PI and ferrous discrimination

[Teleno]

I'm more of a designer than a user of metal detectors. I don't own or haven't ever used an IB type, so I have a question. What's the usual dostance for ferrous discrimination of a beer cap with an IB machine?

With my latest PI based on a relaxation oscillator the discrimination dostance I get is about 4-5 cm for a beer cap with a 19 cm diameter coil.

How would you rate this feature?

[waltr]

With the TGSL I built I can detect a beer cap at 10-15cm with a 25cm DD IB coil.
Discrimination of beer caps as iron does not always work. Many times they ID closer to a US Quarter (Clad copper).

It does pretty good ID'ing nails and other random iron but iron rings fools it.

[Teleno]

With the TGSL I built I can detect a beer cap at 10-15cm with a 25cm DD IB coil.
Discrimination of beer caps as iron does not always work. Many times they ID closer to a US Quarter (Clad copper).

It does pretty good ID'ing nails and other random iron but iron rings fools it.

At what distance does your detector id the beer cap correctly as iron?

Mi PI can see the caps at 15 - 20 cm but id'ing as iron kicks in at 4 cm.

[Teleno]

Been testing on the beach today.

[Teleno]

[waltr]

At what distance does your detector id the beer cap correctly as iron?

Mi PI can see the caps at 15 - 20 cm but id'ing as iron kicks in at 4 cm.

Iron ID is not reliable at any distance on beer bottle caps.

[Carl-NC]

Beer caps are flat steel and therefore have a much stronger eddy response than a magnetic response. On the vast majority of VLF detectors they appear to be non-ferrous and where they ID varies a lot with detector model. If you turn the beer cap on-edge then suddenly the magnetic response is much stronger than the eddy response and it IDs as iron.

If you were to follow a (flat-buried) beer cap response as the coil sweeps over it you will see an initial iron response, followed by a coin response, followed by an iron response. This is due to the curvature of the TX field and how it cuts through the beer cap. If you produced audio through the whole response (instead of just at the peak) then you would hear a bonk-beep-bonk response that could tell you it's a beer cap. Back in the 80s the 4-filter designs (esp from White's) did an extra double-derivative on this response and a bottle cap gave a rat-a-tat-tat response. Easy to hear.

Teleno, you probably are getting this effect at close range where the TX field curvature is more pronounced. 4-5cm is better than nothing but would not be much use at the beach where beer caps are typically 10-20cm.

[Teleno]

Teleno, you probably are getting this effect at close range where the TX field curvature is more pronounced. 4-5cm is better than nothing but would not be much use at the beach where beer caps are typically 10-20cm.

My approach is to measure the on-period of a relaxation oscillator around the Tx coil using the 16MHz crystal clock of an Atmega328P. Rather than measuring a single period I do a rolling sum of 64 to average out the error caused by the overlap of the clock and pulse periods.

Another issue is the jitter of the oscillator. It can be ninimized by certain design measures explained in this paper: https://picture.iczhiku.com/resource...pPTZRTUBBv.pdf namely maximizing the dV/dt of the ramp and the speed of the comparator. My design uses an op-amp instead (I hadn't red the paper yet) but since then I've increased dV/dt from 250mV/ms to 1V/ms and replaced the op amp by an LMH6622 with a slew rate of 80V/us and that's how I got a period stable enough to detect the caps at 4 - 5 cm. The jitter I'm getting is around 80ns.



The clock period is 62.5 ns. My next step is to try and make the jitter less than the clock period by replacing the op-amp by an LM311 comparator and to reduce the noise at the ramp and the reference voltage.




Also I believe that keeping the LM317 of the current source cool also reduces the ramp noise. My solution was to add R8 and R11 in parallel so that the bulk of the current flows through these resistors and the LM317 only has to manage a percent just high enough to keep the current constant. The noise caused by two cool 100 Ohm resistors in parallel carrying 50 mA each and a cool LM317 at 50mA should be smaller than a hot LM317 carrying a full 150mA at 60 degrees Celsius, correct me if I'm wrong.

Perhaps I can add a couple of cms. to iron discrimination. 5 - 6 cm is better than nothing and as far as I know nothing is what we already get in PI, right?

[waltr]

Interesting circuit. Resonance is a way to get some discrimination info out of a PI.

Good to read Carl's explanation of why beer caps don't always ID as iron.

And yes, any iron disc with a PI is good.