I've searched, but could not find an example of this...
Is it possible to actively drive the OFFSET NULL pin of an opamp like the LT1028, with another opamp acting as DC servo?
Could you just stick the output of your comparator into pin 1 or 8, to push the voltage closer to, or further from VCC? Obviously a rail-to-rail op-amp is required.
LTSpice doesn't have any opamp models with offset pins. Hmmph.
Is it possible to actively drive the OFFSET NULL pin of an opamp like the LT1028, with another opamp acting as DC servo?
Could you just stick the output of your comparator into pin 1 or 8, to push the voltage closer to, or further from VCC? Obviously a rail-to-rail op-amp is required.
LTSpice doesn't have any opamp models with offset pins. Hmmph.
OK, I just tried this with Pete Millet's GFA-555 spice model, in lieu of an opamp with exposed offset null pins, ...and it seems to work! I just used the ideal opamp model in spice, this is not a practical circuit. (Opamp is putting out 74V! haha)
I've stuck a low Hfe power transistor on one side of the LTP so the offset is exaggerated. The output is brought down from 106mV to 0.7mV.
Any reason not to do this? It will inject noise I suppose, but the servo has a super slow time constant, so maybe not much?
I've stuck a low Hfe power transistor on one side of the LTP so the offset is exaggerated. The output is brought down from 106mV to 0.7mV.
Any reason not to do this? It will inject noise I suppose, but the servo has a super slow time constant, so maybe not much?
I would follow the note in the data sheet & not do that, most likely you will make the noise and offset worse over temperature. One would need to be very careful about the null adjustment.
An option is to use a pot to find the best null point then use two proper value 1% low Tempco metal film resistors in place of the 1k ohm pot.
Any amplifier servo you used to null the offset would need to have lower offset & drift than the LT1028 & would need to have very low noise (you do not want to inject noise in the front end).
The null offset option balances the two sides of the input differential pair (already laser trimmed).
Unless you add real artifacts and proper non-idealities to the simulation you will not see the errors in the simulation. Often the macro model model approximations are NOT that accurate for noise and offset effect.
An option is to use a pot to find the best null point then use two proper value 1% low Tempco metal film resistors in place of the 1k ohm pot.
Any amplifier servo you used to null the offset would need to have lower offset & drift than the LT1028 & would need to have very low noise (you do not want to inject noise in the front end).
The null offset option balances the two sides of the input differential pair (already laser trimmed).
Unless you add real artifacts and proper non-idealities to the simulation you will not see the errors in the simulation. Often the macro model model approximations are NOT that accurate for noise and offset effect.
Last edited:
Hi,
an idea just coming to my mind is .... using a PNP in parallel to the collector resistor of the input stage. This PNP could take over a fraction of the IP-stage´s idle current.
The resistor needed to be incresased in value accordingly.
The PNP could be part of a current mirror driven by the dc-servo OPamp to preserve some PSRR (using a base resistor to Vpos to drive the PNP would spoil it).
So basically it´d be performing kind of level-translation, thereby keeping the OPAmps output close to 0V.
If the output of the OPAmp needed to settle first on a voltage that differs alot from 0V You will certainly run into switch-on and -off issues.
At start-up the OPAmp might even not settle at its idle value but start into the wrong direction.
jauu
Calvin
an idea just coming to my mind is .... using a PNP in parallel to the collector resistor of the input stage. This PNP could take over a fraction of the IP-stage´s idle current.
The resistor needed to be incresased in value accordingly.
The PNP could be part of a current mirror driven by the dc-servo OPamp to preserve some PSRR (using a base resistor to Vpos to drive the PNP would spoil it).
So basically it´d be performing kind of level-translation, thereby keeping the OPAmps output close to 0V.
If the output of the OPAmp needed to settle first on a voltage that differs alot from 0V You will certainly run into switch-on and -off issues.
At start-up the OPAmp might even not settle at its idle value but start into the wrong direction.
jauu
Calvin
Yeah, I know about the built-in offset null adjustments, but a servo is nice because it will account for different offsets caused by different cart impedances. It doesn't need to absolutely accurate, just keep it under 1mv, good to go.
One problem with DC servos is that they inject noise. Maybe it's not much, I haven't done the maths. I'm just wondering if this might be a lower-noise way to accomplish a servo...
...but now maybe I'm talking myself out of it. If one sticks the servo output, via a 47K resistor or something, into the inverting input of the last opamp stage, it will see the impedance of the "R1 || R2", which in the example below is 100 ohms to ground || 3.92K to ouput... Maybe it's nothing.
Plus the servo is moving really slow. Perhaps one could put an additional RC shunt to get rid of more noise.
Here's how Walt Jung did it in this two-stage passive EQ phono stage...
One problem with DC servos is that they inject noise. Maybe it's not much, I haven't done the maths. I'm just wondering if this might be a lower-noise way to accomplish a servo...
...but now maybe I'm talking myself out of it. If one sticks the servo output, via a 47K resistor or something, into the inverting input of the last opamp stage, it will see the impedance of the "R1 || R2", which in the example below is 100 ohms to ground || 3.92K to ouput... Maybe it's nothing.
Plus the servo is moving really slow. Perhaps one could put an additional RC shunt to get rid of more noise.
Here's how Walt Jung did it in this two-stage passive EQ phono stage...
Last edited:
If you are going to do a general DC servo, it is better not to use the LT1028 null pins, they are only for minor Vos adjustment. The front end of the LT1028 is already laser trimmed.
Did this decades ago on an OP-27 (AFAIR), which has a similar input stage topology as the LT1028. Servo opamp may have been an AD711, operating in inverting mode.
How the servo was configured to interact with the signal opamp was different from the schematic that you showed, but the concept was identical.
Once the correct polarity was established, the circuit worked fine, there were never any startup glitches AFAIR.
Low-pass filtering both the input and output of the servo opamp is recommended, to reduce injection of noise and distortion.
For example, use two resistors each with one-half the target value, then connect an appropriately-valued capacitor from the resistor-resistor node to ground.
hth, jonathan
How the servo was configured to interact with the signal opamp was different from the schematic that you showed, but the concept was identical.
Once the correct polarity was established, the circuit worked fine, there were never any startup glitches AFAIR.
Low-pass filtering both the input and output of the servo opamp is recommended, to reduce injection of noise and distortion.
For example, use two resistors each with one-half the target value, then connect an appropriately-valued capacitor from the resistor-resistor node to ground.
hth, jonathan
Thanks Jonathan! Interesting. Yeah, the servo filtering you suggest should work for lots of servo scenarios. I guess you could even go LRC? The noise of the op-amp could be made practically nothing.
LRC will increase the possibility of resonances, and may require the addition of damping resistors (but you are free to try anything).
Also, adding lowpass filters will increase the number of time constants in a limited frequency range. Therefore when designing the filters, be sure to space apart the corner frequencies appropriately.
A separate technique is to attenuate the output of the servo opamp.
This works because the servo opamp's output level (null signal) will increase to compensate, while the servo opamp's noise will remain more or less the same.
hth, jonathan
Also, adding lowpass filters will increase the number of time constants in a limited frequency range. Therefore when designing the filters, be sure to space apart the corner frequencies appropriately.
A separate technique is to attenuate the output of the servo opamp.
This works because the servo opamp's output level (null signal) will increase to compensate, while the servo opamp's noise will remain more or less the same.
hth, jonathan
If you strive for very low distortion, you have to keep in mind that the relation between correction signal and offset is non-linear; there is an inverse hyperbolic tangent relation between the servo signal that you inject at the collector of the input differential pair transistor and the pair's offset voltage. Therefore, this kind of servo will aggravate low-frequency distortion to some extent - no idea to what extent, as I never made an attempt to calculate that.