Segmented PI coils

[Carl-NC]

It has been mentioned that breaking up a PI coil into a series of smaller coils (windings, not diameter) and individually damping each segment results in a faster coil response. The basis for this claim seems to be Spice sims, since I don't think I've seen anyone try it yet.

So I threw it into Spice and, sure enough, 2 segments is faster than a mono coil, and 4 segments even faster, and 8 segments wicked fast. It appears that as you approach a series of individually-damped single-turn coils the whole thing gets infinitely fast, and we should have the Holy Grail of PI coils.

This is one of those "something for nothing" deals, and smells funny.

Right away I saw a glaring fallacy: my Spice sims did not include all the mutual couplings between all the coil segments, which makes the sims completely bogus. Since each segment couples to all the other segments, getting it correctly set up in sims would be a nightmare. So I just built the durn thing.

My design, as mentioned in this thread, is a 12" diameter 20T coil @ 300uH. I built a mono version for comparison. The segmented version has 4 sections of 5T each, measured to 19.7uH. A small perf board holds the 4 damping resistors and wires all the coils together.

In the end, when the coil segments are critically damped, the overall settling of the coil is identical to the mono coil. The required Rd for each segment is exactly 1/4 the Rd for the mono, which is what I expected.

In order to get the results seen in Spice, I suspect all I need to do is spread the coil segments (axially), and re-damp for their individually smaller (non-mutual) inductances.

Sorry guys, there is no Free Lunch. Pay attention to the details, and don't believe everything Spice tells you.

- Carl

[Aziz]

Hi Carl,

It has been mentioned that breaking up a PI coil into a series of smaller coils (windings, not diameter) and individually damping each segment results in a faster coil response. The basis for this claim seems to be Spice sims, since I don't think I've seen anyone try it yet.

So I threw it into Spice and, sure enough, 2 segments is faster than a mono coil, and 4 segments even faster, and 8 segments wicked fast. It appears that as you approach a series of individually-damped single-turn coils the whole thing gets infinitely fast, and we should have the Holy Grail of PI coils.

This is one of those "something for nothing" deals, and smells funny.

Right away I saw a glaring fallacy: my Spice sims did not include all the mutual couplings between all the coil segments, which makes the sims completely bogus. Since each segment couples to all the other segments, getting it correctly set up in sims would be a nightmare. So I just built the durn thing.

My design, as mentioned in this thread, is a 12" diameter 20T coil @ 300uH. I built a mono version for comparison. The segmented version has 4 sections of 5T each, measured to 19.7uH. A small perf board holds the 4 damping resistors and wires all the coils together.

In the end, when the coil segments are critically damped, the overall settling of the coil is identical to the mono coil. The required Rd for each segment is exactly 1/4 the Rd for the mono, which is what I expected.

In order to get the results seen in Spice, I suspect all I need to do is spread the coil segments (axially), and re-damp for their individually smaller (non-mutual) inductances.

Sorry guys, there is no Free Lunch. Pay attention to the details, and don't believe everything Spice tells you.

- Carl

could there be a general problem of the SPICE simulators with the mutual inductances? If so, then it would be quite difficult to simulate any transformers.

The critical coil damping Rd shouldn't end up with the 1/4 Rd on the segmented coil due to lower capacitances on each coil segments.
Are the critical damping Rd's calculated or adjusted so to have the scope result of critical damping curve?
If calculated, which formula are you using for the critical coil damping?

Aziz

[Carl-NC]

could there be a general problem of the SPICE simulators with the mutual inductances? If so, then it would be quite difficult to simulate any transformers.

In my original sim I did not have mutual inductances entered. I have not tried it, as I would need to manually edit the netlist, plus I'm not sure about coupling amongst multiple coils.

The critical coil damping Rd shouldn't end up with the 1/4 Rd on the segmented coil due to lower capacitances on each coil segments.

It's hard to say what the coil capacitance is really doing, since it's distributed in an incredibly complex way. To be sure, the overall coil capacitance does not change, as it's still 20 turns bundled together. None of that capacitance has magically vanished just because I've added taps.

The Rd's I used were what I found to be necessary to achieve critical damping. I took them very slightly higher, and got undershoot.

- Carl

[Aziz]

Hi Carl,

the theory states, lowering the coils capacitance and increasing the damping resistor to get critically damped, the flyback voltage should increase. When exposing same magnetic energy on two systems, the integrals of the flyback voltages (Uflyback(t)*dt) should then be same.

To see distinct results, other capacitances should not dominate much (cables, mosfet, diode). As the flyback voltage is limited by the mosfets tfall-time, it can not go to infinite voltage level. It reaches the breakdown voltage of the mosfet. When this happens, during the time until the voltage goes below the breakdown voltage, the current slope is very low compared to not clipped periods. The flyback voltage clipping should be avoided to have more distinct effects.

We can only damp faster, when there is a higher flyback voltage, that is not clipped by the diode and fet. Then a higher current can be dissipated on the damping resistor (P = U*U / Rd ) . So the center-tapped coil will use anyway this feature due to doubling the flyback voltage and breakdown voltage.

Aziz