Linear sensor

If the movement is not too long, with differential coils at the ends, so the metall enters accordingly (if it has magnetic properties) ciao Ban

Wow. Big telescope? Amusement-park ride? Unless it's a very stiff structure, it'll deflect a lot more than that from gravity.

Cheers

Phil Hobbs

Il 24/02/2010 14.58, Phil Hobbs ha scritto:

Something like that :) It's a big tank and it can rotate on its rails. It's made of concrete.

Anyway, it don't deflect a lot, and the current way to read the position works quite well. But they ask me if it's possible to remove any mechanical stuff.

Bye! Marco

Il 24/02/2010 7.39, Ban ha scritto:

The full span is about 5 m cw and 5 m ccw.

Marco

=20

=20

ef.

Put some reflective decals on the rail; track them optically with a=20 camera or like a bar code scanner?

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Marco Trapanese a écrit :

Maybe have two piezos? One at the reference point, one at the measurement point. Pulse one and detect on the other. That'll give you the rail length between the points. Monitoring the rail end reflections will calibrate the propagation velocity if it's not consistent enough from rail to rail (I had to say that :-)

So the tank pivots at the ends and the curved rail is at the equator, like a premixed concrete truck? Or are there two rails? And the problem is dirt, rainwater, or what?

A picture would help a lot--maybe post it on a.b.s.e. if you can.

Generally in really grimy applications, your best bet is some sort of magnetic encoder, e.g. variable reluctance.

You're going to need something like 10,000 encoder lines, which isn't trivial in that sort of application. How about putting a resolver on one of the pivots?

Cheers

Phil Hobbs

Il 24/02/2010 18.32, Fred Bartoli ha scritto:

Good idea. The best would be two piezos at both ends and another at the reference point. So I pulse one end and detect on the reference and on the other end. In this way I have a ratio, no matter about propagation velocity (that may change with temperature, I guess). Then I pulse the other end and repeat the measurement in the opposite direction.

And what about to fix the piezo on the reference point? You remember the rail will move so I need a brush or a wheel.

Thanks! Marco

Il 24/02/2010 18.45, Phil Hobbs ha scritto:

Yeah, the environment could be very dirty with sand and rain.

I have no picture of the actual plant but I bet I can improve my description. Please, look at this picture:

Imagine our tank is like that. The first rail is located around the front end of the cylinder. The second one near the rear face. The open ends of the rails are upwards, that is the cylinder can rotate cw or ccw for over 100°.

Of course there is a concrete structure to sustain the whole tank and the movement gears.

I hope now it's more clear.

For some reason the customer ask to eliminate any mechanical stuff such as encoders (currently used) or pivots.

Thanks! Marco

Il 24/02/2010 18.18, Vladimir Vassilevsky ha scritto:

I think there are several problems:

- to achieve the resolution of 1 mm we need 10000 signs, not so trivial to put on the rails, I bet.

- any optical system will be degraded over the time by dust and rain.

- in this way we have a relative position. With the acoustic method we have the absolute one.

Marco

Il 24/02/2010 18.59, Marco Trapanese ha scritto:

With this configuration, assuming a propagation velocity about 5000 m/s for a generic metal, the minimum time I have to detect is 1 ms:

The problem is the resolution. To detect changes of 1 mm I need a time resolution of:

That is I need a counter with a clock of 5 MHz. That's not a big deal. I'm concerned about the detector. The received pulse will be distorted, I bet. And may be very difficult to detect the exact time of the reception.

Marco

So you want to encode position with no encoder? To 10000 steps? Sounds hard.

Cheers

Phil Hobbs

Il 24/02/2010 19.53, Phil Hobbs ha scritto:

To be honest, they want ;)

Anyway, yes, that is the goal! Marco

Marco Trapanese a écrit :

I didn't get through the numbers. Then you could try sine wave excitation and measure phase. Being a tuned system, you can workout noises,... and also do swept measurements to have more info about the system and get rid of some parasitic modes. (Well I don't really know, but I'm throwing ideas as they come). AKO acoustic VNA.

Also have a look at FDR and WFDR (frequency domain reflectometry and wideband FDR). I once looked a this for a project that didn't come out and I've found a thesis about this with interesting results that I can send you if needed.

But all that is quite a bit of work.

Il 24/02/2010 20.04, Fred Bartoli ha scritto:

Yes, I know the VNAs. I will investigate on the acoustic version.

Thanks for the input. Marco

So sort of "U" shaped, except it's a smooth circular shape?

Where is the pivot point? Or does it rotate staying "in-circle"? ...Jim Thompson

Il 24/02/2010 20.14, Jim Thompson ha scritto:

Exactly, it follows the circumference of a cylinder for about 210°.

The pivot point is at the center of the cylinder. This cylinder rotates on its own axis; along the external circumference there is the rail.

Marco

Marco Trapanese a écrit :

Hmmm, you can run a pipe along with the rail, which will enclose your measurement medium (air for ex.) That'll greatly reduce your speed requirements and will also nicely uncouple the medium from external conditions (dirt or whatever).

Il 24/02/2010 20.20, Fred Bartoli ha scritto:

And what about the reference point? It's hard to couple the transducer located at the reference point to this moving pipe...

Marco