PI metal detector for really small nuggets

[Carl-NC]

The other point which KingJL may have alluded to is what is called 'shoot through'. When, say, Q2 and Q4 turn off the high voltage will appear on the drain of Q3. This will be coupled via Cgd and turn Q3 on. Quite what will happen with the avalanching noise is anyone's guess, but likely the flyback voltage, will find its way to pin 7 of the driver. Same situation for the other side. Would not suitable zener diodes across all 10K resistors be an idea worth trying?

Shoot-through is where, say, Q2 & Q3 are on at the same time and you have current getting dumped straight to ground. This can happen in a continuous-wave driver if timing edges are too close and, say, Q2 hasn't quite turned off when Q3 turns on. It's not likely to happen in a PI transmitter where there is so much dead time between alternating pulses.

But the issue with parasitic current injection is real. I would expect that a low-impedance gate driver could handle this but these FET drivers were mostly developed for motor control where flybacks are minimal.

[Ferric Toes]

Shoot-through is where, say, Q2 & Q3 are on at the same time and you have current getting dumped straight to ground. This can happen in a continuous-wave driver if timing edges are too close and, say, Q2 hasn't quite turned off when Q3 turns on. It's not likely to happen in a PI transmitter where there is so much dead time between alternating pulses.

But the issue with parasitic current injection is real. I would expect that a low-impedance gate driver could handle this but these FET drivers were mostly developed for motor control where flybacks are minimal.

This paper is relevant for those who want the details of shoot-through.
https://archive.eetasia.com/www.eeta..._a753769f.HTM# On the coloured graph the Vgs low side is hard to see but it is yellow.

Is there any benefit in using an H bridge instead of the half bridge? I have used half bridge just once and it seemed very straightforward in that one end of the coil can be grounded. However the primary supply then has to be centre tapped to ground. I use two 12V rechargeable packs to do this. It makes the 5V supply rails for the receiver easy; just a +5V and -5V regulator in each respective line, which eliminates the need for an inverter.

Eric.

[kt315]

sch of Carl

[Carl-NC]

This paper is relevant for those who want the details of shoot-through.
https://archive.eetasia.com/www.eeta..._a753769f.HTM# On the coloured graph the Vgs low side is hard to see but it is yellow.

Now I understand... Cdv/dt shoot-through is a particular kind of shoot-through. Shoot-through in general is when the P & N devices are both 'on' and current gets dumped straight to ground. It can happen in a simple CMOS inverter if one device is not quite 'off' when the other device is turning on. CMOS chip designers go to great pains to minimize this through non-overlapped timing because it can be a significant portion of the power consumption.

With Cdv/dt shoot-through I was thinking that the current through the parasitic C was what they were calling the shoot-through current. Instead, the parasitic C current partially turns on the device and that's what creates the shoot-through current through the drain-source channel. Never considered this before, good to know.

Is there any benefit in using an H bridge instead of the half bridge? I have used half bridge just once and it seemed very straightforward in that one end of the coil can be grounded. However the primary supply then has to be centre tapped to ground. I use two 12V rechargeable packs to do this. It makes the 5V supply rails for the receiver easy; just a +5V and -5V regulator in each respective line, which eliminates the need for an inverter.

You can use a half bridge with a single supply if you add a big cap to the coil. The old Fisher Impulse did it that way. I like the full bridge for 2 reasons. First, it pretty much guarantees 100% symmetrical pulses; the half-bridge doesn't because PMOS controls one pulse, NMOS controls the other. Second is that for the same supply voltage full-bridge produces twice the current.

[KingJL]

... I like the full bridge for 2 reasons. First, it pretty much guarantees 100% symmetrical pulses; the half-bridge doesn't because PMOS controls one pulse, NMOS controls the other. Second is that for the same supply voltage full-bridge produces twice the current.

My sentiments exactly!

[Carl-NC]

sch of Carl

No, not what I had in mind. I prefer PMOS devices for high-side switches to avoid having to use boosted gate drives. Also, your schematic still needs flyback diodes somewhere, probably on Q2/Q4.

Here is a simplified H-bridge:

[Xtrem]

No, not what I had in mind. I prefer PMOS devices for high-side switches to avoid having to use boosted gate drives. Also, your schematic still needs flyback diodes somewhere, probably on Q2/Q4.

Here is a simplified H-bridge:

Smart, i was thinking more complicated, but is there no problem with switch off ??
10Kohm ??

[green]

Tried bipolar awhile back. Ended with more diodes. Wondered if is necessary. D1 and D2 are MUR460, D3, D4, D5 and D6 are 5EWH06FN.
First column, with D1, D2, D3, D4, D5 andD6. Resonance with Rd disconnected, about 1MHz.
Second column, D3 and D4 removed and D5 and D6 shorted. Resonance with Rd disconnected, about 600kHz. Needed a lower Rd.
Coil(133mm fig8 ).
With extra diodes, resonance higher and could use a shorter delay. Not sure it's better yet. Need to do some more testing.
Top four pictures are with either M1 or M2 turning off.

[green]

Tried some more testing. Four micro second delay, four micro second sample.
Circuit with extra diodes seemed to have half the noise at integrator out. (bipolar Tx)
I can disable half the Tx to make it unipolar. Lot less effect when placing hand near coil at integrator out when unipolar.
Don't know why and if it will repeat tomorrow?

[Carl-NC]

Smart, i was thinking more complicated, but is there no problem with switch off ??
10Kohm ??

No, at switch-off the body diodes of MN1,2 provide ground clamping while the opposite side has a positive flyback. So if MN1,2 are a little lazy to turn off it's no big deal.

[green]

Been trying to determine best settings for Tx time, target delay, target sample, ground sample delay and ground sample times to cancel ground signal for small nuggets(maybe less than 2 grains)for the 133mm fig8 coil. Larger coil would be better for larger nuggets. Testing for ground slope decay again this morning. Shorter Tx times have a steeper slope. Was wondering if added resistance in series with Tx coil would help. Was surprised slope increased, maybe it shouldn't? Anyone have thoughts on timings?

Ground in a quart zip lock bag. About 6x7x1 inch thick.

[waltr]

Been trying to determine best settings for Tx time, target delay, target sample, ground sample delay and ground sample times to cancel ground signal for small nuggets(maybe less than 2 grains)for the 133mm fig8 coil. Larger coil would be better for larger nuggets. Testing for ground slope decay again this morning. Shorter Tx times have a steeper slope. Was wondering if added resistance in series with Tx coil would help. Was surprised slope increased, maybe it shouldn't? Anyone have thoughts on timings?

Ground in a quart zip lock bag. About 6x7x1 inch thick.

Interesting results Green.
My guess is that the shorter TX pulses are not fully exciting the test soil sample you have.
Maybe try with different size bags of test soil, volume and thickness, and see if the decay slope changes.

[green]

Interesting results Green.
My guess is that the shorter TX pulses are not fully exciting the test soil sample you have.
Maybe try with different size bags of test soil, volume and thickness, and see if the decay slope changes.

Have recorded ground decay with sample in a 45mm film canister with a smaller coil. Similar results(shorter Tx, steeper slope). Have been thinking shorter Tx time for small nuggets, 40us lot longer than 5 times small nugget TC. Makes sense when GEB off, maybe not if GEB on. Charted amplitude vs target TC for the different ground slopes. The 100us ground sample increases noise. Steeper slopes require shorter target samples to ground balance. Less target signal with a lower TC hole.
Tried adjusting GB today with my unipolar detector. 40us Tx appears to be steeper than -1.5 slope. Target sample was less than 5us to ground balance with 40us Tx. 160us Tx was close to Excel calculation.

[green]

Been thinking about bipolar Tx vs unipolar Tx again. Adjusted V2 for 2A peak Tx(unipolar), adjusted V1 for 1A peak(bipolar). Both have same average Tx current, Think S/N ratio should be close to same both circuits. Bipolar has less avalanche time(V7 and D8 were added to simulate avalanche for unipolar circuit). Spice doesn't avalanche MOSFETS so a or b exceed avalanche volts. Unipolar Tx current would change as 12V Tx battery changed. A LDO regulator could be used to lower 12V PS to 6.9V for bipolar Tx keeping Tx current the same as battery changed. A switching regulator would be more efficient, but wondering if it might cause a noise problem. Anything I'm missing or should be looking for?

D3, D4, D5 and D6 could be removed, a connected to e and b connected to f for the bipolar circuit like Carl posted. .9MHZ instead of 1.05MHz resonance with Rd removed with simulation. Need to try with real circuit.

[green]

Tried the unipolar and bipolar circuit again. Bringing my hand next to coil caused a problem with the bipolar circuit. Thought coil volts not decaying to power supply volts might be causing problem. Tried next day and a few times after, no problem with bipolar circuit. Don't know why. Next is adding regulator to Tx power supply so can adjust for 1A peak at 160us Tx on time. Starting to like the bipolar circuit.