PI metal detector for really small nuggets

[green]

I have the same simulation except d3 and d4 which i dont have and a board design waiting to be finalized. In what way it wasnt better? The coil settle time is not improved?

I added D3 and D4 so M3 and M4 didn't avalanche. Didn't seem to make much difference in spice. I used AO3407A for M3 and M4 because that is what I had, couldn't find any avalanche specs. Looking at circuit think I see why when adding D5 and D6 it doesn't help with D3 and D4 added. Spice doesn't simulate avalanche and I'm wondering about Coss. Am using 11N60 Nfet's.

I'm trying to detect a #9 lead shot so settling time is the main problem. Need minimum delay time and detecting or not detecting my hand seems to change with circuit change.

[eclipse]

How about lower inductance like 50 - 100uH ?

[green]

How about lower inductance like 50 - 100uH ?

I've wondered the same thing. Used a coil calculator to design one but haven't made one yet. 75uH is 1/4 of 300uH(1/2 the number of turns)1/2the Tx strength? Might help for #9 lead shot, probably not larger nuggets?

Haven't tried the added diodes D5 and D6 without D3 and D4 to prevent avalanching the Pfets. Am I going to damage the AO3407's if I try?

Noticed a 100p capacitor source to drain while looking at KingJL's latest schematic. Didn't have a 100p so tried a 330p(no added diodes, just the two between the P and N fets). Didn't decay faster but seems to work better, no problem bringing my hand near the coil. I could adjust damping to minimize the hand effect without the added capacitors but the added capacitors seem to work better. Don't know why or if 100p would work better.

[green]

https://www.geotech1.com/forums/atta...0&d=1611122616

Not thinking. Should have connected cathode of D3 and D4 to c and d not a and b to prevent avalanche of Pfets. Tried it. Coil resonance around 1MHz where it should be but no improvement when adjusting damping resistor. Still playing.

[green]

Zapped a couple fet drivers yesterday. After repair I decided to do a couple tests for reference as I try to improve the detector. I've been lighting a led when signal exceeds a threshold for detection distance. Recorded signal to compare S/N if noise changes later when I try to improve the circuit.

A plastic ruler was spaced 1.25 inches above the coil and targets where swept across the ruler. End to end across center and across center of one end.

Noise appears to be random. Wondering if the no target(noise)trace looks like what I should expect.

Tx_40us about 1A peak, delay_4us, sample_10us, 2000 samples/sec

Added another test. #9 lead shot end to end across center(forgot to label picture). 5usec sample instead of 10us. S/N is better but not as much as I was expecting?

[green]

I see I labeled bipolar Tx 10 pictures wrong. Left shot pictures should be across center one end, right shot pictures should be end to end across center(higher amplitude).

[green]

Another test. Varied the pulse rate(sample rate). S/N looks higher for higher sample rates. Would be better if noise was more consistent and lower.

[6666]

Another test. Varied the pulse rate(sample rate). S/N looks higher for higher sample rates. Would be better if noise was more consistent and lower.

Good work, for what its worth recently I am finding in noisy environments a CA3140 less noisy than a 5534 as a single pre amp, not trying for huge gains .

[green]

Good work, for what its worth recently I am finding in noisy environments a CA3140 less noisy than a 5534 as a single pre amp, not trying for huge gains .

Some of my noise is coil pickup.

[green]

Increasing sample rate has been suggested for increasing S/N. Doubling sample rate doubles integrator gain(integrator gain=Rfdbk/Rin*sample rate*sample time). To increase S/N, noise signal would have to increase less than double. Don't know what increase in noise to expect if sample rate is doubled. Tried a spice simulation awhile back with white noise, appeared square root of 2(1.414)might be close. Test bipolar 12, doubling rate didn't double signal gain so I tried another test using #8 lead shot(better signal to noise) and held target on center of one of the coils and lifted quickly(hopefully more repeatable). Looks better, signal increase closer to double. Noise increase less than double, maybe close to 1.5.

[Ferric Toes]

Have you tried doubling the sample rate plus doubling the integrator time constant?

Eric.

[green]

Have you tried doubling the sample rate plus doubling the integrator time constant?

Eric.

Not sure I'm correct at what you are suggesting. I use a integrator with the switches out side the feedback loop. Gain=Rfdbk/Rin*sample rate*sample time and TC=Rfdbk*Cfdbk. Doubling sample rate doubles gain and TC remains the same.

With switches inside the loop. Gain=Rfdbk/(2*Rin) and TC=Rfdbk*Cfdbk/sample rate/(2*sample time).
With switches inside the fdbk loop, doubling sample rate should halve the TC. Doubling Rfdbk should increase the TC back to where it was and double the gain. Same as integrator with switches outside the loop.

Hope I didn't make any errors.

[green]

More thinking for bipolar Tx for small nuggets. https://www.geotech1.com/forums/atta...7&d=1606753902 comparing 133mmRx with 300mmRx.
Looks like 133mm might be better for small nuggets. Especially if detecting #8 and #9 lead shot is important. Thinking probably need GEB. Made a chart comparing different GB sample times. Ground slope and target delay would effect target sample time to ground balance(a guess what they might be). Hole TC is bottom of hole(minimum signal). Thinking 50us GEB sample time for first try. 40us Tx time gives about 1A peak. Could Tx 8000pps, thinking 4000 or 5000 to start with. Any thoughts appreciated.

Tried GB today. Couldn't GB, not enough adjustment range. I set ground sample time then adjust target sample time with 4us delay between to minimize ground signal. Calculated adjustment ranges based on my guess for ground slope. bipolar Tx1.png. Slope a lot steeper with short Tx times. Been using 160us Tx lately. Need to change my adjustment ranges. Thinking 40 or 80us Tx would be good when using the 133mm fig8 for small nuggets. Larger coil could use a longer Tx pulse(maybe 160us)with a 6us instead of 4us first delay. Would need different adjustment ranges. Maybe there is a better way to do ground balance?

Ground target_ quart zip lock bag filled with some hot dirt from California about 6x7inches 1in thick.

[bklein]

Good work, for what its worth recently I am finding in noisy environments a CA3140 less noisy than a 5534 as a single pre amp, not trying for huge gains .

In experimenting with Music Man guitar amp designs years ago I found bifet op amps would not rectify CB radio interference. Maybe this plays a part here too?

[Ferric Toes]

Not sure I'm correct at what you are suggesting. I use a integrator with the switches out side the feedback loop. Gain=Rfdbk/Rin*sample rate*sample time and TC=Rfdbk*Cfdbk. Doubling sample rate doubles gain and TC remains the same.

With switches inside the loop. Gain=Rfdbk/(2*Rin) and TC=Rfdbk*Cfdbk/sample rate/(2*sample time).
With switches inside the fdbk loop, doubling sample rate should halve the TC. Doubling Rfdbk should increase the TC back to where it was and double the gain. Same as integrator with switches outside the loop.

Hope I didn't make any errors.

Just double the capacitance and keep resistors the same. Response speed should be the same as you had at the lower sample rate, but better noise averaging.

Eric.