UAF42 - offset problems when designing according to appl. bulletin SBFA005 / AB-062

Hello Lars-Olof,

The UAF42 has undergone one, possible two, fabrication site moves since about 2006. Even though the processes are basically the same, parameter distributions can take on a new mean values and distributions. The voltage offset and input bias currents, although still within specification, may lie in a different place than they did in 1999 when the product was developed. A parameter distribution changes might not only occur for the UAF42, but the DAC8221 and OP37 as well. All of these devices might have distribution shifts that could affect the offset.

The frequency of the digitally programmed filter is changed by changing the equivalent integrator resistance. Increasing the equivalent resistance lowers the filter's cutoff frequency. If the input bias current is higher than it has traditionally been, the voltage offset - especially at the lowest frequencies - would be higher.

About the only idea I have for injecting a counter offset without affecting the filter characteristics would be to do so using the UAF42 auxiliary amplifier. The non-inverting amplifier input could be configured with a potentiometer and resistors to each supply creating an adjustable voltage divider. The offset would be adjusted for the best compromise level over the frequency range.

Regards, Thomas

PA - Linear Applications Engineering

Hello Thomas!

Thank you for the answer!

But unfortunately, the aux amplifier is allready in use as a notch filter summing amp.

I have allready inserted offset potentiometers around each OP. They are connected in series with fixed resistors (for minimizing the adjusting range) between pin 1 and 8 on every OPA37 (center tap of the pot to +V). The fixed resistors are manually selected during the first offset calibration of the unit. The pot's are in the middle and suitable resistors on each side of the pot's is selected. One OP affects the HP filter output offset, the other affects the LP output. The BP offset output is a result from HP and LP together.
This is performed for each OP, that is two OP for each filter (there are four filter sections in the product) when the filter sections are set to the lowest frequency 3Hz. The offset can be positive for the HP filter output  and negative for LP filter, or vice versa.  Since all three outputs is used in the design (selected by a MUX), I think each OP have to be trimmed. One adjustment for the complete filter section is not possible.

The offset is very sensitive for the temperature, so for accurate trimming at the lowest frequency, the unit must raech the working temperature.

The offset is declining very fast with increased set frequency, so if the filter only would be used with the lowest set-abel frequency at ~20Hz, trimming wouldn't be necesarry.

I was thinking if there could be some way to stabilize and secure that the offset at least is trim-able, with maybe a high resistance bias resistor on a inteligent place in the design?
Or a more suitable OP?
I was thinking if that was possible, this could be an universal course solution and on top of that, I would be continuing with the manual individual fine trimming.

Can you see such solution is possible? Suggestions? Other type of OP?

Regards,

Lars-Olof

Hello Lars-Olof,

When I review the OP37 data sheet I see that the trimmed offset value is less than +/-100 uV for most grades. These are very low levels so I think that you may be compensating for any offset that the Vref and DAC that precedes the op amp contribute to the circuit. Using the OP37 offset null capabilities to minimize the offset does introduce offset drift as the AD data sheet describes on page 11. The TI OPA37 data sheet indicates this as well on page 9. Thus, the offset will move in response to a temperature change. The drift behavior for any given amplifier isn't predictable and any two amplifiers may have very different drift characteristics.

The TI OPA32 data sheet states, "Large system offsets can be nulled without drift degradation by input summing." When I review the UAF42 digitally programmable notch filter circuit it does look like there are nodes where a counter offset could be summed into the circuit and not affect the filter characteristics. The 4th op amp, the auxiliary amplifier, is used as a summing amplifier in the notch circuit. I believe a dc offset could be summed into its summing node pin 5, or applied to the non-inverting input pin 4, without altering the filter's ac characteristics.

Another possible point where a dc offset could be applied to the UAF42 is pin 11. This pin is common to the non-inverting inputs of the two integrators. Their outputs sum at the auxiliary amplifier's input. Connecting an offset source to any other points on the UAF42 would likely result in the filter's ac characteristics being altered.

Regards, Thomas

PA - Linear Applications Engineering 

Hello Thomas!

I would like to do tests for this, but due to lack of time and budget, I'm not allowed to. So I'm only able to "dry swim" with theroretically discussions. If I would find a very promising idea, I maybe can get funds to evaluate this.

I have been considering your thoughts and maybe you are right that adjusting at the OP-amps is not a good idea in the aspect of adding of errors. But... if I use the design without trimming at all, as in manufacturing adjustment procedure before the external fixed serial resistors is mounted, usually the outputs are saturated at +/-15V when setting the lowest frequency 3Hz. So maybe the pot's at the OP's are contributing with thermal errors, but the absolute offset errors is till there and they are really big.

Since I have made all outputs from UAF42 available by MUX switching all three outputs needs to be adjusted. Today, as I described, there are two adjustments for the three outputs; HP and LP and the BP offset is a result of the two previous and this use to be good enough,that is, if the two adjustments can reach and capture the offset.
The notch filter is available as an option by a strapping selection between BP/Notch. The analog MUX that selects outputs and by-pass has only four channels, so that's why the Nocth/BP is selected by strap. As standard only HP/LP and BP are used. So pins 4 and 5 are not useful for the main filter outputs.

My concern is, if I use the remaining possible trim option pin 11 as a summing point where I can introduce a DC-level, will this be ok for all three standard outputs HP/LP/BP? Or will this be an insuffiecient compromise for all three outputs? My feeling is that an individually adjustments at least as today are needed. I mean, if all three outputs shall be adjusted 1V at the same direction, it would work with a common adjustment. But if one output has to be adjusted in one direction and the other at the other directiion and some other amount of Volts, there will be a problem.

What do you think?

Hello Lars-Olof,

I understand your concerns about injecting an offset and how it will affect the different filter outputs. I have put together a "discrete" UAF42 TINA simulation circuit to help evaluate offset adjust possibilities. It isn't the full digitally adjustable notch filter, but rather a fixed 60 Hz notch filter to keep the simulations reasonable. You can find the TINA circuit file in the attachment.

The simulation circuit shows two points where an offset injection circuit is connected. One is connected at pin 6, the summing junction of the UAF42 auxiliary amplifier, and the other at pin 11, the common non-inverting inputs of the two integrators. Adjusting the Setting (%) of the potentiometers changes the injected offset.

The effect of the offset applied at the auxiliary amplifier is limited to just that stage output; the band-reject filter. However, if the offset is applied at non-inverting integrator inputs the offset effect is observed at all of the filter outputs. You will note that the may take on a different sign depending on where it is observed in the circuit. Adding the offset at either of these two circuit points doesn't appear to change the filter responses. I don't know if that will remain the case for the digitally programmable circuit.

There are likely other summing junction points in the UAF42 where an offset could be injected. The simulation circuit could be used to evaluate the effects.

Regards, Thomas

PA - Linear Applications Engineering