First drsstc project, looking for recommendations

[costas_p]

Hello HVF community,

I am building my first DRSSTC, and I am looking for some recommendations.

Coil specs:

Secondary 75mm diameter, 30 cm height

Toroidal top load D_major 250mm / D_minor 60mm

Primary 8 turns of 1/4 inch copper tubing (~6.35mm)

loneoceans inspired H bridge of FGH60N60SMD + x2 820uF bus caps

THT version ud2.7c from profdc9

Greece power line 240 V ac

Secondary is raised around 15mm from primary for a ~0.12 coupling

Also I made a relay + 150 ohms resistor pcb with around 1sec delay to precharge the caps when it detect mains power applied to the bridge rectifier.
(tested and charged the 2 820uF parallel caps at 325 Vdc at 600mA)


JavaTC reports that my secondary+topload resonant frequency to be around 200kHz.

Adding around 13% detuning , i should make a MMC for the primary to run around 173kHz.
JavaTC says that for 173kHz and 8 turns of primary my MMC should be around 45-55 nF.

I don't have access to an oscilloscope yet in order to measure the real numbers yet (I plan to use 80-100 cm wire attached to the topload to find the "running" resonant frequency of the coil , in order to determine the actual primary frequency)

Playing with Mad's MMC calculator x10 22nF WIMA MKP1T022205B00MSSD configured 2s 5p give around 55nF

I am planning to run the coil around 280-300Apk.

If my calculation are correct each cap has Ipeak= C * V/uS = 0.022uF * 3kV/uS = 66A, so the MMC has 330A
Please correct me if I am wrong.


Does the numbers seems right so far ??

[davekni]

Does the numbers seems right so far ??

Looks like a great build.  Much cleaner mechanically than anything I construct.  I presume all the black structure above the vertical 80/20 pieces is plastic (something non-conductive).

I'd consider going a bit higher coupling factor, closer to 0.14 instead of 0.119, by dropping secondary mounting slightly.  Performance may be a bit better there.  Risk is racing sparks.  Depends on your tolerance for risk.

Have fun finishing and testing your coil!

[costas_p]

Hello davekni, thanks for your answer.

All the primary coil supports/bolts are nylon/plexiglass, and all hardware used to connect the aluminum extrusion is made from non-anodized aluminum or brass.

I will make testing about the coupling.

Although I spend a significant amount of time reading about MMC design guides etc, I still cannot fully understand 1 thing.

Example: a single 100nF/1kV cap in 2x2 config is a 100nF/2kV  MMC, twice the current ability (parallel) and twice the voltage rating (series) - at least in paper.

How a voltage rating affects the MMC regarding H-bridge transitions - what is the limiting factor to not exceed voltage rating, the on time ?

[klugesmith]

>>I am planning to run the coil around 280-300Apk.
>>If my calculation are correct each cap has Ipeak= C * V/uS = 0.022uF * 3kV/uS = 66A, so the MMC has 330A.

If the MMC is in series with primary coil, then MMC and coil have the same current at every instant.
In 2x2 configuration, each individual capacitor has half of the primary coil current.

The capacitor current and voltage build up in each cycle of the H-bridge.   To levels that depend on how many RF cycles the H-bridge has been switching.   For capacitor datasheet limits that depend on heating, we might figure the RMS value including time off between RF bursts.  Maybe one of the coilers here can speak about that.

[davekni]

All the primary coil supports/bolts are nylon/plexiglass, and all hardware used to connect the aluminum extrusion is made from non-anodized aluminum or brass.

I was mostly concerned about the horizontal structure joining the tops of the aluminum extrusion posts.  Want to avoid the issues Fredrick had with his large coil and arcs due to voltage around his top anodized extrusion structure, somewhere in this long thread:
https://highvoltageforum.net/index.php?topic=1430.msg10798#msg10798
A fully-conductive ring at the top of your vertical posts wouldn't make arcs, but is still too close to primary, shorting out much of its magnetic field, lowering inductance and coupling and Q.

Although I spend a significant amount of time reading about MMC design guides etc, I still cannot fully understand 1 thing.

Example: a single 100nF/1kV cap in 2x2 config is a 100nF/2kV  MMC, twice the current ability (parallel) and twice the voltage rating (series) - at least in paper.

Works in practice as well as on paper.  That presumes bleeder resistors to keep voltage sharing equal and sufficient cooling (since arrays reduce surface-area/volume ratio).

How a voltage rating affects the MMC regarding H-bridge transitions - what is the limiting factor to not exceed voltage rating, the on time ?

Driver OCD shut-down should limit current to a value that keeps peak voltage under MMC rating.  (Though given the relatively short time TCs are ran compared to commercial electronics, peak voltage ratings can often be exceeded by 20-50%.)  If coupling factor is high enough and arc loading sufficient, secondary arcs will drain power as fast as the H-bridge is adding power without ever hitting OCD limit.
As Kludgesmith pointed out, the other limit is RMS current, to avoid overheating capacitors.  RMS current is within-burst RMS current times sqrt(duty cycle).  Again, it is possible to exceed limits to some degree, especially with a fan blowing across/through MMC.

[costas_p]

The contraction of my DRSSTC is almost complete.

Thanks to davekni for pointing out that the top horizontal bars should not make a complete circle from any "metal" material - i used a cnc and a scrap plexiglass piece 20mm height

Have not done the MMC yet, but I am aiming for 3s - 7p of wima FPK1 22nf 2kvdc, for a total ~52nf 6Kvdc MMC (which I suppose gonna be ok for my design - considering Kaizer MMC calculator)

So question:
I am using ud2.7c THT from profdc9

- in SMD version from loneoceans, OCD can be adjusted from pot R20 and TP1 - the THT version has RV1 and J9-pin2 - what is the purpose of J9 pin1 and pin3 ??

-also regarding UVLO - SMD version has R29 and TP2 - the THT version has RV2 and J16 (+/gnd) I guess thats my test point for uvlo, correct?

schematics in attachments

[davekni]

Thank you for including schematics.  That helps with specific questions.

- in SMD version from loneoceans, OCD can be adjusted from pot R20 and TP1 - the THT version has RV1 and J9-pin2 - what is the purpose of J9 pin1 and pin3 ??

-also regarding UVLO - SMD version has R29 and TP2 - the THT version has RV2 and J16 (+/gnd) I guess thats my test point for uvlo, correct?

I believe those test points are all for measurement, not jumpers.  J9 allows measuring both OCD setpoint (pin 2) and actual primary current (pin 3).  J16 allows measuring ULVO threshold voltage.

[costas_p]

Hello again,

I just measured my actual secondary + topload resonant frequency through a borrowed scope, to be around 255khz.

but still have not managed to choose what MMC config to use: in order for primary to run around 255-10% kHz javaTC reports that a ~40nF MMC should be used (considering the full 8 turns of my primary (~14 uH)

What I can buy (because less and less caps are on stock from major sellers in Europe) is 22nF/2kVdc wima fkp-1

4s-8p in Mads MMC calculator is 44nF

Should I make a more capacitance mmc and just use less primary turns ? (in kaizer website a secondary my size says that mmc should be 150 to 450nf.

Also I am using profdc9 interrupter based on Steve Wards, I scoped its output of IC U3 to be 100-800Hz BPS and max ~250uS on-time (schematic in attachments). I would like the MMC to be able to go 250uS on time, else i guess I have to experiment on the RC network of U3 in order to have the on-time to a lower maximum.

Sorry if there are so many questions (or break any forum rule)

[davekni]

4s-8p in Mads MMC calculator is 44nF

Should I make a more capacitance mmc and just use less primary turns ? (in kaizer website a secondary my size says that mmc should be 150 to 450nf.

44nF seems reasonable for your coil.  At 230kHz (255kHz-10%), resonant impedance is 15.7 ohms.  At 300A current limit, that is 300 * 15.7 = 4.7kV peak MMC voltage.  So even 3S6P would work for peak voltage.  (Not sure about RMS, as that depends on actual primary current and interrupter duty cycle.  Might need 4S8P to avoid overheating MMC, or 3S6P with fan cooling.)  If you increased capacitance, impedance goes down.  Net energy storage in primary goes down for given 300A max current.  (Energy is L*I^2.  Lower primary inductance decreases primary energy capability for given max current.)

Also I am using profdc9 interrupter based on Steve Wards, I scoped its output of IC U3 to be 100-800Hz BPS and max ~250uS on-time (schematic in attachments). I would like the MMC to be able to go 250uS on time, else i guess I have to experiment on the RC network of U3 in order to have the on-time to a lower maximum.

Again, depends on actual primary current in operation to avoid excessive heat in MMC, primary coil, and H-bridge, and avoid tripping breaker in your 240V line power source.  Just start testing at low duty cycle and monitor temperature after short runs.  Slowly increase duty cycle and length of runs and keep monitoring temperatures (and total line power being drawn).

[Mads Barnkob]

You are properly not going to use the full 8 turns of primary inductance. You need to put in some head room for detuning primary below secondary resonance. So that arc loads pulls secondary resonance into primary resonance.

Try to calculate it for 6 turns (75% primary coil used).

High impedance coils needs longer on-time to produce the same arc length, thus you just move the losses from your MMC to the IGBTs. I would personally consider 0.044 uF to be too low (for a regular DRSSTC where more MMC capacitance have generally meant better spark performance) seen purely from a ratios perspective and not only based on math on the MMC itself. Not to de-value Daves calculations

You are not breaking any rules with questions that are based on your own research and where you supply all the data in question. You do it just perfect

[davekni]

22nF/2kVdc wima fkp-1

BTW, I purchased a large lot of 22nF/6kVdc WIMA FKP-1 caps, mostly for my Marx generator.  Tested them on my DRSSTC at +-4kV peak.  They did not last long.  However, for FKP-1, 6kVdc and 2kVdc caps are rated for the same 700Vac (for <=1kHz).  For the short bursts of a Tesla coil, they are likely fine for 1.5kV or 2kV peak.

For 230kHz, 22nF is 31.45 ohms impedance, so 1.5kV peak is 47.7Apeak.  Six parallel (6P) would be slightly over 1.5kV at 300A OCD (50A peak per cap).  7P would be under 1.5kV peak, and 8P well under.  Thus, if you made a 4S8P array (44nF), you can easily tap it at 3S8P and 2S8P and 1S8P for 58.7nF, 88nF, and 176nF respectively.  Test those with correspondingly less primary inductance.  (Of course, you can hit values between the above by further MMC rearrangement, such as 3S9P for 66nF etc.)

Or, you could start by purchasing fewer caps to save cost, perhaps enough for just the 3S8P option.  At 44nF you may never hit 300A even with long on-times.  I think that is what Mads is indicating.  More capacitance and fewer primary turns (lower primary impedance) may make better use of your 300A bridge capability.  If you hit 300A OCD too often, then increase primary impedance.

High impedance coils needs longer on-time to produce the same arc length, thus you just move the losses from your MMC to the IGBTs.

I'm a bit puzzled by this statement.  For a given total line (wall-plug) power, lower on-time (lower interrupter duty cycle) requires higher H-bridge current.  Higher current at lower duty cycle is generally higher IGBT power dissipation.  If duty cycle is reduced without increasing current, then both MMC and IGBT power are reduced.

[Mads Barnkob]

22nF/2kVdc wima fkp-1

BTW, I purchased a large lot of 22nF/6kVdc WIMA FKP-1 caps, mostly for my Marx generator.  Tested them on my DRSSTC at +-4kV peak.  They did not last long.  However, for FKP-1, 6kVdc and 2kVdc caps are rated for the same 700Vac (for <=1kHz).  For the short bursts of a Tesla coil, they are likely fine for 1.5kV or 2kV peak.

For 230kHz, 22nF is 31.45 ohms impedance, so 1.5kV peak is 47.7Apeak.  Six parallel (6P) would be slightly over 1.5kV at 300A OCD (50A peak per cap).  7P would be under 1.5kV peak, and 8P well under.  Thus, if you made a 4S8P array (44nF), you can easily tap it at 3S8P and 2S8P and 1S8P for 58.7nF, 88nF, and 176nF respectively.  Test those with correspondingly less primary inductance.  (Of course, you can hit values between the above by further MMC rearrangement, such as 3S9P for 66nF etc.)

Or, you could start by purchasing fewer caps to save cost, perhaps enough for just the 3S8P option.  At 44nF you may never hit 300A even with long on-times.  I think that is what Mads is indicating.  More capacitance and fewer primary turns (lower primary impedance) may make better use of your 300A bridge capability.  If you hit 300A OCD too often, then increase primary impedance.

High impedance coils needs longer on-time to produce the same arc length, thus you just move the losses from your MMC to the IGBTs.

I'm a bit puzzled by this statement.  For a given total line (wall-plug) power, lower on-time (lower interrupter duty cycle) requires higher H-bridge current.  Higher current at lower duty cycle is generally higher IGBT power dissipation.  If duty cycle is reduced without increasing current, then both MMC and IGBT power are reduced.

About the MMC, that was what I was trying to say

About the power dissipation. I got no data to back it up, but I think there is a thread somewhere with Uspring showing some calculations on why fewer switching cycles at higher current has less losses than many switching cycles at low current.

[Uspring]

About the power dissipation. I got no data to back it up, but I think there is a thread somewhere with Uspring showing some calculations on why fewer switching cycles at higher current has less losses than many switching cycles at low current.

I hope I didn't say this. Losses in the primary tank, which are probably larger than the IGBT losses, are roughly proportional to current squared. Power input is proportional to current linearly. So loss/power ratio becomes larger for larger currents.

The OPs question was about the choice of primary impedance.

If you have a primary tank with a low impedance, the losses will be lower since the primary coil will have less turns and less resistance dissipating energy. Also a low impedance primary will ramp up current faster up to the tolerable limit.
But: A low impedance primary will more easily hit the overcurrent limit, which should either stop the burst or go into a pulse skipping mode. Either will decrease the power input level below the capability of the bridge. The coil won't reach its full potential this way.
Better is a choice of a primary impedance high enough, so that a decrease of power input to avoid overcurrent isn't necessary at all or only for a few cycles.
A larger coupling and also optimised tuning permit the usage of lower impedance primary tanks.

[Mads Barnkob]

About the power dissipation. I got no data to back it up, but I think there is a thread somewhere with Uspring showing some calculations on why fewer switching cycles at higher current has less losses than many switching cycles at low current.

I hope I didn't say this. Losses in the primary tank, which are probably larger than the IGBT losses, are roughly proportional to current squared. Power input is proportional to current linearly. So loss/power ratio becomes larger for larger currents.

The OPs question was about the choice of primary impedance.

If you have a primary tank with a low impedance, the losses will be lower since the primary coil will have less turns and less resistance dissipating energy. Also a low impedance primary will ramp up current faster up to the tolerable limit.
But: A low impedance primary will more easily hit the overcurrent limit, which should either stop the burst or go into a pulse skipping mode. Either will decrease the power input level below the capability of the bridge. The coil won't reach its full potential this way.
Better is a choice of a primary impedance high enough, so that a decrease of power input to avoid overcurrent isn't necessary at all or only for a few cycles.
A larger coupling and also optimised tuning permit the usage of lower impedance primary tanks.

I should not have trusted my leaky memory on that, it is full of holes that it will fill up with information that "makes" sense

I did however think I found the threads I mentioned, where you state the same as above, so you were in the right.

We started discussing optimal MMC capacitance here https://highvoltageforum.net/index.php?topic=90.msg462#msg462 and ended up splitting it to a separate topic, where you maybe has some answer to my last post? https://highvoltageforum.net/index.php?topic=113.0

[davekni]

About the MMC, that was what I was trying to say

Mads has more experience than I do with typical coil parameters that work well in practice.  Thus a lower impedance primary (more than 44nF MMC) may be optimum.  Measuring primary current and/or seeing how often OCD trips is the way to tell for sure with your coil.
I found a coil that seems rather similar to yours:
https://highvoltageforum.net/index.php?topic=1705.msg12944#msg12944
Alan has a 16-ohm primary impedance, the same as yours would be with 44nF.  He is hitting 300A at 200Vbus, and concerned about raising Vbus to maximum because of resulting excessive primary current.  Coupling factor was changed during the above thread, from 0.2 to 0.12 (a bit too low).  Low coupling increases required primary current.  If you want to get to 340Vbus (peak of 240Vac), Alan's example indicates 16ohms may barely be high enough (44nF barely low enough) even with a more typical 0.14 coupling factor.

Will be great to add your results to our collective knowledge when you start testing!

[costas_p]

Thank you all for your replies so far,

I calculated for 6 turns primary: inductance is  8.16uH so the MMC should be around 60nf for 230kHz (3s-8p as davekni suggested for 58nF) and then i need to find the exact tap point on my primary

Meanwhile i adjusted OCD on the UD:

CT is 300:1 so for 280Apk there is 0.84A (maybe I should start lower OCD for initial tests)
Burden resistor is 5.1 ohms

Ohm's law V = I * R : 0.84 * 5.1 = 4.284V on J9-pin2

[Uspring]

Mads wrote:

We started discussing optimal MMC capacitance here https://highvoltageforum.net/index.php?topic=90.msg462#msg462 and ended up splitting it to a separate topic, where you maybe has some answer to my last post? https://highvoltageforum.net/index.php?topic=113.0

I've added a reply to this thread https://highvoltageforum.net/index.php?topic=113.0

[davekni]

I've added to this other thread too, a simulation case intended to roughly match this coil:
https://highvoltageforum.net/index.php?topic=113.msg14999#msg14999
This simulation is for 44nF MMC, 11uH primary (for 229kHz primary frequency), 257kHz secondary frequency w/o arc, and coupling factor of 0.14 (slightly above your 0.12).  With optimum arc load, primary impedance gets to 1.56 ohms.  Any real arc load that differs from ideal will reduce primary impedance.

For your coil, 1.0 ohms should be about optimum.  1 ohm is 280A at 280Vrms.  280Vrms matches the fundamental sine-wave frequency component of a 311V square wave (280V = 311V * 2 * sqrt(2) / PI).  311V seems reasonable for a loaded 340Vpeak line including IGBT Vce etc.  Note that 1.0 ohms is optimum for a real arc load.  Therefore the ideally-matched arc load case needs to be something above 1.0 ohms.  However, I have no guesses as to how far above (how close real arc loads get to optimum for any specific real coil).

58nF instead of 44nF should lower max primary impedance to 1.17ohms.  If you keep coupling at 0.12, that lowers max primary impedance further down to almost exactly 1.0 ohms.  This is likely too low for optimum (maximum) resonant impedance given that arc loading will never be perfectly matched to your coil.  Thus I'd recommend at least planning options to increase coupling slightly and decrease MMC capacitance, perhaps back to 44nF.  These options would be to use if you hit 280A OCD too frequently.

[costas_p]

Caps for MMC finally arrived,

the total capacitance measured is 53.5 nF  for 4s - 10p

I will be using a coffeepot with water as a static load for the initial tests

So I have to ask: what is the preferable way to power up the coil?

Apply desired variac voltage to the bridge, then start the interrupter and slowly play with BPS/on time? (while measuring current/temperatures)

OR

start the interrupter and then apply voltage through the variac

this question also applies on later stages of the coil where hopefully everything goes well and I apply power from the precharge pcb where the bridge will have ~325V before start

[davekni]

start the interrupter and then apply voltage through the variac

Or part way between: Start with some bus voltage (variac voltage), then start interrupter at low BPS.  UD2.7 will not start oscillating unless the first half-cycle generates enough primary current through the feedback CT to trigger then next half-cycle.  You could start at 0V, then see what bus voltage is required to get oscillation started.  (Anyone had issues with UD2.7 outputs being stressed by the low frequency signal of non-oscillation resulting in GDT saturation?)

[costas_p]

First test is completed , it wanted around 70V ac from variac to start oscillating, there is a buzzing sound coming from the coffeepot, after 1 minute of run on 70V ac everything was cool to the touch,
actual measurements not done cause my tools are on work place at the moment. I am verry happy nothing exploded yet

https://youtube.com/shorts/fxLiGhJB-ZI?feature=share

[AstRii]

(Anyone had issues with UD2.7 outputs being stressed by the low frequency signal of non-oscillation resulting in GDT saturation?)

While we're at it, how does one recognize if GDT is saturating?
I usually check primary current of a transformer to see if it doesn't peak way above nominal current, but current flowing to transistor's gates is by definition a high current peak, rendering my method useless in this case.
How do output waveforms of a GDT look like if it's saturating?

[davekni]

While we're at it, how does one recognize if GDT is saturating?
I usually check primary current of a transformer to see if it doesn't peak way above nominal current, but current flowing to transistor's gates is by definition a high current peak, rendering my method useless in this case.
How do output waveforms of a GDT look like if it's saturating?

Normal peak GDT input current is during voltage switching.  Saturation current is highest just before voltage reversal, and typically lasts longer, from the time of saturation until voltage reversal.  So, time of peak current is the key.  Saturation current will show up as relatively-rapid charging of coupling capacitor in UD2.7 output circuit and in higher output FET voltage drop.

[costas_p]

I finally managed to assemble the coil, and make some tests, coil achieved first light.

https://youtube.com/shorts/_saYnn-gCiA?feature=share

Phase lead is not yet tuned.
I managed to supply full ac mains power throught a variac without blowing anything, however the spark length did not had big difference from eg 70v (start oscillating) to full 220v.
What is the recommendation: make coupling factor higher, or start adjusting primary coil tap point ?

[davekni]

What is the recommendation: make coupling factor higher, or start adjusting primary coil tap point ?

If coupling is still at your original 0.12, I recommend increasing it to at least 0.14.  Even higher might be good, especially for initial testing.

Tuning would be easier with a scope.  Adjusting tuning will likely help too.  Higher coupling will allow success with a bit more detuning.  Thus I'd increase coupling first.

[costas_p]

By the way, when we talk about coupling factor, what is the correct way?

In my case I am using 1/4" copper tubing for primary coil, so to measure coupling I measure distance from top/bottom or middle of primary and the start of secondary winding ?

[davekni]

In my case I am using 1/4" copper tubing for primary coil, so to measure coupling I measure distance from top/bottom or middle of primary and the start of secondary winding ?

I'd guess that JavaTC is based on middle of primary winding.  However, someone else will need to answer this more definitively.

[costas_p]

I did all calculation from the beginning, my coupling factor as it is right now is 0.126, due to construction  I can remove another 5mm of secondary coil height for a maximum 0.138 coupling.

While playing midi (6 turn tap + 0.126 coupling) the coil happily output around 20cm sparks at 70-80v input , increasing the voltage from variac the OCD led starts flashing like crazy so I decrease voltage to lower.

I will experiment further in order to be able to use full main voltage.


Mean while doom theme song:

[davekni]

While playing midi (6 turn tap + 0.126 coupling) the coil happily output around 20cm sparks at 70-80v input , increasing the voltage from variac the OCD led starts flashing like crazy so I decrease voltage to lower.

Increased coupling will likely reduce OCD events, allowing higher voltage.