On Fri, 23 Feb 2018 17:14:28 -0000
"David C. Partridge" david.partridge@perdrix.co.uk wrote:

The HP journal does only mentions a single opamp integrator. The Art
of electronics only deals with the close relative used in the HP 34420,
which is similar, but not the same. It uses a two opamp structure,
which TAoE3 explains to be to increase amplification and slewrate.
I guess that can be said about the HP3458 as well, but that begs the
question why did they chose a 3 opamp structure? And why are U2 and U3
wired up the way they are?

			Attila Kinali

--
It is upon moral qualities that a society is ultimately founded. All
the prosperity and technological sophistication in the world is of no
use without that foundation.
-- Miss Matheson, The Diamond Age, Neil Stephenson

On Fri, 23 Feb 2018 18:31:47 +0100
Attila Kinali attila@kinali.ch wrote:

Let me expand here a little bit to make it clear what puzzles me.
The HP3458 circuit uses 3 opamps: LT1001, LM6361 and AD848.
The LT1001 is there for its low input bias current of typ 1.2ns,
but has a GBW of typ 0.8MHz, SR of 0.25Vµs and A_vol of 700 V/mV.
The LM6361 comes with GBW of 40MHz, SR of 200 V/µs and A_vol of 0.55 V/mV.
The AD848 has a GBW of 175MHz, SR of 200 V/µs and A_vol of 13 V/mV.

Just by judging from the slew rate, the LM6361 alone should be enough
to get the ciruit up to speed, and the AD848 wouldnt be needed.
The AD848 has a much higher GBW and A_vol and I can see that
both would be beneficial for the integrator. But why not just use
a AD848 and leave the LM6361 out? Why use both?

The jfet is obviously there to decouple the high speed opamps with
their high bias current inputs from the sensitive summing point.
C2 is probably to limit the slew rate of U2 in order to keep the
integrator stable. But this means that U2+U3 together have some
stability problem that could be solved by removing one (preferably
U2 as it's the slower of the two).

		Attila Kinali

--
It is upon moral qualities that a society is ultimately founded. All
the prosperity and technological sophistication in the world is of no
use without that foundation.
-- Miss Matheson, The Diamond Age, Neil Stephenson

Looking your schematic, I see that you are using LT1225 for both U2 (U111) and U3 (U112)

Why did you use those rather than LM6361 and AD848?  Was it just because LTSpice didn't have models for them?  It can make a difference ...

As for why they chose to use 3 op-amps rather than 2?  I don't know except that when dealing with low level signals maybe they needed the extra gain.

BTW LM6361 is 550V/mv not .55V (I assume that was a mistake in your post) ... Slew rates seem a bit different at +/- 15V supply as well.

David

-----Original Message-----
From: volt-nuts [mailto:volt-nuts-bounces@febo.com] On Behalf Of Attila Kinali
Sent: 23 February 2018 18:02
To: Discussion of precise voltage measurement
Subject: Re: [volt-nuts] HP3458 ADC integrator

On Fri, 23 Feb 2018 18:31:47 +0100
Attila Kinali attila@kinali.ch wrote:

Let me expand here a little bit to make it clear what puzzles me.
The HP3458 circuit uses 3 opamps: LT1001, LM6361 and AD848.
The LT1001 is there for its low input bias current of typ 1.2ns, but has a GBW of typ 0.8MHz, SR of 0.25Vµs and A_vol of 700 V/mV.
The LM6361 comes with GBW of 40MHz, SR of 200 V/µs and A_vol of 0.55 V/mV.
The AD848 has a GBW of 175MHz, SR of 200 V/µs and A_vol of 13 V/mV.

Just by judging from the slew rate, the LM6361 alone should be enough to get the ciruit up to speed, and the AD848 wouldnt be needed.
The AD848 has a much higher GBW and A_vol and I can see that both would be beneficial for the integrator. But why not just use a AD848 and leave the LM6361 out? Why use both?

The jfet is obviously there to decouple the high speed opamps with their high bias current inputs from the sensitive summing point.
C2 is probably to limit the slew rate of U2 in order to keep the integrator stable. But this means that U2+U3 together have some stability problem that could be solved by removing one (preferably
U2 as it's the slower of the two).

		Attila Kinali

--
It is upon moral qualities that a society is ultimately founded. All the prosperity and technological sophistication in the world is of no use without that foundation.
-- Miss Matheson, The Diamond Age, Neil Stephenson _______________________________________________
volt-nuts mailing list -- volt-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts
and follow the instructions there.

Moin attila

I'd start with simulating and understanding the integrator in the 34401A, its a little simpler but still has a coupe of opamps in the integrator. Then analysing the GBW requirements of the various opamps used in the HP3458A shoul become somewhat clearer.

Bruce

Hint: start by looking at the summing junction voltage of a single opamp integrator with R = 50k, C = 330pF.

NB the LT1225 isnt unity gain stable so its use for any of U2, U3 and U4 is questionable.

They will all likely oscillate on their own.

Bruce

On Fri, 23 Feb 2018 18:48:24 -0000
"David C. Partridge" david.partridge@perdrix.co.uk wrote:

Yes. LTSpice doesn't have an LM6361 or AD848. Linear lists the LT1225
as replacement for LM6361, so I thought this was safe. I deemed the AD848
to be close enough.

But as Bruce mentioned, the LT1225 is not unity gain stable, while the
LM6361 is. That might explain the oscillations.

The datasheet I have found (Natsemi Sept 1995) lists 550 V/V = 0.55 V/mV.
Which I think is a bit low, but then, the opamp is old.

		Attila Kinali

--
It is upon moral qualities that a society is ultimately founded. All
the prosperity and technological sophistication in the world is of no
use without that foundation.
-- Miss Matheson, The Diamond Age, Neil Stephenson

Attila

Some insight can be gleaned by using a single pole model for an opamp configured as an integrator and deriving the expression for the output signal.
Look at the non linearity in the output.
Look at the voltage at the inverting input.
The effective impedance at the inverting input due to feedback action can be modelled as a simple serries network. What are the network parameters?
What happens to the non linearity if the positive input is driven in such a way that the inverting input is held at ground?
How would one do this?
What does the resultant signal at the inverting input look like with real opamps?
Whats the maximum GBW of the corrector amp so that the phase margin of the combination isn't reduced too much?

~30 years ago when I first analysed the limitations of opamp analog integrators I didn't have access to Spice and had to resort to Laplace transforms and analytical models of opamps.

Bruce

Check other datasheets, that's definitely wrong

David