Ground Balance and ID

[Daves]

bidirectional signal = for that we using DC bias, or you can use abs() in program

Can you share a "dc bias" usage example for direct sampling?
I use the ABS function.
Problem = When I process the signal to get double peaks or 3 peaks in I and Q signals, I get more than one target id for a target. Generating an audio signal becomes a problem.

To fix this, I am trying to understand whether I am making a mistake on the analog side or whether there is a deficiency on the digital side.

In addition = it works fine in the air test, as an example I get the target id of coin = 82.
When I bury the coin in the ground and pass over it, I get about 7-8 id. What could be the reason for this?

[pito]

bias

What could be the reason for this? = wrong reference signal for phase detector
the analog side or whether there is a deficiency on the digital side. = we don't see what you are doing

[Daves]

bias

what do you think other problems

[pito]

what do you think other problems

we don't see what you are doing, no schematic, no program

[Carl-NC]

I agree, there's not much we can offer without knowing what you are doing. For example:

1. What is the TX frequency? Is it a normal sinusoid or something else?
2. What is the ADC sample rate?
3. How are you creating the X & R signals?
4. Do you have an R (or G) signal that is completely ground balanced?
5. What do the X & R signals look like as you sweep a target across the coil?

And so on...

[Daves]

I agree, there's not much we can offer without knowing what you are doing. For example:

1. What is the TX frequency? Is it a normal sinusoid or something else?
2. What is the ADC sample rate?
3. How are you creating the X & R signals?
4. Do you have an R (or G) signal that is completely ground balanced?
5. What do the X & R signals look like as you sweep a target across the coil?

And so on...

I agree, there's not much we can offer without knowing what you are doing. For example:

1. What is the TX frequency? Is it a normal sinusoid or something else?
2. What is the ADC sample rate?
3. How are you creating the X & R signals?
4. Do you have an R (or G) signal that is completely ground balanced?
5. What do the X & R signals look like as you sweep a target across the coil?

And so on...

What is the TX frequency? Is it a normal sinusoid or something else?

• TX frequency is 15kHz, single tone normal sinusaoid.

What is the ADC sample rate?

• I use an external ADC which is 16 bit with sampling rate of 250 kHz.

How are you creating the X & R signals?

• I use direct sampling. I filter and amplify the RX signal, than I sample it with the ADC. The ofset voltage of the ADC is 0 DC.
• I than buffer the sampled signal, 512 samples. I take FFT, and take the complex value corresponding to the bin of 15 KHz. At that bin, I get the highest amplitude as expected.
• The complex value at the bin of 15 kHz is our target value, whose real value corresponds to X channel, and imag value corresponds to R channel. I also make filtering to smooth the signals.
• I do buffering and FFT operation 100 times per second (so the X and R channels have a sampling freq of 100Hz). Then I remove the DC of the X and R channels by taking derivative. It is just having the same effect with high pass filtering in order to remove DC. I am also working on taking the second deriv but I am not sure how to proceed.
• I than make the decision whether target is present by thresholding the derivative of R. When the threshold is passed, I calculate atan(X/R) to find the angle (id of the target). Since R channel is ground balanced, I dont get any response from soil.

Do you have an R (or G) signal that is completely ground balanced?

• I do ground balacing by rotating the X+jR complex signal so that the R channel will be zero when the target is ground/soil. For other targets, i get a non zero value of R channel.

What do the X & R signals look like as you sIep a target across the coil?

• The X and R signals are seen in the following figure. When I pass a non-ferro metal, I got both of them going high, when I pass a ferro metal, one channel goes high and the other goes down. The representation of the signals are given here
• 

I did many tests in the air, the Id’s are stable enough. However when i test in the field, i have a very specific problem: For the unburied coins, i get correct ID. But when ı bury and cover the coin with soil, i get different ID when passing the coil over the coin in different directions. When ı pass right to left, i get one ID (17), when i pass left to right, i get a different ID (88 ) IDs are stable for a same scenerio in different passings. (IDs are mapped to 0-100 scale)

I observe the same phenomena when ı use other detectors in “all metal “mode. But when i switched to “discrimination” mode, the ids are becomming same for different direction passings for buried coins.

[Carl-NC]

Please re-post the image, thanks.

[Daves]

Please re-post the image, thanks.

[pito]

what about ?
Serial.print(X);
Serial.print(R);

R and X


magnitude and VDI, gold ring


I take FFT, and take the complex value corresponding to the bin of 15 KHz. At that bin, I get the highest amplitude as expected. = fft bin 98 15.2 kHz



I get about 7-8 id. What could be the reason for this? = phase shift of your ground is - 74 deg.

[Carl-NC]

The basic signal directions are correct. If the TX is 15kHz and the ADC samples at 250kHz for 512 samples then you are not synchronous. Are you applying a window filter before the FFT? I'm not sure what you will get from this but I would suspect a phase-rotating vector. That is, the FFT phase is not usable. I would suggest stepping back and doing a simple software-multiply demodulation using (+1,-1). To make it even simpler, synchronize the TX and ADC by either increasing the TX to 15.625kHz or decreasing the ADC to 240kHz. That way you have exactly 16 points per cycle.