here.

Certainly. I'd do all that if I had a reason to, and a target spec to
meet. At present, I don't have either.


You can do that math as well as I can.

Thanks for overestimating me. I'm not that quick
in head calculations on things where I have little
training.
Do you always rain on ideas?

No offense intended. Sorry.

Rene
 
R

Rene Tschaggelar

John said:
John Larkin wrote:
[snip]
John,
and the switches for the high voltage being ?
There are several conditions to it. Such as
low leakage, low charge injection. Win was
talking about an electrometer input, meaning
there cannot 1uF at the input.

Rene


Some people can riff on ideas, and some people
stamp them down as soon as they appear.

Not necessarily. I consider the capacitive divider
idea interesting. Well, it is a sampling solution.
Bigger caps are a bit hard on the switches, and
smaller caps are sensitive to effects of semiconductor
switches. Beside that the switches need a supply
range at least covering the signal range.

Rene
 
F

Fred Bartoli

Le Wed, 10 Oct 2007 18:08:18 -0700, Winfield a écrit:
John said:
Rene said:
Winfield Hill wrote:
Here's a nice example of a circuit that pulls itself up by its
bootstraps - a high-voltage opamp follower featuring
electrometer-style high input impedances.
My design is meant to replace expensive high-voltage opamps, like
the PA97 offered by Apex, with low-cost superior-performance
low-voltage opamps, like those offered by Analog Devices (Stephan
Goldstein should enjoy this one). My goal is for an amplifier with
an offset voltage under 65uV and less than 1pA input current, yet
operating over a 430-volt signal range.
+420
_|
IN ___| \___ OUT - within
<1pA |__/ 65uV of input
|
-15

Well, whaddya think?
Win,
I'm interested. Considering the cost of an APEX part, there is some
slack in building the function from scratch with 35 parts.

OK, what am I missing?
http://s2.supload.com/free/WinsHVamp.JPG/view/

My input current-limiting circuit, transient protection for the
MOSFETs, a small isolated capacitor to absorb fault charge, a protected
guard circuit, and series resistors to limit the fault currents if a
MOSFET fails.

You have an extra zener with cap, unless you want to add an active PMOS
pulldown circuit to increase negative slew-rate capability. That's my
next version and pushes the parts count up to 38.

Nothing to take care of stability issues?
I mean bootstrapping the supply rails from the output defeats the opamp's
compensation capacitor and will almost surely get you into trouble if
this is not taken care of.

Did you actually build this or is it just a back of the envelop idea?
 
W

Winfield Hill

Fred said:
Winfield a écrit:
John said:
Winfield Hill wrote:
Here's a nice example of a circuit that pulls itself up by its
bootstraps - a high-voltage opamp follower featuring
electrometer-style high input impedances.
My design is meant to replace expensive high-voltage opamps, like
the PA97 offered by Apex, with low-cost superior-performance
low-voltage opamps, like those offered by Analog Devices (Stephan
Goldstein should enjoy this one). My goal is for an amplifier with
an offset voltage under 65uV and less than 1pA input current, yet
operating over a 430-volt signal range.
+420
_|
IN ___| \___ OUT - within
<1pA |__/ 65uV of input
|
-15
[snip]
Well, whaddya think?
My input current-limiting circuit, transient protection for the
MOSFETs, a small isolated capacitor to absorb fault charge, a
protected guard circuit, and series resistors to limit the fault
currents if a MOSFET fails.
You have an extra zener with cap, unless you want to add an active
PMOS pulldown circuit to increase negative slew-rate capability.
That's my next version and pushes the parts count up to 38.

Nothing to take care of stability issues?
I mean bootstrapping the supply rails from the output defeats the
opamp's compensation capacitor and will almost surely get you into
trouble if this is not taken care of.

Did you actually build this or is it just a back of the envelop idea?

Yes, like I said, it actually has 40 parts, including
active pulldown. Sorry for the over-simplification,
I hadn't wanted to make a big ASCII drawing, but with
these subsequent posts, some of the detail is coming
out bit-by-bit anyway. feel free to ask away.

Here's the opamp compensation part. As you can see
the opamp has its own local HF feedback control, but
at a slower rate it can do what it wants to control
the MOSFET, which in turn positions the stack.

+420
|
|--'
,----||<-,
input | |--+ dn3145
protection _____|________|
\__ _|_ | |
IN --|__|---|+ \ | _|_,
| >--+ /_\ 5.1V
,--|-__/ _|_ | 100k load
| | --- +---- OUT ---/\/\--- gnd
| | | |
'----| ----+--/\/\--+
'--------------+
|
1.5mA
sink

As before, there are missing parts, like the MOSFET
resistors and beads to suppress RF oscillation, etc.
 
J

John Larkin

John said:
John Larkin wrote:
[snip]
You could also do a flying-capacitor thing, with a dpdt relay, if
static measurements were enough. It could report the differential
input almost 1:1, or you could dump into a bigger cap and get a
voltage-divider effect.


John,
and the switches for the high voltage being ?
There are several conditions to it. Such as
low leakage, low charge injection. Win was
talking about an electrometer input, meaning
there cannot 1uF at the input.

Rene


Some people can riff on ideas, and some people
stamp them down as soon as they appear.

Not necessarily. I consider the capacitive divider
idea interesting. Well, it is a sampling solution.
Bigger caps are a bit hard on the switches, and
smaller caps are sensitive to effects of semiconductor
switches. Beside that the switches need a supply
range at least covering the signal range.

Rene

Real relays are still neat for a lot of things, and there are some
very good ones around, telecom rated for 1500 volts. A little series
resistance will protect the contacts from arcing when the cap is
switched, without affecting charge distribution.

John
 
R

Rene Tschaggelar

John said:
John Larkin wrote:

John Larkin wrote:
[snip]

You could also do a flying-capacitor thing, with a dpdt relay, if
static measurements were enough. It could report the differential
input almost 1:1, or you could dump into a bigger cap and get a
voltage-divider effect.


John,
and the switches for the high voltage being ?
There are several conditions to it. Such as
low leakage, low charge injection. Win was
talking about an electrometer input, meaning
there cannot 1uF at the input.


Some people can riff on ideas, and some people
stamp them down as soon as they appear.

Not necessarily. I consider the capacitive divider
idea interesting. Well, it is a sampling solution.
Bigger caps are a bit hard on the switches, and
smaller caps are sensitive to effects of semiconductor
switches. Beside that the switches need a supply
range at least covering the signal range.
Real relays are still neat for a lot of things, and there are some
very good ones around, telecom rated for 1500 volts. A little series
resistance will protect the contacts from arcing when the cap is
switched, without affecting charge distribution.

Yes, indeed. Then we're rather close to DC. A reed
relay is specified for 10^7 cycles. I admit a relay
is a part I try to avoid.

Rene
 
R

Robert Baer

Rene said:
John said:
John Larkin wrote:

[snip]
You could also do a flying-capacitor thing, with a dpdt relay, if
static measurements were enough. It could report the differential
input almost 1:1, or you could dump into a bigger cap and get a
voltage-divider effect.



John,
and the switches for the high voltage being ?
There are several conditions to it. Such as
low leakage, low charge injection. Win was
talking about an electrometer input, meaning
there cannot 1uF at the input.

Rene



Some people can riff on ideas, and some people
stamp them down as soon as they appear.

Not necessarily. I consider the capacitive divider
idea interesting. Well, it is a sampling solution.
Bigger caps are a bit hard on the switches, and
smaller caps are sensitive to effects of semiconductor
switches. Beside that the switches need a supply
range at least covering the signal range.

Rene
...and relays have their own problems...
 
W

Winfield

Winfield said:
Fred said:
Winfield a écrit:
John Larkin wrote:
Winfield Hill wrote:
Here's a nice example of a circuit that pulls itself up by its
bootstraps - a high-voltage opamp follower featuring
electrometer-style high input impedances.
My design is meant to replace expensive high-voltage opamps, like
the PA97 offered by Apex, with low-cost superior-performance
low-voltage opamps, like those offered by Analog Devices (Stephan
Goldstein should enjoy this one). My goal is for an amplifier with
an offset voltage under 65uV and less than 1pA input current, yet
operating over a 430-volt signal range.
+420
_|
IN ___| \___ OUT - within
<1pA |__/ 65uV of input
|
-15
[snip]
Well, whaddya think?
OK, what am I missing?
http://s2.supload.com/free/WinsHVamp.JPG/view/
My input current-limiting circuit, transient protection for the
MOSFETs, a small isolated capacitor to absorb fault charge, a
protected guard circuit, and series resistors to limit the fault
currents if a MOSFET fails.
You have an extra zener with cap, unless you want to add an active
PMOS pulldown circuit to increase negative slew-rate capability.
That's my next version and pushes the parts count up to 38.
Nothing to take care of stability issues?
I mean bootstrapping the supply rails from the output defeats the
opamp's compensation capacitor and will almost surely get you into
trouble if this is not taken care of.
Did you actually build this or is it just a back of the envelop idea?

Yes, like I said, it actually has 40 parts, including
active pulldown. Sorry for the over-simplification,
I hadn't wanted to make a big ASCII drawing, but with
these subsequent posts, some of the detail is coming
out bit-by-bit anyway. feel free to ask away.

Here's the opamp compensation part. As you can see
the opamp has its own local HF feedback control, but
at a slower rate it can do what it wants to control
the MOSFET, which in turn positions the stack.

+420
|
|--'
,----||<-,
input | |--+ dn3145
protection _____|________|
\__ _|_ | |
IN --|__|---|+ \ | _|_,
| >--+ /_\ 5.1V
,--|-__/ _|_ | 100k load
| | --- +---- OUT ---/\/\--- gnd
| | | |
'----| ----+--/\/\--+
'--------------+
|
1.5mA
sink

As before, there are missing parts, like the MOSFET
resistors and beads to suppress RF oscillation, etc.

Then there's the "output compensation", which provides
for rolling-off the huge effective gain of the output
MOSFET stage in a controlled way, with capacitor poles
and series-resistor zeros.

+420
|
|--'
,----||<-,
input Rg |--+ dn3145
protection _____|________|
\__ _|_ | |
IN --|__|---|+ \ | _|_,
| | >--+ /_\ 5.1V
| ,--|-__/ _|_ | 110k load
| | | --- +---- OUT ---/\/\--- gnd
| | | | |
Cc | '----| ----+--/\/\--+ <-- Xo
gnd --||-+-/\/\--+-------+------+
| |
,--||-- GUARD ---/\/\---' 1.5mA
| cable cap, etc sink
gnd MOSFET

The small compensation capacitor, Cc, is also coupled to
the input protection circuit with two back-back diodes
(that normally have no voltage across them) to insure it
rapidly follows input transients, and thanks to the input
resistors, helps slow them down. Cc is a high-voltage
part.

Likewise, if the GUARD connection is used, its capacitance
also contributes to controlling the output gain, but with
a series resistor as a pole, said resistor limiting its
effect on the output and helping to limit any transient
voltages that might arise from faulty use of the GUARD.

Note the output node, Xo. When a huge external transient
imposed on the circuit makes it to that point, all the
fragile items, like the opamp, with its supplies and
protection diodes, simply follow the node. The MOSFETs
have zener diodes with their gate resistors, not shown, to
insure that their Coss + Crss charge can't cause trouble
if (when) this happens.

These little tidbits help push up the parts count.
 
Top