## Finding inductance of an inductor at high speed (~10KHz)

23 posts / 0 new
Author
Message

Hi there - I've been looking at using something along the lines of this bad boy. It's a variable inductor, essentially. The inductance changes as you move the core over the coil, allowing you to get a contactless linear position measurement. It's a lot like an LVDT, only it's not a transformer - just an inductor.

How would you suggest instrumenting such a beast? They say the excitation frequency is 112KHz. I'm guessing they mean that that is the maximum frequency that you should excite it at - but I could be missing something.

So here's my first pass at a plan for it:

Connect it to a full H bridge. Do a differential voltage measurement across it. Run the H-bridge at 112KHz with a square wave. Measure the current and voltage 4x per cycle, two times when it has positive current flowing through it, two times when it has negative current flowing through it. Find the di/dt of the first set of currents and of the second set of currents. Divide the voltage measured by these di/dts and you have two measurements of your inductance. Repeat this over and over and average the hell out of it.

Seem reasonable?

Total votes: 0

You can drive it with a sine and measure phase shift.

You can put a voltage step and measure the rate of change of the current (V = L dI/dt), You can put a current ramp on it, and measure the voltage.

Thats is about it.

Jim

Until Black Lives Matter, we do not have "All Lives Matter"!

Total votes: 0

If its an inductor, just hook it in parallel with a known value cap, hook a signal generator to this tank circuit thru a 1k resistor... sweep freq, look for peak on scope f=1/(2pi sqrt(LC))

Imagecraft compiler user

Total votes: 0

ka7ehk wrote:
You can drive it with a sine and measure phase shift.

You can put a voltage step and measure the rate of change of the current (V = L dI/dt), You can put a current ramp on it, and measure the voltage.

Thats is about it.

Jim

Hi Jim - but remember that I'm looking to get 10KHz of bandwidth out of this device... So this'd be a 112KHz (or some other pretty quick speed) sine wave. I've never tried to generate a sine wave - but generating a 112KHz sine wave strikes me as being difficult due to the speed. And you're talking about a phase shift in the current, right? So I'd have to measure the current at an even faster speed than my sine wave...

A voltage step is what I was describing.

A current ramp would again be pretty tricky at the speeds I'm looking at.

This is for an embedded application - probably will be supervised by a DSP or ARM. So I don't have any fancy equipment to do it for me.

Total votes: 0

bobgardner wrote:
If its an inductor, just hook it in parallel with a known value cap, hook a signal generator to this tank circuit thru a 1k resistor... sweep freq, look for peak on scope f=1/(2pi sqrt(LC))

I should have been more clear... I'm looking to sample the inductance at 10KHz or faster... and I'm looking to have the whole system be embedded on a very small PCB. So no fancy equipment and no human watching over it.

Total votes: 0

Not sure what you mean by "10KHz of bandwidth". Did not find any reference to bandwidth or to the end application beyond contactless position sensing.

Rather than all those current measurements, just measure the peak-peak current. That uniquely identifies dI/dt if dI/dt is anywhere close to constant. Further, you know exactly when the peaks occur (at the edges of the square wave) so two samples per cycle should do it.

Jim

Until Black Lives Matter, we do not have "All Lives Matter"!

Total votes: 0

These are all rather complicated solutions. Just put the inductor as one arm of a voltage divider. Feed the sine wave to the voltage divider and measure the voltage at the output of the voltage divider. Best to use AC (DC blocking capacitor) as the DC current might disturb the functioning of the inductor (core saturation).

If you need temperature compensation etc. then you can use a full or half bridge configuration, with two similar inductors (one moving, one stationary). This assumes that both inductors will be similarly affected by temperature changes and variations due to that will cancel out.

If you think education is expensive, try ignorance.

Total votes: 0

nleahcim wrote:
bobgardner wrote:
If its an inductor, just hook it in parallel with a known value cap, hook a signal generator to this tank circuit thru a 1k resistor... sweep freq, look for peak on scope f=1/(2pi sqrt(LC))

I should have been more clear... I'm looking to sample the inductance at 10KHz or faster... and I'm looking to have the whole system be embedded on a very small PCB. So no fancy equipment and no human watching over it.

Perhaps you could hook up an 555 as free runing an replace the R-s with the inductor. It is easy to try, and it vil giv a variabel frequency, perfect for a micro to be eaten.

HM

HM

Total votes: 0

Similar to what Mossige describes - make a colpitts oscillator and measure the frequency.

Total votes: 0

Ooo one other thing I should mention... I am hoping to get 12 bits of resolution minimum out of these, preferably more. I talked with the manufacturer - the interface electronics that they sell are purely analog. The parts are highly non-linear, so the interface electronics linearize them. The inductance ranges from about 0 to 500 microhenries.

Total votes: 0

Non-linear in which sense? With respect to voltage or current? Or, with respect to core position?

Put it in an oscillator, count the frequency, then use a linearization lookup table. Could be a problem with 0uh, however! Any measurement method will have problems with even 500:1 inductance ratio, let alone infinite inductance ratio!

Jim

Until Black Lives Matter, we do not have "All Lives Matter"!

Total votes: 0

Study the specifications of the linearising circuits very carefully. You don't want to lose accuracy because of imperfections and drift in the linearising circuits themselves.

If you are not completely happy, you can do the same in code.

What is the transfer function of the coil? Can you post a graph showing that?

If you want to get 12 bits out of these, you will probably have to use a bridge configuration that will at least swamp out the effects of temperature changes on the coils themselves.

If you are feeling inventive, you could experiment with both inductors operated by the same movement, but with one inductor inverted (kind of like a class A push pull amplifier). Maybe if you do that the non-linearity will cancel out.

If you think education is expensive, try ignorance.

Total votes: 0

Using the oscillator method can give you pretty high resolution, especially for small changes. This could be just enough to get the 12 Bits from a frequency measurent via the ICP funtion of an AVR. For 0.1 ms you can have about 2000 cycles and about 10 slopes of each type. So the limit would be close to 12 Bits, really keep the controller bussy.
The "analog" alternative would be to have an PLL follow the frequency and take out the VCO voltage. This should also be possible with 12 bit resolution, but is limited by the VCO quality.

The bridge Circuit can more easy handel large changes. If the other side is an inductor too, its essentially like a resistive divider. So Its just measureing the amplitude of the 100 kHz Signal. Not a big deal.

So there a re lots of options, choose one.

Total votes: 0

ka7ehk wrote:
Non-linear in which sense? With respect to voltage or current? Or, with respect to core position?

Put it in an oscillator, count the frequency, then use a linearization lookup table. Could be a problem with 0uh, however! Any measurement method will have problems with even 500:1 inductance ratio, let alone infinite inductance ratio!

Jim

Non linear with respect to core position.

The guy said 0uH, but I don't buy it. I mean, it's a coil of wire. Of course it will always have an inductance. I'm planning on correct for the linearity problems in software. As much as I'm impressed by them doing it all in hardware - I just don't see the need.

Total votes: 0

emuler wrote:
Study the specifications of the linearising circuits very carefully. You don't want to lose accuracy because of imperfections and drift in the linearising circuits themselves.

If you are not completely happy, you can do the same in code.

What is the transfer function of the coil? Can you post a graph showing that?

If you want to get 12 bits out of these, you will probably have to use a bridge configuration that will at least swamp out the effects of temperature changes on the coils themselves.

If you are feeling inventive, you could experiment with both inductors operated by the same movement, but with one inductor inverted (kind of like a class A push pull amplifier). Maybe if you do that the non-linearity will cancel out.

I wish I had the transfer function. That datasheet and what I've stated in this thread is all I know.

What sort of bridge are you talking about? Also - you referred to coils (plural) - I'm just talking about one coil...

Total votes: 0

OK, suppose that you do it with peak current measurement. If you operate at fixed frequency square-wave drive and minimum inductance is 1uh, then you will have a 500:1 variation in peak-peak amplitude. You can't detect that with 10bit accuracy using an internal ADC by any realistic means. Its even difficult to do ANY peak-peak amplitude measurement to that resolution, let alone accuracy.

How about using a comparator with hysteresis (note - hysteresis can be generated outside the comparator using positive feedback) to "clock" the H-bridge? This will give you a frequency that is proportional to 1/L rather than 1/sqrt(L) in an LC-oscillator. Measuring frequency to 12 bits should be a lot easier than with other oscillators.

The idea is that current ramps up, trips the comparator at a fixed current level. The comparator output IS the H-bridge input, so the current starts ramping down. When it reaches the lower fixed level, it toggles, starting the cycle over again. The result is fixed delta-I, variable frequency.

The comparator and the feedback net would have to be carefully designed so that the trip-point amplitude does not vary over the operating frequency range. But, that is known, straight-forward technology. Not easy, but known. You would also have to be careful about comparator response speeds relative to the minimum period.

Jim

Until Black Lives Matter, we do not have "All Lives Matter"!

Total votes: 0

Quote:

What sort of bridge are you talking about? Also - you referred to coils (plural) - I'm just talking about one coil...

Quote:

If you need temperature compensation etc. then you can use a full or half bridge configuration, with two similar inductors (one moving, one stationary). This assumes that both inductors will be similarly affected by temperature changes and variations due to that will cancel out.

I'm pretty sure you will have to use two coils. :( Otherwise the temperature drift alone will kill you.
Quote:

If you are feeling inventive, you could experiment with both inductors operated by the same movement, but with one inductor inverted (kind of like a class A push pull amplifier). Maybe if you do that the non-linearity will cancel out.

I'm talking about connecting two coils back to back (in opposite directions). Mount them on the same shaft/moving part in such a way that when one core moves in, the other moves out. Put both coils in a bridge. Now when the cores move, the inductance of one coil will increase while that of the other will decrease. This will make the system more sensitive to movement, but if you are lucky, the non-linearity of the two will cancel out. Give it a shot.

If you think education is expensive, try ignorance.

Total votes: 0

Now, you are talking about "LVDT". For some reason, he is constrained to a single coil. I'll bet its because "boss" said so.

Jim

Until Black Lives Matter, we do not have "All Lives Matter"!

Total votes: 0

Quote:

Now, you are talking about "LVDT". For some reason, he is constrained to a single coil. I'll bet its because "boss" said so.

I agree with you. He should tell the boss that using two coils will be a lot cheaper than trying to do the compensation and linearisation with circuitry.

If you think education is expensive, try ignorance.

Total votes: 0

emuler wrote:
Quote:

Now, you are talking about "LVDT". For some reason, he is constrained to a single coil. I'll bet its because "boss" said so.

I agree with you. He should tell the boss that using two coils will be a lot cheaper than trying to do the compensation and linearisation with circuitry.

We're looking at LVDTs as well. However, variable inductors like these get better performance and are smaller, so they're very attractive.

Total votes: 0

With two of these you could make an LVDT. Or try to, at any rate. :P

If you think education is expensive, try ignorance.

Total votes: 0

Sometimes, there is a key technology "gotcha" that really prevents you from realizing the perceived benefits (size, low power, speed, accuracy, whatever).

LVDTs do not have to be huge. You CAN make your own or have them made to your specs. They are not magic and way simpler than, say, a custom IC. Only slightly more complex than a custom bolt.

Jim

Until Black Lives Matter, we do not have "All Lives Matter"!

Total votes: 0

ka7ehk wrote:
Sometimes, there is a key technology "gotcha" that really prevents you from realizing the perceived benefits (size, low power, speed, accuracy, whatever).

LVDTs do not have to be huge. You CAN make your own or have them made to your specs. They are not magic and way simpler than, say, a custom IC. Only slightly more complex than a custom bolt.

Jim

We're going to have these professionally made to our specs. The vendor is able to make the variable inductor parts smaller as there are less coils.