LDC1000 detection distance and accuracy

Hi Hidetoshi,

there is no linearity in proximity data versus distance, you have to do some kind of approximation.

As to me, it looks very similar to exponent: prox. data = a*e^-b*t + c. It would be nice to get more pricise

explanations from TI-specialists.

And what about the coil size: we are trying here a 1.6 m coil (diameter) to measure a 0.8 m distance.

Best wishes,

      Igor Gorbounov

HI,

To make rough estimation you could use LDC1000 WebENCH:

http://www.ti.com/lsds/ti/analog/webench/inductive-sensing.page

Hello Igor and Oleksiy

Thank you for giving us a hand in answering this question, truly appreciated!

Igor, we have not found an exact formula yet for different coil and target sizes and shapes. Exponential approximation may work well for certain distances, indeed.

Hi, Evgeny,

we use here a target between two identical sensor coils. It has been a great surprise for me as the difference of

their proximity data appeared to be equal to hyperbolic sinus of diplacement:

                 prox1 - prox2 = a* sinh(b*displacement),

which is quite linear in a range of b*displacement from -1 to 1, that corresponds to decrease of proximity data difference

for e times from the central position value. Therefore I need only two points to calibrate our device.

It's a pity that the whole arrangement is yet quite noisy for me to prove this relationship in a proper way.

Best wishes,

     Igor

Hi Igor,

Just to clarify - you use two LDC1000 for each coil, not that your coils are connected to each other, either is series or parallel config?

Great find with Sinh!

Hi, Eugeny,

yes, we use two independent LDC1000 (actually, four) with sensors opposite each other. Now we've encountered

a problem that sensor coils interfere with each other - apparently frequence differencies affect the data: our target

(a weld rod) is a rather small object in diameter, but if a plate is used instead the noise becomes significally

reduced.

      Best wishes,

      Igor Gorbounov

Hi, Ryuji,

it seems to be a mere geometry issue: the farther from the conductor with eddy current are the coil turns positioned,

the less magnetic field lines are crossing the coil and hence the electromotive force in the coil is weaker. The general

rule is that the reasonable distance should be less then half of a coil diameter.

Perhaps if we could somehow distort the electromagnetic field of eddy current, to focus it somehow, then this

sensing distance would be more proximate. Does anybody know whether I am right?

Best wishes,

    Igor

Hello Igor,

You can avoid cross-talk by either shifting the frequency of oscillations (15-20% difference is usually sufficient), or acquire data from a sensor while the rest of them are shut down.

Hi Ryuji,

Just as Igor said, it's all about magnetic lines. Generally speaking, the magnetic field becomes very weak coil-diameter away from the coil, and thus sensitivity is very low. It's basic physics limitation, and there's very little we can do about it.

The only practical way of increasing sensing distance is adding a ferrite between your target and the coil.

Flat ferrite plate on the back of the coil will slightly help as well.

Ryuji-san,

We do not have a document dedicated to a ferrite. This is too application-specific.

Just add the ferrite that bridges the air gap between the coil and your target, diameter of a ferrite should be half-coil diameter.

Hi, Eugeny,

"... or acquire data from a sensor while the rest of them are shut down...." - it would be a nice solution

if it were not so slow. 2 ms is a tremendous time and those welding rods move fast.

Thanks for your answer,

     Igor

Hi Igor,

What is the timeline for your project?

Hi, Evgeny,

I'm not quite sure what do you mean by "timeline"... We've got moving 10 welding rods per second, about 40 cm each, that means about 4 meters per second...Resonance frequency of coils is about 750 kHz and there are about 11 kHz of ready pulses for every sensor, there are 4 of them. So I have about 44 kilosamples of data. The task is to measure the position of central metal rod at about 5-10 um accuracy. I have about 1000 points of measurements (4 numbers for each) for every rod - it is about 0.4 mm between them. There is a gap between rods - several mm, it should be recognized, because some rods vibrate in the end while moving.

Now I have variations of measured distance at about 50 um, that is an average of 16 samples (6.4 mm), and the frequencies of the tested pair of coils are 750 kHz and 950 kHz. Now still are used our standard sensors with coaxial cables, no special coils are disigned yet. Between these sensors and LDC1000 pulse transformers are used to protect the inputs of LDC1000.

Thanks for your attention,

       Igor Gorbounov

Oh, sorry for ambiguity.

I meant the schedule of the project. When do you need to have proto, go to production?

Hi, Evgeny,

never mind for ambiguity, it's usefull to write down the whole conceipt.

I'm not quite sure for the schedule: part of the team is on vacation now, and then we will demostrate
the prototype in a month or two. And it is not clear now whether there will be production stage at all.

Best wishes,

     Igor

Hi Igor,

Where are you located?

I suggest you get in contact with our marketing, and we'll work from there. There might be better solution for your app.

Hi, Evgeny,

we are in the south of  Russia, near Rostov-on-Don.

Best wishes,

     Igor Gorbounov,

    seniour software developer,

    "Topaz-Service", Volgodonsk

Thank you, Igor,

I have referred our marketing team to this thread, they should get in contact with you soon.

Hi Igor,

Thanks again for reaching out to us!

Please reach out to me offline through my email at tchang@ti.com

Warm regards,

Tarig