Pardon me if this topic has been discussed. I did a search and didn't find a specific thread for it. I'll appreciate it if you can direct me to one.
I was on a site to calibrate a GCV with MKVIe. This was my first time. The customer engineer chose to take the lead on the job. During the polarity check, the valve moved as it should (Open & Close.) However, during null bias measurements, he reported that the servo voltages recorded were positive.
1.Can we have correct polarity and still measure a positive voltage value? Or could this have been because the wires had been interchanged?
2. As part of the checks, we're told to look out for smooth movement of the valve. Isn't it enough that the valve responds appropriately to the command (Open and close)?
3. To check the polarity, the customer gave valve commands of +10% and -10%. I thought 100% would have been better. Does this matter?
Describe the method used to check/verify BOTH servo-current polarity AND to adjust the null bias.
Describe how the servo current(s) was(were) being measured.
Was the servo-valve new, refurbished, or in-service for some time prior to this "calibration?"
What was the null bias set to prior to the start of this "calibration?"
GCV current--when the null bias is properly set--should be between -2.67% and -4.0% per processor, which means a negative servo current because negative servo current is required to increase fuel flow--and the null bias is required to overcome the failsafe spring tension in the servo-valve which is trying to reduce the flow of air (or flow for IGVs).
If the null bias is incorrect, then it's entirely possible the servo current could be positive.
I realize I'm one of the only people on the plant Earth saying that the null bias current calculation which is in the Control Specifications provided with GE-design heavy duty gas turbine control systems is incorrect--but that doesn't make me wrong (or even correct, I guess...!). But, in my three-plus decades of experience, it's simply NOT possible to calculate the null bias current required for proper operation of a servo-valve as used on a GE-design heavy duty gas turbine. There is a RANGE of allowable values (-2.67% to -4.0%), and--trust me--there are a LOT of turbines operating inside this range with absolutely NO problems, and there are even many units operating outside this range with no perceived problems (but probably with several ignored Diagnostic Alarms).
It should NOT be necessary to adjust the null bias spring tension of a GE-design heavy duty gas turbine control system if the servo polarities have been properly verified to be correct, AND the servo-valve is in good working condition with the failsafe spring is properly adjusted (from the factory!), and the hydraulic actuator is working correctly. I've done literally hundreds of LVDT calibrations without ever having to adjust the null bias setting. And, when I've had to adjust the null bias setting it's either because the servo was recently refurbished, or it had been in service for as long as 20 years (and needed to be replaced), or someone had mucked with failsafe spring tension (which rendered the servo useless). A few times I've tried adjusting the null bias settings to overcome a worn hydraulic actuator, but that's always come back to haunt me--ALWAYS. A worn actuator can't really be "fixed"--even temporarily--by adjusting the null bias settings. Because, people forget to change the null bias settings back, or they think they can "prolong" the time to change the actuator by continually changing the null bias settings.
And, trying to adjust the failsafe spring tension in the field without proper test equipment is just asking for (more) trouble.
So, tell us how this "calibration" was being done. Actually, tell us first why it was felt it was necessary to do a "calibration"--was a check of the existing calibration performed and found to be inaccurate? Was the GCV replaced or worked on? Was one or both of the LVDTs replaced?
LVDTs are NO different from pressure switches or temperature switches or any kind of pressure- or temperature or current- or voltage- transmitter--when performing a "calibration" one usually checks and records the as-found condition, and if no adjustment is required none is done. When people say they are "calibrating" devices during an outage what they're really doing is verifying the devices are calibrated--and working properly. A few devices have usually either failed, or the calibration may have drifted--and "calibration" finds these things and appropriate action is taken. BUT, ONLY AFTER a check of the as-found condition is made--and LVDTs are NO different from any other field device or instrument.
Lastly, it's simply false and a fabrication (widely-held fabrication, but still false) that "calibrating" the IGVs or a fuel control valve has ANY effect on the servo-valve. NONE. ZERO. ZILCH. NADA. ZIPPO. NIENTE. The ONLY thing that gets calibrated is LVDT feedback. Full stop. Period. End of discussion.
There have been increasing reports of the failsafe spring in electro-hydraulic servo-valves which are new from the manufacturer (Moog) being set incorrectly. This should be fairly obvious--if the null bias has to be anything more or less than -2.67% to -4.0% then there's something wrong with the failsafe spring tension and/or the servo internals.
>2. As part of the checks, we're told to look out for smooth
>movement of the valve. Isn't it enough that the valve
>responds appropriately to the command (Open and close)?
"Smooth" operation DOES NOT ensure proper polarity. In fact, during manual positioning of servo outputs some of the GE-design turbine control systems have slower-than-normal output changes--which can lead to "jumpy" operation. And, isn't "smooth" a relative term which can mean something different to just about every person?
>3. To check the polarity, the customer gave valve commands
>of +10% and -10%. I thought 100% would have been better.
>Does this matter?
+10% is not a bad, reference, but it's certainly not a good one either. Certainly, -10% is NOT a proper valve command (because the valve shouldn't go to -10 of stroke....).
Polarity is NOT a function of smoothness--it's a function of whether or not a single coil, with a slightly negative current (-2.67% to -4.0%) can keep a control valve or the IGVs open to some mid-stroke position (say aproximately 50%). Of course, it won't hold the exact reference being used, because the null bias currents from the other processors/coils aren't being applied--but if the reference is 50% the device should move to and maintain some mid-stroke position when under the control of a single processor/coil. If it doesn't open, or it slams shut--then the servo current being applied is incorrect. And, this needs to be verified for ALL outputs. (SIMPLEX and DUAL REDUNDANT control systems use two coils, driven by a single output--but there are either three or four wires at the TSVO or PCAA terminal board, for the two coils.)
The subject of how to verify polarity on a TMR turbine control system (and we don't know if this was a SIMPLEX, DUAL REDUNDANT, or TMR control system....) has been covered many times before on control.com. Many times. It should only be done under the control of one processor, or one coil, at a time. And polarity checks are to be done BEFORE null bias current is checked, or adjusted--if even necessary.
Provide more details, and we can provide more information. Search the forum for procedures for verifying servo current polarity.
So, I need to make a correction. The allowable range of null bias current is -0.13% to -4.00%, or -2.67% +/-1.33%--for a TMR control system. That corresponds to -0.8 mA, +/0.4 mA TOTAL for a TMR control system, where 100% equals 10 mA (and -100% equals -10 mA).
And, it may be helpful to understand WHEN it's necessary to change the null bias setting (within the allowable limits).
First1, let's look at the null bias calculation procedure in the GE Control Specification. It says (in part):
"--Determine the null bias voltage by measuring the voltage across each servo coil at the TCAS cards. Place the positive lead on SQnH (where Q = R,S,T) and the negative lead on SQnL. Refer to the I/O Elementary Diagram for terminal locations.
"--Calculate the individual servo currents by dividing the servo voltage by the servo coil resistance (1000 ohms for gas turbine servos). The null servo current should be -0.267+-0.13 mA per controller. Notify Controls Engineering through customer service if the sum of the currents is not within specifications. An adjustment to the mechanical null bias, located on the side of the servo, may be necessary. Note that a Normally Closed valve should be opened with negative current because we want the valve to fail null bias closed."
That's a direct copy-and-paste--from a Mark VIe Control Specification for a GE-design heavy duty gas turbine. (And the last sentence of the above passage is just completely whacko.)
If anyone can tell me how to relate resistance (ohms) to failsafe spring tension, I could understand how this might work. Servo-valve coils are not always exactly 1000 ohms. The length and size of the wiring between the Mark* terminal boards and the coil leads is not always exactly the same and should factor into the equation. The temperature of the coil should also factor into the equation.
Nowhere in the "instruction" does it say to determine what the 100% stroke position is, but it does say if the null bias calculation is not correct an error could result between the GCV reference and the GCV actual position could result if the "...bias is off." The "instructions" go on to say to "...Measure the GCV position. The GCV should be at the 25% stroke position as indicated by the dial indicator..." But--how does one know what 25% of actual, physical travel is--unless it's specifically stated in the Control Specification (which it used to be decades ago) or unless it's physically measured before doing any verifications after making changes to the null bias current? (25% was from the example in the Control Specification.)
EVERY time I have seen people use this "procedure" they NEVER measure the physical stroke. NEVER. EVER. So, even if the calculation was correct or the procedure was correct, what good does it do if the actual, physical stroke is compared to the reference position? The idea behind changing/adjusting the null bias current setting is to make the actual, physical position (AFTER the calibration has been completed and verified!) to be as close to the reference position as possible. People assume the LVDT calibration is PERFECT and that the indicated position on the screen is EXACTLY EQUAL to the physical position of the device--the measured, physical position with respect to the full stroke travel (for a GCV). So, why use a procedure if one is only going to follow it partially by not physically measuring the actual position and making sure the indicated position (the "calibrated" position feedback) is equal to the actual, measured position??? Can one just decide which portions of a procedure to use?
And, I have seen the null bias current settings changed to +10 and -10 (which, I believe are the programmed limits of adjustment in the Mark* turbine control system) because people don't understand what changing the null bias current is supposed to do. I've been told of sites taking two or three DAYS to try to calibrate "the IGVs" (or the SRV, or the GCV, or the LFBV) and adjust the null bias current--only to be unable to start the unit.
And, again--if someone can tell me how typical coil resistance (and not actual total circuit resistance at the time the calculation is being performed)--can be correlated to failsafe spring tension then I might think the entire procedure may have some merit.
The process should be:
--Check current calibration against actual, measure, physical position (as related to total physical stroke--for a GCV). This means that someone needs to know what the actual, physical stroke of the device is in order to be able to calculate percentages when making measurements--which means that one needs to measure the actual physical stroke prior to beginning any calibration procedure, and then measure the actual stroke during any verification or after any calibration of LVDT feedback.
--Perform an AutoCalibration of LVDT feedback, if necessary--AFTER resetting the null bias current back to the default value of 2.67 (which is inverted in the Mark*) before beginning the AutoCalibration procedure.
--Verify the accuracy of the LVDT calibration by comparing the reference and indicated positions to the actual, measure physical position. If the indicated LVDT position is approximately equal to the actual, measured physical position--BUT the actual, measure physical position is not approximately equal to the reference position, then, and ONLY then should any change to the null bias current setting be made.
A proper LVDT calibration (using AutoCalibrate) is one that proves the indicated position on the screen is approximately equal to the actual, measured physical position of the device. There may be a slight difference between the reference position and the indicated/actual position--but that's not the "fault" of AutoCalibrate. Calibrating LVDT feedback means verifying the indicated position on the screen (after the feedback has been scaled) is approximately equal to the actual, measure physical position--NOT that the indicated position is equal to the reference.
Once the indicated position on the screen is proven to be approximately equal to the actual, measured physical position, if there is a "large" difference between the reference and the indicated/actual position THEN--AND ONLY THEN--should one even think about adjusting the null bias setting from the 2.67 value (-2.67%, in reality).
One DOES NOT change the null bias current setting to make the indicated LVDT position (on the screen) equal to the reference position--one changes the null bias current setting to make the actual, measured position (which should be approximately equal to the indicated LVDT position on the screen!) closer to the reference position. Unless the actual, measured physical device position is approximately equal to the indicated LVDT position on the screen, the LVDT calibration was not done correctly. People mistakenly believe that the indicated LVDT position is ALWAYS exactly equal to the actual, physical position--and they never measure to verify.
If the actual, measured physical position is equal to the indicated LVDT position AND there is a "large" difference between the reference position and the indicated/actual position--there is NO NEED to change the null bias setting from the typical 2.67 (which is inverted in the Mark*). And, if a change is necessary, for example if the indicated position and the actual position are nearly equal but less than the reference position then a small amount of current should be 'added' to the null bias current setting.
For example, let's say the indicated position and the actual, measured physical position are both approximately 22.9%, and the reference position at the time of the verification is 25.0%, then I would say a small change in the null bias current setting is warranted. If the existing null bias current setting was 2.67 (-2.67, in reality), then I would increase the setting to 3.00 (-3.00, in reality), download and re-boot and check the indicated/physical position against the reference again. The "addition" of a small of null bias current should result in an increase of the indicated/physical position--as well as the current displayed on the screen (which should have been approximately -2.67% per processor prior to the change, and which should have changed to approximately -3.0% after the change).
Changing the null bias current should cause the indicated/actual position to more closely match the reference position. BUT, unless the indicated position is nearly equal to the actual, physical position (by measurement), then what's the use of changing the null bias position?
Now, the curious reader is asking, "How can a turbine run if the calibrations aren't usually done correctly--including adjusting the null bias current?!?!?" Well, it's because most of the control (after initial firing and warm-up) is done closed-loop--so, during acceleration there is an accelerate rate reference, which is compared to the actual acceleration rate, and the GCV reference/position is adjusted as required to make the actual acceleration rate equal to the acceleration rate reference. And, if the valve has to go to 19.8% indicated position (when it's really only at 17.6%) to make the actual acceleration rate equal to the acceleration rate reference--then the Mark VIe will do what's necessary--because, the "loop" is acceleration rate, and the valve is moved to whatever position (indicated/actual) that's required to make the actual equal to the reference.
Exhaust temperature control is exactly the same--the GCV is moved to whatever position is required to make the actual exhaust temperature equal to the exhaust temperature reference. The "loop" is exhaust temperature--not gas valve position. The gas valve gets moved to whatever indicated/actual position is required to make the actual exhaust temperature equal to the exhaust temperature reference.
So, hopefully this helps to understand why it might be necessary to adjust/change the null bias current setting. Either the servo is bad, or the hydraulic actuator is bad, or there's a hydraulic flow problem, etc. It's not to make the indicated position on the screen equal to the reference--UNLESS the indicated position has been verified to be approximately equal to the actual, measured physical position. A proper LVDT calibration (because one is NOT calibrating the GCV, or the IGVs, or the SRV--one is ONLY calibrating the LVDT feedback from these devices, and nothing more) is one that has been verified to prove the indicated position (after the calibration) is approximately equal to the actual position, by measurement.
And, if the required adjustment causes the setting to be less than 1.33 (-1.33%) or more than 4.00 (-4.00%)--to make the actual position equal to the reference position--then something else is wrong.
If there's a "large" difference between the indicated/actual position and the reference, then--and only then--should the null bias be adjusted from the normal, typical 2.67 (-2.67%, in reality). And, the limits of adjustment are from 1.33 (-1.33%) to 4.00 (-4.00%).
Whenever an LVDT calibration is checked/verified (such as during or after a maintenance outage) and found to be out of calibration, the null bias current should be reset to 2.67 before anything else is done. And, again, a PROPER LVDT calibration is when the actual, measured physical position is approximately equal to the indicated position on the screen--NOT when the indicated position is equal to the reference position. When the indicated position is equal to the actual position BUT not equal to the reference position, then one can adjust the null bias current--within the limits of adjustment--to make the indicated/actual position equal to the reference position.
LVDT calibration is NOT necessary after a servo-valve replacement. (Because replacing the servo-valve does nothing to the physical stroke of the device, nor does it change the adjustment of the LVDT(s).) AutoCalibration ONLY calibrates LVDT feedback; it has nothing to do with the servo-valve stability or gain.
Servo coil polarity should be checked whenever a new (or refurbished) servo-valve is installed, AND, whenever there is a question about the polarity. Smoothness has nothing to do with polarity.
And, whenever a new or refurbished servo-valve is installed, the null bias current setting should be reset to 2.67 as part of the procedure. LVDT calibration should be verified after the installation, but an AutoCalibration is NOT required, and ONLY changes the LVDT feedback scaling.
And, servo polarity has nothing to do with "smoothness," either. That's another false part of the Control Specification.
Thank you so much for your elaborate explanation.
I've always felt the Control Spec provides little information on calibration. But I had no idea some of the info could be false. I've had to reread your explanation a couple of times to grasp everything and unlearn what I know. Yes, I have been adjusting the null bias to match the LVDT feedback with the reference, without verifying! And I didn't know calibration doesn't affect the servo either. Thanks again for giving me a better understanding of what I've been doing.
(Apologies for the late response.)
My confusion has been cleared by your in-depth explanation.
But just in case, here are answers to provide more detail:
*Method for Servo Polarity: Valve was commanded open and closed and monitored to confirm it followed command.
*Servo voltage measurements: I was not present when they were taken but since valve responded to command (open and close), I think it's possible the person interchanged the probes.
*Null Bias change: By erroneously matching LVDT feedback to reference
*The site was installing a new valve so I assume new servo.
*Null bias was set to 2.5%. I know now I should have set to 2.67%
before beginning calibration.
*The system was a TMR.
Thanks for your help.
Thanks for the feedback! I'm a terrible proof-reader of my own writing, and there are a couple/three errors in my writing which I hope didn't cause a great deal of confusion.
Servo current polarity and "smoothness" of travel are NOT related. The device will move if a single coil is receiving the wrong polarity. And, the device will move under the control of a single coil--that's why it's important to verify servo polarity under the control of a single coil at a time to make sure the polarity of the current being applied to all three coils is correct. It's not hard, it doesn't even take that much time. And, it's absolutely the final and decisive indicator of proper servo polarity. Any other method is "smoke and mirrors" based on an incorrect understanding of how servos work.
The purpose of AutoCalibrate is ONLY to calibrate (scale) LVDT feedback. The feedback must be linear over the range of travel of the device (fuel control valve; IGVs), so the zero-, or minimum-, stroke voltage must be set correctly per the LVDT being used. (Typical LVDTs used for MOST--but NOT ALL--applications on GE-design heavy duty gas turbines is 0.700 VAC RMS, +/-0.020 VAC RMS, or from 0.680 VAC RMS to 0.720 VAC RMS).
Calibrated (Scaled) LVDT feedback (also referred to as "indicated" LVDT feedback) should always be compared to actual, measured physical position of the device. And once the indicated LVDT feedback is approximately equal to the actual, measured physical position AutoCalibrate is deemed to be successful.
An LVDT (or multiple LVDTs) only measure the physical travel of the device which is being moved by hydraulic pressure (or, sometimes pneumatic pressure, or sometimes an electric actuator, an "electraulic" actuator, or even a manual positioner). The LVDT just measures "position"--which can be scaled in per cent (%), or mm or cm or inches or mils, etc. AutoCalibration is used to scale the LVDT feedback so that the indicated position (on some screen) is approximately equal to the actual, measured physical position.
AutoCalbrate DOES NOTHING to the operation or "calibration" or "adjustment" of an electro-hydraulic servo-valve or any of its control parameters--NOTHING. Before "calibrating" ANY LVDT on a running unit, one should determine the as-found condition of the LVDT feedback. If the as-found condition is out of tolerance with the actual, measured physical position--then a re-calibration of LVDT feedback is necessary. If the as-found condition is withing acceptable tolerance, then no calibration is necessary! Just note it so on the calibration data sheet for the LVDT(s) (every site uses them, right?)--and move on to the next device to be verified ("calibrated").
But, since LVDTs are used on aircraft flaps and ailerons--specifically because they DON'T drift very much, if at all, when properly adjusted in position, would one expect LVDT calibration to suddenly be incorrect on a gas turbine application? And if one is not physically measuring and observing the device movement when checking (verifying) LVDT calibration) If one ONLY changes a servo-valve, it IS NOT necessary to calibrate the device--because one IS NOT calibrating the IGVs, or the fuel control valve using AutoCalibrate. One is ONLY calibrating LVDT feedback, which is a function (in most cases!) of the physical range of travel of the device (the IGVs; the fuel control valve; etc.).
Changning a servo-valve has nothing to do with changing the range of travel of the device; NOTHING. So, it's unnecessary to use AutoCalibrate after a servo-valve is replaced. In fact, when people say, "Using it [AutoCalibrate] can't hurt!" after replacing a servo-valve, it actually can hurt if not performed correctly--especially on units with DLN combustors. So, why do it if it does nothing to affect the servo-valve, and, can in fact, cause problems with operation? It's just a long-held--and FALSE--belief that AutoCalibrate affects servo-valve operation in ANY way. Completely false. Unless, the servo-valve stability is caused by faulty LVDT(s), in which case AutoCalibrate may not solve that problem, either! Or a wiring problem with the LVDTs. Or, an improper zero- (minimum-) stroke LVDT adjustment. And, replacing the servo-valve is not the proper solution for those problems, either.
Servo null bias current is only used when there is "large" discrepancy between the reference position and the indicated position (IF and ONLY IF the indicated position has been proven to be approximately equal to the actual, measured physical position). THEN, AND ONLY THEN should the null bias ever be adjusted to something other than 2.67, and the limits of adjustment should always be between 1.33 and 4.00. If it's necessary to adjust the null bias outside either of those limits, there's something else very wrong with the circuit/system/actuator.
And trying to use coil resistance to change current values as some kind of correlation to failsafe spring tension is just, well, it's not really relative. Until such time as someone can prove otherwise--it's just incorrect to relate current calculated from coil resistance to failsafe spring tension.
I hope this helps--and doesn't add to any misunderstanding. AutoCalibrate is a VERY misunderstood application/tool, and servo-valves (as used on GE-design heavy duty gas turbines) are also very misunderstood. And, there are LOTS of false beliefs ("tribal knowledge"; "wives' tales" (gossip)).
I want to add one more clarification. On a working three-coil servo-valve if the polarity of current being applied to one coil is wrong then the valve will still move and allow flow (of fuel or air). The error between the reference and the indicated (actual) will be a little larger, and the three currents will be unbalanced--usually very unbalanced. [NOTE: An incorrect servo current is not the ONLY reason for unbalanced servo currents on a TMR control system, but is a fairly common reason.]
However, if the polarity of two coils is incorrect the valve will not move--because those two coils are working WITH the failsafe spring to shut off the flow of fuel or air. Null bias current is what is added to the servo-valve output to overcome failsafe spring tension--which is ALWAYS working to try to shut off the flow of fuel or air. So, when two of three coils are working against the failsafe spring and one is working with the failsafe spring the device will move and control flow--but it won't closely match the reference, and one or both of the processors which ARE correctly connected to the servo-valve coils will be putting out a LOT more servo current to overcome the failsafe spring tension and the incorrect coil's action.
If a TMR control system (with no relevant Diagnostic Alarms!!!) is running and one processor shuts down or has to be shut down, if one of the two remaining servo coils has the current being applied incorrectly, the unit will trip--because one of the remaining coils is working WITH the failsafe spring to shut off the flow of fuel or air. If it's a fuel control valve, the likely cause of the trip will be "LOSS OF FLAME TRIP" (everyone's favorite trip!); if the device with an incorrect remaining servo coil polarity are the IGVs then there will likely be an "IGV POSITION TROUBLE TRIP." Again, this is if there are no other relevant Diagnostic Alarms trying to alert the operations/technicians to problems with the servo or the LVDTs or the associated card(s).
When two coils have the incorrect current polarity applied to them along with the failsafe spring tension--the device will move to shut off the flow of fuel or air.
BUT, when performing servo-valve polarity checks under the control of one processor (two of three servo-valve coils are disconnected--which is the PROPER way to check servo current polarity) one processor has the capability to put out sufficient current (when connected so the polarity is correct) to overcome the failsafe spring tension and cause the device to move to increase the flow of fuel or air. The total amount of current which will be put out by the control processor will be nearly equal to -0.8 mA, +/-0.4 mA (-8%, +/-4%)--as it has to be putting out sufficient current to overcome the failsafe spring tension. And, if the connection to the servo coil is correct, a single Speedtronic servo-valve output can move and control the device properly. The error between the reference and the indicated (actual) will be large--but that's to be expected.
The important thing is that when under the control of a single processor the device will move to the position that increases the flow of fuel or air when a positive reference is applied. If the current being applied is incorrect, the device will not move when a positive reference is applied. So, for example, with a reference of 50% the device would move to nearly 50% when all three coils are connected to the Speedtronic control processors. When one coil is disconnected the device should remain somewhere near 50%, but if the polarity of both remaining coils is correct the device will lose a little bit of position with respect to the reference, but not much. If the device moves to shut off the flow of fuel or air under the control of only two coils, the polarity of one of the coils is incorrect and needs to be changed.
When the second coil is disconnected, the device will lose a little more position with respect to the reference, but it should remain near mid-stroke (with a reference of 50%). If it moves very quickly to shut off the flow of fuel or air the polarity being applied is incorrect.
Once the polarity of one coil is deemed to be correct, reconnect one of the two remaining coils--the device should remain at near mid-stroke (with a reference of 50%). It if moves quickly to shut off the flow of fuel or air the current being applied to the second coil is incorrect and the connections of that coil need to be reversed. If the device stayed at near mid-stroke when the second coil was connected, then disconnect the first coil--and the device should remain at near mid-stroke.
Finally, reconnect the third coil while the second coil is still connected and the device should remain at near mid-stroke (with a reference of 50%). If it stays near mid-stroke, disconnect the second coil and the device should continue to remain at near mid-stroke--if it doesn't, reverse the connections to that coil and the device should return to near mid-stroke again.
And, that's it--re-connect each of the two disconnected coils again, one at a time. The device's position should get closer to reference when two coils are connected, and even closer to reference when the third coil is connected.
This is to be done BEFORE an AutoCalibrate (during commissioning), and whenever a servo-valve is replaced, or the servo-valve coil connections are disturbed (such as during maintenance or refurbishment of the servo-operated device or nearby devices). It has been reported, and I have experienced, new, out-of-the-box servo-valves with coil lead colors that were NOT correct (did not match the servo-valve coil lead colors being removed). DO NOT rely on coil lead colors when installing a servo-valve as verification of coil polarity. The existing servo-valve may not have its coil polarity properly verified, and the leads could be incorrectly terminated (with respect to colors). ALWAYS check servo-valve polarity whenever installing a new servo-valve, or when the servo-valve wiring has been disturbed--anywhere along the circuits.
Finally, it is NOT necessary to perform an AutoCalibrate when a servo-valve is replaced--but it is VERY important to perform a servo-valve polarity check. No matter what anyone tells you, or what you've observed someone else do--it's not necessary, and can cause problems if not done correctly. Not with the servo-valve--because the servo-valve is UNAFFECTED by AutoCalibration (no matter what anyone tells you what you have observed others doing or saying). ONLY LVDT feedback is affected by AutoCalibrate. ONLY devices with LVDT feedback can be "operated" manually using AutoCalibrate. (If one or both LVDTs are replaced at the same time as a servo-valve, then, of course, it will be necessary to perform an AutoCalibration. But, if ONLY the servo-valve was replaced performing an AutoCalibrate does NOTHING to the servo-valve, only LVDT feedback scaling (calibration).)
Hope this helps!