Calling LVDT (Testing) Gurus

You either didn't spec the test bench properly before you bought it or didn't take the sensitivity of the test rig and LVDTs into account before purchasing, or before accepting it.

Or, your test is flawed.

Or, you don't know how the perform the test to the controlling specification.

In any case, you have a test bench that renders the test invalid.

Not sure there is anything the forum can do to solve this self inflicted problem.

If it were mine, I'd make all the critical wires the same length. (Not really. If it were mine, I'd have spec'd it properly in the first place or been prepared to modify it to suit my test sensitivity after I accepted it.)

Unfortunately, there's never enough time to do it right in the first instance, but there's aways be time to re-do it.

I am not an LVDT test person. I just have a lot of opinions about things.

Oh, never mind--I thought you said LGBTs.

I used LVDT-s but never was confronted with the phase shift. The only lab action was centered on the calibration with their signal conditioner (mod/demod) in order to obtain their transfer function distance-signal after conditioning. This was necessary to define their linear range which can be shorter or longer depending on the acceptable degree of nonlinearity. In general the carrier frequency is such that the linear movement even at "high frequency" does not lead to an error of importance. The movement is in general in the 10th to 100th Hz range and the carrier in the range of several kHz thus no interference.

Why is for you the phase shift important ?

I guess you could rewire the test bench (preferred) or add about 40 feet of the same cable to the primary side.

Actually, you could measure the propagation delay of the 40 feet of cable and simply subtract that from the results. You should know the signal frequency. Calculating the phase angle is easy when you know the delay time and frequency.

If I remember correctly, I used to employ a balun transformer for balancing or matching impedances in transmission lines. By inserting a balun in one of the lines used in the OP's test rig may solve his dilemma? Balun transformer comes in several pre-configuration or format depending on the intended application... Depending on the unit to be tested, The OP can insert or remove the balun in-between his test lines..

G.Coastal -

Maybe its not too late yet to modify your testing rig? How about incorporating a variable impedance in your jig to take care of the varying length of lines of DUT? Balancing (creating a Null) the input and output impedance of the system maybe enough to cancel any phase shifts before doing your test analyzes?

I am interpreting this as more than a RCL circuit induced phase shift, but also the overall length of the cable and its propagation delay.

Maybe a simple cable analyzer will give you the true specifications of the test cables, then you could mathematically normalize the results or simply redefine the original phase requirements to account for the disparity of the test cables.

In other words, rewrite the test to account for the phase shift of the cables.

AH -

It appears to me that the OP failed to properly spec'd out all the requirement that will interfere or may alter the results both electrically and environmentally, for the pre-ordered test rig. My initial impression was they're more on the manual side (turning knobs & reading meters) of testing rather than the analytical side. So incorporating the mathematical aspect maybe out of the picture?

Your suggestion is good and maybe easier to incorporate though!

I've been wrong so many times, also for 70 years of making assumptions!

Anonymous Hero (#4 below) has it right. You can short out the LVDT under test and measure the delay due to input and output cables. I always use a step function (a low frequency square wave will work) as the test signal, with the scope connected to the input cable (use a Tee), and channel two connected to the end of the output cable. The cable delay is now easy to measure.

Now put the LVDT back into the circuit and measure delay again (use a hundred averages, by the way). The difference is the delay of the LVDT.

Ten degrees at forty feet happens at over 600 kHz. I don't know of any LVDT that functions anywhere near that frequency.

Yes, you are right. It is not propagation delay, but cable capacitance and inductance that are actually inducing the phase shift.

It must be really crappy cable if it introduces large phase errors in forty feet at audio frequencies. If LVDT's are being tested at RF, or, even, 20 kHz, the test procedure is wrong.

Are you capacitively coupled?

If so it would induce some issues.

"You either didn't spec the test bench properly before you bought it or didn't take the sensitivity of the test rig and LVDTs into account before purchasing, or before accepting it."

You? You act like I wrote the spec, RFQ, and did all the design reviews and acceptance of the bench. No, this is one of those fun cases where the Federal Government didn't write the spec and trusted the private contractor to fill in the knowledge holes only to find out later the private contractor didn't have the knowledge either...but stands ready and eager to suck up more of our tax dollars to fix the problems they caused in the first place.

"Why is for you the phase shift important ?"

Same reason it would be important for any LVDT I suppose, large enough shifts between primary->secondary would cause the demodulator to produce incorrect positions when it converts.

"I am interpreting this as more than a RCL circuit induced phase shift"

This is what I was suspecting and was hoping to see someone mention same. I set up a real basic SPICE simulation using ballpark figures for cable and it seems like the inductance of the cable actually has quite a large effect. When I use numbers I think are close for the cable we are using (I'm away from work on holiday at the moment so I'm not sure) I get phase shifts in both magnitude and direction similar to what I measure on the bench.

"I guess you could rewire the test bench (preferred) or add about 40 feet of the same cable to the primary side."

This is likely what we'll have to do. We haven't tried it yet because the wiring is a bit of a beast to do this and would take the bench down for a few days that we don't have to spare at the moment.

Sorry, In ALL the applications I had, the electronic module for signal conditioning was NOT so far from LVDT, some times it was even incorporated, and even if there is a phase shift problem if the physical displacement and the signal output are correlated the error is compensated. Even if the phase shift is variable along the stroke the calibration eliminate the problem. It could be a problem if you do not make a calibration, if you only base your measurements on the signal and a sensitivity from the manufacturer. Thus my question. The calibration should be made with complete system between input (displacement) and output (already conditioned signal).