Opamp w/ Vsupply >36V

[Tom Del Rosso]

The FTP thing didn't work well, so I'm using Dropbox now. Here's the
HV amp idea:

https://dl.dropboxusercontent.com/u/53724080/Circuits/Optos/HVamp.JPG

I actually did it once, for the atom probe project, and it worked
fine.

What do you do with FB?

 

[Tom Del Rosso]

Whatever you like. What you do with feedback is application dependant.

Of course, but how? You must have had an equally clever way to isolate it.

 

[Glenn]

On 18/07/13 15.51, John Larkin wrote:
....

The FTP thing didn't work well, so I'm using Dropbox now. Here's the HV amp
idea:

https://dl.dropboxusercontent.com/u/53724080/Circuits/Optos/HVamp.JPG

I actually did it once, for the atom probe project, and it worked fine.

How about cascode coupling the phototransistors?

* The phototransistors will get more bandwidth.

* The phototransistors only have to withstand a few volts Vce.

* The two high-voltage output transistors can be complementary - and
common emitter or collector.

/Glenn

 

[George Herold]

Opamp Vsupply >36V



Hi all, I’ve got a circuit with that uses an opa2134. I find myself pushing the 36V supply limit and I’m a little worried. So I went lookingfor a similar opamp but with a bit more headroom.

So specs might read,



Dual opamp

unity gain stable

4(+) Mhz BW (8MHz or more would be nicer)

10 V/us slew (again more is better)

20mA current (positive, again more would be nicer)

small input C ( <5pF would be nice.)

8 pin dip preferred



Trolling digikey I found a few possible candidates.



TLE2142 (6 MHz GBW)

TLE2072 (only 38 V supply, 11 pF input C)

OPA2604 (nice but 10 pF input C)

(I’ll order a few of each.)



There was also the LF412A, but a bit lacking is positive supply current.



Any others that I might have missed?



Thanks,

George H.

Hi all, I finally had time to try out these other opamps.
(it was a fail on each one.)
The TLE2142 was a sweet opamp. But I forgot about the bias current.
I've got a variable source impedance (100k ,33k and 10k ohm) and the ~1uA of bias current means 10mV to 100 mV of DC offset. (not acceptable.) Too bad 'cause otherwise it was nice.

The TLE2142 has an issue running near it's negative rail. It needed the rail to be -3.0 Volts to keep the DC offset below 5mV (When operated from -1.5 V's the offset was ~120mV!)

The OPA2604 oscillated out of the box and needed a bit of feedback C to tame it. (not a problem.) But it also had a lot of 'issues' (ringy dingies) when driving a coax cable. I mucked around with termination resistors (source and end.) but never got the 'ringies' down to an acceptable level.

So I figured I'd try over voltaging the opa2134's I've got the supply voltage up to 40V (36V max) and no magic smoke yet. :^)
I tired a few different pieces, and did a bit of turn on/off torture.
The circuit has now been running on the lab bench for several hours. Should I test with some elevated temperatures?
I'm thinking this opamp will be fine. (I'll design for 35 V supplies and worse case it'll only be a tad above 36 volts.) (tad = 400mV in this case)

George H.

 

[Frank Miles]

The OPA2604 oscillated out of the box and needed a bit of feedback C to
tame it. (not a problem.) But it also had a lot of 'issues' (ringy
dingies) when driving a coax cable. I mucked around with termination
resistors (source and end.) but never got the 'ringies' down to an
acceptable level.

Did you try the compensation circuits given in the spec sheet? BB (TI)
shows some input RC circuits that help when driving capacitive loads with
this amplifier.

So I figured I'd try over voltaging the opa2134's I've got the supply
voltage up to 40V (36V max) and no magic smoke yet. :^) I tired a few
different pieces, and did a bit of turn on/off torture. The circuit has
now been running on the lab bench for several hours. Should I test with
some elevated temperatures? I'm thinking this opamp will be fine. (I'll
design for 35 V supplies and worse case it'll only be a tad above 36
volts.) (tad = 400mV in this case)

What kind of reliability do you need? If this is a short-term one-off, you
can probably get away with this kind of brinksmanship. If it has to
work for others over a longer time period - do everyone a favor and either
use a part rated for higher voltage or modify your circuit (e.g. composite
amplifier) such that the amplifiers aren't overstressed.

 

[George Herold]

Did you try the compensation circuits given in the spec sheet? BB (TI)
shows some input RC circuits that help when driving capacitive loads with
this amplifier.

No. I'm not sure it was the cable C that was the issue. (Only 3 foot of coax and shrinking that to a few inches didn't change a thing.) I did put 50ohms in series.

What kind of reliability do you need? If this is a short-term one-off, you
can probably get away with this kind of brinksmanship. If it has to
work for others over a longer time period - do everyone a favor and either
use a part rated for higher voltage or modify your circuit (e.g. composite
amplifier) such that the amplifiers aren't overstressed.

Brinksmanship(?).. It's only 0.4V?.. and that's for worst case.
We'll sell maybe 100 or so of these over several years.

So can you say anything more about the stress? What's the cause?
I tried hitting the one on the bench with a heat gun, and shorted the output. (The case rose to ~100 C. I should see what happens with only 36V.) Hey, I'll stick it in a socket. That way if it does fry in the future it can be replaced. (opamps are cheap.)

George H.

 

[Frank Miles]

No. I'm not sure it was the cable C that was the issue. (Only 3 foot of
coax and shrinking that to a few inches didn't change a thing.) I did
put 50 ohms in series.

Ah - you hadn't mentioned that before. Hmmn - could you describe the ringing
(amplitude, duration, ringing frequency)?

Brinksmanship(?).. It's only 0.4V?.. and that's for worst case. We'll
sell maybe 100 or so of these over several years.

I'm pretty conservative when it comes to semiconductor voltages - I'd call
0.4V _less_ than rated voltage was brinksmanship. But I expect my designs
to have "reasonably high" reliability (not medical/life support level).
If you look at failure rate curves with voltage (and I admit I don't recall
seeing any for more modern lower voltage opamps) they are climbing rapidly
and smoothly in the vicinity of the rated voltage. Of course there is a
significant amount of variation.

So can you say anything more about the stress? What's the cause?
I tried hitting the one on the bench with a heat gun, and shorted the
output. (The case rose to ~100 C. I should see what happens with
only 36V.) Hey, I'll stick it in a socket. That way if it does fry
in the future it can be replaced. (opamps are cheap.)

If your customers are savvy enough to diagnose the problem, and replace
the amplifier without damage to themselves or to the device or to whatever
function they were meant to provide - sure. My users, while often PhDs,
are most often not that electronically savvy. And down-time is expensive.

 

[George Herold]

Ah - you hadn't mentioned that before. Hmmn - could you describe the ringing

(amplitude, duration, ringing frequency)?

Better than that here's some 'scope shots!

So I’m looking at avalanche breakdown voltages with different quenching resistors (100k and 10 k) the first two ‘scope shots are with the opa2134
100k ohm
//bayimg.com/MaoKGaaEA
10k ohm
//bayimg.com/NAoKDAAea

Oh and here’s the TLE2142 with 10 k ohm
//bayimg.com/nAOKGaaEa

And now the opa2604 first 100 k
//bayimg.com/naokHaAEa
and 10 k
//bayimg.com/naOkjaaea

This is before I added a 50 ohm series resistor. That gave the bumps a longer period but a bit higher amplitude.

I'm pretty conservative when it comes to semiconductor voltages - I'd call

0.4V _less_ than rated voltage was brinksmanship. But I expect my designs

to have "reasonably high" reliability (not medical/life support level).

If you look at failure rate curves with voltage (and I admit I don't recall

seeing any for more modern lower voltage opamps) they are climbing rapidly

and smoothly in the vicinity of the rated voltage. Of course there is a

significant amount of variation.










If your customers are savvy enough to diagnose the problem, and replace

the amplifier without damage to themselves or to the device or to whatever

function they were meant to provide - sure. My users, while often PhDs,

are most often not that electronically savvy. And down-time is expensive..

Grin.. well PhD's are often less of the 'hand's on' type. But the circuit is pretty easy. I made it through hole, so if someoen wanted to change things it would be easier.
http://bayimg.com/PAoKiAaEa

The ugly pot in the middle will go away. (just changing the supply voltage..) I let it cook over night at ~100C. no issues in the morning.

I'll test the heat again with lower supply voltage (but with the palstic cover in place.)

So what type of PhD's do you service? Can you tell us where you work?

George H.

 

[George Herold]

Sounds OK, but it would be better/funner to test the part to failure.

Maybe it fails at 41, maybe it fails at 70.

Well I could go a bit higher. (power from Phihong 48V wall wart.)
But other parts might start to break too. (There's an ref102 that goes to 40V so I might loose that before the opamp.)

George H.

 

[Frank Miles]

Better than that here's some 'scope shots!

So I’m looking at avalanche breakdown voltages with different quenching
resistors (100k and 10 k) the first two ‘scope shots are with the
opa2134 100k ohm
//bayimg.com/MaoKGaaEA
10k ohm
//bayimg.com/NAoKDAAea

Oh and here’s the TLE2142 with 10 k ohm //bayimg.com/nAOKGaaEa

And now the opa2604 first 100 k
//bayimg.com/naokHaAEa
and 10 k
//bayimg.com/naOkjaaea

This is before I added a 50 ohm series resistor. That gave the bumps a
longer period but a bit higher amplitude.

Hmmn, just noticed that the GBW of the TLE2142 is about 1/4 of the OPA2604
GBW. That might be part of why you see the problem with the OPA but not the TLE.
The 'ring', after all, is near the unity-gain BW of the TLE. And you only
see the problem with some length of coax attached? Does it occur with no coax,
just probing with a good low-Cin 10x or FET probe? Could you plug in a much faster
amplifier while temporarily running at a lower voltage to see if there's a problem
with your circuit or its implementation?

If you really like the TLE2142 - how about compensating for the input bias?
Yes, it would cost a tweak.

Grin.. well PhD's are often less of the 'hand's on' type. But the
circuit is pretty easy. I made it through hole, so if someoen wanted to
change things it would be easier. http://bayimg.com/PAoKiAaEa

The ugly pot in the middle will go away. (just changing the supply
voltage.) I let it cook over night at ~100C. no issues in the morning.

Ok, that seems simple and your users are presumably knowledgeable about
antistatic issues, etc.

I'll test the heat again with lower supply voltage (but with the palstic
cover in place.)

So what type of PhD's do you service? Can you tell us where you work?

George H.

Ah... the devices we design and build are for a collection of neuroscientists.
Many have some electronics skills but most of them have enough on their
minds trying to understand and/or manipulate the brain.

 

[George Herold]

Hmmn, just noticed that the GBW of the TLE2142 is about 1/4 of the OPA2604

GBW. That might be part of why you see the problem with the OPA but not the TLE.

The 'ring', after all, is near the unity-gain BW of the TLE. And you only

see the problem with some length of coax attached? Does it occur with nocoax,

just probing with a good low-Cin 10x or FET probe? Could you plug in a much faster

amplifier while temporarily running at a lower voltage to see if there's a problem

with your circuit or its implementation?

Thanks for all the advice. I just can't 'spend' anymore time on this project. (As much as I'd like to chase down the different opamp weirdnesses.)
RE the TLE2142 and bias current. It's a simple buffer, with switched inputresistors to ground on the front end. I don't know how I could compensatefor the current.. would I have to put switched resistors in the feedback? 100k ohm of feedback may come with it's own problem.

For the opa2604 ringies it's there with no coax. (just a double male barrel connector into scope.) Alas I have no low C probes. (The 'best' x10 has16pF)

If you really like the TLE2142 - how about compensating for the input bias?

Yes, it would cost a tweak.









Ok, that seems simple and your users are presumably knowledgeable about

antistatic issues, etc.







Ah... the devices we design and build are for a collection of neuroscientists.

Many have some electronics skills but most of them have enough on their

minds trying to understand and/or manipulate the brain.

Thanks, sounds interesting. There was a research/ medical doctor/ biologist who came visiting. I was looking at the step response of an amplifier on the 'scope. And I told him it was a two pole butterworth low pass.
"Hmm", he says, "the delay looks a bit long there." pointing at the trace.
"Oh there's also a series 1 k ohm resistor in the circuit so maybe that slows it down.", I respond.
I shorted out the 1 k ohm. And he was happy.
Now that's something you don't see every day. A doctor who *really* knows his step response. (Needless to say, they do a lot of electronics in his lab.)

George H.

 

[Frank Miles]

On Wed, 31 Jul 2013 10:17:28 -0700, George Herold wrote:

[snip]

Thanks for all the advice. I just can't 'spend' anymore time on this
project. (As much as I'd like to chase down the different opamp
weirdnesses.) RE the TLE2142 and bias current. It's a simple buffer,
with switched input resistors to ground on the front end. I don't know
how I could compensate for the current.. would I have to put switched
resistors in the feedback? 100k ohm of feedback may come with it's own
problem.

You only need to provide a current source at the opamp input. According
to the datasheet this current is pretty stable so you might be able to get
away with a resistor between amplifier and tweak wiper, and the tweak element
ends between ground and the appropriate supply voltage. If you need
a higher impedance value you can always connect the resistor to a bipolar
transistor's emitter, and take the compensation current from the transistor's
base, using the beta to reduce the current and increase the impedance.

But it's always important to recognize when to stop working on something.

 

[George Herold]

On Wed, 31 Jul 2013 10:17:28 -0700, George Herold wrote:



You only need to provide a current source at the opamp input. According
to the datasheet this current is pretty stable so you might be able to get
away with a resistor between amplifier and tweak wiper, and the tweak element
ends between ground and the appropriate supply voltage. If you need
a higher impedance value you can always connect the resistor to a bipolar
transistor's emitter, and take the compensation current from the transistor's
base, using the beta to reduce the current and increase the impedance.

Ahh, (silly me.) Well that's an easy tweak. I'll give it a try. (The board needs another spin to make some solder holes bigger anyway.)

But it's always important to recognize when to stop working on something.

Sometimes that's the hardest thing to know. (When it's time to stop.)

George H.