Any cunning way to zero-shift an LVDT output?

Four wire LVDT and you want this adjustable?

Take a second LVDT, connect the LVDT primaries in parallel and the secondaries in series. Adjust the second LVDT core in/out to zero the first.

Cheers

Without thinking this through,...add a small amount of AC to the sensor coil's output? It's fixed frequency, so you can shift phase as needed.

How about a transformer in parallel; just couple excitation into output ? A few resistors to adjust offset... Let us know how you do it, Best Regards, Dave

Stick in a small transformer, like a MET-37, and a trimpot maybe. Pick off the excitation, scale as required, add in series with the secondary signal.

What is the configuration? 5-wire? I could sketch it up if I knew the details. I'm deep into an LVDT simulation project right now, been thinking about this stuff a lot.

You want a ratio transformer, as used at National Standards Labs. The individual taps can be identical to about one part in 10^7 and you can use them to slice and dice AC signals with remarkable precision

If you've got access to a good technical library, you may be able to get hold of a copy of Rayner and Kibble's book

For your application you'd set up the ratio transformer to give you a well controlled fraction of the AC signal being used to excited the LVDT and you'd use this signal to back off or augment the signal coming out of the sensing winding on the LVDT.

-- Bill Sloman, Nijmegen

Basically you inject an AC signal into the mix, to add to one pickoff coil and (possibly) subtract from the other. How you do this depends entirely on -- you.

But you need to be very careful about linearity. LVDTs come in a variety of flavors, designed for specific methods of converting to a position measurement, and the result may not be linear when you go and mess with them by adding in signal.

(This came as quite a surprise to me when I discovered it: the two main flavors of 5- (or 6-) wire LVDTs are one where x = k * (|A| - |B|) / (|A|

• |B|), where |A| and |B| are the rectified and averaged coil outputs, and another where x = k * (|A| - |B|). In general, each of these is only linear in the service for which it is designed. Much careful reading of LVDT data sheets is necessary to get the desired outcome.)

Great suggestion, THANK YOU. Should have thought of that ;)

I think you are interfacing to slightly more than one LVDT ;)

My one is here

The two AC signals come out from the LM358 outputs.

John Larkin wrote

That, however, appears to do just a zero shift in one direction.

If I wanted to shift the zero in both directions around the current zero, then I need to have a way of adding variable amounts of the excitation signal to one output or the other.

It seems easy enough with op-amps, and done with just a trimpot, no transformers.

Make two copies of the excitation, one the other inverted, and connect the ends of the pot to the copies, take the correction signal from the middle.

(Just wondering what is so difficult here ...)

Trimpots are fine, until you start needing real stability and accuracy, and you can't put a turns counting dial on the front to find out where you have set them. Ten-turn servo-quality potentiometers are much better, if still not as good as ratio transformers - but they are much cheaper, if not exactly cheap.

-- Bill Sloman, Nijmegen

Have you tried looking at the circuit on Page 337 of Walter G Jung's "IC Op- Amp Cookbook" (third edition). This is an instrumentation amplifier with a zero off-set adjust and differential input. A bit more digging and you will get the differential output configuration as well.

If the eventual destination is into a micro-controller, you might also consider some mathematical corrections. That I will leave as your exercise.

We do it on sensors at work, a pot across the works with the center tap (wiper) to slightly drain off one side which offsets the balance.

All units I've used is doing a basic level null mix from a single frequency source, you only need to offset the high sides of either the primary or secondary and we do it with a pot with some added wing R's so that the effects are only marginal and the wiper hits the common.

Jamie

You want to go both ways? A transforrmer with two secondaries would work. They are common too.

Sure, if you have power available. I was thinking that you might want a "black box" hack to the LVDT.

Your schematic already has two pots. Just turn them!

John Larkin wrote

No; that won't shift the zero.

Imagine you are at zero output already and you want to shift it. The pots won't help. One has to inject an AC signal in there.

"Paul E. Bennett" wrote

I found some appnote from TI on how to do it all with a DSP. That would be the proper way in production, of course.

Bill Sloman wrote

In this application, one would not want to shift the full scale values. One would want to do a very small zero adjustment, of the order of 1-2% of FSD.

I think a transformer, whose secondary is in series with one of the output coils, and whose primary is fed from the wiper of a trimpot connected across excitation and /excitation, will do this fine.

Only if the impedance of the trimpot (or potentiometer) is a lot lower than that of the transformer. You could buffer the output of the trimpot with an op amp, but the offset in the op amp then applies a persistent DC voltage across the transformer and if you don't take care to deal with this, the current through the transformer can build up to a level which can force the op amp output into saturation.

Once the op amp output is close to saturation, the current through the transformer will clip at one extreme or the other, compensating for the off-set, so you may not notice, but it's not a good way of dealing with the problem.

-- Bill Sloman, Nijmegen

Are you sure? Skewing the gains will simulate core motion. LVDTs usually operate at a small fraction of their total motion transfer curve, so that things stay linear, so a gain skew should work. It's not much different from injecting a pair of AC voltages, at least for modest simulated offsets.

Connect a trimpot across the secondary in the obvious manner. Try 10K to start.