I'm going to split some hairs here (be very, very detailed) and make some corrections to what you're describing. This topic can't really be covered completely in a forum like this; it's complicated. But, we're going to take a short swing at the high points.
In the Speedtronic turbine control panel, the error between a servo-valve output's regulator feedback and its reference is converted into servo current. When the feedback of a regulator is equal to the reference the error is zero, so zero error would mean zero current. But, if zero current is applied to the servo-valve, the fail-safe spring in the servo-valve will make the hydraulic actuator move to shut off the flow of fuel or air or steam.
Null bias current is the value of current that is added to the servo-valve output current to overcome the tension of the fail-safe spring in the servo-valve. So, some amount of current must be added to the output when the regulator error is zero (when the regulator feedback is equal to the reference) to provide sufficient current to overcome the fail-safe spring to keep the device in position to maintain a steady flow of fuel or air or steam. In the Mark V, null bias current is a *fixed* value of current, defined in the I/O Configurator, that is *added* to the output to overcome fail-safe spring tension.
In the Mark V, servo current is expressed as a percent of full-scale servo current. 100% servo current is equal to 10.0 mA, so 0.1 mA equals 1.0%.
The servo current values you see when the unit is running or when you are manually positioning a device are almost never the null bias currents. The servo current values you see when the unit is running are the total servo current being put out by the control processor, *including* the null bias current value. You can't really see the null bias current portion of the total current that's being applied to the servo coils *unless the feedback is nearly exactly equal to the reference.* Then and only then is the servo current value being displayed equal to the the null bias current and only the null bias current.
And this is done by each control processor independent of the others in a TMR control panel. So, if one control processor thinks the feedback for some device is different than the reference and different than another control processor's or processors', the total amount of current from the control processor will be different than the current from the other control processor(s). Each control processor will have the *same amount* of null bias current added to its output, but each control processor's output can be different if each control processor thinks its feedback is different than the others'.
When the regulator feedback is different from the reference, then the control processor will adjust its total current output *which includes the fixed null bias current value* to try to make its feedback value equal to the reference value. And each control processor is doing this for every servo-valve output. Again, the null bias current is a fixed value, defined in the I/O Configurator, which is always added to the total current output of each control processor.
When something like what you are asking about happens, you need to find out what the feedback values are for all three control processors for the servo output and you will likely find that one or two of them are very different from the other(s). If all three control processors don't think the feedback is the same and equal to the reference (the reference should be the same for all three control processors), then each control processor will adjust its servo output current to try to make its feedback equal to the reference.
For example, consider the GCV servo output. The GCV regulator feedback is the high-selected value of LVDT feedback from the two LVDTs on the GCV. Let's say that <R> thought the GCV position was 57.8% and <S> thought the GCV position was 55.4% and <T> thought the GCV position was 54.8%, and the reference position for the GCV was 55.2%, then the servo currents would likely be unbalanced. And probably by a fairly large amount. <R> might be putting out -3.9% servo current, and <S> might be putting out - 1.9%, and <T> might be putting out -2.5%. Those values are *not* null bias currents, but each one includes the fixed value of null bias current which is defined in the I/O Configurator.
In this example, the problem is *not* the fixed null bias current value. The problem
is that the three processors have fairly different ideas about the position of the GCV and each one is trying to move the valve to the reference position, and they all have to work together (and that means that one or two are trying to overcome the other) to keep the valve at a steady state position. The bigger the discrepancy in what each control processor believes the feedback to be, the bigger the discrepancy in the servo output currents (which include the fixed null bias current value).
Now, let's talk specifically about the null bias current value. Let's say that the value of null bias current defined in the I/O Configurator and that was downloaded to and being used by all three control processors was 2.667 % (the Mark V automatically inverts the value in the I/O Configurator!). Further, let's say all the control processors were indicating a GCV position of 49.7%, the measured position was approximately 49.8%, and the reference was 50.0% and the three servo currrents were all indicating about -2.9% per control processor, or thereabouts.
If you changed the null bias current value in the I/O Configurator to approximately 3.0 (which would correspond to -3.0%; remember: the Mark V automatically inverts the value from the I/O Configurator!), downloaded that value to all three control processors, and re-booted all three control processors, you'd probably find that the indicated valve position feedback for all three processors was nearly 50.0%, the measured position would be about 50.0%, and the servo currents would be almost exactly -3.0% per control processor. In this case, the amount of current being displayed for each control processor would be nearly equal to the null bias current amount, because each control processor thought the feedback was nearly identical to the reference *AND* because the amount of null bias current was exactly equal to what was required to overcome the fail-safe spring tension.
But it should be clear that unless all three control processors believe their regulator feedback values to be nearly identical to each other, the servo currents being put out by each control processor will not be the same. And it has nothing to do with the fixed value of null bias current being applied to the servo-valve output. The value of current that is displayed when the unit is running is not just the null bias current unless all three control processors are using nearly the same value of feedback for the device and the feedback is very nearly identical to the reference.
The amount of null bias current required to overcome the fail-safe spring is actually a range: -0.267 mA, +/- 0.133 mA, or, -0.133 mA to -0.400 mA (-1.33% to -4.00%). So, the actual amount of null bias current required for a particular servo may be anywhere between -0.133 mA to -0.400 mA and still be within spec. The value of *null bias current* doesn't have to be exactly equal to -2.667%, but 2.67% is a fairly good value and works for the majority of servo-valves in use on the majority of GE-design heavy duty gas turbines. About the only time that null bias servo currents need to be adjusted is for some DLN valves, and even then, it's questionable whether or not it's really required.
The regulator feedback is compared to the reference 128 times per second, and the total servo current output is adjusted as necessary to try to make the feedback equal to the reference. *BUT* the value that's shown on any display or in any VIEW tool capture or output is only updated four times per second. In other words, the value of servo current written into the control signal database is only updated 4 times per second, even though it could be changing at the rate of 128 times per second. (I think that's different for Mark V LM panels, by the way.)
Lastly, the LFBV uses Liquid Fuel Flow Divider Feedback as its primary control feedback and the SRV uses P2 pressure feedback as its primary control feedback. So, feedback is not always position. Some LFBVs have LVDTs as another stabilizing element of the control loop.
I am in the process of checking the calibration of PM1, 2, 3 and SRV control valves, and have run in to exactly the difficulty you describe in trying to measure null current. Since we are currently offline, I tried disconnecting two of the three servo outputs and driving the servo to 50% stroke, to eliminate the competition between the controlers. It seems that this would allow the one controller to match reference and feedback, and then go to zero output (except for the null current).