LVDT for feedback?

[dmills]

Just a thought, but has anyone tried building a LVDT on the front of the drive coil for feedback?

Idea would be to glue a short ferrite bead onto the push rod right in front to the drive coil, and wind the three coils of the LVDT on a little plastic or tufnol former fixed to the PCB with the sense buffer amp (a jfet), then drive and sense can be sent over something like a sata cable to the signal processing (Two screened 100R differential pairs, practically made for this sort of thing).

LVDT are position not velocity sensing, but that is easy to fix, and the major upside is that if you run the electronics at a few MHz, a synchronous detector will completely ignore the fields from the drive coils, and the LVDT coils will only need to be a few turns or so.
The other win is that the sense coils are on the stator so you don't have issues with flexing connections.

Anyone out there played with this sort of thing?

Regards, Dan.

[markrob]

Not an LVDT, although that should work in theory. I played with eddy current sensors like were used in the Fairchild stereo head and with commercial Kaman sensors. I picked up a Kamen KD2300 on eBay cheaply, but even though it runs its sensing at 1 Mhz, it was still affected by the drive coil fields. As is, its useful up to about 5 khz before the crosstalk makes the measurement void. I think the issue is, even though its senses way above baseband, the tiny signals get swamped by the high magnetic fields that the driver produces and the demod is not totally immune stray pickup. I was going to roll my own, but never got back to it. I'm sure in the LVDT case, there will be some issues to deal with. I'll be interested to see if you are able to implement that.

Mark

[dmills]

L/C BPF in the input stage would fix the LF pickup I expect, or mix it up to 10.7MHz and use an FM radio IF ceramic filter or two to strip off the out of band the low frequency signal.

Going to give this a go, I figure a H mode mixer will work as a demodulator for this, probably based on those Fairchild bus switches that PA3AKE characterised for his HF set design.
It is a bit of a pity that Toko are out of the tuned IF transformer business which would have been an obvious approach to the filters.

The eddy current sensors are an interesting approach, but I could see them adding an unfortunate amount of moving mass given that the target really needs to be a few skin depths thick. Your problem may well have been the core acting as a sort of saturatable reactor and running up and down the B/H curve in response to the drive fields which is annoying as that is a harder fix then just fitting a LC HPF in the signal wiring.
I wonder is a solonoidal eddy current sensor with a copper core machined as a cone would work? Could do something ironless to eliminate the BH curve.

I have wondered about an optical approach, led plus large area photodiode and a triangular flag on the drive rod?

Capacitive sensor is tricky, I have worked on condenser microphone design, not sure I fancy trying it in that environment unless doing it as an RF sensor, and even that is disturbingly tricky to pull off really well, all sorts of fringing field bullshit.

I really need to pick up a watchmakers lathe to make playing with this stuff less annoying.

[markrob]

I did a quick try of the optical approach using a old OPB813-S7 I had in my shop (7 mil aperture) and it worked better than I expected, but it was still was seeing the driver EM field. Also, in my crude setup, the noise floor was pretty high.
I did not try modulating the LED with a carrier to see if that would have killed the baseband. That would have required a faster sensor rather than the transistor output in the OPB813. This is also an area that should be looked into more.

I also tried to do a capacitive sensor, but the high impedances and small capacitance changes to be sensed seem too much for me to play with in depth.

Mark

[dmills]

Yea, those slotted opto switches are not exactly designed for high speed linear operation.

Hamamatsu make some interesting linear position measuring photodiodes where the ratio of the outputs depends on the linear position of the light spot, annoyingly high resistance however.

Cap sensors are a pain in the arse, in basically all ways.

[markrob]

I picked up some faster BPW34 photodiodes with the intent to try them in that mode. I am also looking at trying a self mixing laser interferometer as a possible approach. The issue there is the need for a Bragg cell or other AOM to be able to see direction.

[dmills]

Would a AOM work?
I think I may have some kicking around somewhere coated for HeNe (I know I have at least one rather butch example normally used as a Q Switch).
Might not really have the bandwidth you would want however.

[markrob]

I love the idea of using low cost solid state laser diodes in a self mixing mode as its basically self contained, small, and very cheap. I'm no expert on this stuff by any means, but from the papers I've read, the issue is the need to do some sort of cycle counting to unwind the motion as you move multiples of a wavelength. I think that would make the bandwidth pretty low due to the processing needed. In the self mixing case, I don't know how an AOM would fit in since there are not two separate light paths. But I was considering getting the same effect by varying the diode current to provide the frequency shift. Haven't given that a try yet. Too much other stuff going on right now.

[flozki]

hello
caruso head was designed in 2009/2010 with optical feedback. test where very promissing.
the torque tube had reflectors that took the movment directly close to the stylus. way better than far away...
main reason was the idea, that the users can quickly swap a burned drivecoil. would be so easy if no tiny feedback is involved.
all was designed for optical pickup. even a small clamp to loose the links...
but it turned out that its not that easy. so i had to decide to bring the head out and sadly i used plan b with
standard feedback coils.

so if i read all this high tech ideas my tech heart says yes. but hey. just wind a coil and the problem is solved.
i know not rocket science.
i wish i would have more time to play and find alternatives.
i hope one day someone comes up with a nice solution.
looking forward to see a working version.
close the loop and start real cutting. you will find out the difficulties.

but even on the pickup front... a good old moving coil still rocks the house....


happy research.happy cut
f.