Eddy currents are causing major delay and attenuation

[Kevinh]

Hello everyone.
I am trying to measure the position of a magnet moving inside a steel tube using a magnetic sensor on the outside. Eddy currents are causing errors due to phase shifts and amplitude changes. Any ideas? Can eddy currents be re-directed using selective copper paths? Since eddy currents are a function of speed, can a derivative function be implemented and used to counter? My gut says there must be a way.
Thanks

 

[marcusl]

Are you sure about eddy currents? You can correct for them, by the way, if you know the characteristics of your tube and magnet like conductivity, diameter, wall thickness and speed. A bigger problem is steel is ferromagnetic and hysteretic, so its magnetic behavior depends on the history.

 

[Kevinh]

Thanks Marcus,
I know it is eddy current because at higher speeds I get more delay and attenuation. You are correct about hysteresis, but its magnitude (mesuered at very low speeds) is small compared with eddy current effects.
The problem is that I do not know speed. If I infere speed from position and time, then my speed calulations are suspect since my position information is already no good.

 

[marcusl]

Ok. The good news: a hollow cylinder is one of the few geometries that is soluble analytically. The bad news: the solution is a set of equations that fills a page or two.
You'll find it in Tegopoulos, Eddy Currents in Linear Media. I got rid of my copy last year during an office move, so I can't help you out there.

What are the dimensions of the magnet and tube? If the pole faces are separated by enough distance, you'll might deduce the speed from the + and - impulses as it moves by.

 

[Kevinh]

Found the book and ordered it, thanks ( a bit pricy, you should have saved your copy ). The pole spacing is around 50 mm and the entire cylinder is around 700 mm with an inside diameter of 80 mm and a cylinder wall thickness of 5 mm. You can imagine that I am using pretty powerfull Neo magnets given this massive structure.
Using the north and south peaks as they go by is a definite possibility except the velocity can change between the peaks. There is also acceleration to some extent.

 

[berkeman]

(Takes a step back...) Can you explain why you are needing to sense a cylindrical magnet's position inside a ferrous tube? Can you just sense the cylinder's position some other way? (like via ultrasound reflections from one end of the tube).

 

[Kevinh]

Good question. I am trying to measure position of a piston in a hydraulic cylinder that has a steel barell. This is a special project for me and being a magneto-haulic, I am taking this approach.
There has been numerous attempts using ultrasonics, microwave,...(not by me, but to my knowledge non has been really successful and / or cost effective). Every technology has its pro's and con's. Ultrasonic will probably suffer from enviromental conditions like misture and temperature. I am sure there are ways to correct for it, but...

 

[berkeman]

Good question. I am trying to measure position of a piston in a hydraulic cylinder that has a steel barell. This is a special project for me and being a magneto-haulic, I am taking this approach.
There has been numerous attempts using ultrasonics, microwave,...(not by me, but to my knowledge non has been really successful and / or cost effective). Every technology has its pro's and con's. Ultrasonic will probably suffer from enviromental conditions like misture and temperature. I am sure there are ways to correct for it, but...

Interesting problem. If it's a piston, can you sense the stroke of the rod instead?

 

[Kevinh]

The rod and piston are attached to each other. If you sense one, you have sensed the other. The magnets are attached to piston since there is more real estate.

 

[berkeman]

The rod and piston are attached to each other. If you sense one, you have sensed the other. The magnets are attached to piston since there is more real estate.

But can't you get a sensor right up against the rod inside the cylinder? The piston has to be flush with the wall, but the rod is smaller diameter, no?

 

[dlgoff]

Yep. Put marks on the rod (encoder) and count them with an optical emitter/detector.

 

[Kevinh]

I need to measure absolute position and not incremental so in the case of loss of power, i still know where I am. Optical encoders will not survive the dusty / nasty enviroment. Putting a magnetic sensor looking at the rod was one of the original ideas but did not work due several reasons including Earth's field interfering with measured output after being concentrated in this massive steel structure. The current with moving piston magnets solution works except the eddy current issue.

 

[marcusl]

After reading the recent posts, I'm confused about your application. Are you trying to sense the piston as it passes a fixed point (which you could do with eddy currents) or monitor its position anywhere in the stroke? I don't see how to do the later magnetically...

 

[Kevinh]

The piston is carrying a set of magnets and as it moves, it creates a magnetic signature (kind of like sine waves) outside of barell. If you can imagine magnetic field actually goes through the barell (after some of it is shunted). I have an array of magnetic sensors outside of barell(along the entire length of cylinder) that sense the magnetic field signatures. The problem is that as the magnets move by fast, the sensed field is attenuated / delayed as mentioned previously.
As a side, I came across a technique of shaping current waves to counter the eddy currents in MRI systems. I think somehow I can take the derivative of the position and then add it to the signal??

 

[dlgoff]

I've got a question. Has anyone tried to look at the temperature of the steel tube with an IR TV camera? I would think that heat would be generated as the piston moves hence creating a visual map of its position. Using pattern reconizition software, you could determine the position accurately.

Regards

 

[marcusl]

The piston is carrying a set of magnets and as it moves, it creates a magnetic signature (kind of like sine waves) outside of barell. If you can imagine magnetic field actually goes through the barell (after some of it is shunted). I have an array of magnetic sensors outside of barell(along the entire length of cylinder) that sense the magnetic field signatures. The problem is that as the magnets move by fast, the sensed field is attenuated / delayed as mentioned previously.
As a side, I came across a technique of shaping current waves to counter the eddy currents in MRI systems. I think somehow I can take the derivative of the position and then add it to the signal??

Are the magnets mounted so the field outside is radial, or tangential (parallel to the cylinder bore)?

As for MRI eddy current compensation, it is a time-dependent wave shaping applied to the electromagnet coils inside, and you have permanent magnets.

 

[Kevinh]

I like the out of the box thinking dlgoff. Unfortunately in this case absolute position sensing is needed. This means that if you loose power and then the piston moves to a new location, upon power up an hour later, you must know where you are.

 

[Kevinh]

Marcus,
The field is radial outside of the barell.
I realize that the MRI has a time based wave shaping. In my case, I only have eddy current issues when magnet moves. When it does move, then I should be able to take a derivative of magnetic field with respect to position (time). I would think this will be somewhat proportional to the fields generated by eddy currents which are dependant on speed. Therefore I maybe able to use this added information to compensate / add to the sensor output??
Happy new year!

 

[dlgoff]

"... if you loose power and then the piston moves to a new location, upon power up an hour later, you must know where you are."

Well if you are talking about knowing where it is when the machine is stopped, then couldn't you use some sort of simple hand-held flux meter to find it location? Or do you need to have some sort of computer input?

 

[berkeman]

Good question. I am trying to measure position of a piston in a hydraulic cylinder that has a steel barell.

Since the hydraulic fluid is incompressible, and since you know the cross-sectional area of the cylinder ID and shaft OD, can you just use a quickly-responding flow sensor at the hydraulic input coupling to the cylinder, and use the delta-Volume(t) to tell you the position of the piston at the end of the shaft? You will need to use some sort of limit switch to get an absolute reference position of course, but from then on (until you need to re-cal), you should get pretty accurate position information.

Or maybe combine the flow meter approach with your on-piston magnets to improve the accuracy of each method by itself...

 

[Kevinh]

I like to thank everyone for their suggestions and questions. Sometimes a good enduring question sets a framework for new possible solutions. This was my first time using this forum and I am already enjoying it.