Inductive Sensing Design Challenge using TI LDC devices.

Heater.

Haven't noticed if anyone has mentioned this here before but I cam across these interesting "Inductance to Digital Converters" (LDC) from TI and an associated design competition:

http://www.element14.com/community/roadTests/1273?ICID=sensor-inductchallenge-panel

They have a webinar coming up heralded as "A Revolution in Sensing—The World's First Inductance-to-Digital Converter" Certainly I have never heard of such devices before.

http://www.element14.com/community/groups/sensor-technology?CMP=KNC-EU-InductiveSensing-D

Clearly the competition winner should be using a Propeller to get functionality the other entrants will be lacking...

Beau Schwabe

Honestly I think they are trying to hash an old idea that has been around for quite some time and fool everyone into thinking it's the next thing since sliced bread.

If you oscillate an inductor and count the number of pulses within a given amount of time, the result is a digital value.

mindrobots

Heater. wrote: »

Clearly the competition winner should be using a Propeller to get functionality the other entrants will be lacking...

I'm betting the winners are either Arduinos or TI products.

Beau Schwabe

Would this be considered prior art? ...uses a Propeller, and is Digital ... they claim sub micron resolutions ... I know in the past, I have achieved a resolution of 1000 within 1/4 of an inch (1/4000th of an inch). That was 15+ years ago using a BS2 and an external (4040B Counter IC)

http://nutsvolts.texterity.com/nutsvolts/200903/?folio=36#pg36

Heater.

Beau,

No doubt measuring inductance and getting a digital result has been around for ages. Heck I was doing it back in Uni with tubes!

But if they can package it into something the size of a gnats...err...eyeball, make it low power, calibrated, robust, easy to use, and cheap then that is cool.

If you bust open a traffic light controller cabinet you will find the electronics connected to the inductive vehicle detector loops generally comes on a Eurocard and costs hundreds of dollars.

Tracy Allen

The LDC1000 did show up on our sensors forum back when. At that time there was a 2013 inductive design contest sponsored by edn magazine. The winners and links to their entries are listed here. Most interesting to me is the second place winner, the device that acts kind of like of stud finder. Beau's design could do that easily, or really, hold its own against any of the winners.

I had submitted a 1st phase entry, and they sent me the development kit, but I had trouble making it work. The demos worked great, of course, but, for my application I was trying to use it at the lower end of its frequency range, around 20kHz, with a high mu toroidal core, and I couldn't get a stable oscillation out of it. I came back to it several times but no joy. I'd pursue it some more, but at this point I would agree with Beau, it is probably better just to start with the raw materials and do the i/o direct from a Propeller. My biggest complaint (not just mine, BTW) is that the LDC1000 is a black box and TI is not revealing their IP of exactly how it does its magic. So, when you try to do something further removed from the demo-board, you are left hanging wondering what is going on inside and what to do about it. It is a nicely integrated chip though. I see that two of the main contest winners used the coil and demo board exactly as provided by TI. The 2nd place guy actually did experiment with coil design.

Heater.

I guess if you need a micro-controller to use the the thing anyway and your micro-controller can easily do the job itself, like say a Propeller, then there is not much point.

Unless perhaps the output is already nicely calibrated. Anyone know if it is, I have not checked?

trooks

Over four decades ago I built an inductance to audio device. There is even a bronze spike from a 15th century sailing ship in the UofF museum that I donated after having found it on Singer Island. It was back in the dunes and not out on the actual beach.

You still see inductance devices once and a while especially if you go to the beach very much. That is what the old geezers are sweeping back and forth as they cruise the beach.


DUUH,

Tim

Heater.

Seeing those guys with metal detectors on the beach always makes me wonder how come nobody has made a robot to do all that walking and swiping.

trooks

Heater. wrote: »

Seeing those guys with metal detectors on the beach always makes me wonder how come nobody has made a robot to do all that walking and swiping.

A robot would have no appreciation at all for the scenery on the beach(or so I suppose)<G>.

Tracy Allen

Unless perhaps the output is already nicely calibrated. Anyone know if it is, I have not checked?

It returns the resonant frequency of an LC circuit. Calibration depends on the choice of C and the time base. A parameter sets the duration of the sampling window, and within the window it is precise to 16 bits. It also measures the equivalent parallel resistance to 16 bits, as it servos the oscillation to a constant amplitude and monitors the energy required to keep it there. The dev kit comes with a GUI that graphs both inductance and resistance and shows the values. The only external parts required, other than the LC tank circuit, are an integration capacitor and a bypass capacitor, and the SPI and timebase interface to the controller.

jmg

Tracy Allen wrote: »

It returns the resonant frequency of an LC circuit. Calibration depends on the choice of C and the time base. A parameter sets the duration of the sampling window, and within the window it is precise to 16 bits. It also measures the equivalent parallel resistance to 16 bits, as it servos the oscillation to a constant amplitude and monitors the energy required to keep it there.

All of that makes it a little tricky to clone in software.
A Prop can do a decent job of measuring frequency of an external LC osc, but measure the equivalent parallel resistance to 16 bits would seem more of a challenge.

Tracy Allen

That's very true, and a good reason to appreciate the integration and IP in the LDC1000.

Do you have an idea of how you might approach that measurement of the equiv. parallel resistance with the Prop? The usual methods that are done with a Q meter or via bandwidth measurement or decay time might work, but that does not seem to be what is done in the LDC1000. It should be a primarily digital method. How to hold the oscillation amplitude close to a steady amplitude? It could be detecting the peak of voltage and at that point depending on the level either inject a packet of charge or not. From that, calculate the energy input required to counteract dissipation. It would have to account for the charge redistribution loss.

jmg

Tracy Allen wrote: »

That's very true, and a good reason to appreciate the integration and IP in the LDC1000.

Do you have an idea of how you might approach that measurement of the equiv. parallel resistance with the Prop? The usual methods that are done with a Q meter or via bandwidth measurement or decay time might work, but that does not seem to be what is done in the LDC1000. It should be a primarily digital method. How to hold the oscillation amplitude close to a steady amplitude? It could be detecting the peak of voltage and at that point depending on the level either inject a packet of charge or not. From that, calculate the energy input required to counteract dissipation. It would have to account for the charge redistribution loss.

The most sensitive and fastest reading would be a current meter, in a external amplitude-regulated oscillator.
However, that's quite a lot of extra analog parts.

'Mostly Digital' could involve an external 1GU04 or 74AUP1Z04 and 1G97 as Mux.

It would run in LC Osc mode for Precision Frequency measure mode, and then flip to drive from a Prop 1 pin, which would need to sweep Frequency to measure Q.
Fine frequency sweep is not easy, and even NCO mode would need some LC limits, but may be practical with sub MHz Osc and high Q's ..

Pin threshold as comparator may be enough here, to pick up two skirt points, but I think for best Q results the amplitude ratio is needed, and that is more complex, and all this is looking slow...

An alternative would be a Controlled Osc, with a simple amplitude detector, and a controlled shunt resistor.
The Prop would have a loop that checks amplitude, and too-high lowers shunt resistor, and too-low raises it.

As the Q externally falls, that shunt setting would reduce, and highest Qs would need most added damping.
Possibly a Open drain pin switching a resistor on a percentage basis - but the beat frequencies here could need attention.

Perhaps an opto-fet or other voltage controlled resistor would avoid too many frequencies getting into the mix, and even a Operational Transconductance Amplifier (NE5517, LM13700N ) or Variable Gain Amplifiers would give DAC to Gain/damping controls.

Tracy Allen

That's tricky all right! Makes it all the more remarkable that the LDC can cover an resonant frequency range of 5kHz to 5MHz with 24 bit resolution of inductance and 16 bit resolution of Rp (per data sheet specsmanship anyway) . Worth the $5 price tag.

The range of Rp that it covers is limited. 768Ω on the low end, and some MΩon the high end. LC has to oscillate, no critical or subcritical circuits, not too much energy to build up the oscillation.

I'm thinking of something like your controlled oscillator with fixed threshold and switched shunt resistance. Instead, a gate that can add energy to the tank. Coupled in through a small capacitor perhaps, so on a detected rising slope, the gate would go from low to high, then, on the falling slope, back low. Like pushing a swing. The processor detects whether or not the peak of the tank waveform has gone above the threshold, and if it does, it foregoes the push. Again, just like the kid on the swing--You know when they are at their limit (as if!). Count the ratio of pushes to total cycles, that is a function of the energy dissipation per cycle, the friction in the chain and bearings. That might work on a Prop at low or audio frequencies. I'm thinking in particular of a situation where the resonant frequency is pretty much known and constant, and the variable is dissipation. Loops to servo both amplitude and frequency would complicate it some, not too much, maybe...

jmg

Tracy Allen wrote: »

I'm thinking of something like your controlled oscillator with fixed threshold and switched shunt resistance. Instead, a gate that can add energy to the tank. Coupled in through a small capacitor perhaps, so on a detected rising slope, the gate would go from low to high, then, on the falling slope, back low. Like pushing a swing. The processor detects whether or not the peak of the tank waveform has gone above the threshold, and if it does, it foregoes the push. Again, just like the kid on the swing--You know when they are at their limit (as if!). Count the ratio of pushes to total cycles, that is a function of the energy dissipation per cycle, the friction in the chain and bearings. That might work on a Prop at low or audio frequencies. I'm thinking in particular of a situation where the resonant frequency is pretty much known and constant, and the variable is dissipation. Loops to servo both amplitude and frequency would complicate it some, not too much, maybe...

The limitation there is likely to be the granularity of that push, which would put a ceiling on the Frequency, but maybe you could flip this a little, so you have a width-modulated amplitude control via analog*, and then the Prop measures Freq and Average Width, which it can do averaging over a number of cycles.

The oscillator would need to start in a simpler amplifier mode, because with no swing, there is no trigger into.

* A 4046 does Voltage to time quite well, so it could manage the push-control modulator.