AMX TX

[Willy Bayot]

This feature has suddenly appeared in this project even if it was already discussed in details elsewhere in the forum.
I think we should first list the benefits we expect from this new TX approach compared to the traditional unipolar TX for PI systems.
I can list them as far as I understand the details (to be commented and possibly completed and/or corrected):

1. Intrinsically BIPOLAR (gets rid of the Earth Magnetic Field effects)
2. Low Energy requirement from Battery --> Most of the energy injected in coil is recycled on consecutive pulse periods . The only energy losses are due to resistive coil circuits (constant) and to the absorption by ground and targets (highly variable). There is no need of damping the TX circuit. Note : Traditional PI's loose all energy injected in coil at end of each pulse period through the DAMPING --> Traditional PI's are large consumers of battery energy and generate a lot of heat inside the box)
3. No detrimental (negative) Eddy Currents generated during the long growing phase of the coil current (CONSTANT CURRENT vs sawtooth-shaped current). Targets with TC longer than the pulse half-period can still generate those.
4. Internal capacitance of the coil is not an obstacle to low pulse delays as in traditional PI's. Thus, higher inductance coils (more turns and lower current) is a possible option. Less coil current-->simpler Power supplies, lower cable and coil wire diameter, higher amp/turn. BUT Higher flyback--> needs the most modern MOSFET's with high VDS and low RdsON, 1200V, 60mOhm are widely available now
5. Only slightly more complex TX circuit than traditional single MOSFET.

PROBLEM TO BE SOLVED: Needs compensation of energy losses to guarantee the constant current. Question is: How do we compensate the constant and variable energy losses?

6. Traditional PI's generate an effective action on the targets during the second half of each period, the first half being the building-up of the ramp of coil current. This approach (Constant Current or Square current wave) generates a variable magnetic field and eddy currents during EACH half period. For the same value of KPPS, we get twice the decay slopes.
7. No need for heatsink over the damping resistor since there is NO damping resistor at the XMIT level. No heatssink over the MOSFET's if low RdsON.

Condtions under which this circuit should work:

1. All MOSFET's should be selected for a MINIMAL RDsON
2. The two main switching MOSFET (on both sides of the coil) should be selected with a MAXIMUM VDS to sustain the high flyback voltage without avalanching. The other MOSFET's can be low volatae type.
3. The small capacitor making the resonant circuit with the inductance of the coil should also be of high voltage type.
4. The two triggering signals should be strictly complementary and preferably generated both by hardware and not complemented by software (if both ON at the same time--> BOOUM)
5. The two triggering signals should have a dead time at each side. The dead time duration should be set depending on the switching speed of the main MOSFET's

[Willy Bayot]

Did you post something about this before? I've poked around various threads but can't find anything.

Then, after solving the precise measurement of the ramp of current, we should also solve how to apply the corrections!!!

[Willy Bayot]

This where the above power argument can take a right turn. Instead of 2A through a 24T coil, you push 1A through a 48T coil. Now that 3W drops to 1.5W. I like it! However, that 48T coil needs to have wire with 1.4x the diameter, meaning the weight goes up by 4x. There may be particular advantages in different coils of different inductances so you can probably accept a wide variety of turns, with accompanying trade-offs.

Let's take 50 turns of AWG22 wire and compare with the same # of turns with AWG24 wire.
Delta Weigth is 21 gr.
DELTA RESISTANCE : AWG24 : 2.6 ohm, AWG22 : 0.848 ohm

Thus, accepting a delta weight of 21gr, we could have a coil of 50x3=150 turns with about the same resistance of 2.5 ohm

[Idefix]

Actually, I would recommend the schematic and Gerbers of post #16 of the referenced thread as the schematic and Gerbers were updated to accommodate on board selection of CC and half-sine operating modes.

It would be good if everyone used to updating the version number and the date inside the files. All industrial files uses them in general. (Rev1 Rev2 and so on ), See for example the Kicad cartridge.
The file could also usefully have its name completed by the release number.
This avoids opening all files and looking for differences. And at the end of project, it s possible to have a "list of files" with their last release.

[Willy Bayot]

Let's take 50 turns of AWG22 wire and compare with the same # of turns with AWG24 wire.
Delta Weigth is 21 gr.
DELTA RESISTANCE : AWG24 : 2.6 ohm, AWG22 : 0.848 ohm

Thus, accepting a delta weight of 21gr, we could have a coil of 50x3=150 turns with about the same resistance of 2.5 ohm

This post should be ignored

[Willy Bayot]

Quick calculations comparing a traditional PI and the potential AMX in terms of power consumption, AmpTurns pushed into the ground and coil weight.

Carl's example (traditional PI) : Let's take a Coil 25 turns of AWG24 over a diameter of 8" with peak coil current = 2Amp giving consumption of 3.5Watt and generates 50AT

That gives a resistance of 1.3 ohm and a weight of 29 gr

AMX : In order to get the same effect than the PI, the AMX has only to generate a constant current of 1 Amp (delta coil current = +1Amp to -1Amp = 2 Amp peak-to-peak) and generates 50AT

• Keeping the same wire and the same number of turns, we get 1.3 ohm x 1 Amp x 1 Amp = 1.3 Watt
• Doubling the number of turns (doubles the resistance) would increase the coil weight (only +30gr) and double the power consumption (2.6Watt, still less than PI) and could generate up to 100AT provided the flyback does not avalanche the MOSFET's.

[KingJL]

Did you post something about this before? I've poked around various threads but can't find anything.

No, it comes from a design in progress using the Moodz TX. I "somewhat" alluded to it when I discussed the Moodz TX eval/test board using a 100K pot that in a developed system is replaced with a digipot. I use a MCP4552-104E_MS (digipot, I2C controlled) and a TPS56339DDC (buck converter/regulator) to generate the TX voltage.

[KingJL]

...The two triggering signals should be strictly complementary and preferably generated both by hardware and not complemented by software (if both ON at the same time)...

Agree! That is where the selection of the processor becomes extremely important... for example if the right processor with the necessary hardware peripherals, all timing signals can be generated with no interrupt servicing or hardware intervention.

A real world example ( I have actually used this in another project except that I was only using 4 RX samples):
Using peripheral modules (especially the Configuable Logic Cell (CLC) of the PIC32MK0512MCJ064 to generate the TX and RX timing with no interrupt servicing or software intervention required (after setting up and initializing the peripherals)… i.e. runs autonomously.

TX timing signals:
TMR2→CLC1(D_FF)→A/B signal
A/B→CLC2(AND_OR)→TXA signal
A/B→CLC3(AND_OR)→TXB signal

RX sample timing signals (up to nine samples)
TMR2→OC1→SMP1
TMR2→OC2→SMP2
TMR2→OC3→SMP3
TMR2→OC4→SMP4
TMR2→OC5→SMP5
TMR2→OC6→SMP6
TMR2→OC7→SMP7
TMR2→OC8→SMP8
TMR2→OC9→SMP9


edit: forgot to mention that OC1-OC9 operating in the Dual Compare mode.

[waltr]

A real world example ( I have actually used this in another project except that I was only using 4 RX samples):
Using peripheral modules (especially the Configuable Logic Cell (CLC) of the PIC32MK0512MCJ064 to generate the TX and RX timing with no interrupt servicing or software intervention required (after setting up and initializing the peripherals)… i.e. runs autonomously.

Yes, this is one reason the PIC32 family can work well for this detector.
The internal OC modules are perfect for timed pulse generation with NO Software over head.

Add that it has a 120MHZ processor core, an FPU and a DSP core and DMA ADC to memory it is more than capable to handle the hardware timing as well as almost any software processing needed.

In the Software section of forum I posted the setup code for PIC32 OC modules.
https://ww1.microchip.com/downloads/...S60001570D.pdf

[Willy Bayot]

I've actually used a 0.033 ohm in the supply line to the high side switches and a 100x current sense amp IC (INA293B3-Q1) for the exact same purpose. I try to minimize any additional resistance in the low side

It's not enough to measure the coil current, we must measure the slope of the current. How do you measure the two sides of the ramp?

[Olly]

The deadband between complementary signals will probably need to be very precisely controlled. I’m not sure about about a PIC32 but I do know that with an FPGA this could be controlled with a resolution of 5 to 10 nanoSeconds.

[KingJL]

It's not enough to measure the coil current, we must measure the slope of the current. How do you measure the two sides of the ramp?

Actually you can measure the slope of the current if you need or want to... it is all in the way you configure the input to the sense amp (hardware selection) and how you configure the (cpu) ADC capture timing (software).

[KingJL]

Yes, this is one reason the PIC32 family can work well for this detector.

I wholeheartedly agree, but selection within the family is important. Not all of the PIC2's (very few actually) provide for CLC's, which I find very handy and satisfies the needs I previously could only satisfy with a FPGA or external discrete logic.

[Carl-NC]

AMX : In order to get the same effect than the PI, the AMX has only to generate a constant current of 1 Amp (delta coil current = +1Amp to -1Amp = 2 Amp peak-to-peak)

Yup, I forgot about that!

[KingJL]

I wholeheartedly agree, but selection within the family is important. Not all of the PIC2's (very few actually) provide for CLC's, which I find very handy and satisfies the needs I previously could only satisfy with a FPGA or external discrete logic.

... PIC2's... I must be lost back in my bit slice and microcode days!!