How can I improve this switching circuit for PI?

[CrizzyD]

As mentioned in some other posts of mine I have inherited a metal detector design and I want to make some improvements to the pulse circuitry. Currently I have a P-Channel FET configuration that works well but gets very hot. I want to change it so the main switching element is an N-Channel FET, so I can choose better avalanche and RDSon characteristics (amongst other things). The original circuit looks like this:




My new design is the following:


My question: Can improvements be made to the switching topology? L1 and R5 are limitations as they are intrinsic to the coil and cannot be changed. L1 (the search coil) must also have one side connected to ground as a further limitation. There is no push-pull circuit this time, would this possibly cause an issue regarding switch-off time? Would an N-channel driver improve the performance of the circuit? I also intend to use a different MOSFET, I have only chosen the IRF840 as it seems ubiquitous - what's the best option FET if budget is not a limitation?

[Carl-NC]

You can try this but often the polarity of the TX affects the timing polarity, the charge pump polarity, and the signal response polarity so these items might also need to be corrected. It might be a bigger job than you expect.

The turn-off speed of the new design is limited by R6. An AB driver as in the old design is preferable (Q3 is upside-down BTW). You want to choose an NMOS that doesn't avalanche, and that will depend on the peak current and turn-off speed. SiC goes up to 2000V but likely you can get by with 500-900V, which can be found in normal Si devices. This will make the FET run cooler but the damping and clamp resistors will get hotter. Potentially you can sample earlier and get better sensitivity.

[CrizzyD]

Good catch for Q3 (the real schematic has it in the correct orientation!). This will be a huge job as you rightly say, everything is reversed so the signal conditioning will need to be redesigned too. At present, the back EMF generated is close to 1kV and the IRF9640 is only rated at 200V so there's a fair amount of thermal energy generated. Would there be use for an N-channel driver chip rather than a discrete push-pull bjt design? If so, is there anything particularly suitable? If I can get the thermals down enough I might experiment with a higher transmit voltage and wider pulse width for higher sensitivities. Annoyingly there's a lot of ringing that can't be compensated for (another design limitation) so the sample delay minimum is 20us which seems like quite a long time....

[Carl-NC]

Yes, a FET driver chip will work well. Tons to choose from, I used the UCC27517 in the AMX.
Keep in mind that the total heat dissipated depends on the energy in the coil at turn-off and the frequency. If you eliminate the FET avalanche, the heat will show up somewhere else (resistors).

[CrizzyD]

The issue with FET drivers is that they mostly have a Vdd of 3-20V from what I've seen and so switch a positive voltage only. I need -12V at the FET gate. I could put a cap and a diode between the driver and the FET but I think that would defeat the point of the fast switching driver. For example, your suggestion of the UCC27517 wouldn't be possible in the circuit above as it would switch a positive voltage only unless I'm mistaken?

[Altra]

Most PI circuits use a low current dc to dc converter to create a bipolar voltage supply. PI power supplies can be confusing for people new to this type of circuit. You should study some proven designs on the forum before proceeding. Search GS4, GS5, Surf pi, hammerhead pi, sd2000 barracuda etc

[Carl-NC]

The issue with FET drivers is that they mostly have a Vdd of 3-20V from what I've seen and so switch a positive voltage only. I need -12V at the FET gate. I could put a cap and a diode between the driver and the FET but I think that would defeat the point of the fast switching driver. For example, your suggestion of the UCC27517 wouldn't be possible in the circuit above as it would switch a positive voltage only unless I'm mistaken?

Everything is relative. If the FET runs from 0V to -12V, then also run the driver chip from 0V to -12V. Just keep in mind that the input logic must also be referenced to -12V. If needed, this is where you put the cap and diode clamp. I do this on the AMX transmitter to AC-couple the TX drive so if the micro locks up the bridge doesn't burn.

[CrizzyD]

I actually did try that in LTSpice but I couldn't get it to work. I'll see if I can figure it out and I'll report back.

[Altra]

I do this on the AMX transmitter to AC-couple the TX drive so if the micro locks up the bridge doesn't burn.

That's a great idea.

[CrizzyD]

Classic example of running the sim incorrectly I think!! I was pulling my hair out wondering why the logic reversal wasn't working as I had a voltage drop of -12V across Vdd and Vss by connecting Vdd to ground and Vss to -12V . Well it turns out that If I use a voltage source in LTSpice instead of the ground symbol and add a low voltage value, it works perfectly. Lesson learned.

For anyone that's interested:




This results in the following output:



Which is looking pretty good and the current draw is pretty much identical with the high side driver configuration. I guess the next step is sourcing a suitable SiC mosfet and playing around with the transmit voltage.

[CrizzyD]

Most PI circuits use a low current dc to dc converter to create a bipolar voltage supply. PI power supplies can be confusing for people new to this type of circuit. You should study some proven designs on the forum before proceeding. Search GS4, GS5, Surf pi, hammerhead pi, sd2000 barracuda etc

All good designs, but most are pretty ancient by this point

[Altra]

True, but the power supply configurations are the same for the most current designs. Just updated components. Good luck with your developments.

[Carl-NC]

Wow, that's gonna be a seriously hot circuit. As in, don't burn your fingers.

[CrizzyD]

Yeah, I'm not limited by battery power so I can power this from a DC supply. I need as strong a magnetic field as possible as I'm looking to detect small metallic objects > 1m away. I can't really find a way to deal with the heat as the energy has to go somewhere. Dissipation through the coils would be ideal but it's not really possible in this application.

[Tinkerer]

Yeah, I'm not limited by battery power so I can power this from a DC supply. I need as strong a magnetic field as possible as I'm looking to detect small metallic objects > 1m away. I can't really find a way to deal with the heat as the energy has to go somewhere. Dissipation through the coils would be ideal but it's not really possible in this application.

You can get the same Ampere turns magnetic field strength with a coil of 500mm diameter and 7 turns x11Amp =77 Ampere turns or 36 turns and 2.2Amp = 79 ampere turns.