How many ways can you process a PI target signal with a Micro ?

[6666]

Not sure how to ask this question, but how many ways are there to process a PI target signal with a micro ?
There are a few micro projects that use what I think is called direct sampling, such as PICKINI, and Felezjoo,
The inventor of the PICKINI has already given good explanations of how his system works, temporal sampling,
but I am just wondering how many other ways there are to use a micro and what the circuit function name is ?

[homefire]

Other then Detecting the Targets signal in the Time Domain it's self the only other information that would be available I know of would be Signal Intensity or Strength. That would be variable with target proximity and size.

If you could measure the Target and correlate the Rx Signature with known Signatures you may be close to Approximating what the target material is ? Some what like VLF TID gustimation numbers! LOL Don't see that option with the Processor Speeds we have to play with.

[rickb]

Another problem is the dynamic range of the signal. Even a 16 bit adc/cpu will lose some information near the noise level. To me, an ARM with a 18-20 bit adc would be the minimum. Accuracy is not real important but repeatability is.

[Teleno]

Another problem is the dynamic range of the signal. Even a 16 bit adc/cpu will lose some information near the noise level. To me, an ARM with a 18-20 bit adc would be the minimum. Accuracy is not real important but repeatability is.

Dynamic range can be largely improved by numerical integration. No need to use extraterrestrial ADCs.

[rickb]

Dynamic range can be largely improved by numerical integration. No need to use extraterrestrial ADCs.

That only works if the signal is buried in noise. If it is clean and below the value of the lsb on your 16 bit adc, you won't even see it.

[Teleno]

That only works if the signal is buried in noise. If it is clean and below the value of the lsb on your 16 bit adc, you won't even see it.

That will never happen, but if it does then you add the noise.

[rickb]

If the entire premise of your argument was correct, we would not have 24 bit adcs. The technique is used to gain a bit or 2 of signal resolution out of an existing design, it is not a substitute for a wide dynamic range signal.

[F117]

Dynamic range can be largely improved by numerical integration.

This is exactly the major advantage of using a simple micro with limited resolution ADC or timers: you can accumulate samples over time in very controlled and configurable way.
In the Pickini project, the number of samples that are accumulated determines the sensitivity and is controlled by a simple potmeter that gives a value between 0 and 5V. This value is read via an ADC channel - 8 bits is more than sufficient for this. When you would use this value without any processing, this would mean that you would add up 1 (not 0) to 255 values to obtain more resolution of the result.
One thing to consider here is the time it would take to accumulate 255 samples every 2ms ( pulse frequency ) = 500 ms. This is too coarse - 2 values per second. So, the amount of samples to accumulate should be limited to 32 or something..

Accumulation / addition of samples over time is probably much harder to achieve with discrete components.

Another major advantage of using a micro is the audio signal. You can basically generate whatever you want: fixed frequency with PWM = volume changes or changing frequency with signal strength...
This is also harder to make with discrete electronics.

Another major advantage of using a microcontroller: price - 1 to 2 $ for an 8 bit Microchip PIC controller with lots of goodies on board.

As for the original question:

Not sure how to ask this question, but how many ways are there to process a PI target signal with a micro ?

I would say 2:

• pulse width measurement
• voltage measurement

Processing the entire slope of the feedback by taking voltage / pulse width samples all the way back to steady state is not another way, but an extension of one of these 2. It may add some indication of the conductivity of the target ( slope direction coefficient ). I experimented with this but never got a clear result from it. Probably more extensive processing is needed than what I had available in a simple 1$ microcontroller

- Bernard

[Teleno]

If the entire premise of your argument was correct, we would not have 24 bit adcs. The technique is used to gain a bit or 2 of signal resolution out of an existing design, it is not a substitute for a wide dynamic range signal.

24bit ADCs have a conversion times of milliseconds, which is too slow. More bits won't help you get below the noise floor, you just digitize the noise. You need multiple samples in order to improve the signal/noise ratio. Integration does exactly that.

[kosacid]

time delay after the pulse then sample is a good way, you can take multiple with a different delay stepping up each time, take the stress out of taking multiple at one shot, that way you can use a basic ADC

[KingJL]

24bit ADCs have a conversion times of milliseconds, which is too slow...

The LTC2380-24 conversion time is 343 ns min, 392 ns max.

[green]

The LTC2380-24 conversion time is 343 ns min, 392 ns max.

Did a test to get a number for how many bits might be needed. Target(US nickel), 8 inch coil, 1 amp peak current, sample taken at 10usec after coil turn off. Change in coil volts with target at 12 inches(1.3uv), at 16 inches(.23uv). Wondering if the numbers sound reasonable.

[green]

The important part is detecting the target, how important is cancelling EF and the ground signal be included in the process?

[green]

Wanting to learn something about using a ADC for a PI frontend. Including a graph from the 3280 spec sheet. Can someone explain what it means and if it is the best graph to try and understand for the PI application?

[Teleno]

Wanting to learn something about using a ADC for a PI frontend. Including a graph from the 3280 spec sheet. Can someone explain what it means and if it is the best graph to try and understand for the PI application?

More averages => less noise,
More averages => larger dynamic range.

The noise floor of a PI is the sum of the noise of the damping and limiting resistors. For a carefully designed, standard PI a noise floor of 3uV is already a feat. If you want to directly sample this then your 3dB signal above noise is 6uV, you need an ADC resolution of 19 bits. You spend $50 on a 24bit ADC to waste 21% of its resolution.