AMX TX

[Carl-NC]

I agree, it's been slow. I don't know what to do about it.

[ivconic]

Moodz experiences can point to an interesting doubt about dead time.
I checked some facts.
There are many inexpensive H-bridge configurations on the market in the form of modules, adapted to connect to Arduino.
Mostly they all work with Arduino without problems, although the following data shows some values that would immediately
discredit the listed platforms for use, because you can't achieve a very short dead time period of a few nS with them.
Which further indicates that the dead time is not necessary to be in the range of a few nS; it can however be a longer period.
In these cases; the use of one pin and an inverter greatly simplifies the job.
I still insist on these questions for purely educational reasons.
Some examples:
Atmega328P one clock cycle is equivalent to approximately 62.5 nanoseconds when running at 16MHz.
STM32 Blue Pill (STM32F103C8T6 variant) approximately 125 nanoseconds when running at 8MHz.
STM32 Blue Pill (STM32F103C8T6 variant) approximately 13.9 nanoseconds at 72MHz.
LGT8F328P at maximum clock speed of 32MHz, and a minimum instruction cycle time of 31.25ns.

[ivconic]

Second, what I've noticed with most mosfets I've come across to date; is that they are highly resistant and durable to short current "excesses".
In this case, the AMX TX h-bridge configuration; if the dead time is exceeded; all "dissipation" will first be reflected on the coil and only then on the mosfets.
The coil has a low resistance, with a relatively high inductance (1mH if I understood the inscription on the schematic correctly).
The excessive current flow can cause voltage drops and fluctuations, which can affect the performance of other components in the circuit.
Both the high and low side transistors of the H-bridge will be on at the same time, leading to a short circuit across the power supply rails.
From all of the above; if it is a very short "problem"; the effects will first be visible on the coil itself, in the form of heating.
If the wire is thick; the mosfets will burn out first. If the wire is thin; the coil will blow like a "fuse".
Moodz, if you didn't take into account the dead time, and your prototype works fine; two questions:
1) Coil properties, wire thickness, resistance, inductance?
2) Did you let it run longer, say a few minutes, and did you check that the coil is heating up?
If the current is constant all the time; the whole process will be somewhat "controlled" and "limited".
But at the same time, this possibly "excessive" situation can be monitored through power withdrawal.
We will agree that the quality of work of other circuits will largely depend on the quality of work of TX.

[ivconic]

However, a similar h-bridge can be safely controlled with the Arduino UNO, if the simplest "nop" operation is used.

/*
 * the avr assembly code for a no operation "nop" is simple to use if you can tolerate a blocking delay
 * At 16MHz, there are 16 nops per microsecond.
 */
void setup()
{

}

void loop()
{
  //*****************************************************************************************************
  __asm__ __volatile__("nop\n\t""nop\n\t""nop\n\t""nop\n\t");  // Four "nops" will give the 0.25us delay
  
  // or
  
   __builtin_avr_delay_cycles (4);
  //*****************************************************************************************************  

   __builtin_avr_delay_cycles (12); // 12 "nops" will give the 0.75us delay

   __builtin_avr_delay_cycles (1); // 1 "nop" will give the 62.5 nanoseconds delay
    
}​

Otherwise ideal for IRQ use (note that it is still a blocking delay)...

[ivconic]

... If the wire is thick; the mosfets will burn out first. If the wire is thin; the coil will blow like a "fuse"...

Carl I saw your pcb design and it seems correct.
Maybe you should have put a few mm very thin connection to the coil connector, in the testing phase, for the above mentioned.
Or some very fast SMD fusible fuse.
White smoke is always there to call when you least expect it!

[moodz]

Moodz experiences can point to an interesting doubt about dead time.
I checked some facts.
There are many inexpensive H-bridge configurations on the market in the form of modules, adapted to connect to Arduino.
Mostly they all work with Arduino without problems, although the following data shows some values that would immediately
discredit the listed platforms for use, because you can't achieve a very short dead time period of a few nS with them.
Which further indicates that the dead time is not necessary to be in the range of a few nS; it can however be a longer period.
In these cases; the use of one pin and an inverter greatly simplifies the job.
I still insist on these questions for purely educational reasons.
Some examples:
Atmega328P one clock cycle is equivalent to approximately 62.5 nanoseconds when running at 16MHz.
STM32 Blue Pill (STM32F103C8T6 variant) approximately 125 nanoseconds when running at 8MHz.
STM32 Blue Pill (STM32F103C8T6 variant) approximately 13.9 nanoseconds at 72MHz.
LGT8F328P at maximum clock speed of 32MHz, and a minimum instruction cycle time of 31.25ns.

No need for dead time as this is not a conventional H bridge ... the flyback return current blocks the supply current during flyback ( examine the supply current during flyback period )
Applying deadtime slows the coil transitions and adds tilt to the current peaks. What is flowing in the coil ???? Current ... if you add deadtime the current cant flow ???

[moodz]

Second, what I've noticed with most mosfets I've come across to date; is that they are highly resistant and durable to short current "excesses".
In this case, the AMX TX h-bridge configuration; if the dead time is exceeded; all "dissipation" will first be reflected on the coil and only then on the mosfets.
The coil has a low resistance, with a relatively high inductance (1mH if I understood the inscription on the schematic correctly).
The excessive current flow can cause voltage drops and fluctuations, which can affect the performance of other components in the circuit.
Both the high and low side transistors of the H-bridge will be on at the same time, leading to a short circuit across the power supply rails.
From all of the above; if it is a very short "problem"; the effects will first be visible on the coil itself, in the form of heating.
If the wire is thick; the mosfets will burn out first. If the wire is thin; the coil will blow like a "fuse".
Moodz, if you didn't take into account the dead time, and your prototype works fine; two questions:
1) Coil properties, wire thickness, resistance, inductance?
2) Did you let it run longer, say a few minutes, and did you check that the coil is heating up?
If the current is constant all the time; the whole process will be somewhat "controlled" and "limited".
But at the same time, this possibly "excessive" situation can be monitored through power withdrawal.
We will agree that the quality of work of other circuits will largely depend on the quality of work of TX.

... again the CCP is not a H bridge ..

Its using a 2 volt supply .. the load resistance is about 1 ohm ( coil and mosfets ) ... so the maximum current under fault conditions is 2 amps = 4 watts.

4 watts will not blow your nose let alone blow any mosfets. Bang head on wall. repeat.

I have run versions of this circuit for many 100's of hours ... no heating / no smoke.

The TX coil is a bundle wound 10 inch 300 uH 0.5 ohm.

The current peak to peak is 1 amp so average is 0.5 amp so the power is I2R = 0.5x0.5x0.5= 125 milliwatt ... Hell will freeze over before the coil gets warm.

[ivconic]

Don't blame me!
"The wall" from your smiley is built by Carl and his scary stories here!

[moodz]

..actually there is a potential for shootthru on the low voltage fets from the 2 volt supply ... taking everything into account including gate capacitances etc the power saving was 5 milliwatts with the deadtimes in place.
However the Tilt was approx 1% worse.

you know what the benefit of banging your head against a wall is ?

It feels so good when you stop !

Not worth the effort.

[ivconic]

Nucleo dev boards haven't arrived yet. That's why I have extra time to deal with this in detail.
At the end; TX out and driving is a circuit independent of mcu selection. So this is not a waste of time.
In fact, despite everything; it turned out to be very easy to write adequate code on the arduino, which drives the TX circuit perfectly well.
So I'll be testing the TX with the Arduino until the Nucleo arrives.

[green]

Playing with my simulator some more. Can adjust for very low tilt with no target. When I add a target, I see target on the Tx coil and tilt is no longer low. Wondering if something wrong with my simulator. Does anyone doing the simulation not see a change in tilt when adding a target? Doesn't look like I could do tilt correction after adding a target.

[Carl-NC]

The TX coil & target form a transformer so while we always think about the TX coil causing the target response, the target response also gets reflected back on the TX coil. This happens with every metal detector. What you should see on the TX coil is not a linear tilt but rather an exponential that matches the tau of the target. No, you cannot correct for this.

[ivconic]

I usually don't feel like messing around with PC power supply type crap.
But today I was dealing with two Thermatake power supplies, one 400W and the other 700W.
And inside I found the following mosfet transistors:

SVF20N50F N-channel 20A,500V,RDS(on)(typ.)=0.20@VGS=10V
SVF18N50F N-channel 18A,500V,RDS(on)(typ.)=0.26@VGS=10V
SVF13N50F N-channel 13A,500V,RDS(on)(typ.)=0.44@VGS=10V
W26NM60 N-CHANNEL 600V - 0.125ohm - 26A
20N60C3 N-Channel IGBT 45A, 600V, UFS Series

Interesting, isn't it? Would they be suitable for TX?

[moodz]

I usually don't feel like messing around with PC power supply type crap.
But today I was dealing with two Thermatake power supplies, one 400W and the other 700W.
And inside I found the following mosfet transistors:

SVF20N50F N-channel 20A,500V,RDS(on)(typ.)=0.20@VGS=10V
SVF18N50F N-channel 18A,500V,RDS(on)(typ.)=0.26@VGS=10V
SVF13N50F N-channel 13A,500V,RDS(on)(typ.)=0.44@VGS=10V
W26NM60 N-CHANNEL 600V - 0.125ohm - 26A
20N60C3 N-Channel IGBT 45A, 600V, UFS Series

Interesting, isn't it? Would they be suitable for TX?

Those supplies are a goldmine ( some of the parts sometimes contain gold ) .. for various experimenter power parts. ( FETS / high voltage caps / inductors / heatsinks / fast diodes .. all types of goodies.
The DELL server supplies are really good.
Also solar inverters that are sold second hand or new ( obsolete for current efficiency requirements ) are good parts source too.

They would work no problems in the TX must avoid avalanche though.

moodz

[ivconic]

I have been dealing with computers since 1982, and with PC computers since 1989.
At the beginning, I realized that switching power supply technology is a serious science, a very special branch of electronics.
But at the very beginning I had the great misfortune of being hit and stressed by one of the first such power supplies... and then I avoided dealing with it for years!
I had dozens and dozens of old PCs in my attic for scrapping.
I don't know the number of power supplies I've thrown away so far!
Now that I see what a mistake I made...