Offset voltage nulling on transimpedance amp

[CC]

Greetings:

I am building a photodiode amp for fun based on OPA657 transimpedance
stage followed by an AD829 second stage. I don't want to be amplifying
the up to 2.2mV offset voltage from OPA657 in the second stage, so I
need to null it.

Both stages will have widely variable gains. A range of
500,1k,2k,5k,...,500k on the first stage, and 2x,20x,200x,2000x on the
second.

The OPA657 doesn't have nulling pins. Should I inject a few mV with an
AC bypassed voltage divider pot at the OPA657 non-inverting pin (which
has a gain of unity due to transimpedance configuration) or inject into
the AD829 as a summing amp?

I like the former idea better since when changing gains on the 2nd
stage, it won't require constantly readjusting the zero. I'd prefer to
put my gain setting resistors on the feedback path of the second stage,
so that I can have a cap tailored to each feedback R. But if I do this,
the gain would also change for the offset nulling voltage.

Any ideas?

 

[Rene Tschaggelar]

Greetings:

I am building a photodiode amp for fun based on OPA657 transimpedance
stage followed by an AD829 second stage. I don't want to be amplifying
the up to 2.2mV offset voltage from OPA657 in the second stage, so I
need to null it.

Both stages will have widely variable gains. A range of
500,1k,2k,5k,...,500k on the first stage, and 2x,20x,200x,2000x on the
second.

The OPA657 doesn't have nulling pins. Should I inject a few mV with an
AC bypassed voltage divider pot at the OPA657 non-inverting pin (which
has a gain of unity due to transimpedance configuration) or inject into
the AD829 as a summing amp?

I like the former idea better since when changing gains on the 2nd
stage, it won't require constantly readjusting the zero. I'd prefer to
put my gain setting resistors on the feedback path of the second stage,
so that I can have a cap tailored to each feedback R. But if I do this,
the gain would also change for the offset nulling voltage.

Any ideas?

Yes, are you sure you need a DC amp ?
You may get far more offset from the ambient light
or from dark currents.

Rene

 

[Pooh Bear]

Yes, are you sure you need a DC amp ?
You may get far more offset from the ambient light
or from dark currents.

Can you still get those integrated chopper amps like the one Intersil ? used to
make ?

Graham

 

[Ken Smith]

Greetings:

I am building a photodiode amp for fun based on OPA657 transimpedance
stage followed by an AD829 second stage. I don't want to be amplifying
the up to 2.2mV offset voltage from OPA657 in the second stage, so I
need to null it.

Both stages will have widely variable gains. A range of
500,1k,2k,5k,...,500k on the first stage, and 2x,20x,200x,2000x on the
second.

The OPA657 doesn't have nulling pins. Should I inject a few mV with an
AC bypassed voltage divider pot at the OPA657 non-inverting pin (which
has a gain of unity due to transimpedance configuration) or inject into
the AD829 as a summing amp?

When you vary the gain of the first stage, how does the offset in the
output change? I suspect that a part of the output's offset voltage will
be proportional to the gain and part will remain constant.

I think you will need at least two offset trim inputs to have to offset
not change with the gain setting. You may need a seperate bias current
trim too.

I like the former idea better since when changing gains on the 2nd
stage, it won't require constantly readjusting the zero. I'd prefer to
put my gain setting resistors on the feedback path of the second stage,
so that I can have a cap tailored to each feedback R. But if I do this,
the gain would also change for the offset nulling voltage.

What I don't think I understand is why you are changing gains and not the
feedback resistor

Rfb
-------------/\/\/-----------------------
! !
! ---/\/\--!!----+
-----+---!+\ ! !
! >-----+----/\/\/--+----!-\ !
---!-/ ! ! >--------+-- Vout
! ! GND--!+/
+--/\/\--!!--
!
/
\
/
!
GND(or offset null)

Only when the op-amp gains become a problem do you have trouble
maintaining the phase margin. Changing the gain of the stage doesn't
effect that point.

 

[CC]

Yes, are you sure you need a DC amp ?
You may get far more offset from the ambient light
or from dark currents.

Yes, DC! I plan a few different variations. The first will be for a PV
diode, UDT PIN-10DP/SB, an 8800pF 100mm^2 beast. Since it is PV, there
will be no dark current, although I may experiment with a very small
2.5V bias on this one. Ambient of course is an issue, at higher gains.

I'd like to keep offsets less than 0.1% of the signal for low gains,
anybe tolerate up to 1% at higher gains. So for 5V output, that means
5mV or less.

I think I will have to throw out AD829 since it has very high biad currents.


Thanks for your input.

 

[CC]

Can you still get those integrated chopper amps like the one Intersil ? used to
make ?

I could look again, but I don't recall choppers as being suitable for
wideband applications.

I'd like to hit 10MHz at a gain of 2 for the second stage, with the
ability to output +/-5V. That needs a slew rate of at least 314V/us.

Certainly can't use a chopper on the front end, that positively has to
be super wideband to deal with the diode capacitance.


Thanks for input.

 

[CC]

When you vary the gain of the first stage, how does the offset in the
output change?

Don't know yet.

I suspect that a part of the output's offset voltage will
be proportional to the gain and part will remain constant.

That's strange. Isn't input offset voltage always accountable as a
voltage source into one of the inputs? I have a classic transimpedance
stage with PD connected to the -input and a resistor from output to
-input. The gain with respect to the input offset voltage should always
be unity, no?

Thus, I can either inject a few mV into the +input to cancel the input
offset, or deal with it in the next stage. But in either case, the
input offset from the first stage should never change due to gain
settings. Just to be clear, gain changes in the first stage are
implemented by changing the feedback resistor. But DC gain for a
voltage present at either opamp input is always unity.

I think you will need at least two offset trim inputs to have to offset
not change with the gain setting. You may need a seperate bias current
trim too.

Bias current isn't much of a problem in the 1st stage because OPA657 has
20pA max at 25C. Although at the highest gain settings for both stages,
this would exceed my desired <1% offset. Maybe I can ignore this since
it will be within spec at all other settings than 200k*2000 and 500k*2000.

What I don't think I understand is why you are changing gains and not the
feedback resistor

I will be changing the FB R in the first stage of course. In the second
stage (let's say it's a vanilla inverting opamp configuration) it
depends. It is possible to leave the FB resistor constant, and change
the input R, or change the FB R. What are the advantages and
disadvantages of each?

Changing the FB R makes it possible to keep BW flatter over the range of
gains, by simultaneously switching the capacitor across the FB R. Ie,
an RC pair can be selected for each gain setting.

Also, this presents a constant load R to the input source.

Disadvantage is only that if the 1st stage offset is not canceled and
must be dealt with by a summing R placed on the 2nd stage -input, then
the gain for that offset correcting voltage would change with 2nd stage
gain changes.

If the input R to the 2nd inverting stage was instead varied, then the
offset injection would not see its gain change. But this would require
the FB resistor to be value needed to acheive the max desired gain of
2000 which isn't desirable because this would seriously limit BW at
lower gains.

So I think it is really best to fix the offset in the 1st stage.

The question is then, is there anything wrong with injecting a few mV
into the +input of the 1st stage? Does it seriously degrade noise
performance, for instance?

Rfb
-------------/\/\/-----------------------
! !
! ---/\/\--!!----+
-----+---!+\ ! !
! >-----+----/\/\/--+----!-\ !
---!-/ ! ! >--------+-- Vout
! ! GND--!+/
+--/\/\--!!--
!
/
\
/
!
GND(or offset null)

What is that thing?

Only when the op-amp gains become a problem do you have trouble
maintaining the phase margin. Changing the gain of the stage doesn't
effect that point.

I'll have to put this in the "for future study" box.


Thanks for input.



Good day!

 

[John Perry]

Yes. There are many available now, such as the National LTC1050, ADI
AD8551, Maxim MAX430, TI TLC2654. Google "chopper stabilized amp" for
more, and for info.

I could look again, but I don't recall choppers as being suitable for
wideband applications.

You don't use them as amps in a high-speed application. You use them to
servo the high-speed amp's offset to near zero. In this mode, they're
very effective.

Google "composite amplifier".

John Perry

 

[John Perry]

Yes. There are many available now, such as the National LTC1050,

--duh. Linear Technology, of course.

ADI

 

[CC]

Yes. There are many available now, such as the National LTC1050, ADI
AD8551, Maxim MAX430, TI TLC2654. Google "chopper stabilized amp" for
more, and for info.


You don't use them as amps in a high-speed application. You use them to
servo the high-speed amp's offset to near zero. In this mode, they're
very effective.

Google "composite amplifier".

Oh I see. Ok, more things to learn. How can one ever get anything done
like this? There's always a better way that comes along after spending
a whole munch of time on the preceding way...


Thanks for the tip!

 

[Ken Smith]

I suspect that a part of the output's offset voltage will
be proportional to the gain and part will remain constant.

That's strange. Isn't input offset voltage always accountable as a
voltage source into one of the inputs? I have a classic transimpedance
stage with PD connected to the -input and a resistor from output to
-input. The gain with respect to the input offset voltage should always
be unity, no?[/QUOTE]

In theory, the offset voltage can all be assigned to the input. In
practice, the completed circuit will have other offsets in it and some of
the offset voltage will be really due to bias currents.

These other effects may be changed when you change the gain. I assume you
are intending to null to output for all of these effects combined.

[....]

Bias current isn't much of a problem in the 1st stage because OPA657 has
20pA max at 25C. Although at the highest gain settings for both stages,
this would exceed my desired <1% offset. Maybe I can ignore this since
it will be within spec at all other settings than 200k*2000 and 500k*2000.

You could add a bias current trim to the first stage using a pot and a
high valued resistor.

--- order changed ---

What is that thing?

That is a transimpedance amplifier using two op-amps. The
feedback networks on each of the op-amps is used to make the whole system
stable. Rfb is the feedback of the transimpedance amplifier. This was
the circuit (or something like it) I was assuming.

I will be changing the FB R in the first stage of course. In the second
stage (let's say it's a vanilla inverting opamp configuration) it
depends. It is possible to leave the FB resistor constant, and change
the input R, or change the FB R. What are the advantages and
disadvantages of each?

The biggest problem in making a variable gain amplifier is avoiding adding
extra capacitance to ground at the inverting input of the op-amp.
Changing the input resistor often works better from this point of
view.

The down side of changing the input resistor is often seen at the previous
op-amp. It sees more capacitance and loading.

You can also change both resistances to change gains:


---/\/\/----+---/\/\/---+---/\/\-------
! ! !
! ! !
! --O !
! <---!-\ !
--------------O ! >--
--!+/

This has the advantage of not needing as wide of a range of resistors to
get a wide range of gains.

Changing the FB R makes it possible to keep BW flatter over the range of
gains, by simultaneously switching the capacitor across the FB R. Ie,
an RC pair can be selected for each gain setting.

Yes, but you are ending up with a lot of parts being switched. This can
make practical problems.


[....]

The question is then, is there anything wrong with injecting a few mV
into the +input of the 1st stage? Does it seriously degrade noise
performance, for instance?

You can make the impedance of the offset circuit low enough that noise
from it is not going to be a problem. You want to divide down from a
quiet reference voltage.