Waveform Questions

[alan sailer]

I am building a small DRSSTC and am in the beginning stages of test/tuning. A few things about the
wave form confuse me and I would like to see if anyone can clarify.

The tesla runs at ~220kHz and uses parallel TO247 devices in an H-bridge.



Red is tank current at 1V/100A. The green is a sample of the secondary voltage, 10X probe with short
piece of antenna. The blue is VCE on a low side device using a 500X diff probe.

The first question is about the strange shape of the ringdown waveform. On my first DRSSTC using
SKM100 devices the ringdown did not have the large voltage overshoot that this unit has.

The second question is the high slope on the Vce waveform toward the end of the interrupter pulse.
I saw this on the SKM100 build but it mostly went away as I tuned it up.

Finally I am curious about the voltage overshoot on the first few Vce waveforms. The UD I am using
(a SimpleTesla) has a feature for starting the coil. It supplies a start waveform, three cycles of square
wave at the beginning of the interrupter to get the coil ringing.

Any insight to these questions would be great. I can also supply much more information about the coil
if needed.

Cheers.

[Mads Barnkob]

The first 2-3 transients on the Vce (blue) would properly be hard switching transients from the start-up oscillator not hitting the right resonant frequency or phase for ZCS.

Could the large transient on reverse recovery through diode be because of a slow diode? Maybe you need a faster anti-parallel diode instead of just using the internal?

[alan sailer]

Mads,

Thanks for the input.

The IGBJT bridge are IRGP50B60PD1 that were previously used in a QCW at about 350kHz.
So I think they should be fast enough. Don't know if there is enough room to put in an external diode, it's a pretty
 filled up circuit board.

I bet you are right about the start-up signal. I may try adjusting the frequency control pot while the coil is running
and see if a different setting could minimize the transients.

Cheers.

Addendum: I was also just looking at the ringdown and the voltage peaking occurs ever other cycle. This is a clue but I have no idea what it means.

[davekni]

The first 2-3 transients on the Vce (blue) would properly be hard switching transients from the start-up oscillator not hitting the right resonant frequency or phase for ZCS.

I'd guess the same.

The first question is about the strange shape of the ringdown waveform. On my first DRSSTC using
SKM100 devices the ringdown did not have the large voltage overshoot that this unit has.

I've seen this several times before.  It is the resonance of local bridge snubber capacitance with wiring (or bus bar) inductance to the bulk capacitance.  Yours appears to be at about half of the operating frequency.  At the end of the interrupter pulse, DC current from the bulk capacitor suddenly switches from discharging to charging bulk capacitance (reverses direction).  The funny shape is because bus voltage is ringing from this sudden current change.  Scope bus voltage at the bridge (across snubber caps) if you want to see this more clearly.  (I had to redo my snubber-to-bulk connection for lower inductance to limit this ring since using 650V IGBTs on a 450V bus.)

The second question is the high slope on the Vce waveform toward the end of the interrupter pulse.
I saw this on the SKM100 build but it mostly went away as I tuned it up.

I see this too when scoping my bridge outputs.  I think it is due to parasitic inductive drops across bridge wiring.  For me, the shape changes depending on where on my bridge I connect a scope probe.  (My DRSSTC is also paralleled TO247 devices, in case that matters.)

[alan sailer]

David,

Thanks for the reply.

The snubber/wire resonance is an interesting idea. I did a quick measurement of the wire length and it gives an
inductance of 250uH. With the snubber capacitance I get a resonant frequency of 110KHz which is very close to
half my operating frequency. Looks like another right answer from you. Amazing.

I have two theories about why you are so on top of tesla technology. The first is that in building coils every possible
thing than can go wrong has gone wrong so you have gained much painful knowledge.

The second theory is that you are just real good with theory and can reason these things out.

Which is it? :-)

Cheers.

[davekni]

I have two theories about why you are so on top of tesla technology. The first is that in building coils every possible
thing than can go wrong has gone wrong so you have gained much painful knowledge.

The second theory is that you are just real good with theory and can reason these things out.

Which is it? :-)

Thank you for the compliment.  I've designed and built analog circuits for many decades, and a few Tesla coils starting 7 years ago.  Much of my Tesla coil "experience" is from virtual debugging of other people's designs here on the forum.  This particular snubber resonance I've seen only once in person on my one DRSSTC.  However, I've seen it several other times here virtually.  The worst (and most interesting) case was when the snubber resonance hit almost exactly twice operating frequency.  H-Bridge bus ripple current is at twice operating frequency (and harmonics of that), so VBus ripple was HUGE during the enable pulse.

[alan sailer]

David,

And thanks for the information. I was talking about your skill at diagnostics with a technical friend and he
wanted to know what your answer to my question would be. I'll tell him.

My electronics, I am sorry to say, never got much beyond the transistor is a switch level. Kind of first order.
I've always been impressed with people who go beyond that.

Thanks again for the help. I plan to rewire the buss cap today and maybe get to see what happens.

Cheers.

[alan sailer]

So I rewired the buss cap. Lowered the inductance (calculated) to about 80uH.

The trace is shown below.



There is still an initial peak but the alternate stuff is gone. I also adjusted the initial drive pulse but was unable to eliminate the
(suspected) non-ZVS switching transient.

If the overshot doesn't scale with increasing buss voltage I should be OK. I'll just have to watch it as I tune things up.

Cheers.

[davekni]

Looks exactly as expected, shorter overshoot ring period and lower overshoot amplitude.  This end-of-enable overshoot voltage (voltage above previous normal Vbus) is directly proportional to primary current.  The added overshoot voltage does not depend on initial Vbus (except to the extend that Vbus effects current).  Looks like about 30V overshoot at 180A peak primary current.  So, for example, if your OCD is set for 600A, Vbus overshoot could hit 100V.  IGBTs need to have voltage rating to handle nominal Vbus plus this Vbus overshoot plus diode snap-off switching transients (the very short spikes).

If you need to reduce this snubber resonance overshoot more, increasing snubber capacitance helps.  Resonant impedance goes as sqrt(L/C).  Reducing L by 1/2 reduces impedance by 1/sqrt(2), about the improvement it looks like you achieved.  Doubling snubber capacitance would provide another factor of 1/sqrt(2).

There is still an initial peak but the alternate stuff is gone. I also adjusted the initial drive pulse but was unable to eliminate the
(suspected) non-ZVS switching transient.

It is often impossible to eliminate these initial transients.  ZCS switching is ideally switching slightly before zero current, when remaining current is sufficient to make a gentle voltage transition within the dead-time between IGBT turn-off and turn-on.  When current is very low, there is insufficient current to make this gentle voltage transition.

[alan sailer]

David,

As you have probably experienced, space gets tight in a TO247 paralleled layout. getting more (low inductance)
snubber is pretty much out of the question.

One of the joys of my only brick style IGBJT coil build was the amazing amount of room to do stuff. That and the 1200 volt
rating. When you only plan to go to 300 volts that's a lot of grace.

I'll just turn up the voltage and see what happens. The maximum current I plan on is 300 amps.

Cheers.

[alan sailer]

I have been messing with a few variables but the major one was to eliminate even more of the buss
cap to Snubber inductance. It had good effects on the ring-down overshoot. the overshoot also doesn't
get worse with higher voltage.



Voltage is about 150 volts current is 1v/100amps.

I am having problems trying to control the current but I'll start another thread to complain about that.

Cheers

[davekni]

Alan,
Not sure about forum etiquette, but I find it easier when the same thread is extended for the same coil rather than starting new ones.  That way it is easy to refer back to relevant waveforms etc.

DRSSTC current is not directly controlled other than when OCD is reached.  If primary current is rising too far before getting arcs, increasing coupling and/or tuning primary frequency closer to secondary frequency (less detuning) should help.

[Mads Barnkob]

I definitely also prefer that there is one long thread for a single projects different problems, because with so many different threads, I got no idea what coil this is really about.

Alan, if you list the threads that are all related, I could merge them into one thread.

[alan sailer]

My apologies if this sounds incoherent but I am having difficulty even explaining my
ignorance.

The coil I am messing with is based on a loneoceans design DRSSTC 1. Alex Yuan has
a very similar coil.

https://www.loneoceans.com/labs/drsstc1/

Both coils appear to be very tightly coupled, with Gao's set at k~0.2. The major difference
between these two coils and mine is that they both use a TO247 half bridge and I am using
a full bridge with parallel devices.

Early in the tuning process I was raising the voltage and at about 150V got a rather scary flash-over
from part way up the secondary to the primary. Apparently the Steve Ward technique of putting
a small capacitor form the negative side of the bridge to ground saved my devices. The lexan tube
that I used to support the primary actually helped the flash over by allowing surface tracking for the spark.
At least that is what the spark looked like.

So I lowered the coupling to about 0.12. At the same time I decided that I was uncomfortable
with the magnitude of the primary current and decided to make up a larger toroid top-load. My
reasoning was this; larger top-load equals lower frequency. Lower frequency means that I need
to lower the primary frequency. If I increase the inductive portion of the tank to lower the frequency
I will be increasing the impedance of the tank which should lower the primary current.

All this was great in theory but if there was an effect on the tank current I did not notice it. When
 I calculated the primary impedance, it went from 16 to 19 ohms, not a large difference.

David seems to say that primary current is really not easy to control other than by OCD. This
kind of agrees with my experience.

So here are some questions;

To first order, can a half bridge running at 300 volts get similar performance to a full bridge running at
150 volts? The reason I ask is that the primary current in both of my DRSSTc get too high for my comfort
when I get above 200 volts. I am running doubled 120AC for 310 volts maximum.

I have run across threads where people are discussing the merits of a high impedence primary
vs a low impedence primary. What the heck are they talking about?

Is there any first order way to estimate what primary currents will be in a coil design?

And the most nebulous question of all, is there any way to systematically tune a coil? I have tuned
several tesla coils (SSTC, QCW, DRSSTC) and it always feels totally chaotic. I never reach a satisfying
conclusion, at some point I just say, good enough and quit.

Cheers.

[davekni]

So I lowered the coupling to about 0.12.

0.12 is probably too low.  I also had trouble initially on my larger DRSSTC with arcs from part way up the secondary.  Made several fixes.  In particular, added more layers of urethane to the secondary and reduced coupling from 0.157 to 0.141.  You may get away with a bit more than 0.14 on this smaller coil.  Higher coupling will help reduce peak primary current.  Arc loading will take energy out of the primary at lower current with higher coupling.

To first order, can a half bridge running at 300 volts get similar performance to a full bridge running at
150 volts?

Yes.  Both produce 300Vpp square waves into the primary L+C circuit.

Is there any first order way to estimate what primary currents will be in a coil design?

I spent a lot of time with LTSpice simulation before building my DRSSTC.  The hardest part is figuring out how to model arc loading - at what voltage the arc starts and how it grows with time.  (Successive arcs in a burst are different.  Later arcs start at lower voltage and extend farther given the hot ionized channel left by previous arcs.)  Scaling from an existing similar coil is probably the best option.

When I calculated the primary impedance, it went from 16 to 19 ohms, not a large difference.

Scaling from half-bridge to full bridge depends on your goals.  If you wanted to run at the same 300Vbus at the same wall-plug power, you need 4x primary impedance (64 ohms rather than 16).  4x primary impedance is twice the turn-count in the same geometry (smaller wire to fit in the same height) for 4x inductance, and 1/4th capacitance at 2x voltage for MMC.

Since you have a full bridge and parallel parts, it would theoretically be possible to achieve 4x wall-plug power (2x primary voltage and 2x primary current).  In that case, primary impedance should stay the same at 16 ohms.  One way to look at the partial-secondary arcs is to consider them a result of pushing a small coil to higher power.

Reducing coupling is one piece of a fix (but I'd go back up some from 0.12).  Another is to increase primary impedance some.  (In other words, don't go all the way for 4x peak power.  Use some of your increased power capability for longer on-times instead.)  Perhaps 32 ohms would be a good target between 16 and 64.  Will require re-configuring your MMC for lower capacitance and higher voltage, along with more primary turns (and/or larger primary diameter).

And the most nebulous question of all, is there any way to systematically tune a coil?

I don't know of any cook-book answer.  Tuning depends on performance goals too.  There's a thread somewhere about tuning for music vs. tuning for max arc length.  For my coil, optimum MMC capacitance is about 1.4x higher for max arc length (3m ground strikes) vs. good music performance (2m ground strikes).  The lone-oceans link you included has some tips about measuring secondary frequency with wires to simulate arc loading.  For your case of peak primary current problems, I'd suggest starting with a sharp (and moderately long) breakout point to get arcs started early.  Measure secondary frequency with the breakout, then tune primary for about 10% lower.  If primary current is still too high, increase frequency to only 5% below secondary.

[Uspring]

Is there any first order way to estimate what primary currents will be in a coil design?

It is possible, although tedious to derive an equation relating the effect of arc loading on the primary tank. Power consumption by the secondary tank will show up as power consumption in the primary tank, because that is, where the power comes from. More formally, a secondary Q value, which is caused by the arc, will result in a primary Q value. The primary tank will look lossy. That will limit the primary current.

Primary and secondary Q are approximately related by this equation:

Qpri = (Qsec/k^2) * (1 - f^2/fsec^2)^2 + 1/(k^2 * Qsec),

where f is the frequency the coil runs at, fsec the resonance frequency of the secondary and k the coupling. Say, we have tuned the coil nicely, i.e. primary and secondary resonance frequencies are equal and we are running at the lower pole, i.e. f=fsec/sqrt(1+k). We can then insert this f into the equation and find (approximately for not too large k):

Qpri = Qsec + 1/(k^2 * Qsec)

A low primary Q is an advantage. If you think about e.g. a resistor in series with the primary coil, the primary Q will be low if the resistor is large. The power dissipated in this resistor is proportional to its value for a given primary current. That implies, that the power transferred to the secondary is larger for a lower Q. Keep in mind here, that we are talking about an effective Q caused by secondary load not a real resistor in the primary circuit.

Qpri is minimal in the above equation, if Qsec = 1/k. The secondary tank is designed optimally, if this relation holds. I've written a bit about this here: https://highvoltageforum.net/index.php?topic=1538.msg12055#msg12055 . Note that Rarc and Zsec mentioned there are related to Qsec by Qsec = Rarc/Zsec.

So assume, we have the secondary built optimally, i.e. Qsec = 1/k. Then we get:

Qpri = 2/k

This corresponds to a effective loss resistance in the primary tank of

Rloss = k/2 * sqrt(Lpri/Cpri) .

Also Rloss = Vmax/Imax, where Vmax is the bridges max output voltage and Imax the bridges output current. We get then:

Vmax/Imax = k/2 * sqrt(Lpri/Cpri)

That is similar to the equation proposed here: https://highvoltageforum.net/index.php?topic=113.msg658#msg658

There are a lot of assumptions going into this:
1. The derivation of the first equation is for a steady state, i.e. sinusoidal voltages and currents of constant amplitude. That is not really the case of DRSSTCs running for short bursts.
2. I've assumed the coil to be perfectly tuned. That is nearly impossible, since the secondary resonance frequency changes due to arc loading, while the primary resonance frequency stays put.
3. A matched arc loading is assumed. That cannot be the case, when the arc is growing, since the load increases with arc length.

The equation provides just a rough estimate. But it points to the parameters to adjust if the primary current becomes too large:
a) increase k
b) increase primary L/C ratio, or equivalently primary impedance.

[alan sailer]

David,

Thanks much. I understand some of what you are saying and will study the rest. I'm kind of slowed down on
tuning right now because the garage is getting to be warmer that I like. Soldering and tinkering is finicky work
and I find sweat makes me less able to deal with it. There is also the safety aspect, feeling to warm distracts me.

Uspring,

Thank you also. I will try messing with the equations that you showed. I'm not after a close estimate of expects
tank current just something closer than the nearly two orders of magnitude difference between my (obviously)
wrong estimate and reality.

Cheers.

[Chrisader]

I suppose but not sure that spikes and oscillations at first cycles are depends from comparator. Sometimes they need more attention, because when inputs are not properly balanced then comparator start to oscillate. Check by scope output/s of Your comparator and then should be clear...

[alan sailer]

Uspring,

So I tried your equation and it looks like it works well and is much closer than I expected.
 I am measuring 310amps at 150 Vbuss and the equations says 280.

Thanks again.

Your screen name is interesting. What's the meaning if any?

Cheers.

[Uspring]

That is also closer, than I expected 

The U is my first name initial and my last name is a word in German for the spring season.
But also: We use hereabouts U as a symbol for voltages and a spring looks similar to a coil. A U-spring is a voltage-coil. I thought that to be appropriate for my interests and this forum.