A dynamical arc model
Uspring, Thu Aug 08 2013, 10:05AMBased on a recent measurement
and some theory I've created an LTSpice model to predict arc current from arc voltage. Here are some of the results: Below are diagrams of a measurement of the difference (blue) of the secondary top current and the arc current. The current in blue shows therefore only that part going into the toroid capacitance. It is thus proportional to the top voltage. The measured arc current is shown in red. The measured voltage served as an input to the simulations.
High power measurement, a vertical unit is 1A for the toroid current (blue) and 0.5 A for the arc (red). Horizontal unit is 0.33us.
High power simulation
Medium power measurement, a vertical unit is 1A for the toroid current (blue) and 0.5 A for the arc (red).
Medium power simulation
Low power measurement, a vertical unit is 1A for the toroid current (blue) and 0.25 A for the arc (red).
Low power simulation
Waveform amplitudes and shapes agree quite well, although there is a tendency to underestimate the current at high power. Possibly this is due to the way I made my measurements: At high power about half of the current going from the secondary into the toroid is due to the arc. The other half comes from the capacitance of the toroid itself. I single-triggered the DSO on a secondary current threshold, so a single very conductive and branching arc will make the DSO trigger first, as I ramp up the bus voltage. At lower power, i.e. trigger thresholds set lower, the arc current is smaller compared to the current due to the toroids capacitance and the effect of particularly conductive arcs on the triggering is less.
The circuit for simulation is this:
On the left is a DRSSTC circuit. The right is the arc model. The basic idea is, that an arc can be modelled by a resistive wire. The capacitance is kept constant per unit length and the resistance depends on the power being dissipated in the wire. The power will heat up the wire (aka arc channel) and also create free electrons and ions, which makes it more conductive.
When you run the model, you can watch the arc grow by looking at the voltages down the arc chain. Every resistor is about 8cm of the arc. Actually you will observe small voltages arbitrarily far down the chain. Steve Ward has made the observation, that arcs require at least 40kV to grow, so that the arc probably stops growing, when the voltage at its end drops below that value.
You can also simulate ground arcs by shorting the chain to ground after the voltage level has reached a certain point along the arc.
The heart of the model is the sub circuit shown here:
The arc resistor is the current source G1. The current is given by
I = (V(N1)-V(N2)) * (4e-7 + 6e-11*V(N4)*V(N5))
The first factor, V(N1)-V(N2) is the voltage across it, so the current source acts as a resistor with a resistance of
R = 1/(4e-7 + 6e-11*V(N4)*V(N5))
The voltage source B2 calculates the power, which is averaged by 2 RC circuits. R1 and C1 have a long time constant of 300us. R2 and C2 have a short time constant of 20us. They were inspired by the fact, that the arc conductivity seems to have a quickly responding part as seen in the wiggles of current in the high power measurement and a slowly responding part at the start of the arc. Probably the long time constant accounts for the arc channel heat up and the short one reflects ion lifetimes.
V(N4) and V(N5) are the time averaged powers.
The constant 4e-7 allows the arc to start growing. Initially the power averages are 0 and with an infinite initial resistance, there would be no power dissipated and the arc model would not start. The bigger this value is, the earlier the arc will start growing. The constants 4e-7 and 6e-11 were chosen to get a close as possible match to the measurement. The choice of the arc capacitances depends on the calculated length of the arc. Each RC element in the arc stands for a certain piece of arc length. A bigger C will make the arc shorter. So basically the choice of C will change the scale of the simulation, i.e. the amount of arc length one RC element stands for.
Below is the result of a complete simulation, i.e. arc model and DRSSTC circuit. Note, that the diagrams above have not been made with a DRSSTC circuit, but with a voltage input to the arc derived from the measurements.
The initial rise and the later dip in voltages and currents are similar to measurement. The later oscillations in the amplitude are missing, though, and also the max current is too low, 1.8A instead of the measured 2.4A. I believe the discrepancies are partially due to the incorrectly simulated phase shifts between arc voltage and current. In reality, the phase changes between 60 and 45 degrees, while in the simulation, they stay in the vicinity of 60 degrees. In the simulation, the phase shift is a bit higher during peak currents, just as in the measurement but the predicted variations are much smaller. Also the arc seems to acquire strength too early. This is probably caused by too much conductivity at low power
The model here fails drastically when applied to the data of a measurement made on the Electrum by Greg Leyh.
The current in the simulation is much too high and the simulated arc length too long.Generally resistive wire type arc models show a voltage to current ratio coming from the total arc capacitance, i.e. length multiplied by capacitance per unit length. This is only weakly dependent on how arc resistance varies with power dissipation. I believe, that the capacitance per unit length of the Electrum arcs must be considerably lower than those of my measurements. I've tried using smaller capacitances in my model, but the arcs became even longer. That can be traced to too much arc conductivity in the model. The model is based on the conductivity being proportional to the square of the power dissipated. Probably the conductivity rises much slower with power for large arc currents. Measurements at higher powers may show a way in which to extend the model.
I've included the schematics here in case you want to experiment with this.
]arc3circuit.zip[/file]
For higher power levels additional arc RC elements should be added in order to accomodate longer arcs. Also note, that the measurements, the model is based on, used a breakout point. Breakout depends heavily on this.
Re: A dynamical arc model
Steve Conner, Thu Aug 08 2013, 10:22AM
Very interesting! Thanks for sharing
I'll try running it against the simulations I've already done for my upcoming DRSSTC.
I thought the ion lifetimes would be tens of milliseconds, though. By changing the break rate of a Tesla coil, you can see that above about 100bps, the streamer is mostly reusing the channel from the previous bang, instead of starting with a fresh one. This shows that it must still be at least somewhat conductive after tens of milliseconds.
I would be interested to see how your modelled streamer length varies with break rate for constant bang energy. The experiments I've done show a "knee" around 100Hz and traditionally the most efficient coils in terms of spark length per watt input seemed to run at 100 or 120Hz. Higher break rates than this tend to make the arcs brighter rather than longer. You gain a lot more spark length going from 25Hz to 100, than you do going from 100 to 400.
Steve Conner, Thu Aug 08 2013, 10:22AM
Very interesting! Thanks for sharing
I'll try running it against the simulations I've already done for my upcoming DRSSTC.I thought the ion lifetimes would be tens of milliseconds, though. By changing the break rate of a Tesla coil, you can see that above about 100bps, the streamer is mostly reusing the channel from the previous bang, instead of starting with a fresh one. This shows that it must still be at least somewhat conductive after tens of milliseconds.
I would be interested to see how your modelled streamer length varies with break rate for constant bang energy. The experiments I've done show a "knee" around 100Hz
and traditionally the most efficient coils in terms of spark length per watt input seemed to run at 100 or 120Hz. Higher break rates than this tend to make the arcs brighter rather than longer. You gain a lot more spark length going from 25Hz to 100, than you do going from 100 to 400.Re: A dynamical arc model
Dr. Dark Current, Thu Aug 08 2013, 10:40AM
Great post, thanks. I wonder if the arc model would work also for "QCW"-type sparks.
Dr. Dark Current, Thu Aug 08 2013, 10:40AM
Great post, thanks. I wonder if the arc model would work also for "QCW"-type sparks.
Re: A dynamical arc model
Uspring, Thu Aug 08 2013, 11:44AM
Thank you for the encouragement.
@Steve:
The BPS dependency you observed is probably coming from the heat remaining in the arc channel between bursts. Heat thins air and this causes lower breakdown voltages. I tried to make this part of the model with the longer (300us) time constant. Actually 300us is too short to account for the knee at 100 BPS. I'm not too sure about that 300us value. I've used 600us bursts in my measurement and I would need much longer bursts to pin that down.
Thinking about this, I'm noticing an error I've made. My simulations were single shots, while the mesurement was made at 100 BPS. Probably the simulation will change somewhat, if I simulate 100 BPS, but as having said, the 300us time constant will create a knee likely only at much higher break rates.
@Dr. Dark Current:
Uspring, Thu Aug 08 2013, 11:44AM
Thank you for the encouragement.
@Steve:
I'll try running it against the simulations I've already done for my upcoming DRSSTC.I'm very interested in how your simulations compare to mine. Please follow up on that.
I thought the ion lifetimes would be tens of milliseconds, though. By changing the break rate of a Tesla coil, you can see that above about 100bps, the streamer is mostly reusing the channel from the previous bang, instead of starting with a fresh one. This shows that it must still be at least somewhat conductive after tens of milliseconds.The wiggles in current amplitude in the middle and at the end of the burst follow the voltage changes rather rapidly. In my measurement post, I had also plotted the conductivity of the arc and it seems to change closely following the voltage wiggles. Whether this is can be due to ion lifetimes I don't know, but it was the only explanation I could think of. I should check the literature on that.
The BPS dependency you observed is probably coming from the heat remaining in the arc channel between bursts. Heat thins air and this causes lower breakdown voltages. I tried to make this part of the model with the longer (300us) time constant. Actually 300us is too short to account for the knee at 100 BPS. I'm not too sure about that 300us value. I've used 600us bursts in my measurement and I would need much longer bursts to pin that down.
Thinking about this, I'm noticing an error I've made. My simulations were single shots, while the mesurement was made at 100 BPS. Probably the simulation will change somewhat, if I simulate 100 BPS, but as having said, the 300us time constant will create a knee likely only at much higher break rates.
@Dr. Dark Current:
I wonder if the arc model would work also for "QCW"-type sparks.In principle it should work. Most QCWs run at higher frequencies, though, and I'm in doubt how well the capacitance values I've chosen fare there. The Electrum e.g., running at 35kHz, seems to be different. My capacitance values seem to be due to heavily branching arcs, since they are much higher than that of a straight wire. The branches are just lumped into one arc in the model. QCW's can make nice straight arcs.
Re: A dynamical arc model
Goodchild, Thu Aug 08 2013, 03:23PM
This is very cool, I have been working to get a decent spark model for some time now, you seem to have done this in spades!
Just a thought, I noticed that with most QCWs that the voltage and primary current grow up to a large amount just before they break out due to the toroid at which point they fall drastically and then start growing again, but at a slower rate. The toroid seems to stand off the voltage until it reaches a point in which it's high enough to ionize the air. In my simulation I modeled this as a TVS type setup with a breakdown voltage similar to what would be seen as the breakdown voltage at the toroid. In a SGTC or DRSSTC this effect is not really noticeable due to the fact that the first cycle has enough energy to break out from the toroid immediately, where a QCW needs 10’s to 100’s of cycles to get to the same level.
I'm also interested in what Steve is talking about. I have noticed this in my DRSSTCs for years. Low break rates tend not to generate as large of sparks. But there is a sweet spot, Steve's is at 100Hz, but I have noticed it from 80Hz up to 250Hz depending on the size of the coil. My larger coils tend to have a lower sweet spot than my smaller higher frequency coils. Another example is DR. Spark’s FATBOY, that coil performs better at lower break rates than any coil I have ever seen...
4hv really needs more post like this, thanks for posting!
EDIT:
One last thought I wanted to add in regards to your phase shift. Is it possible that maybe the arc in real life has some amount of inductance that is contributing to create a larger phase shift? This may explain why you see a shift in real lift but not in simulation. Just a thought.
Goodchild, Thu Aug 08 2013, 03:23PM
This is very cool, I have been working to get a decent spark model for some time now, you seem to have done this in spades!
Just a thought, I noticed that with most QCWs that the voltage and primary current grow up to a large amount just before they break out due to the toroid at which point they fall drastically and then start growing again, but at a slower rate. The toroid seems to stand off the voltage until it reaches a point in which it's high enough to ionize the air. In my simulation I modeled this as a TVS type setup with a breakdown voltage similar to what would be seen as the breakdown voltage at the toroid. In a SGTC or DRSSTC this effect is not really noticeable due to the fact that the first cycle has enough energy to break out from the toroid immediately, where a QCW needs 10’s to 100’s of cycles to get to the same level.
I'm also interested in what Steve is talking about. I have noticed this in my DRSSTCs for years. Low break rates tend not to generate as large of sparks. But there is a sweet spot, Steve's is at 100Hz, but I have noticed it from 80Hz up to 250Hz depending on the size of the coil. My larger coils tend to have a lower sweet spot than my smaller higher frequency coils. Another example is DR. Spark’s FATBOY, that coil performs better at lower break rates than any coil I have ever seen...
4hv really needs more post like this, thanks for posting!
EDIT:
One last thought I wanted to add in regards to your phase shift. Is it possible that maybe the arc in real life has some amount of inductance that is contributing to create a larger phase shift? This may explain why you see a shift in real lift but not in simulation. Just a thought.
Re: A dynamical arc model
Uspring, Thu Aug 08 2013, 06:34PM
Hi Eric,
thank you for your kind words.
You write:
The model probably won't describe BPS dependence the way it is currently parametrised. A 300us decay time amounts to about 30 decay times between the bursts at 100 bps. Since I've used an exponential decay of energy (RC circuit), there will be not much energy left at the beginning of the next burst. I believe, that the assumption of an exponential decay of heat is inaccurate. Temperature probably drops slower at the end of the cooling period.
Inductance probably has some effect on phase. I've omiited it, since arc resistances are on the order of a few 100 Ohms/cm. To make a significant effect on phase shift. the inductively caused impedance must be of similar magnitude. But that would imply several hundred uH/cm, which doesn't seem plausible.
Uspring, Thu Aug 08 2013, 06:34PM
Hi Eric,
thank you for your kind words.
You write:
In a SGTC or DRSSTC this effect is not really noticeable due to the fact that the first cycle has enough energy to break out from the toroid immediately, where a QCW needs 10’s to 100’s of cycles to get to the same level.With SGTCs and with your monster DRSSTCs you are certainly right. If you look at the measurement with my little coil, arc current begin is delayed several cycles after begin of secondary rampup. This is probably due to my rather low bus voltages and the high impedance of my primary tank.
The model probably won't describe BPS dependence the way it is currently parametrised. A 300us decay time amounts to about 30 decay times between the bursts at 100 bps. Since I've used an exponential decay of energy (RC circuit), there will be not much energy left at the beginning of the next burst. I believe, that the assumption of an exponential decay of heat is inaccurate. Temperature probably drops slower at the end of the cooling period.
Inductance probably has some effect on phase. I've omiited it, since arc resistances are on the order of a few 100 Ohms/cm. To make a significant effect on phase shift. the inductively caused impedance must be of similar magnitude. But that would imply several hundred uH/cm, which doesn't seem plausible.
Re: A dynamical arc model
HV Enthusiast, Thu Aug 08 2013, 10:28PM
I've noticed this as well. Around 100Hz is a sweet spot indeed. Interestingly enough, you would think that you would simply get longer arcs at the lower PRFs because the capacitor droop isn't as much as the bus voltage (average) tends to be a somewhat higher than at higher PRFs (dependent on how much recharge current you are getting from the line and size of your bus capacitors. But seems the opposite is true as you mentioned.
HV Enthusiast, Thu Aug 08 2013, 10:28PM
Goodchild wrote ...
I'm also interested in what Steve is talking about. I have noticed this in my DRSSTCs for years. Low break rates tend not to generate as large of sparks. But there is a sweet spot, Steve's is at 100Hz, but I have noticed it from 80Hz up to 250Hz depending on the size of the coil. My larger coils tend to have a lower sweet spot than my smaller higher frequency coils. Another example is DR. Spark’s FATBOY, that coil performs better at lower break rates than any coil I have ever seen...
I'm also interested in what Steve is talking about. I have noticed this in my DRSSTCs for years. Low break rates tend not to generate as large of sparks. But there is a sweet spot, Steve's is at 100Hz, but I have noticed it from 80Hz up to 250Hz depending on the size of the coil. My larger coils tend to have a lower sweet spot than my smaller higher frequency coils. Another example is DR. Spark’s FATBOY, that coil performs better at lower break rates than any coil I have ever seen...
I've noticed this as well. Around 100Hz is a sweet spot indeed. Interestingly enough, you would think that you would simply get longer arcs at the lower PRFs because the capacitor droop isn't as much as the bus voltage (average) tends to be a somewhat higher than at higher PRFs (dependent on how much recharge current you are getting from the line and size of your bus capacitors. But seems the opposite is true as you mentioned.
Re: A dynamical arc model
Goodchild, Thu Aug 08 2013, 10:46PM
In regards to what you are talking about with the heat keeping the channel ionized. Maybe it still is exponential but the arc is just a lot hotter than we may think it is. If it is indeed very very hot say 5000c or more maybe this time constant can be extended leading to the factor we see in tesla coils. It is possible to say it could get this hot, in a lot of cases we are putting a lot of energy into that sparks( 10's of Kw in a lot of cases) and only a fraction goes to producing light or EM field, so the rest has to be in heat.
Take what I'm saying with a grain of slat however, I am only speculating off the top of my head.
Regardless I think you have a great model here. I may just use it in the near future for Calculating a secondary MMC in out new supper large QCW. It's difficult to calculated frequency drift in the secondary circuit without an accurate model for sparks. Mainly what I was looking for was a capacitance per unit length of sparks or an equation perhaps modeling this capacitance. I'm wondering to how you came up with the value for your simulation? I'm assuming it was a measure and match sort of thing?
For QCW spark capacitance, my plan was to get an accurate read of secondary L and C with an LCR bridge. After this run the coil for some burst length and record the change in frequency at several sampled points along a single burst. Assuming L in the sparks is negligible, a value for added C per unit length may be found using the original L,C and the shift in frequency. It bet it turns out this value is none linear in nature.
Fruit for thought!
Goodchild, Thu Aug 08 2013, 10:46PM
In regards to what you are talking about with the heat keeping the channel ionized. Maybe it still is exponential but the arc is just a lot hotter than we may think it is. If it is indeed very very hot say 5000c or more maybe this time constant can be extended leading to the factor we see in tesla coils. It is possible to say it could get this hot, in a lot of cases we are putting a lot of energy into that sparks( 10's of Kw in a lot of cases) and only a fraction goes to producing light or EM field, so the rest has to be in heat.
Take what I'm saying with a grain of slat however, I am only speculating off the top of my head.
Regardless I think you have a great model here. I may just use it in the near future for Calculating a secondary MMC in out new supper large QCW. It's difficult to calculated frequency drift in the secondary circuit without an accurate model for sparks. Mainly what I was looking for was a capacitance per unit length of sparks or an equation perhaps modeling this capacitance. I'm wondering to how you came up with the value for your simulation? I'm assuming it was a measure and match sort of thing?
For QCW spark capacitance, my plan was to get an accurate read of secondary L and C with an LCR bridge. After this run the coil for some burst length and record the change in frequency at several sampled points along a single burst. Assuming L in the sparks is negligible, a value for added C per unit length may be found using the original L,C and the shift in frequency. It bet it turns out this value is none linear in nature.
Fruit for thought!
Re: A dynamical arc model
Uspring, Fri Aug 09 2013, 09:32AM
From a theoretical point of view, there shouldn't be a sweet spot for some bps.Arc length should increase for higher values, although slowly beyond 100 bps if burst length and bus voltage stay the same. I think, though, that these conditions usually are not met. Arc length won't profit much from higher bps beyond 100Hz, while power consumption really goes up.
The effect of remaining heat between bursts is quite strong. If the arc channel cools say to 600K between bursts, the air will be half as thick as at 300K, reducing breakdown voltage by a factor of 2 by Paschens law. The arc temperature still has to be brought up again to about 6000K (plasma temperature) but the initial heat will give it an advantage. If you run your coil on top of a big mountain, you will probably also get longer arcs, provided you find a wall plug.
Eric wrote:
If in the simulation your arc gets too long, you can add some RC elements. On the other hand the length of arc per element should not get too big, i.e. less than 10cm. If that happens, you should decrease the capacitance in the element and start over adjusting the parameters.
Anyway, I'm hungry for data to improve the model on. If you have some, I will gladly try to adjust it and share the results.
Uspring, Fri Aug 09 2013, 09:32AM
From a theoretical point of view, there shouldn't be a sweet spot for some bps.Arc length should increase for higher values, although slowly beyond 100 bps if burst length and bus voltage stay the same. I think, though, that these conditions usually are not met. Arc length won't profit much from higher bps beyond 100Hz, while power consumption really goes up.
The effect of remaining heat between bursts is quite strong. If the arc channel cools say to 600K between bursts, the air will be half as thick as at 300K, reducing breakdown voltage by a factor of 2 by Paschens law. The arc temperature still has to be brought up again to about 6000K (plasma temperature) but the initial heat will give it an advantage. If you run your coil on top of a big mountain, you will probably also get longer arcs, provided you find a wall plug.
Eric wrote:
If it is indeed very very hot say 5000c or more maybe this time constant can be extended leading to the factor we see in tesla coils.That's a good idea. I have to be careful, though, since a longer time constant will also affect arc growth at shorter time scales. Likely I will need another time constant somewhere to account for the bps effect..
It is possible to say it could get this hot, in a lot of cases we are putting a lot of energy into that sparks( 10's of Kw in a lot of cases) and only a fraction goes to producing light or EM field, so the rest has to be in heat.Yes it is definitely heat. In terms of a light source, a TC is a poor performer.
Mainly what I was looking for was a capacitance per unit length of sparks or an equation perhaps modeling this capacitance. I'm wondering to how you came up with the value for your simulation? I'm assuming it was a measure and match sort of thing?I started off with some arbitrary arc capacitance. Then I adjusted the parameters (i.e. 4e-7 and 6e-11) so they would fit the measured currents. Then I followed the voltage amplitude down the chain to get the arc length. Comparing that to the measured arc length gives the value how much arc length a single RC element stands for.
If in the simulation your arc gets too long, you can add some RC elements. On the other hand the length of arc per element should not get too big, i.e. less than 10cm. If that happens, you should decrease the capacitance in the element and start over adjusting the parameters.
Anyway, I'm hungry for data to improve the model on. If you have some, I will gladly try to adjust it and share the results.
Re: A dynamical arc model
Goodchild, Fri Aug 09 2013, 03:41PM
When I finally get to testing this 1/8th scale QCW in about a week or two I will forward on any data I collect. I plan to use the process I outlined above and we will see how that turns out.
Goodchild, Fri Aug 09 2013, 03:41PM
When I finally get to testing this 1/8th scale QCW in about a week or two I will forward on any data I collect. I plan to use the process I outlined above and we will see how that turns out.
Re: A dynamical arc model
Steve Ward, Sat Aug 10 2013, 05:38AM
For what its worth, i push the knee well below 100hz on my coils now by operating with pulse durations of 500-1000uS. I can achieve full length arcs (10 feet) at 50hz in this manner. its pretty awesome. I need about 500uS to get long arcs at these very low rep rates, it seems to be a matter of pumping enough heat into the arc. I wouldnt recommend this type of operation unless you know your system can handle it, the longer pulses can heat up the IGBT chips pretty significantly in just 1mS.
Anyway, nice spice model! I wish i was that clever. The 20uS figure is familiar, ive read about similar time constants for ion recovery time in plasmas before.
I think your model might do better if you put branches in it to simulate "bushy" versus uhhh... "non bushy" sparks. BIG tesla coils will often make huge arcs with not a whole lot of branches.
Do you suppose any of the high-rate wiggling of power levels in the high-power measurement is due to the low coupling of tesla coils in general? Id expect there to be some back and forth exchange over 1/k cycles or so, which may modulate the streamers power.
Steve Ward, Sat Aug 10 2013, 05:38AM
For what its worth, i push the knee well below 100hz on my coils now by operating with pulse durations of 500-1000uS. I can achieve full length arcs (10 feet) at 50hz in this manner. its pretty awesome
. I need about 500uS to get long arcs at these very low rep rates, it seems to be a matter of pumping enough heat into the arc. I wouldnt recommend this type of operation unless you know your system can handle it, the longer pulses can heat up the IGBT chips pretty significantly in just 1mS.Anyway, nice spice model! I wish i was that clever
. The 20uS figure is familiar, ive read about similar time constants for ion recovery time in plasmas before.I think your model might do better if you put branches in it to simulate "bushy" versus uhhh... "non bushy" sparks. BIG tesla coils will often make huge arcs with not a whole lot of branches.
Do you suppose any of the high-rate wiggling of power levels in the high-power measurement is due to the low coupling of tesla coils in general? Id expect there to be some back and forth exchange over 1/k cycles or so, which may modulate the streamers power.
Re: A dynamical arc model
Goodchild, Sat Aug 10 2013, 07:09AM
Finally took the high impedance DR path 'wink' 'wink'. My big DR run in this manner. In MIDI operation we run well into 500uS down near 30Hz to 50Hz on our big CM600 coils.
What coil did you get up to 1mS? That's crazy indeed!
Goodchild, Sat Aug 10 2013, 07:09AM
Steve Ward wrote ...
For what its worth, i push the knee well below 100hz on my coils now by operating with pulse durations of 500-1000uS. I can achieve full length arcs (10 feet) at 50hz in this manner. its pretty awesome. I need about 500uS to get long arcs at these very low rep rates, it seems to be a matter of pumping enough heat into the arc. I wouldnt recommend this type of operation unless you know your system can handle it, the longer pulses can heat up the IGBT chips pretty significantly in just 1mS.
For what its worth, i push the knee well below 100hz on my coils now by operating with pulse durations of 500-1000uS. I can achieve full length arcs (10 feet) at 50hz in this manner. its pretty awesome
. I need about 500uS to get long arcs at these very low rep rates, it seems to be a matter of pumping enough heat into the arc. I wouldnt recommend this type of operation unless you know your system can handle it, the longer pulses can heat up the IGBT chips pretty significantly in just 1mS.Finally took the high impedance DR path 'wink' 'wink'. My big DR run in this manner. In MIDI operation we run well into 500uS down near 30Hz to 50Hz on our big CM600 coils.
What coil did you get up to 1mS? That's crazy indeed!
Re: A dynamical arc model
Steve Ward, Sat Aug 10 2013, 08:19AM
Im using my new digital drivers so i can do tuning tricks* and primary current regulation. The coils are 30" tall secondaries, running ~50khz. They have made 11.5 foot arcs to the ceiling so far, with 350V-560V bus and 600A peak in the primary with a 0.3uF primary cap. They have a generous 78pF of secondary capacitance (they use a secondary-side MMC at 300kV). These coils managed to essentially "drill a hole" in a concrete floor when it found some rebar just 2" below the surface. There was a 1/4" diameter hole after the ground spewed orange sparks. These coils make incredibly bright arcs to ground. A pair of them will throw 10 foot arcs (each) with a 10kW power supply thats shared. Might be some of the best coils ive ever worked with, mainly because they maintain better tuning with big spark loads. The low primary amps is mainly due to this good tuning effect, the streamer resistance is reflected to the primary side as a significant resistance which keeps the primary amps reasonable with very long pulses. And if that doesnt work out, the driver knows how to step on and off the gas to keep it going at whatever primary amps we want.
I think my latest driver is the most like Steve Conners PLL drivers, except it doesnt use a PLL and uses a PSoC instead. But, in terms of how it drives a coil, they are very similar. So i recon his All-Fragger ought to be able to kick out some fat bassy sparks just fine
.
* the tuning trick is something you can only do if your driver has a start-up oscillator or some other frequency forcing mechanism. You tune the coil however necessary to achieve good tuning with a streamer present. The key is how the driver starts excitation of the coil. The self-oscillating drivers (like the UD2) will ring up both resonant frequencies, which might not be the best thing to do. Rather, its better to drive it at the desired pole frequency with the oscillator for several cycles to establish operation at that frequency only.
For QCW coils: If your coil has a high enough coupling coefficient, you can tune the primary lower than Secondary Fres but force the thing to oscillate at the higher pole frequency, which will has some benefits for growing arcs. It provides a positive feedback mechanism between the TC loading and tuning that causes the coil to draw more power even in the CW state as the arc capacitance looks larger and larger. In this way, it aids the driver at ramping up the coil power while driving the arc. You can observe a similar behavior in Udo's measurements as his coil comes in tune with the streamer capacitance, however he continues driving even though the tuning seems to have degraded, as evident by reduced output current. In the QCW coil, the output needs to constantly rise, so going out of tune makes this become excessively difficult, hence the "tuning trick". My original QCW design ran up to 270A primary current to create a 6 foot spark (it was perating out of tune by this point). After implementing my tuning trick and lowering the secondary impedance to maintain better tuning, i make bigger sparks (almost 7 feet now) with 150A (same driver voltage).
Steve Ward, Sat Aug 10 2013, 08:19AM
Im using my new digital drivers so i can do tuning tricks* and primary current regulation. The coils are 30" tall secondaries, running ~50khz. They have made 11.5 foot arcs to the ceiling so far, with 350V-560V bus and 600A peak in the primary with a 0.3uF primary cap. They have a generous 78pF of secondary capacitance (they use a secondary-side MMC at 300kV). These coils managed to essentially "drill a hole" in a concrete floor when it found some rebar just 2" below the surface. There was a 1/4" diameter hole after the ground spewed orange sparks. These coils make incredibly bright arcs to ground. A pair of them will throw 10 foot arcs (each) with a 10kW power supply thats shared. Might be some of the best coils ive ever worked with, mainly because they maintain better tuning with big spark loads. The low primary amps is mainly due to this good tuning effect, the streamer resistance is reflected to the primary side as a significant resistance which keeps the primary amps reasonable with very long pulses. And if that doesnt work out, the driver knows how to step on and off the gas to keep it going at whatever primary amps we want.
I think my latest driver is the most like Steve Conners PLL drivers, except it doesnt use a PLL and uses a PSoC instead. But, in terms of how it drives a coil, they are very similar. So i recon his All-Fragger ought to be able to kick out some fat bassy sparks just fine
. * the tuning trick is something you can only do if your driver has a start-up oscillator or some other frequency forcing mechanism. You tune the coil however necessary to achieve good tuning with a streamer present. The key is how the driver starts excitation of the coil. The self-oscillating drivers (like the UD2) will ring up both resonant frequencies, which might not be the best thing to do. Rather, its better to drive it at the desired pole frequency with the oscillator for several cycles to establish operation at that frequency only.
For QCW coils: If your coil has a high enough coupling coefficient, you can tune the primary lower than Secondary Fres but force the thing to oscillate at the higher pole frequency, which will has some benefits for growing arcs. It provides a positive feedback mechanism between the TC loading and tuning that causes the coil to draw more power even in the CW state as the arc capacitance looks larger and larger. In this way, it aids the driver at ramping up the coil power while driving the arc. You can observe a similar behavior in Udo's measurements as his coil comes in tune with the streamer capacitance, however he continues driving even though the tuning seems to have degraded, as evident by reduced output current. In the QCW coil, the output needs to constantly rise, so going out of tune makes this become excessively difficult, hence the "tuning trick". My original QCW design ran up to 270A primary current to create a 6 foot spark (it was perating out of tune by this point). After implementing my tuning trick and lowering the secondary impedance to maintain better tuning, i make bigger sparks (almost 7 feet now) with 150A (same driver voltage).
Re: A dynamical arc model
Uspring, Sat Aug 10 2013, 04:37PM
Steve Ward wrote:
I do have big difficulties with the Electrum measurement, mainly due to its much lower capacitance per unit length. Your remark about big TCs likely is part of an explanation.
As for tuning, I think it's a great idea to run at the upper pole while tuning the primary low. As Steve said, this keeps primary current peaks low.
Uspring, Sat Aug 10 2013, 04:37PM
Steve Ward wrote:
I think your model might do better if you put branches in it to simulate "bushy" versus uhhh... "non bushy" sparks. BIG tesla coils will often make huge arcs with not a whole lot of branches.I tried that once in order to find out, if the incorrectly predicted phase shifts between arc voltage and current would improve. It didn't help. The capacitance per unit arc length needed for a fit did increase.
I do have big difficulties with the Electrum measurement, mainly due to its much lower capacitance per unit length. Your remark about big TCs likely is part of an explanation.
Do you suppose any of the high-rate wiggling of power levels in the high-power measurement is due to the low coupling of tesla coils in general? Id expect there to be some back and forth exchange over 1/k cycles or so, which may modulate the streamers power.That could well be so. Another idea is, that the voltage drop after the initial surge puts the secondary out of tune again, so that primary current will rise. Since there is some delay in the response of the arc to the then rising voltage, it will "overshoot", which then leads to another surge similar to the initial one. That produces oscillations of a frequency dependent on the rise speed of primary current and the delay between arc voltage and current.
As for tuning, I think it's a great idea to run at the upper pole while tuning the primary low. As Steve said, this keeps primary current peaks low.
Re: A dynamical arc model
Steve Conner, Sat Aug 10 2013, 07:40PM
Steve, so you're saying you finally got round to trying a PLL driver with current control and you like it? Only took you like 7 years
Yes, I am looking forward to doing some serious damage with Odin.
Steve Conner, Sat Aug 10 2013, 07:40PM
Steve, so you're saying you finally got round to trying a PLL driver with current control and you like it? Only took you like 7 years
Yes, I am looking forward to doing some serious damage with Odin.
Re: A dynamical arc model
Dr. Drone, Sun Aug 11 2013, 07:02PM
Dr. Drone, Sun Aug 11 2013, 07:02PM
Re: A dynamical arc model
Uspring, Mon Aug 12 2013, 09:57AM
Eric wrote:
Uspring, Mon Aug 12 2013, 09:57AM
Eric wrote:
When I finally get to testing this 1/8th scale QCW in about a week or two I will forward on any data I collect. I plan to use the process I outlined above and we will see how that turns out.Thank you, I'm looking forward to that.
Print this page