AMX TX

[moodz]

Signal at coil is rock solid and super clean ... ONLY with both pulses set with 50% duty cycle.
But when I set the pulse width lower then 50%; at one or both the pulses, the same: signal at coil is "tilting" in a sequences!?
Carl... this will be the problem if you plan adjustable pulse width... or not?
Too many things still not clear to me...

The intent is to adjust the frequency ... the duty cycle will have to be 50/50 or the energy conservation does not balance ...

For bipolar pulses varying the duty cycle from 50/50 is probably not good for lots of reasons.

[moodz]

I got same 20uS as proper timing difference for dead time.
I can't measure the TX amplitude, it is way over the virtual oscilloscope range, I tried with attenuation, voltage splitter, first with 1M/9M and then 10M/90M, no results though.
Amplitude is still over the roof!
Also it was the doubt where to connect the bootstraped VS from the driver. It works this way though:

your voltage divider is the wrong way round ... the 10M should be on the ground leg and the 90M should be on the series path.

[ivconic]

I thought the electronic part will be more difficult... dang!
Now I am struggling with Arduino code!
It works, but I see on the oscilloscope exactly what I don't want to see!

[ivconic]

your voltage divider is the wrong way round ... the 10M should be on the ground leg and the 90M should be on the series path.

Dang!
Pfff...
I am totally out of focus last couple days...

[ivconic]

1M/9M and 10M/90M don't solve the problem , amplitude still way over the roof.
But!
1M/90M finally attenuates it so I can catch it whole:

[ivconic]

The intent is to adjust the frequency ... the duty cycle will have to be 50/50 or the energy conservation does not balance ...
For bipolar pulses varying the duty cycle from 50/50 is probably not good for lots of reasons.

Thanks! Now I have written proof from you that I wasn't gone mad!

Ok.
So we will not have option to adjust pulse width...

[Altra]

You can get fifty volts/div on your proteus scope by turning the inner knob counter clockwise. Set both arrows to 20.

[ivconic]

You can get fifty volts/div on your proteus scope by turning the inner knob counter clockwise. Set both arrows to 20.

[ivconic]

Present "attenuator" is 1/100.
So the math gives ~650Vpp actual amplitude.
I used IRF840 for simulation.
In reality; aren't we gonna need mosfet with a higher BVDSS?

[ivconic]

Ok, finally, according to simulator; I got winning solution with Atmega328P, usin ONLY ONE digital pin!

[ivconic]

Frequency of the pulse is 24.96kHz, coming from Arduino UNO digital pin.
So... I think the work arround alternative TX driving is done here.

[Carl-NC]

1. You want to measure the coil current, not the coil voltage.
2. The circuit will work with any BVDSS but the available transition slew rate will be limited due to the BVDSS. Remember C_extra? That's what it's for, to slow down the edges.
3. The TX current duty cycle does not need to be 50%, it can be anything. However, the P-clk and N-clk edges need to be very closely aligned. For sims, I use exactly aligned edges. When I mentioned earlier about including the option to create non-overlapping clocks using dead time, I meant on the order of nanoseconds. If you put in too much dead time then energy will dissipate in the RLC coil structure and increase tilt.

[ivconic]

That's why I gave up of second pin on Arduino and used nand. Kinda tired to trimm the timer and adjust both pins.
Now I skipped that, I have now one pin, one pulse source, and nand provides very closely aligned edges with dead time exactly in order of nanoseconds.

[Carl-NC]

The NAND gate doesn't provide a dead time, it provides a delay. For one transition this will be what you want, for the other it is the opposite of what you want. However, if it's small enough it probably doesn't matter.

[ivconic]

The NAND gate doesn't provide a dead time, it provides a delay. For one transition this will be what you want, for the other it is the opposite of what you want. However, if it's small enough it probably doesn't matter.