Re: Increasing QCW streamer length
Mads Barnkob, Mon Jul 14 2014, 07:25AM

I have not yet built a QCW coil, but you are making sparks that are over 13 times your secondary length, I do not even think this ratio have been achieved by anyone else.

Do you use a secondary mmc as Steve Ward did?

Re: Increasing QCW streamer length
BSVi, Mon Jul 14 2014, 07:58AM

I'ts easy to acheive high leader:secondary ratio in QCW coils by shrinking secondary, but it's much harder to increase streamer.

As for secondary MMC - I opted to use relatively big toriod. My theory is that toroid stabilizes and even "pushes" streamer up by it's E-field, so the bigger toriod the better results are.

I think Steve had to use small toroid and secondary MMC in his famous gun due to aesthetic reasons.

Re: Increasing QCW streamer length
Steve Conner, Mon Jul 14 2014, 10:00AM

Well done, this is very impressive! :D I think it must be a new record for streamer length vs. coil length. Are you using high level modulation with a buck converter, or phase shift modulation?

You found that increasing the DC bus voltage made a longer streamer. So, reducing the impedance of the primary (fewer turns, larger tank capacitor) should have the same effect.

However, this will mean higher primary currents than if you increased streamer length by boosting the DC bus voltage.

One possibility would be to salvage a PFC front end from some big SMPS and use it to boost your DC bus to 400V.

Re: Increasing QCW streamer length
BSVi, Mon Jul 14 2014, 11:01AM

Are you using high level modulation with a buck converter, or phase shift modulation?

I'm using phase shift modulation

However, this will mean higher primary currents than if you increased streamer length by boosting the DC bus voltage.

Can you explain why is it so? As I undestand, Ohms law applys here, I=V/Z. And it's indifferent wthat to change - increase V or decrease Z.

I actually tried to decrease turns count by one and compensate with added capacitance. Current increased but streamer didnt changed much. Maybe I have to try to unwind more turns.

Oh, and have to mention - this coil has very high coupling - 0.39. Changing it to 0.2 decreases perfomance by about a half. Another idea to try is to increase coupling even more.

ne possibility would be to salvage a PFC front end from some big SMPS and use it to boost your DC bus to 400V.

I can actually build one myself with IR1155 or something like that, but I want to pack this thing in a box and dissapating another 200w from PFC could be troublesome. So, if possible I'd like to go without additional converter :)

Re: Increasing QCW streamer length
Uspring, Mon Jul 14 2014, 11:34AM

However, this will mean higher primary currents than if you increased streamer length by boosting the DC bus voltage.

Can you explain why is it so? As I undestand, Ohms law applys here, I=V/Z. And it's indifferent wthat to change - increase V or decrease Z.

I believe Steve meant, that in order to increase streamer length you need more input power, either by increasing bus voltage or primary current or both.

Re: Increasing QCW streamer length
Steve Conner, Mon Jul 14 2014, 12:28PM


Energy is equal to V*I*t. If you have 300V*200A*20ms that is 1200J of energy. Increase the DC bus voltage to 370V, and the current (assuming a resistive load) will increase somewhat due to Ohm's law. You have a current of 200*(370/300) = 246.7A, and an energy of 370*246.7*20m = 1825J.

If we want to achieve the same energy by leaving the DC bus voltage at 300V and decreasing the load resistance, then we must have a current of 1825/(20m*300) = 304A.

A QCW with IGBTs has huge switching losses and these increase at higher voltages, so it's not immediately obvious which of the above options is best. Are you using any sort of Class-DE arrangement with deadtime and snubber capacitors?


I'm guessing that by unwinding turns you are also decreasing the coupling. These have equal and opposite effects. You need to decrease the turn count without decreasing the coupling. This might involve widening the spacing of the remaining turns or using several wires in parallel.

Re: Increasing QCW streamer length
Uspring, Mon Jul 14 2014, 04:07PM

I'm guessing that by unwinding turns you are also decreasing the coupling. These have equal and opposite effects.

If we're talking about the effect on primary current at constant Vbus, I'd expect both reducng coupling and reducing primary inductance to increase current. The former would increase Qpri and the latter would imply more current at constant Qpri.

I'd guess the reason for better performance at larger coupling to be the OCD to kick in later. Also the copper resistance of the primary might make the coil less efficient wrt to the power transferred to the secondary.

Re: Increasing QCW streamer length
teravolt, Mon Jul 14 2014, 05:29PM

if you want to increase your bus voltage can you use a buck transformer on your 220. it may not be very big. you may need a dual primary so maybe 220 or 110 if you have neutral to 24 and 24 at 10 amp is only a 500va transformer

Re: Increasing QCW streamer length
Thomas W, Mon Jul 14 2014, 08:18PM

This stuff sounds really interesting... does anyone have any links to QCW design theory and how it works & so on?

Re: Increasing QCW streamer length
BSVi, Tue Jul 15 2014, 07:06AM

Steve Conner

You are right, P=U^2/Z. Its seems increasing U is really the best way to add length.

As for losses, they are almost independent of voltage as bride is soft-switched (half-zvs, half-zcs). But they are directly proportional to current.

I also have 450v bus capacitors, so with PFC I could add even more to streamer length.

Are you using any sort of Class-DE arrangement with deadtime and snubber capacitors?

I can say so, but without explicit deadtime generator or capacitors. Deadtime is formed by high gate R (10 Ohm at the gates + 1.8 Ohm in primary of GDT). Gate voltage rise looks pretty exponential, fall is, of course, forced by diodes. As for capacitors, frequency is high and output capacitance of IGBTs is large enogth to acheive ZVS.

My power stage is as simple as this:

A QCW with IGBTs has huge switching losses

They are not as huge as you tell :) Conduction loses are major source of loses as I think (because of soft switching). 150A*2V (C-E drop) is 300W. This 300W is spread betwen two paralleled devices (of course, with some error). 150W/die is well within specs even with heatsink at 100 celsius and even for CW operation.

Heatshik heats up to 40 celsius at 1bps without any cooling with ribs facing down. I think there is enough headroom with this thermal design to increase BPS to my target of 5.

I'd guess the reason for better performance at larger coupling to be the OCD to kick in later.

No OCD doesn't trigger at all. You all know the universal soulution in tesla coil building: If OCD kicks in, simply increase it's treshold.

Re: Increasing QCW streamer length
Dr. Dark Current, Tue Jul 15 2014, 09:23AM


Why would frequency have anything to do with the turn-off energy loss? It's specified in the data sheet for a given turn-off CURRENT, and is, of course, independent of the switching frequency.

At hundreds of kHz, even for the fastest IGBTs, the switching power losses largely dominate the conduction losses.

So, I'm sorry, but your argument is invalid.

Re: Increasing QCW streamer length
BSVi, Tue Jul 15 2014, 09:58AM

Why would frequency have anything to do with the turn-off energy loss?

Maybe I'm wrong, but:
Higher frequency->lower primary inductance->higher primary current (~V/L)->more inductive energy (~LI^2)->Easier to charge body capacitance (which independent of freq) druing deadtime.

Of course, my thoughts applys to this topology only.

the switching power losses largely dominate the conduction losses

This coil operates in soft switching mode, so switching loses are not dominant (in my opinion, I don't actually measured them). Of course, they are not non-existent, but heatsink temperature is pretty close to what I calculated it should be taking into accuunt conduction loses only.

Re: Increasing QCW streamer length
Dr. Dark Current, Tue Jul 15 2014, 10:36AM

The inductance of the primary winding or energy stored in it has absolutely no effect on the switching loss, my friend. Only the turn-off current is what matters and the datasheet values simply hold true.

Your biggest enemy here is the transient thermal imepdance, not a steady-state (DC) dissipation. The temperature ripple of the die should be small. The transistor will work for a while even if it is large (lets say above 30 °C), but it WILL fail sooner or later.

Re: Increasing QCW streamer length
Steve Conner, Tue Jul 15 2014, 12:17PM

Yes, if you turn off an IGBT at a given current, without a snubber capacitor across it, you will get the switching loss given in the datasheet for that current.

Re: Increasing QCW streamer length
Uspring, Tue Jul 15 2014, 12:29PM

I'd guess the reason for better performance at larger coupling to be the OCD to kick in later.

No OCD doesn't trigger at all. You all know the universal soulution in tesla coil building: If OCD kicks in, simply increase it's treshold.

Oh, and have to mention - this coil has very high coupling - 0.39. Changing it to 0.2 decreases perfomance by about a half. Another idea to try is to increase coupling even more.

Decreasing coupling usually leads to a higher primary Q, i.e. more current for a given bus voltage. I'd expect more input power because of this, leading to longer streamers. Even if current didn't increase (and didn't drop), the input power level would stay the same. So where does all the power go, if not into the streamer?

Re: Increasing QCW streamer length
Steve Conner, Tue Jul 15 2014, 01:07PM

I believe the increase in primary Q is equal to the decrease in coupling, in other words the effective amp-turns of MMF seen by the secondary doesn't change.

If the OCD isn't triggering, then on loosening the coupling, the spark length stays the same and the primary current goes up.

If the OCD is triggering, then the primary current stays the same (as the OCD is controlling it) and the spark length goes down.

If the primary circuit is lossy, the increase in Q won't be enough to counterbalance the decrease in coupling, and the spark length will go down.

Re: Increasing QCW streamer length
Uspring, Tue Jul 15 2014, 03:45PM

Qpri, as by the theoretical equation, can vary in dependence of k somewhere between 1/k^2 or not at all. That hinges on the tuning conditions, Qsec, etc.

My argument was based on power conservation and, as you point out losses in the primary play an important role. I imagine the loss resistance Rl and the coupled in resistance Rc (the one implied by the Qpri equation) to sum up in the primary tank. Lost power would be I^2 * Rl and power transferred to the secondary I^2 * Rc. Rc increases in some way with coupling.

EDIT: The situation is much like a voltage connected to 2 resistors in series.

If Rc < Rl, increasing coupling implies more spark output. If Rc > Rl, it will be the other way around. Since BSVi observed the former, the loss resistance is probably dominant. From this it seems, that winding a thicker primary could increase primary current as the OP wanted and at the same time iimprove efficiency.

Re: Increasing QCW streamer length
BSVi, Tue Jul 15 2014, 03:56PM

The inductance of the primary winding or energy stored in it has absolutely no effect on the switching loss

Yes, I'm agree that at turn off phase shifted IGBTs experiences hardswitching and has very high power dissapation. I calculated it to be around 800W which is much greater than 150W conduction loses. Thaks for pointing this out. I'll calculate transient thermal response and write result here :)

I believe the increase in primary Q is equal to the decrease in coupling, in other words the effective amp-turns of MMF seen by the secondary doesn't change.

Decrease of coupling decreases streamers. I think coupling has much more effect than Q.

Re: Increasing QCW streamer length
Uspring, Tue Jul 15 2014, 06:09PM

Decrease of coupling decreases streamers. I think coupling has much more effect than Q.

Coupling and primary Q are related. A larger coupling will draw more power from the primary. That will show up as a reduced primary Q.

Re: Increasing QCW streamer length
Dr. Dark Current, Tue Jul 15 2014, 07:01PM

I'd say that the spark length should increase if you decrease the coupling IF the over current protection doesn't kick in. Simply the input current and power are larger. The efficiency does decrease, but I'd think not to the point so that the power loss would dominate over the transfered power. That would be one very inefficient coil

Re: Increasing QCW streamer length
redruM69, Tue Jul 15 2014, 07:02PM

Can you upload a video of this beast?

Re: Increasing QCW streamer length
Uspring, Tue Jul 15 2014, 07:20PM

I'd say that the spark length should increase if you decrease the coupling IF the over current protection doesn't kick in. Simply the input current and power are larger. The efficiency does decrease, but I'd think not to the point so that the power loss would dominate over the transfered power. That would be one very inefficient coil.

I completely agree with that. He observed decreasing arc length with decreasing coupling, though, which makes me believe in a lot of losses. On the other hand the coil performs nicely. I'm puzzled.

@BSVi: Can you tell us, what primary wire diameter you use?

Re: Increasing QCW streamer length
BSVi, Tue Jul 15 2014, 08:48PM

I'd say that the spark length should increase if you decrease the coupling IF the over current protection doesn't kick in.

No, it doesnt. After coupling is decreased, streames got smaller. And OCD didn't kicked in.

Can you upload a video of this beast?

Easily:
But video is not as good as stills.

Can you tell us, what primary wire diameter you use?

Its 2.5mm^2 solid copper wire (1.8mm^2). Primary gets warm, but not hot.

Re: Increasing QCW streamer length
Goodchild, Tue Jul 15 2014, 09:50PM


I don’t agree with this statement matter in fact it should be the opposite. Spark length should get larger with an increased coupling. Lowering the coupling only makes the current large because more of the energy is staying in the primary circuit.

The desired outcome is to transfer the energy into the secondary circuit as fast as possible, such that it can be used to make sparks, rather than sloshing around in the primary LC loosing energy to resistive loss.

In a DR coupling is the only thing limiting the flow of energy into the secondary circuit and because we allow all of this energy to be available at once, the coupling needs to be kept low such that we don’t have an excessive dv/dt on the output, causing a flash over.

In a QCW however we are able to control power out of the bridge, only giving the resonator power as we feel fit for optimal spark growth. As such the coupling can be made much larger, because the coupling is no longer the only thing limiting the flow of power. This allows control of the coils output dv/dt to a rather fine degree.

From what I can tell this is why QCWs perform so well. By keeping the output dv/dt low we are able to grow the plasma channel over more time without flashing over at higher couplings. This is all due to controlling the rate at which power is fed to the resonator.

From what I have found the limiting factor is usually how well you can keep the coil in tune. As sparks get bigger and the secondary get more out of tune with the primary and requires more and more power to increase spark length.

The other limiting factor (and it looks like BSVi has hit this one) is the primary tank impedance. With a particular tank impedance configuration you will only be able to grow sparks to some length before you can no longer ring the LC up any further.

There are two solutions to this problem, either you must increase the bus voltage (if your setup can handle that) or you must lower the Z of the LC.

Re: Increasing QCW streamer length
Steve Conner, Wed Jul 16 2014, 07:54AM

I agree with Eric, except in a DRSSTC, the impedance of the primary also limits the rate of energy transfer. I think it's better to limit it by using a higher impedance primary and maintaining tight coupling, than by loosening the coupling.

I'm interested by this spark length limit where the tank circuit refuses to ring up any further. I would like to find a way of predicting it in advance, rather than finding out after the coil is built. Do you have any measurements you could share that might help?

Re: Increasing QCW streamer length
Dr. Dark Current, Wed Jul 16 2014, 10:41AM

I still can't see how a higher input power (resulting from decreased coupling) doesn't result in longer sparks, because I think it does. The efficiency is not really that bad I'd say.

I think that it's best to look at any system involving power transfer from the point of powers and energies.

Re: Increasing QCW streamer length
Uspring, Wed Jul 16 2014, 11:00AM

Its 2.5mm^2 solid copper wire (1.8mm^2). Primary gets warm, but not hot.

I did a quick check on the AC resistance and came up with at most 0.2 Ohm for the primary. That isn't really enough to explain the coils performance dependence on coupling by resistive losses.

Steve wrote:

Energy is equal to V*I*t. If you have 300V*200A*20ms that is 1200J of energy.

300V on a 12000uF cap will store 540J of energy. Can it be, that you run out of cap voltage?

Re: Increasing QCW streamer length
Steve Conner, Wed Jul 16 2014, 11:20AM

DR. DC, your argument would be true if the primary current were allowed to increase. However I think loosening the coupling gives less spark length increase per primary current increase, than maintaining the coupling and lowering the primary impedance.

Looking at it in terms of power and energy, when the coupling is loosened, more energy is stored in the primary as reactive power and returned to the DC bus after the burst ends. That is where the "missing" energy goes.

Re: Increasing QCW streamer length
Uspring, Wed Jul 16 2014, 12:20PM

DR. DC, your argument would be true if the primary current were allowed to increase. However I think loosening the coupling gives less spark length increase per primary current increase, than maintaining the coupling and lowering the primary impedance.

Power transfer is proportional to k * Ipri and other things. For a given bus voltage Ipri is proportional to Qpri omitting losses in the primary. The equation for Qpri predicts a proportionality between Qpri and 1/k^2, if the operating frequency would not be affected by coupling. Since that is the case, also a weak dependence on k is possible. So we have to extreme cases:

a) Ipri ~ 1/k^2. Then power transfer is proportional to 1/k, i.e. it will increase with less coupling, although at the price of higher primary currents.

b) Ipri is constant. Then power transfer is proportional to k and will increase with more coupling.

Mostly the situation is somewhere between.

EDIT: The "other things" mentioned above makes this a lot more complicated as I've written. So please forget about the above. Sorry.

Looking at it in terms of power and energy, when the coupling is loosened, more energy is stored in the primary as reactive power and returned to the DC bus after the burst ends. That is where the "missing" energy goes.

Yes, but we're looking at 20ms bursts. Energy in the primary is a few J compared to hundreds in the burst.

Re: Increasing QCW streamer length
BSVi, Wed Jul 16 2014, 12:35PM

I have measured bus voltage before and after discharge. Before it is 379V and after - 300v. So I dont run out of voltage.

Considering capacitance of 13600uf, bang+losses energy is 364J.

Re: Increasing QCW streamer length
Dr. Dark Current, Wed Jul 16 2014, 01:38PM


Yes, but we're talking about QCW coils, where the bursts are so long that the system can be analysed in a steady state.

Re: Increasing QCW streamer length
teravolt, Wed Jul 16 2014, 04:50PM

you guys know more than I on this but I would think that more coupling is better because I would think that you want to convert your energy storage into spark. Instead of storing it in the primary. Aren't drsstc's secondaries allowed to self resonate a little therefore less coupling.

After the cap sags a little the mains will start taking over with ripple so you probably will get more than the calculated joules. the question is how long dose it take for the caps to sag to that point

Re: Increasing QCW streamer length
Goodchild, Wed Jul 16 2014, 09:31PM


Steve I agree, I usually build higher impedance tanks to help control current in most of my DRs as well.
In a QCW though having the impedance too high will result in not being able to ring up additional energy without added bus voltage. I think this is one of the issues BSVi was having.




QCWs and DRs are the same mode of operation (both transient, not steady state by any means); the only difference is how the energy is added to the tank circuit.
In a regular DR it’s not typically controlled, except by coupling and tank impedance (these are not variable parameters). In a QCW we control power flow via some modulation scheme. This still only controls the rate at which energy is allowed to flow into the LC, it doesn't change the fundamental operation of the Tesla coil resonator.

Think of it this way, if you kept lower coupling you would slowly build up more and more current in the primary LC. Once you had the secondary completely out the current would be massive in the primary and almost all of it would recycle back to the bus caps (some lost to resistive loss). With no secondary in place your coil will still have a massive amount of reactive power flowing in it.

One the other hand the tighter you couple the coil the more of this energy is transferred into the secondary LC. This is what we want! A larger current in the primary doesn’t constitute a larger output in a loosely coupled transformer.

Re: Increasing QCW streamer length
Dr. Dark Current, Wed Jul 16 2014, 11:08PM

Of course the system must be analyzed with the secondary coil in place. The ring up is much shorter than the burst length, in other words, the energy stored is much smaller than the energy transferred. So my argument for larger sparks with looser coupling still holds true... (the ring up will be longer, but still it is comparing numbers like 0.1ms and 20ms).

Re: Increasing QCW streamer length
Goodchild, Thu Jul 17 2014, 06:44AM


I still don't quite believe that, do you have any experimental data that shows this? Because classical transformer theory tells a different story.

As you lower coupling, energy that is normally transferred to the secondary ends up as reactive power in the primary. This is transformer operation regardless of whether or not it's a Tesla coil. Lowering coupling only slows down this process

The only reason we don't normally run higher couplings in Tesla coils is to control the output dv/dt so that the coil won't flash over. However in QCW we have control over the dv/dt and as such we slow it down significantly, thus allowing higher coupling without exceeding the dv/dt that the coil can stand off without flashing over.

Re: Increasing QCW streamer length
Steve Conner, Thu Jul 17 2014, 07:28AM

Jan's point (which is valid) is this:

During the burst, only real power can be transferred into the tank circuit, because we use feedback to get the drive voltage in phase with the current. Therefore the only way for energy to be recycled is by returning to the DC bus after the burst is finished. This can be significant in a DRSSTC, but there is no way a QCW tank circuit can store the amounts of energy we are talking about.

Phase shift modulation might well alter this argument, as the voltage and current are no longer in phase and reactive power flow is possible. Loosening the coupling will change the bandwidth, so you may need a different amount of modulation to get the same current envelope as you had before, and a different amount again to get the increased current required for looser coupling to show increased spark length.

I think loosening the coupling also makes detuning by streamer load more significant.

Re: Increasing QCW streamer length
Uspring, Thu Jul 17 2014, 07:58AM

Eric wrote:

I still don't quite believe that, do you have any experimental data that shows this?

Primary capacitance is 12.5nF, frequency about 400kHz. With 200A current, that would amount to a cap peak voltage of 6.4kV. The energy content of the MMC is then about 0.25J. Compare this to the total bang energy of 364J. Input power mainly goes right through the primary into the secondary.

Steve wrote:

I think loosening the coupling also makes detuning by streamer load more significant.

Yes, and that is the reason why it is a better idea to reduce primary turn count, then to loosen coupling in order to increase primary current.

Still this leaves the question open, why the energy argument does not seem to work to explain the reduction in arc length when coupling is reduced.

Re: Increasing QCW streamer length
Steve Conner, Thu Jul 17 2014, 08:49AM

Maybe he forgot to re-adjust the phase shift modulation as I suggested above. I think on changing the coupling of a phase shift modulated QCW, the optimum modulation envelope would change, possibly by a lot.

Well, that assumes that phase shift modulation is achieved by detuning the frequency. Phase shifting one side of a full bridge against the other can be seen as a form of PWM, I'm not sure if it implies a change in frequency at all.

Re: Increasing QCW streamer length
Uspring, Thu Jul 17 2014, 09:25AM

I was thinking more about the end of the burst, where there is probably no shift at all.

I have a problem of understanding how the phase shift is implemented. Are the two half bridges shifted against each other or a both bridges shifted against the current. I wouldn't know, why one would do the latter. Just to make sure.

Well, that assumes that phase shift modulation is achieved by detuning the frequency. Phase shifting one side of a full bridge against the other can be seen as a form of PWM, I'm not sure if it implies a change in frequency at all.

I tend to believe, that shifting voltages against currents in the tank implies a frequency change. For phase lead to higher frequency and more change for a low primary Q.

Re: Increasing QCW streamer length
Steve Conner, Thu Jul 17 2014, 09:51AM

Yes, but the length of streamer from a QCW depends on what happens throughout the whole burst, not just the end, so you can't understand it by considering just the zero phase shift case.

The only reason to shift "both" bridges against the current would be if you had a halfbridge.

Re: Increasing QCW streamer length
BSVi, Thu Jul 17 2014, 12:40PM

In my opinion, secondary, loaded by streamer this big, is mostly active load. We use it's resonant nature only to ignite streamer, then we simply pump power into streamer.

At this stage we need high coupling to quickly transfer power so streamer stays hot. Hot streamers branches less.

IMHO, the biggest challenge in QCW design is to make so that streamers are not branching.

As for phase shift tuning, I tune it so at lowest frequency (at the end of the bang) it outputs 100% power. I tried different settings and different ramp profiles (exp and log of different powers). Linear ramp and this settings is the best combination as I found out.

Re: Increasing QCW streamer length
teravolt, Thu Jul 17 2014, 02:00PM

BSVI are you doing frequency shifting or phase shift?

Re: Increasing QCW streamer length
BSVi, Thu Jul 17 2014, 03:19PM

teravolt, i'm doing phase shifting

Re: Increasing QCW streamer length
Ash Small, Thu Jul 17 2014, 03:48PM


So if you start off with it 'out of tune' so that, as streamer length increases, in comes into tune, you'd be able to get bigger sparks?.....Or would you not be able to grow the sparks in the first place?....As long as you can get sparks they should grow as it comes more into tune?.......Depends how far out of tune it is to start with, I suppose?

Re: Increasing QCW streamer length
Goodchild, Thu Jul 17 2014, 04:04PM

So to illustrate my point I have generated a simple LT SPICE simulation that models a basic ZCS dual resonant transformer with a resistive load on the output. Yes, yes the resistive load is not perfectly accurate I know, but it will suffice to illustrate this concept. The driving source is a behavioral voltage source running primary current feedback such that it will ZCS. The output is +-100VDC. Also all tests are performed with a 1mS burst.

I have set up LT SPICE to step through several coupling coefficients so that we can observe power at various points in the circuit. No other parameters are changed.



The lower window of traces are output power (as defined by V * I of R1 the load resistance) we will assume that higher power here represent more output power for driving the sparks and as a result longer sparks.

The middle window is input power (as defined as V * I of B2). This is the input power to the system. It should be noted that because my setup is simple and has no bulk capacitor it can be assumed that all reactive AND real power must flow through this behavior voltage source B2.

The final windows at the very top is the difference of input power to output power and is a measure of the reactive power element of the system (i.e. power that doesn’t go towards making bigger sparks but instead just sloshes around in the primary LC and bridge).

The last bit of information to note is the trace colors. Green is the lowest coupling (0.1) blue is the next highest (0.2) red (0.3) and so on as the traces get smaller and smaller.
So for better illustration of this topic I will pick two couplings and compare them side by side: 0.1k and 0.4k

Let’s start with input power:
0.1k = 12.448Kw
0.4k=3.282Kw

Next output power:
0.1k= 6.742Kw
0.4k=3.114Kw

I can see why it could be thought that lower coupling could produce bigger sparks, because input power and output power are both indeed larger with lower coupling! However this can be misleading, we must look at what part of this power is reactive and not used for spark production.

Reactive Power:
0.1k= 5.706Kw
0.4k=0.168Kw

Now it can be seen that the amount of reactive power in the system is much lower with a higher coupling. However this is still not a fair comparison, because a looser coupled system will natural try to draw more power. So we must run an additional simulation.

We will now tweak the drive voltage such that the 0.4K system also draws around 12.4Kw and then look at reactive power again. After increasing the bus voltage to +-195VDC the 0.4K coupled system now draws 12.479Kw making it a valid comparison for the early 0.1K test.

0.4K at +-195VDC:
Input Power = 12.479Kw
Output Power = 11.838Kw
Reactive=0.641Kw

Now 0.641Kw is still much lower than the 5.7Kw seen with the system coupled at 0.1k for the same input power.

Lastly we can calculate the ratio of transferred to total power for both systems:
0.1k = 54.2%
0.4k=94.88% (100V bus)
0.4K=94.86% (195V bus)

So as an extra validation it can be seen that the ratio holds true even at varying power and bus voltage for the same coupling coefficient.

So in conclusion by lowering coupling it can be misleading, because input and output power will grow without varying any other parameter , leading to bigger output, However it is much less efficient from a reactive power stand point.

I hope this helps to clear up any confusion, sorry for the technical rant.

Re: Increasing QCW streamer length
BSVi, Thu Jul 17 2014, 04:21PM

Excelent investigation, Goodchild! This is closely maches what i'm observing in real system. Seems that increasing bus voltage (and coupling) still the best way to increase streamer.

Maybe, it's even feasible to build buck-boost and use 1200v IGBTs to get ultralong streamers :)

As for tune, my experementation shows that you don't need good tune to have good streamers. That profs that secondary is almost resisitve load with low Q.

Re: Increasing QCW streamer length
Goodchild, Thu Jul 17 2014, 04:56PM


BSVi, I agree, this is what I observed with my QCW as well. As an alternative to making the bus voltage higher when raising coupling, the tank can be made lower impedance. This is useful if you are already running near the max bus voltage the system can handle. The trade off however is having to deal with larger currents in the primary LC.

The tuning part is curios; I also observed this with my system. I could change tuning by full turns and still produce much the same output for the same input power. I would assume part of this is due to the tank LC being high impedance, but I imagine coupling also plays a part. I need to think on this some more.


On another topic:

By chance have you measured the frequency/phase shift from X meter of spark from your system? I’m currently working on a spark mode for QCW sparks. And I need more data than what I can gather from my system.

Here is what I’m doing. I measure the L , C, and R values of the secondary using a high frequency LCR bridge at resonance. (Calculated values would also work) This gives me the base resonant frequency of the secondary with no spark loading. Then what I do is look at the frequency/phase shift of the secondary (via a base current CT) with various lengths of spark loading down the secondary.

Using this data we can then graph frequency shift per unit length of spark loading. This will give an approximate mathematical model of the spark’s parasitics. We can then use that model to find approximate L and C values per unit length of spark.

BSVi, would you be interested in doing a measurement of frequency/phase shit for various lengths of sparks on your system? In addition provide your secondary parameters? I would be interested to do a comparison.


EDIT:
I’m not sure if you have accesses to a deep memory O-scope, but this works very well for capturing a whole burst so that frequency/phase shift can be analyzed for the whole period of a single burst.

Re: Increasing QCW streamer length
teravolt, Thu Jul 17 2014, 07:13PM

BSVI, since you phase 2 dual half bridges ( a form of PWM as Steve said) do you try to sync the phasing to the ring up in the primary because when I watch your video some times the spark starts to leave but does not seam to go the full distance? Question, should the energy lead the ring up in the ramp up via PWM or phasing? like in a car you can advance or retard your ignition to affect efficiency. In newer cars they have variable timing to increase efficiency. Can this type of logic be applied to a tesla coil. does that make sense?

Re: Increasing QCW streamer length
Steve Ward, Fri Jul 18 2014, 07:26AM

First, excellent work BSVi .

Second, this whole coupling vs power confusion i think can be answered by saying "the copper losses are significant". I think Usprings analysis agrees completely with my line of thought: we simply get better results from higher coupling because it lowers the primary resistive losses by requiring less stored energy. Storing more energy is required when you want a bigger spark; its just the price you pay for using resonance. If we could avoid the resonant transformer, the efficiency should be better still, but good luck making any other type of HV transformer operate at >300khz and produce the required voltage.

Eric, you missed in your simulation in 2 ways. The first one is that you modeled the secondary circuit incorrectly as a series LC, which confuses the analysis. Secondly, you didnt pay any attention to the resistances of the coils, which i think its equally important to the design as the coupling or impedance of the tesla coil as a whole. Simply ignoring it will be missing the point.

In summary, if you want more power you need to lower resistances by using bigger wire and less of it, which results in lower coil impedances. Want proof?

Re: Increasing QCW streamer length
kilovolt, Fri Jul 18 2014, 07:26AM

Hello BSVi

Unfortunately my knowledge is not enough for discussing your design, but I just wanted to tell you this is one awesome great super-coil!!! I love it!

Best regards
kilovolt

Re: Increasing QCW streamer length
Uspring, Fri Jul 18 2014, 08:24AM

Eric, I think you somehow added input and output power instead of subtracting them. Also, you should run your simulation longer. At the end of the burst you will see, that there are positive and negative components to the difference, which will on the average be zero.
Initially there is a difference between input and output power due the charging up of the tanks. Once they have gotten to their final value, the difference will on the average go to zero.
In BSVis coil, the amount of energy in the primary and secondary tank is small compared to the total bang energy, so most of the time input and out power should be equal. That is, of course, only so if there aren't copper losses in the primary.

EDIT:

Here is what I’m doing. I measure the L , C, and R values of the secondary using a high frequency LCR bridge at resonance. (Calculated values would also work) This gives me the base resonant frequency of the secondary with no spark loading. Then what I do is look at the frequency/phase shift of the secondary (via a base current CT) with various lengths of spark loading down the secondary.

Interesting stuff, do you measure the phase shift between primary and secondary current?

Re: Increasing QCW streamer length
BSVi, Sat Jul 19 2014, 08:21AM

BSVi, would you be interested in doing a measurement of frequency/phase shit for various lengths of sparks on your system? In addition provide your secondary parameters? I would be interested to do a comparison.

I can measure both freq and phase (as i suspect, of secondary current, relatively to what?), but I dont have high-speed camera to compare results to spark length.

do you try to sync the phasing to the ring up in the primary

I don't understand what are you mean by "ring up in the primary". If the question is "do I synchronize phase shift to primary current", then yes, I do. It's also questionable why some streamers are much smaller than others. I suspect it's some glitch in control circutry or, maybe, good streamers happens on the peak of line voltage.

Question, should the energy lead

I have tried different ramp profiles (I suspect that is what you mean by "lead"), and results are pretty obvious - if you raise equalent voltege too fast, streamer explodes into branches. Linear ramp works best.

I just wanted to tell you this is one awesome great super-coil!!!

Thank you!

Want proof?

Outstanding!

By the way, guys. Due to war in the Ukraine, I'm looking for a job abroad. If your company needs full-stack electronics engeneer and consider foreigner, feel free to contact me via PM. I'm 28 y/o and mainly do MCU programming, but I'm also good at hardware design, pcb layout, FPGA and PC programming.

Re: Increasing QCW streamer length
teravolt, Sat Jul 19 2014, 04:56PM

BSVI do you use a FPGA to control the bridge half's and do you have pictures of the driver and its controller. thanks for the reply I would like to build my own QCW I haven't decided whether to phase shift the bridges or use a ramp generator and there advantages.

I live near silicon valley and I would think that there would be a big demand for your skills hear in California. How far do you want to go? Europe or the US. Has any of the fighting affected life in Kiev

Re: Increasing QCW streamer length
BSVi, Mon Jul 21 2014, 06:05AM

BSVI do you use a FPGA to control the bridge half's and do you have pictures of the driver and its controller.

No, I'v decided to move away from FPGAs for this project. You can do a lot of cool things with them, but for hobby use it takes too long to write code/testbenches. Now I'm a little sorry for this decision, because phase shifting requires corrections that you can't do in analog without lots of excess circutry. For example, I think it would be better to adjust QCW ramp for transient frequency. I have pretty much all FPGA code to redo this project with it, but I see no sense.

In this project I'm using xc9572 with analog comparator to do phase shifting. One input of comparator is driven with current source+capacitor to make linear ramp. This current source is reset on feedback edges to synchronize with feedback. Other comparator's input is feed with filtered PWM. In this way comprator changes it's output with phase shift from feedback and this shift is determined by PWM value.

I'v ordered 10 boards from china. When they arrive, I'll sell them for all who interested. Those board should be able to drive regular DRSSTC as well. I also have to write documentation. Hate to write documentation :(

Current setup looks as messy as this:

How far do you want to go?

Ideally, I want to go somewere near - Poland or Czech Republic as an example, but really, you dont think much when enemy is at your borders. Also, the only foreign language I know is english. As I know, it's hard to get US visa, for those who havent travelled alot so I didnt think about US. Anyway, to move I need some kind of job offer with relocation sponsorship. I can't do it myself.

Has any of the fighting affected life in Kiev

There is no actual fighting in Kiev, but economical effects affects life and expectations are really bad. I can tell all the uplesanties in separate thread or via PM if you wish.

Re: Increasing QCW streamer length
Kizmo, Mon Jul 21 2014, 02:06PM


Me me me me me

I would *love* to purchase a board or two :)

Re: Increasing QCW streamer length
Goodchild, Mon Jul 21 2014, 03:11PM

Steve,

What you said agrees with my simulation. Stored reactive energy in the primary circuit = higher losses. Lowering coupling makes these losses higher, because the energy stays sloshing in the primary circuit longer.

Also I didn’t neglect either resistance in the simulation; both are included as part of the inductor models, both primary and secondary. LT SPICE allows you to add these parameters to the inductor it’s self.

Lastly why does modeling the secondary as a series circuit confuse the simulation? I’m genially not sure why you woudn't model it in this way? In real life this circuit apparels like a series circuit with Ctop in series with the secondary. Only the secondary parasitic C should be in parellel with the secondary in my mind. (Which in this simulation it is, also incorporated in the secondary L parameters).

BTW Steve awesome photo!

Udo,

I used the RMS values of the primary and secondary side currents for my calculations.

On the subject of secondary spark parasitic, the idea is to measure phase/frequency shift from the beginning to the end of the burst in secondary current. If the system was coupled loose enough it could also be seen in the phase shift between the primary and secondary as you mentioned.

BSVi,

What I want to look at specifically is the phase/frequency shift of the secondary current in relation to its self over the period of one burst. No high speed photography needed. All that needs to be done is to capture one whole burst (secondary and primary current) on a DSO. Then we can expand and look at the difference in frequency and phase (in relation to phase and frequency at the start of the burst) over the period of one whole burst.

Basically my running theory is that the frequency at the end of the burst should be much lower than when it started, because as the spark gets bigger, it gradually adds more surface area to the toroid side of the capacitor, lowering the resonant frequency. This should be observable as a gradual change in frequency from start to finish. What my aim is to find out how to model this change is it: linear, logarithmic, sqrt, etc.

Now I have yet to see this operation but, it’s consistent with what others have claimed. Maybe Steve can also shed some light on what’s going on, he would probably knows better than I would, as he has many more hours into experimentations and measurement on the QCWs.

Re: Increasing QCW streamer length
Dr. Dark Current, Mon Jul 21 2014, 05:45PM

Just for comparison: In a VTTC running from a half-wave doubled voltage (which is somewhat close in operation to a QCW coil), the unloaded Q of the primary coil is around 200-300 and the loaded Q is around 10 (6x30cm secondary coil produces 60cm sparks). So the losses in the primary circuit are about 5% of the input power or less. Of course here we are running higher peak powers, but I don't believe the loss in the primary could be higher than 10-20%. The loaded Q of the primary tank circuit in a high peak-power coil could actually be lower than 10 and I believe the unloaded Q of our primary coils is higher than 100. I will stop here, take anything you want out of it

Re: Increasing QCW streamer length
Uspring, Tue Jul 22 2014, 09:24AM

Let’s start with input power:
0.1k = 12.448Kw
0.4k=3.282Kw

Next output power:
0.1k= 6.742Kw
0.4k=3.114Kw

Reactive Power:
0.1k= 5.706Kw
0.4k=0.168Kw

I've typed in your schematic and can confirm these values. My interpretation is different, though:
For e.g. the k=0.1 case, the total input energy for the simulation is 1ms * 12.448kW = 12.448J, total output energy 6.742J, missing energy 5.706J.
At the end of the burst, the primary cap has ramped up in peak voltage to about 10kV, so the primary tank contains about 5J. The secondary cap is at 90kV. It contains then about 0.24J. Together this is 5.24J, which accounts for most of the missing energy.

If you run the simulation for 20ms, input energy would be 531J and output energy 507J. The difference is almost negligible and cannot explain the poorer performance for the lower coupling. Adding a significant resistance into the primary tank could explain that, but it must be quite a bit higher, than the 0.2 ohms, that I estimated by wire diameter, length and skin effect for BSVis coil.

You'd need to use BSVis tank parameters for a quantitative estimate when adding primary resistance. Your choice of tank capacitances and inductances is illustrative but BSVi has different values.

Re: Increasing QCW streamer length
Goodchild, Tue Jul 22 2014, 02:39PM


Udo,

I agree with you in the fact that given unlimited amounts of time to transfer energy and no copper losses (DC or AC) coupling probably wouldn’t make any difference, because the reactive energy I talk about in my simulation would eventually be transferred. This is defiantly evident if you look at your interpretation using overall energy rather than energy per unit time like I did.

However, in a practical sense we have to deal with copper losses (both DC resistive and AC). In addition when we make sparks we have to do it in a finite amount of time. QCWs are a great example of this. Making the power ramp longer even by 5ms or so doesn’t make the sparks longer, rather chubbier. This principle of power per time is important.

Just as an example, let’s use the SPICE sim we have been playing with. Let’s say this SPICE coil needed to drive 12.448Kw on the output to produce a spark of X length. Now we can’t change the burst length because the burst length is already optimal for X length of spark. So in order to achieve this we have two options, use 0.1k with a higher reactive power of 5.7Kw or use the higher coupling with a lower reactive power and only have 0.168Kw additional.

Now to your point Udo, yes the remaining energy in both cases (5.7J and 0.168J) would simple recycle into the bus cap at the end of the burst. However during the burst the: bridge, primary capacitor, and other parts all have to deal with that extra 5.7J of energy.

What this means is that if your bridge’s max power handing is say already 12.448Kw then with 0.1k you have reached your limit. Were if you increased your coupling to 0.4K you now have an addition 5.53Kw of headroom, while still transferring the same amount of energy to the output in the same amount of time.

The best thing I can relate this to is PFC, when you use a PFC the total power doesn’t change, your only making the reactive part of it smaller so that you have more room for real power.

Re: Increasing QCW streamer length
Uspring, Tue Jul 22 2014, 04:30PM

I pretty much agree up to this point:

What this means is that if your bridge’s max power handing is say already 12.448Kw then with 0.1k you have reached your limit. Were if you increased your coupling to 0.4K you now have an addition 5.53Kw of headroom, while still transferring the same amount of energy to the output in the same amount of time.

The figure of 5.53kW only applies to the first 1ms of the burst. After about 2ms the primary peak voltage has settled, so there is no additional energy anymore being stored in the tank. At that point input and output power will become equal (provided we omit copper losses). The drawback of lower coupling is, that the higher amount of stored energy in the primary will cause more losses in the copper resistance. But remember, that the OP wanted to have more current in the primary, which also means larger losses.

I'm still puzzled about BSVis observation, that decreasing coupling has such a bad effect on arc length. In the lossless case it should actually increase due to the higher power input. I've done some calculations for a lossy primary and it seems, that you need a loss resistance of about 1 Ohm to explain, what he observed. But that is a lot more than I get by a simple estimate from his wire thickness etc.

Re: Increasing QCW streamer length
Goodchild, Wed Jul 23 2014, 02:01PM

Udo,

You are correct in that with this simulation the primary current stabilizes after 2mS. However this wouldn’t be the case in a real system. In a DR the burst would almost always be shorter than 1mS. And in a QCW power would be added to the tank more controllably than my simulation, making the coil operate in a transient mode for much longer than 2mS.

So in the real world we would never see, operation where the tank current stabilized like it does in that simulation after 1mS. At least not in the way we operate most coils today.

But it brings up a good question! How would the coil operate if it was allowed to operate in the steady state with lower coupling? Personally I don’t know. I have always favored the ability to grow the current if needed, a perk of operating in transient mode.

Re: Increasing QCW streamer length
Uspring, Wed Jul 23 2014, 02:55PM

You are correct in that with this simulation the primary current stabilizes after 2mS. However this wouldn’t be the case in a real system. In a DR the burst would almost always be shorter than 1mS. And in a QCW power would be added to the tank more controllably than my simulation, making the coil operate in a transient mode for much longer than 2mS.

So in the real world we would never see, operation where the tank current stabilized like it does in that simulation after 1mS. At least not in the way we operate most coils today.

I agree with your remarks on both short and longer burst operation.

But it brings up a good question! How would the coil operate if it was allowed to operate in the steady state with lower coupling? Personally I don’t know. I have always favored the ability to grow the current if needed, a perk of operating in transient mode.

Lower coupling will draw more current for a given input voltage as seen in the simulation. It will also make the coil more sensitive to detuning by the arc capacitance. If you can't obtain the desired amount of input current even at max input voltage, it is better to reduce primary inductance. If your current is too high, it is better to increase coupling rather than increase primary inductance. Increasing coupling can be problematic due to flashovers, though.

Re: Increasing QCW streamer length
Ash Small, Wed Jul 23 2014, 03:11PM


It does make sense that if there is no increase in primary current when coupling is decreased then current in the primary is limited by resistance in the primary.

Maybe there are 'other resistances' in the primary circuit?

BSVi has linked to a photo on page two of this thread which he precedes with words something like 'My current setup is as messy as this'.

If those brown capacitors are the primary tank capacitors, then maybe the 'limiting resistance' is there, or maybe elsewhere in the primary tank circuit.

The primary coil itself is not the only resistance in the primary tank circuit.

EDIT: @BSVi, can we have some more photo's of the whole primary tank circuit, please?

Re: Increasing QCW streamer length
teravolt, Wed Jul 23 2014, 04:38PM

BSVI, Ash I think has a point. Have you try adding more capacitance I.E. energy storage for your bridge to convert. If go to 20ms will there be enough gas at the end of the ramp up. How would it be pausible to calculate the amount of energy in joules that is needed for ramp up. if I was going to do it I would just keep adding caps till the spark stops growing or can it be pausible to go to far.

Re: Increasing QCW streamer length
Dr. Dark Current, Wed Jul 23 2014, 04:41PM

So it seems the tank caps could be the limiting factor here. Using better caps, BSVi could probably get longer sparks ( +with lower L and higher C).

Re: Increasing QCW streamer length
Goodchild, Wed Jul 23 2014, 08:25PM


Pardon quoting myself but, I pointed that out a while back. The tank impedance is one of the main factors that limits spark growth when you are running a fixed bus voltage.

And as such lower the tank Z (making the tank cap bigger) allows the current to ring up higher with the same driving voltage.

Re: Increasing QCW streamer length
Uspring, Thu Jul 24 2014, 11:08AM

I believe, that an explanation of the effect of lower arc length with lower coupling may be a mixture of different issues:

First consider a lossless primary tank. Lowering coupling will increase current and therefore power input. Most of this goes to the output. There is also an effect, as Eric pointed out, that more energy will be stored in the primary tank. But this primary energy is about 0.25J for BSVis coil compared to the total bang energy of 364J. So this is a small effect and actually spark length should increase.

Now consider a lossy primary. Current will also increase here with lower coupling, increasing power input, but also energy is burned in the tank. If tank loss resistance is large enough, power output decreases, even though power input increases. That will be at the point, where about 50% of power input is burned up in the primary. This does not really make sense to me, since the coil performs too well to be that lossy. So I don't think, that less spark length with lower coupling can be explained by this alone.

BSVi wrote:

I have measured bus voltage before and after discharge. Before it is 379V and after - 300v. So I dont run out of voltage.

Considering capacitance of 13600uf, bang+losses energy is 364J.

If this is for the large coupling case, I would assume the cap voltage drop during the burst to be considerably larger for the low coupling, since there is an increased current draw. It might be just large enough to stifle spark growth at some time during the burst.

Re: Increasing QCW streamer length
Ash Small, Thu Jul 24 2014, 02:08PM


If, as I suspect, the wiring to the capacitors is the 'limiting resistance' in a lossy primary tank circuit, then decreasing coupling will tend to increase current in the primary tank, which will in turn lead to extra heating, leading to increased resistance, leading to lower current.

My comments are purely based on the way the capacitors are wired up in the photo I mentioned above.

Current will only rise to a finite limit in any tank circuit, due to resistive (ohmic) and heating losses, even if completely uncoupled. Maybe this limit has been reached in BSVi's circuit?

Once this limit is reached, spark length depends entirely on coupling (and secondary losses), I think.

Re: Increasing QCW streamer length
Wolfram, Thu Jul 24 2014, 04:05PM


There will be smoke, fire and burnt components long before the thermal resistance rise of the wiring becomes significant. Also, if wiring resistance was significant enough to limit performance with lower coupling, the dissipation in the wiring would lead to complete destruction in a short time.

Re: Increasing QCW streamer length
Ash Small, Thu Jul 24 2014, 04:21PM


Wolfram, Udo's calculations, as I understand them, suggest that one Ohm of resistance would be sufficient to limit further increase in primary current. I wouldn't expect one Ohm to result in total destruction, but what do I know?

I was just commenting on the way the capacitor bank is wired in, however I will stop posting in this thread if you think I'm on the wrong track.

Re: Increasing QCW streamer length
Steve Conner, Thu Jul 24 2014, 05:38PM

Any info on the caps? What is the dielectric? Do they get hot?

Re: Increasing QCW streamer length
Ash Small, Thu Jul 24 2014, 05:42PM


This is the photo BSVi linked to:



EDIT: If that string of brown capacitors on the left is the MMC then I'm assuming that might have something to do with the problem.

I'm assuming the red wire going off towards the left is to the other end of the primary coil and that those toroidal inductors it passes through form some kind of current/frequency/phase sensor thing.

The only other reference I've found is "MMC=12.5nF"

Re: Increasing QCW streamer length
teravolt, Thu Jul 24 2014, 07:13PM

the way I see it the small electrolytics are in series and that bank is in parallel with the big cap and as for the wire I have seen worse

Re: Increasing QCW streamer length
Ash Small, Thu Jul 24 2014, 07:29PM

Those brown capacitors are definitely a series string, Teravolt. I'm pretty certain that is the MMC. The other end of the series string goes to the primary coil, I assume, and that red wire with the toroidal inductors around it is the return wire from the primary coil, but I'm only guessing.

Re: Increasing QCW streamer length
BSVi, Fri Jul 25 2014, 06:14AM

The boards has arrived. I now in process of soldering. This how bare board looks like:

What I want to look at specifically is the phase/frequency shift of the secondary current in relation to its self

Ok, I'll do that when I get coil running again.

Basically my running theory is that the frequency at the end of the burst should be much lower than when it started

With no streamer system resonates at 430kHz and with full streamer F drops down to 392kHz. So, frequency changes for about 10%.

Making the power ramp longer even by 5ms or so doesn’t make the sparks longer, rather chubbier. This principle of power per time is important

Agreed. I also think an optimal point exist in "power per time" vs "absolute power". Ramping power too fast explodes streamer into branches and pumping more power (more ramp time) stabilizes it.

Another my theory is that in optimal QCW operation we only compensate for streamer power loss. What I mean by this is that before we pump additional power, we have to wait before additional "chunk" of streamer is formed and consumes additional power. If we pump power faster, it will branch. This is consistent with the fact that linear ramp is optimal.

0.2 ohms, that I estimated by wire diameter, length and skin effect for BSVis coil.

it seems, that you need a loss resistance of about 1 Ohm to explain, what he observed. But that is a lot more than I get by a simple estimate from his wire thickness etc.

There is more to losses than resistive copper loss. The toriod is about as warm as primary wire. Secondary is warm too, once its wire even burned out. I think, huge amount of energt is wasted heating up all those things. MMC is cold, which is strange considering thin solder-coated leads. Bridge is cold.

Increasing coupling can be problematic due to flashovers, though.

Unlike drsstcs, voltage in QCW system is low, and you can prevent flashovers with conventional insulation materials.

BSVi, can we have some more photo's of the whole primary tank circuit, please?

I have to take more photos, but I'll do that after coil reconstruction.

Have you try adding more capacitance I.E. energy storage for your bridge to convert

No, unfortinutaley, I don't have access to another big capacitor. I have to get another one or two. I agree that 80v voltage drop is signifacant, throught. But voltage drop decrease logariphimcally with capacitance, so to decrease it much, you need really BIG primary bus capacitor.

I also afraid of big capacitors. With 1kJ energy stored, if circuit explode, IGBT's shrapnel could be dangerous. Eye protection required :)

Now, I think that the best (length-wise) solution to drive QCW is to build buck-boost converter, or some kind of insulated converter with stepup transformer. As we now know, higher voltage gives more bang than higher current. Bus capacitance becames not as important as in phase-shifting version, and you get softswitching everywere.

If this is for the large coupling case, I would assume the cap voltage drop during the burst to be considerably larger for the low coupling, since there is an increased current draw. It might be just large enough to stifle spark growth at some time during the burst.

It's good idea to take measuremet of this voltage drop with lower couping. I'll do this.

Any info on the caps? What is the dielectric? Do they get hot?

It's cheappy CBB-81. Metallized polypropylene dielectric. 8pcs of 2000v/100n caps in series. They are pretty cold.

Those brown capacitors are definitely a series string, Teravolt. I'm pretty certain that is the MMC. The other end of the series string goes to the primary coil, I assume, and that red wire with the toroidal inductors around it is the return wire from the primary coil, but I'm only guessing.

You are right.

Re: Increasing QCW streamer length
Uspring, Fri Jul 25 2014, 08:14AM

Wolfram wrote:

There will be smoke, fire and burnt components long before the thermal resistance rise of the wiring becomes significant. Also, if wiring resistance was significant enough to limit performance with lower coupling, the dissipation in the wiring would lead to complete destruction in a short time.

I tend to agree here. At 1 Ohm primary loss resistance, maybe a 100J or more per bang will get burned somewhere.

BVSi wrote:

No, unfortinutaley, I don't have access to another big capacitor. I have to get another one or two. I agree that 80v voltage drop is signifacant, throught. But voltage drop decrease logariphimcally with capacitance, so to decrease it much, you need really BIG primary bus capacitor.

It's not as bad as that. Doubling the caps will halve the voltage drop.

It's good idea to take measuremet of this voltage drop with lower couping. I'll do this.

Thank you in advance. I'm interested in what you get. If you increase primary current you won't get around beefing up the caps. At higher bus voltages, the drop will be lower in relation to it, so it will be less bothersome.

Re: Increasing QCW streamer length
Steve Ward, Sun Jul 27 2014, 03:25AM

Now consider a lossy primary. Current will also increase here with lower coupling, increasing power input, but also energy is burned in the tank. If tank loss resistance is large enough, power output decreases, even though power input increases. That will be at the point, where about 50% of power input is burned up in the primary. This does not really make sense to me, since the coil performs too well to be that lossy. So I don't think, that less spark length with lower coupling can be explained by this alone.

I think the secondary resistance and impedance plays a role here too. The EMF driving the secondary sees more leakage inductance in the secondary, which has nearly as much impedance as the arc has, so wouldnt this cause a reduction in power as well and help explain the sensitivity to K?

Also, there has been some question about energy storage and i pose this hypothetical to maybe help think about the issue. Imagine you have 2 secondary coils, one with 20pF of effective capacitance and one with 40pF. Assume the Fres is identical. We know from measurements that it takes at least 40kV to begin spark production, so as i see it, the 40pF resonator requires twice the drive current (all else being equal) to achieve just 40kV (which is about half the peak voltage most QCWs will develop). If the coils had the same resistance, then significantly more power would be wasted in the 40pF resonator. However, the winding should only need half the inductance, which means less length and thicker wire. I suspect the resistance could go down sufficiently so that both secondaries have the same losses while producing 40kV. But still, there will be twice the primary amps required to drive the 40pF load. Whats the benefit? The same spark looks like a smaller load to the 40p coil and consequently the inverter can drive more power into the system and make bigger plasma because the source impedance is effectively smaller.

Bigger coils have higher energizing costs, and i think its really quite significant in terms of primary current during operation. Figuring out the right amount of energy storage for your tesla coil design i think is key to optimizing its efficiency. Seems obvious, and overlooked.

Re: Increasing QCW streamer length
teravolt, Sun Jul 27 2014, 05:38PM

BSVI do you plan on selling your board once tested. You could put it in kickstarter if you wanted to develop it and sell QCW kits

Re: Increasing QCW streamer length
Uspring, Mon Jul 28 2014, 11:53AM

I think the secondary resistance and impedance plays a role here too. The EMF driving the secondary sees more leakage inductance in the secondary, which has nearly as much impedance as the arc has, so wouldnt this cause a reduction in power as well and help explain the sensitivity to K?

In a QCW setting, any obstruction of the power flow to the secondary, may it be from lower coupling or higher leakage inductance, will cause the primary to ramp up to higher currents, since it is less loaded. Leakage inductance is not consumptive, so the additional input power will go the secondary. For a short burst DRSSTC a significant part of this extra energy will be stored in the primary and may or may not be returned to the bus at the end of the burst.

...The same spark looks like a smaller load to the 40p coil and consequently the inverter can drive more power into the system and make bigger plasma because the source impedance is effectively smaller.

If the bridge can handle that, yes. Whether it is an advantage depends on whether the bridge has reached its current or voltage limit.

Bigger coils have higher energizing costs, and i think its really quite significant in terms of primary current during operation. Figuring out the right amount of energy storage for your tesla coil design i think is key to optimizing its efficiency. Seems obvious, and overlooked.

I believe, this applies in particular to coils with shorter bursts. Total bang energy and energy in the primary can be of similar magnitude. The energy in the tank does not have to be lost. One mode of operation can be to charge up the tank significantly by avoiding early breakout by detuning. When the arc finally starts, it will pull the secondary into tune, which will dump most of the primary energy into the secondary.

Re: Increasing QCW streamer length
Ash Small, Mon Jul 28 2014, 11:57AM


Your example of two different secondaries with the same Fres reminds me of another thread here some time ago, which I'm not sure you were involved with, Steve.

The subject was the effects of voltage and current on streamer growth. While it seems obvious that higher voltage leads to bigger streamers, the effects of higher current, the thread concluded, can also have a significant effect on streamer growth, due to the heating effect and greater ionization. Put simply, the greater heating and ionization of the air due to higher current makes it easier for the next cycle to ionize more air, leading to increased streamer growth.

This, I assume, further contributes to the load looking even smaller, as far as the higher capacitance, lower inductance secondary is concerned. Also, the capacitance of the streamer has less effect on the Fres of the secondary than the same size streamer would on a lower capacitance, higher inductance secondary, I think.

I hope I'm not going off topic here.

Re: Increasing QCW streamer length
Steve Conner, Mon Jul 28 2014, 02:20PM


Something is wrong with this argument, as it implies that the harder you try to prevent power flowing into the secondary, the more power will flow to the secondary. The logical conclusion is that putting the secondary in the next room would give 20ft arcs. I assume this is not the conclusion you intended, so can you explain?

Re: Increasing QCW streamer length
Goodchild, Mon Jul 28 2014, 03:16PM


You have hit the nail on the head. This is my main reservation on the whole lower coupling = bigger sparks argument.


EDIT:
BSVi, what are you using for an RF ground? I didn’t see a mesh or anything of the sorts in your photo.

Re: Increasing QCW streamer length
Uspring, Mon Jul 28 2014, 06:50PM

Something is wrong with this argument, as it implies that the harder you try to prevent power flowing into the secondary, the more power will flow to the secondary. The logical conclusion is that putting the secondary in the next room would give 20ft arcs. I assume this is not the conclusion you intended, so can you explain?

Steve, Eric, you're being too realistic

The hitch is, that I've neglected the effect of primary copper losses. To quote myself:

In a QCW setting, any obstruction of the power flow to the secondary, may it be from lower coupling or higher leakage inductance, will cause the primary to ramp up to higher currents, since it is less loaded.

In a lossy primary, current won't ramp up forever, even if there is no secondary at all. To quote myself again:

Now consider a lossy primary. Current will also increase here with lower coupling, increasing power input, but also energy is burned in the tank. If tank loss resistance is large enough, power output decreases, even though power input increases.

That is going to happen at some current level. Just for entertainment consider a superconducting primary and caps and fat IGBTs. Set that up in a lab, that is electrically inert and start the burst. After some time you'll have megamps circulating in the primary. Finally the secondary next door will notice and only stop growing arcs until the gigawatts in the other room have found a place to go.

The upshot is, that a low k will increase power throughput only as long that primary losses allow for this, i.e. up to some current level.

Re: Increasing QCW streamer length
Steve Conner, Mon Jul 28 2014, 07:21PM

Well, the other problem is that the lower the coupling, the narrower the system bandwidth and so the more vulnerable it is to detuning by streamer load.

I take this to mean that the lower the coupling, the bigger the topload required for a given size of spark.

I have no idea what happens when you apply infinite power to a superconducting coil that's badly out of tune. Probably a semi-infinite explosion.

Re: Increasing QCW streamer length
Ash Small, Mon Jul 28 2014, 09:03PM


I'm beginning to form the opinion that a larger topload will always result in a larger spark, assuming copper losses in the secondary don't dominate. This was the case with Odin, and we've seen recently in other threads that adding capacitance, in the form of a string of capacitors inside the secondary, also improves streamer length.

This also helps with detuning due to streamer load, and assuming secondary copper losses don't dominate, the voltage will still ring up, regardless of winding ratio.

I'd be interested to hear about any instance when this might not hold true.

Re: Increasing QCW streamer length
Steve Conner, Mon Jul 28 2014, 09:44PM

The way I see it, the topload size is not unlike the "loading" control on the old ham radio antenna tuner. There is an optimal size that extracts maximum power from the driver.

If the topload is too small, the streamers will detune the coil too easily and the primary current will ring up uselessly and trip the current limiter.

If the topload is too big, the primary current will get excessive and trip the current limiter, before the output voltage gets big enough to grow the desired size of streamer.

The optimal size of topload probably depends on several variables.

Re: Increasing QCW streamer length
Ash Small, Mon Jul 28 2014, 10:19PM


Copper losses? (I don't expect an answer, Steve. I think I'm starting to get the idea. It's probably time I built a TC.)

Re: Increasing QCW streamer length
Uspring, Tue Jul 29 2014, 08:12AM

A large topload or equivalently, a secondary MMC, as well as a large coupling are beneficial wrt arc detuning problems.

Wrt excessive primary current, top load might have some effect. But primary current also can be controlled by primary inductance or coupling. For Odin, the extra turn probably was partially responsible for the success. I assume, that you added the extra top load in order not to have to fiddle with the primary MMC (?). Another rationale behind the extra top load is not clear to me.
The upper pole operation of Odin should make it less susceptible to arc detuning.

Maybe a big top load serves as a backing energy store to feed the arc. Mostly, though, the energy stored in the secondary tank is much less than that in the primary.

Edit: In the steady state the energy ratio between secondary and primary tank is Qsec/Qpri, which might be around 1. Initially, during primary rampup, it is likely less than that.

Re: Increasing QCW streamer length
Steve Conner, Tue Jul 29 2014, 09:19AM

In the case of Odin, I added the extra topload because it seemed like the output was being limited by detuning, and the topload was smaller than what I saw used on other coils.

I added the extra primary turn because I felt that the primary impedance was already too low, so I didn't want to lower it still further by adding extra capacitance.

I'm not sure how relevant the upper vs. lower pole distinction is to Odin. The bursts are short enough that the system doesn't really settle to a steady state. My previous coil Mjollnir had at least 3x more cycles per burst, so there was a clear transition into steady-state behaviour and the PLL driver performed better. Odin was supposed to be a straight scaling up of Mjollnir, I'm still trying to figure out why the behaviour is different.

When designing a DRSSTC I would try to get Qsec and Qpri roughly equal. I'm not sure how close it is possible to get, since streamer load varies quite widely over the operating range of the coil.

Re: Increasing QCW streamer length
Uspring, Tue Jul 29 2014, 06:24PM

I'm not sure how relevant the upper vs. lower pole distinction is to Odin. The bursts are short enough that the system doesn't really settle to a steady state.

You're probably right.

If secondary res frequency shifts due to arc load matters, as a consequence the driving frequency also matters. I guess the PLL starts out with a calculated f value and then adjusts f to have ZCS. Initially there won't be ZCS if running at the upper pole, since with strict adherence to ZCS you'll always end up at the lower pole. For short bursts probably performance depends on your PLL start f.

Do you see any PLL f shifts during the burst and if yes, do you know, whether they are due to the effort of the PLL trying to get to ZCS from its initial value or to an actual change of the ZCS f? In the latter case the ZCS f change should have roughly the same magnitude as the shift of secondary fres.

Re: Increasing QCW streamer length
Goodchild, Tue Aug 05 2014, 03:23PM

Hello again everyone,

I will be the first to admit when I’m wrong on something; I absolutely love when it happens because that’s when I learn stuff. I have been talking with another 4hv member over email and during our dialog and SPICE trades a fundamental shift in understanding happened for me.

For starters Steve Ward was correct, my original SPICE simulation was indeed flawed in one way and that was the placement of the lumped capacitance (secondary + topload) capacitance. This should indeed be in parallel with the secondary and this makes a major impact in the operation of the system.

So this time I ran a stepped AC analysis on the revised DR circuit and stepped though couplings 0.1, 0.2, 0.4, and 0.5. Now 0.1k indeed has the largest gain for the system at around 53.3dB while 0.5K is 40.2dB. This is a difference of 13.1dB and I would consider that significant.

Now the next thing we have all been talking about is why this is the case? My explanation is the result of damped oscillators. Both the secondary and primary are damped oscillators. The secondary is damped mainly by the spark loading. The primary is also damped partly by spark loading via the mutual coupling between the two resonators (lower coupling means less influence of Rspark on the primary Q). In addition both circuits have a DC and AC resistance that also damps and limits the max power that can be ringed up in either circuit.



If we take a look at this first simulation that has no limiting resistances (primary or secondary) it can be seen that as coupling is lowered voltage gain of the system increases. As Udo pointed out copper losses will ultimately limit the lower limit you can set the coupling, before gain starts to drop off.



In this second simulation, the primary has a modest 20mOhms of resistance added for AC/DC copper losses and the secondary has 80Ohms added. Now at around 0.1k a drop in gain can be seen and going lower that this would degrade system performance.

The caveat is tuning sensitivity as many have pointed out. I personally think that this is why BSVi experienced poor performance when lowering coupling. To use my simulation as an example; in order for 0.1K to have better or comparable performance to 0.5K the system must stay in tune to within ~350Hz this is incredibly tight! This can be a problem if you have a QCW like BSVi’s that shifted 38KHz during the burst.

Lastly I will present a simulation based on the parameters from BSVi’s setup. I first ran BSVi’s coil through java TC to gets some rough approximations of the electrical parameters of the coil. I used the data provided in the first post on this thread as inputs.




Next I ran the stepped AC analysis with the coil tuned at resonance and found that the optimal coupling for max gain while having a frequency shift of 38KHz is about 0.45k. BSVi’s coupling (calculated) is around 0.534K and this still provides good gain, only differing from the optimal by about 0.7dB. Max optimal gain is around 85.6dB. Anything below this optimal coupling would have degraded the Q of the primary and lowered the overall system gain. This analysis is on par with what BSVi observed when lowering coupling.



So my conclusion: the optimal coupling for the system is all relative and there seems to be a “sweet spot” where going higher or lower would result in poorer performance.

EDIT:
As curiosity got the best of me I went on to collect more data on BSVi’s coil from SPICE and evaluated it over the entire coupling range. I collected and graphed both peak gain along with the +-3dB frequency window in which this gain resided.
The results are very interesting.

Re: Increasing QCW streamer length
Uspring, Wed Aug 06 2014, 11:25AM

For starters Steve Ward was correct, my original SPICE simulation was indeed flawed in one way and that was the placement of the lumped capacitance (secondary + topload) capacitance. This should indeed be in parallel with the secondary and this makes a major impact in the operation of the system.

I also agree with Wards point wrt to how where arc load happens. But from the perspective of a simulation the important parameter is the secondary Q. It doesn't make much of a difference whether it is introduced by a series or a parallel resistance. The values of the resistances are quite different for a given Q, though, since Q=R/sqrt(L/C) for a parallel resistor and Q=sqrt(L/C)/R for a series one.

If you want ZCS, there is not much of a choice, which frequency you run at. Even if you have a large frequency window for a big coupling, it does not necessarily imply, that the coil will run within it. If, e.g. the primary and secondary are tuned to the same frequency fres, the coupling will cause 2 frequencies (actually 3, but I will disregard that here) at which there is ZCS. One is below fres and one above. The larger the coupling is, the further they will move away from fres.

A large coupling will increase performance by making the secondary see more of the primary field, but it will also decrease performance by moving the ZCS frequency away from the optimal point, i.e. the secondary fres. These effects almost cancel each other.

So what is the advantage of a large coupling? Since the operating frequency and the secondary fres are already quite far away from each other, the change of secondary fres due to arc load does not make much difference anymore (relatively). The coil is less sensitive to arc detuning.

Re: Increasing QCW streamer length
BSVi, Wed Aug 20 2014, 06:02AM

SimpleDriver is now up and running. I'v also added QCW mode to my bluetooth interrupter to control it with smartphone. Coil itself is not tuned well yet, but it works. I'm working on documentation now. Here is a short overview video (yes, worst-english-you'v-ever-heared warning):

Re: Increasing QCW streamer length
Steve Conner, Wed Aug 20 2014, 08:46AM

Does the touchscreen on your phone work near a Tesla coil? :O

Uspring: When damping due to streamer loading is taken into account, you do not have 3 discrete frequencies any more. It starts to look more like a filter with a passband.

Re: Increasing QCW streamer length
Kizmo, Wed Aug 20 2014, 09:32AM

At least touch screen on my Xcover2 dies completely if its near drsstc. Phone itself works but touch screen does nothing.

Re: Increasing QCW streamer length
Uspring, Wed Aug 20 2014, 10:05AM

Uspring: When damping due to streamer loading is taken into account, you do not have 3 discrete frequencies any more. It starts to look more like a filter with a passband.

Yes, there's only one ZCS left, if Qsec drops below 1/k, depends somewhat on tuning.
Another advantage of high coupling is, that you can reduce the number of primary turns which lowers primary losses and leads to faster primary current rampup.

Re: Increasing QCW streamer length
BSVi, Wed Aug 20 2014, 11:08AM

Does the touchscreen on your phone work near a Tesla coil? :O

It depends on touchscreen technology. Some screens are suspeciblie to noise from tels coils. For those screens I have special mode that disables touchscreen while coil is running. When this mode is activated, hardware buttons should be used to stop coil.

Older single-touch and virtual multitouch (as in my LG P500) capacitive screens and, especially, resistive touscreens works just fine with coil nearby.