Post by: babass on November 14, 2019, 05:40:58 PM
I'm creating my first SSTC for some school project. However, I have some troubles :
I want to use a full H-bridge. I got some IXGH40N60A IGBTs.
I also got some UCC27321 and UCC27322 to drive them (isn't that the most used driver ?)
For the tests I use a 555 timer that creates a 220kHz square signal with 50% duty cycle (that's about the resonant frequency of my tesla coil), that will be replaced by an antenna later. The 555 timer is powered by a 7805 regulator, and the drivers are powered by a 7812 regulator, pluged on a variable power supply.
Everything is plugged like this schematic found on instructables : https://www.instructables.com/id/Building-a-solid-state-tesla-coil/
Except i'm not using a 555 to modulate my tesla coil, I plug the pin 3 of my drivers to +5V (for now).
When I scope between the 2 outputs of my drivers I got a nice square wave signal, going from +12V to -12V, 50% duty cycle, at 220kHz, so everything is OK, working fine for now.
Time to make a gate driving tranformer. 1 primary coil for 4 secondary coils. ratio 1:1.
My transformer core is kinda big, but it has an Al=4700 approximately.
When I plug the primary of my transformer to the output of my drivers (with a 100nF capacitor in serie, like the schematic) and scope one of the secondary, I still have my nice square wave signal, from +12V to -12V. But when I plug all my IGBTs... the signal voltage at the primary of my transformer drops from +12V/-12V to barely +7V/-7V as you can see on the first picture, and my drivers starts heating (there is some ringing as well, but it's due to my oscilloscope, didn't had them with a numeric one with a different probe).
I'm using 47 ohm gate resistors. On the 2nd picture i'm probing Vge of my IGBTs... again, barely +7V/-7V "square" wave, probably because of too high gate resistors value.
I tried to reduce the amount of turns in my transformer, with no luck.
What should I do ? It will end up killing my drivers, they are heating quite a bit.
BUT... the tesla coil actually works, I can replace the 555 by an antenna, I can get some nice electric arcs (also works with the 555 timer but it's not tuned correctly). But, why are my UCC27321/UCC27322 heating that much ? Why is the voltage droping that much ? Can those drivers handle a full H-bridge ?
Post by: davekni on November 15, 2019, 05:18:19 AM
As with most IGBTs, the IXGH40N60A parts are designed for 15Vge. That's where the current ratings are specified. 12V is a bit low, although probably OK for typical parts. Most silicon FETs are designed for 10Vgs. The instructable uses FETs, so 12Vgs is plenty including margin.
The gate charge for one IXGH40N60A is 200nC for 0 to 12V gate. Although not specified, it's likely another 200nC for the -12Vge to 0Vge half of the gate-drive waveform. So 400nC total per IGBT, or 1.6uC for all four parts of the bridge. That charge is required on every edge of the drive waveform, so 440kHz repeat rate. 440kHz * 1.6uC = 704mA. 704mA * 12V = 8.45W, shared between the two gate drivers and the four gate resistors. Not too surprising that drivers run warm. This is based on continuous operation. Many coils are pulsed (enabled and disabled) to play tunes or whatever. The gate drive power will drop by the duty cycle of the enable pulsing.
My only guess for your +-12V to +-7V issue is that the 12Vdc supply isn't holding up to the 700+mA cuyrrent. Have you tried measuring the 12Vdc supply while running?
Did the scope probe showing the ring have a missing ground connection? I'd be a bit careful of assigning ringing to probes. Gate-drive-transformer leakage inductance and general wiring inductance can cause such ringing. For wiring inductance, the ringing may be present at one end of a cable and not at the other.
Post by: babass on November 16, 2019, 01:21:10 AM
I saw some people using a gate resistor per MOSFET/IGBT on the secondary of the GDT, and some people using one resistor at the primary. Does it make any difference ?
I think I'll go for 2 GDTs, each drived by a pair of UCC27321/27322. I can't check my 12Vdc supply for now.
Post by: davekni on November 16, 2019, 04:56:44 AM
The diodes should be physically close to each other (and to the IGBT) with good thermal contact to keep the diode temperatures matched. Diodes generally have a negative forward-voltage vs. temperature curve, so the hot diode will hog current and get hotter. Most heat is conducted through the leads, so solder the leads all together or to a short piece of heavy copper wire or plane such as a piece of 12AWG wire. (One wire piece for cathodes and another for anodes.)
As with the diodes, most of the driver chip heat is conducted out the leads, especially the negative power lead (usually - not certain about these chips). That makes it hard to add more heat dissipation capability without changing the circuit board design to have larger copper areas for those leads. You can gain a bit of cooling by gluing a piece of copper or aluminum to the top of the packages. A small fan blowing directly at the chips will help too.
Scoping the waveforms on the driver chips, especially the outputs, would be a good idea too. Perhaps some ringing is causing more output switching than the intended 220kHz. Especially with the voltage dropping to 7V, the power shouldn't be all that high. (That's presuming the 12V supply is dropping as the cause for lower gate voltage.)
I've seen gate drive resistors on either side, but by far most often on the secondaries of the gate drive transformer, one set per IGBT. Often the resistors include a diode and second resistor to make turn-off faster than turn-on:
Other members here likely have more experience with gate-drive transformer/resistor combinations than I do. Hopefully you'll get more replies if my suggestion isn't optimum.
Another thought on the driver heating: Each IGBT emitter has two connections, one for the high-current path (to Vbus or MMC or Tesla primary winding), and another for gate-drive transformer. These two connections should separately wire to each IGBT emitter pin, as close to the IGBT case as possible. Larger IGBTs usually have two emitter connections, one for the high-current path and another just for gate-drive return. If the high-current path shares wiring with gate-drive return, then the inductive voltage drop due to the high-current well be added to the gate waveform. I've tried to show that in the above gate-drive schematic, with the emitter connection splitting just under each IGBT.
One final recommendation: Until everything is working properly, run well under 50% duty cycle, say 5%. Mistakes are less likely to fry parts when they are running cooler.
Post by: babass on November 16, 2019, 01:33:17 PM
And why full bridge SSTC shematics aren't that commmon ? Primary voltage amplitude is twice as high than with a half bridge right ? Or maybe it's too hard to drive, and lead to problems like mines.
Post by: thedoc298 on November 16, 2019, 05:19:05 PM
Ok, thanks you for all those information. I'm using 47 Ohm gate resistors, but discharge time is way too high (and charge time too). But if I lower them to, let say, 10-15 Ohm like everybody does, will my drivers heat even more ?
And why full bridge SSTC shematics aren't that commmon ? Primary voltage amplitude is twice as high than with a half bridge right ? Or maybe it's too hard to drive, and lead to problems like mines.
You want to switch your mosfet to switch as fast as possible = minimal heat. Large gate resistors = long switching times and = higher heat.
Post by: davekni on November 16, 2019, 06:55:16 PM
Can you post images of the driver board, or the board layout? Parasitic inductance on board traces, especially ground pins, may lead to local high-frequency oscillations. Such oscillations, which could be 30-300MHz, may explain both problems, hot driver chips and gate drive voltage dropping to +-7V. Depending on the scope and probes, the oscillation may be too high frequency to show up. They'll also be filtered out by the gate resistors. However, the resulting gate waveform will not reach 12V because it is the average of the high-frequency waveform. (I've experienced that issue in several of my earlier hobby circuits, though they weren't Tesla coils.)
Not related to the above speculation of high-frequency local driver-board oscillations, but here's a bit more on IGBT emitter connections: Dual contacts to the emitter are referred to as "Kelvin connections" in case you want to research that more. To be clear, I've updated my half-bridge circuit sketch to show that the diodes across the IGBTs are part of the high-current connections. The gate-drive return shouldn't share any wiring with any of the high-current paths:
[ Invalid Attachment ]
Post by: babass on November 16, 2019, 07:27:30 PM
Yes, my IGBTs are heating quite a bit too.
Is it something common to use a little heatsink on the driver chips ? I also planned to add a bigger capacitor on the top of my tesla coil, to reduce the resonance frequency. Is 220kHz a high frequency for a SSTC ?
And do you have any tips for my circuit design to avoid those parasitic inductance ?
Is it a good idea to power my circuit through a 7812 regulator? Or it add even more potential parasitic oscillations ?
I'm using an old CRT scope for my tests. It can't measure a frequency as high as 30MHz.
Post by: Mads Barnkob on November 16, 2019, 08:03:58 PM
The circuit part with the drivers are on a breadboard, I know those things has a lot of parasitic capacitance, but I don't know about parasitic inductance. I was about to design a real circuit for this, but wanted to fix that heating problem before.
Yes, my IGBTs are heating quite a bit too.
Is it something common to use a little heatsink on the driver chips ? I also planned to add a bigger capacitor on the top of my tesla coil, to reduce the resonance frequency. Is 220kHz a high frequency for a SSTC ?
And do you have any tips for my circuit design to avoid those parasitic inductance ?
Is it a good idea to power my circuit through a 7812 regulator? Or it add even more potential parasitic oscillations ?
I'm using an old CRT scope for my tests. It can't measure a frequency as high as 30MHz.
A SSTC driver on a breadboard will work fine! Steve Wards Mini SSTC / SSTC 5 is a solid build, its easy to get running and very reliable, so we are looking for defective or wrong components here.
First I would like to question the GDT core material and AL value, are you sure that this core is suitable for the job? It is not coated and its material structure looks too coarse to me. A bad GDT core could explain your issues as you are properly driving the IGBTs in the linear region all the time. More GDT waveform troubleshooting: http://www.richieburnett.co.uk/temp/gdt/gdt2.html
220 kHz is far from the limit of this circuit, I built them up to 350 kHz.
Your pictures of the gate resistors on a PCB connector is no-go! You need your gate resistors mounted directly on the IGBT/MOSFET gates.
A 7812 is plenty to drive this coil at 220 kHz, even in CW mode, it might need a heat sink though :)
A good old analog 30 MHz scope is good enough to troubleshoot a SSTC at 220 kHz, you need to tidy up your wires, twisted, short paths etc, that kind of inductance optimization gives results, trying to measure them makes no sense if you are going to measure on long wires anyway. Only measure for high frequency ringing if you did all you could to reduce inductance first.
Ok, thanks you for all those information. I'm using 47 Ohm gate resistors, but discharge time is way too high (and charge time too). But if I lower them to, let say, 10-15 Ohm like everybody does, will my drivers heat even more ?
And why full bridge SSTC shematics aren't that commmon ? Primary voltage amplitude is twice as high than with a half bridge right ? Or maybe it's too hard to drive, and lead to problems like mines.
You want to switch your mosfet to switch as fast as possible = minimal heat. Large gate resistors = long switching times and = higher heat.
Actually you want to "Switch as fast as slowly possible", meaning that if you switch too fast you get ringing, but too slow and you get higher losses, you need to find the golden in-between :)
Another piece of advise, I collected some little information about SSTC design here: http://kaizerpowerelectronics.dk/tesla-coils/sstc-design-guide/
Post by: babass on November 16, 2019, 09:20:41 PM
Then, according to the GDT troubleshoot web site i'm in the "Massively overdamped gate drive waveform" case. Before using 47 Ohm resistors I tried with 10 Ohm resistors and I got something similar to "The perfect gate drive waveform!" with a very little overshoot, but with a even lower peak voltage and my drivers where heating even more.
The core i'm using has an Al of about 4700. I might have some ferrite core laying around, the anti parasitic ones (grey, looks like carbon). Should I use one of them ?
Yes, I'm using quite long cables to connect my IGBTs to my transformer.
What do you mean by "not coated" GDT ? The leakage inductance is quite low (about 0.8 µH, I shorted all the secondaries and plugged the primary in a (quite accurate) RLC meter).
So for my circuit I should try to make all my connections as short as possible ?
Oh, and I'm not using any DC Blocking Capacitor in serie with my tesla coil. I'll add one later.
Post by: davekni on November 17, 2019, 06:26:58 AM
Just realized, 0.1uF is not enough for driving your four IGBTs. Thier total gate capacitance is ~66nF (0.066uF). That explains almost all of the gate-voltage drop when loaded (when IGBTs are connected). It's forming a capacitive divider of 0.1uF and 0.066uF. To avoid much voltage drop, C6 should be much larger than the total gate capacitance, at least 1uF in your case. 2.2uF would be better.
Wiring inductance in the gate circuit effectively adds to leakage inductance, so should be minimized as Mads suggested. Wiring inductance is almost always parasitic, so best to minimize everywhere. (Although not a critical issue at this point, even the 0.8uH gate transformer leakage inductance will limit gate switching time to ~400ns. A common technique for lowering leakage inductance is to wind with four twisted pairs, 8 wires total. Cat5 or similar cable is easy for this. One half of each twisted pair is used for gate-connections. The remain four wires, one from each pair, are connected in parallel and used for the primary. It's similar to what you have, except the primary has four wires instead of one, and each primary wire is tightly coupled to one secondary wire.)
For circuit board layout, the best option is to reserve one layer for a ground plane, with as few gaps as possible if any are needed to make short connections bridging traces on other layer(s). That ground plane will help some with driver heat-sinking too. (There are driver chips in packages designed for heat-sink attachment, but hopefully that isn't necessary for your design. That's presuming the power issue is resolved by eliminating oscillation or whatever the issue turns out to be.) I'd also suggest adding more bypass capacitors. Duplicate C4 and C5. Place one across the input side, pins 1 to 4. Place the other across the output side, pins 5 to 8. Place caps as close to the chip as possible.
Low ground inductance is good for breadboards too. If possible, copper tape strips for ground. Good bypassing is also important. I often solder bypass capacitors directly to the chips, especially ones with high switching current spikes, including the drivers and 555. As a first experiment, I'd try soldering bypass capacitors to your driver chips, 2 per chip as mentioned above. You could leave C4 and C5 in place, and just add more bypass caps directly at the chip pins. They can be soldered to the pins under the breadboard, or directly to the sides of the chips on top (assuming you're using the DIP packages). Just adding those caps might be enough to fix oscillation if that is the reason for excess driver heating.
Another suggestion for breadboarding this circuit in particular: The two driver chips should have Kelvin ground connections much as the IGBT emitters. Hopefully the two driver chips are reasonably close. Wire the two chip's input-side ground pins (pin 4) together. Separately wire the two output-side ground pins (pin 5) together. Connect the two ground wires together at one location. Connect the input circuitry only to the input-side pin-4 ground wire. You could do the same with the +12V power, but that's much less critical. The logic threshold voltage is much closer to ground than to +12V.
One crude method to look for oscillation or rings that are too fast for your scope: Place a small capacitor, 100-1000pF, at the probe tip, from tip to ground. Add a fast signal diode in series with the scope tip. Probe with the open end of the signal diode. Switch the diode direction to probe for peak positive or peak negative voltage. If the gate-driver outputs get much above +12V or below ground, then there is an issue with fast oscillation or ringing.
Post by: babass on November 17, 2019, 04:06:01 PM
I'm connecting the input pin of the 7805 regulator directly on the Power supply out instead of the output pin of the 7812, does it make any difference ?
I also added some 0.1µf in parallel of C9, C10 and C11, is that ok ?
For C4 and C5 I'm also using 0.1µf, is this ok, or should I use higher capacity and specific cap technology ? I'll try to add more, 1 between pin 1/4 and 1 between pin 5/8, as close as possible of the chip.
And yes, haven't seen the voltage divider caused by C6. That may explain why I have barely +8V peak at the input of my GDT.
But it means that my drivers will heat even more, right ?
I'll try to input a square signal with around 30% duty cycle in the enable pin of my drivers, hope it will reduce the heat.
I can only make a single layer circuit, but I'll try to have a "big" ground plane at the ground pin of the drivers to help cooling them, as well as a fan on the top of the circuit.
Post by: davekni on November 18, 2019, 02:53:29 AM
Yes, 0.1 in parallel with C9, C10, and C11 is a good idea.
The TI data sheet for the drivers suggests 0.1uF for the input side (pin 1 to 4) and 1.0uF for the output side (pin 5 to 8). Section 10 of the data sheet shows suggested ECB layout, showing that input circuitry should be referenced to pin 4 and output load (gate drive) referenced to pin 5 ground. If you make a 1-layer ECB, I'd suggest a combination of ECB and proto-board style hand wiring. Use the ECB layer primarily for ground plane, especially under and around the driver chips. Make the signal connections with wires, laid flat against the ground plane. (I haven't looked recently, but there used to be vectorboard available with a ground "plane" included - actually a 0.1" square grid of ground wires running between the 0.1" grid of holes. That's another option rather than making a 1-layer ECB.) I'd still recommend two bypass capacitors per driver chip, soldered directly from the power pins to the corresponding ground pins or to the ground plane.
A side note: Ceramic capacitors are generally best for high-frequency bypassing. However, spec's can be misleading. A "1.0uF" capacitor rated for 25V may drop to 0.5 or even 0.3uF at 12V, and down around 0.1uF at rated 25uF. A few manufacturers are starting to publish at least typical data for this voltage degradation of capacitance.
Concerning the size of the local driver pin 5-8 bypass capacitor, it depends on the wiring inductance back to the 12V bulk capacitor C10. I'd suggest at least the 1.0uF that TI recommends, especially for a hand-wired board that will have more wiring inductance.
Driver heating is the trickiest problem to figure out. If it's caused by high-frequency oscillation, then adding tightly-coupled bypass capacitors and separating the input-side circuitry from the output-side should stop the oscillation and fix the heating issue. (Also need a good low-inductance connection between the two driver chips output grounds - pins 5, since the gate-drive current flows from one chip to the other, so needs a low-inductance return path.)
If the heating is just due to normal gate-drive power, then, yes, it will get worse when C6 is increased to 1-2uF. For that reason, I'd start even lower than 30% duty cycle 10% or less. You could run some initial gate-drive heat testing without the IGBT power connected. (That will change gate-drive overshoot some, but gate-drive power should still be 80-90% of it's full-load condition with 170V on the IGBTs.) Without IGBT power, it will be safer to measure 12V current, waveforms at the driver chip outputs, etc.
The TI UCC27321/2 data sheet lacks information on drive current vs. voltage during the transitions, making it difficult to calculate how the gate power will split between the gate resistors and driver chip. With 47-ohm gate resistors, the drivers should see ~20-30% of the power. With 10-ohm gate resistors, it will be higher, perhaps 40-50% of the power in the driver chips.
To make calculations even harder, this data sheet appears to have some self-inconsistencies. Max die temperature is listed as 150C. Thermal resistance to ambient is listed as 55.9 C/W for the PDIP package. At 2 watts, that's 112C rise, which would allow operation at 38C ambient. However, the power dissipation table lists maximum of 350mW at 25C ambient, roughly 1/6th of what the thermal resistance value would imply. If the correct answer is 2W, you'll probably be fine once any oscillations are fixed. If the correct answer is 350mW, then these drivers aren't sufficient for your application except at low duty cycle. Anyone else see how to reconcile these data-sheet values? Am I making an obvious mistake here?
BTW, there's another apparent mismatch in the data sheet: Figures 7 and 8 show rise and fall times vs. supply voltage. Figures 9 and 10 show rise and fall times vs. load capacitance. For the 10nF load condition of the first two graphs, the ~20ns rise and fall times are not close to the 10nF (right edge) of the other two graphs, which show ~27ns rise and ~170ns fall.
Your diligence on this student project is impressive! This is good experience for the engineering world, where projects are always more complex than they initially seem.
Post by: babass on November 18, 2019, 09:19:26 PM
Hope this will fix all my problems ;).
I have another little problem, concerning the feedback antenna. Well, when I use a 74hc14 hex inverter with Schmitt-trigger inputs, it doesn't work. The tesla coil won't work at all. I tried different cables for the antenna, no luck. But... with an old 4069 it actually works, and pretty good, but the antenna has to be close to the tesla coil, or it will stop working.
Or is there a better way to make a consistant working antenna than an inverter chip ?
Post by: davekni on November 19, 2019, 06:03:37 AM
Post by: babass on November 20, 2019, 10:05:50 PM
So, I tried adding a 1Meg resistor, with no luck ???. I couldn't find anything about your "current-sense feedback". Are you using a current transformer at the "bottom wire" of the secondary of the tesla coil (the one connected to ground) ? Then you use clamping diodes (is 1N4148 any good for this ?) to lower the voltage and send it in the input pin of the inverter chip ?
Concerning my driver heating issue... I added few capacitors, giving the enable pin a 50% duty cycle signal... and the problem is somehow gone. It keeps heating but it's not as fast as before, I can definitly put my finger on it after minutes without any problems. So with a well designed PCB it should be fine.
I need to lower my gate resistors, as you can see on the scope, it takes too much time to load / unload the IGBT gate.
But... it works well, my IGBTs are heating a little under 60V cuz of too high gate resistors but...
I modified my GDT since, about 600µH at the primary, 0.6µH leaking inductance. I think it's fine ;D.
Now I need to design a nice PCB, maybe add a little heatsink on my drivers (I have some very little heatsinks, will do the trick, just need to find out how to stick them correclty).
And I need to fix that antenna issue as well. I'm going to try the current-sense tranformer.
Post by: davekni on November 21, 2019, 06:07:46 AM
Did you try the 1meg with an HC14, or just with the 4069? It should help with either one, but the HC14 with resistor should self-oscillate even with no supply to the IGBTs. If you try 1meg with HC14, measure the frequency with no IGBT power (but with the antenna connected). Change the resistor by the ratio of measured_frequency / desired_frequency, where desired_frequency is 220kHz or whatever your updated resonant frequency may be.
Have you checked the polarity of the H-Bridge? Does it work any better with reversed polarity (reverse gate-drive leads from driver-chips, or reverse H-Bridge output to primary coil).
Yes, I used a current-transformer in the secondary ground lead, but I wouldn't copy my specific implementation. It is self-oscillating, and locked to the secondary in a relatively-narrow frequency range, so needed fairly precise tuning to work well. I think the conventional circuit you have could work with a current transformer in the secondary ground lead, with a load resistor and then capacitor to HC14 pin 1.
Post by: babass on November 22, 2019, 12:15:26 AM
I'll try more experimentations with 74HC14 chip.
Post by: davekni on November 22, 2019, 05:34:08 AM
Post by: babass on November 24, 2019, 12:13:51 AM
I'll try that feedback transformer like this one http://www.loneoceans.com/labs/sstc2/sstc2schematicv10.jpg
They use a core with 50 turns on it. What would be the ideal core (AL value) ?
Post by: davekni on November 24, 2019, 02:56:27 AM
Your comment about noisy oscillation makes me think of another possible issue. Perhaps the antenna is picking up a lot of 50 or 60Hz signal from line power wiring in the vicinity. That's a common situation - line frequency fields everywhere. Anyone know of common circuits to reject low-frequencies from antenna-feedback SSTCs? If not, and if you want to continue down the antenna path, I'll come up with a suggestion.
Probably not directly related to existing issues, but I'll mention anyway: It's normal to tie unused CMOS chip inputs to ground or power or some other signal, not to leave them floating (unconnected). For HC14 and such, I often wire the inverters in a chain to keep it simple. In your case of using only the first inverter (pins 1 to 2), you could wire pins 2-3, 4-5, 14-13, 12-11, and 10-9. That would connect all 5 unused inputs to something.
I'd encourage your switch to current feedback. I have no personal experience with antenna feedback, but it certainly seems more susceptible to noise sources. If you go that direction, however, the schematic you attached appears to have an error. The two 1N4148 clamp diodes on the current transformer secondary should be on the HC74 input (pin 1) rather than directly on the transformer secondary. Even better would be to add a second 1k resistor in series with the HC74 input, then place the clamp diodes between the two resistors:
[ Invalid Attachment ]
Does anyone else have experience with the "http://www.loneoceans.com/labs/sstc2/sstc2schematicv10.jpg" circuit with CT feedback? Anyone have a reason that it makes sense to have a diode directly across the CT, making it carry a net DC current?
Anything around 1uH/turn^2 or more should be fine for the CT core. That would keep the secondary above 2.5uH, which is 3.5k-ohms reactance at your 220kHz, far enough above the 1k load resistance.
Post by: Mads Barnkob on November 25, 2019, 09:05:38 AM
Eh, well, I tried different resistors, I can't get it to oscillate (74HC14). With a capacitor between pin 1 and ground (and much lower capacitance value, like 100k) I got some oscillations, but very inconsistent and very low peak voltage (barely 1V). I don't get it...
I'll try that feedback transformer like this one http://www.loneoceans.com/labs/sstc2/sstc2schematicv10.jpg
They use a core with 50 turns on it. What would be the ideal core (AL value) ?
We typically all make coils in the region of 40 to 500 kHz, here I have found that for all GDT, CT etc, the rule of thumb of somewhere around 5000 AL+/- 1000, will work fine. Just be sure to have the material right aswell, I have used N30 a lot. There is however also other materials in the same size core / permability range.
Post by: babass on November 28, 2019, 01:17:35 AM
However it works fine with a 555 instead of the antenna (very loud and clear buzzing sound, once it's tuned correctly).
By removing R1 from your schematic it's a little bit better.
I also tried inverting the tesla coil primary wires.
I tried to make 2 or 3 turns at the primary of the feedback transformer, but it won't change anything.
Tried with both 4069 and 74hc14.
Post by: davekni on November 28, 2019, 05:36:53 AM
Thinking about Al more, startup is the more stringent condition than full-power. Until the CT output voltage is high enough for the clamp diodes to conduct, the CT secondary is mostly unloaded, so 90 degree phase lead. The higher Al range that Mads suggests (Al around 5uH/T^2, or or 100/2 turns on a lower Al core) will help the CT output get to higher voltage sooner.
Post by: babass on November 28, 2019, 06:10:51 PM
So as I'm using a 1k resistor in series with the transformer, it means that I have a 86,6° phase lead ? The goal here is to be as close as 90° right ? So as my core is quite big, is making much more turns would be a good idea ? And when you say 1 turn at the primary, do i need to make one turn with the ground cable, or just going through the core as I'm doing on the picture (I tried making a turn, wont change anything much).
And what's the purpose of R1 on your schematic ?
But when I try to higher the voltage across the primary the Tesla will start making weird noise and eventually my power supply (some lab power supply, 120V 1A max) will shut down. Cas this be due to bad feedback again, or just my power supply can't handle powering the Tesla coil ?
On the second picture, my scope with a little wire plugged in as an antenna, showing the 220kHz pulses emitted by the Tesla coil (enable pin connected to a 555 generating a 50% duty cycle square wave signal).
Post by: davekni on November 29, 2019, 06:09:07 AM
The goal is to have roughly 0 degrees phase shift for the entire circuit. Positive H-Bridge output voltage when the current is positive and negative voltage when current is negative. Since there are some delays (HC14, gate driver, FETs), a bit of phase lead in the current feedback would be optimum, perhaps 20-30 degrees. That's for optimum output power. If using IGBTs where there's more value (efficiency) to switching just before their zero-current point, the optimum phase shift may be slightly different. Depends on inductance of Tesla primary coil.
With this controller circuit, phase will change as the power increases at the start of each burst. At the beginning of a burst, the current feedback will have almost 90 degrees phase lead. That's because the CT output voltage will be under 5Vpp, so the diodes won't conduct, so infinite load impedance. Current into an inductor creates the 90 degree lead. As the power builds, the CT output increases and the diodes conduct more of the time, so the effective load resistance on the CT drops towards 1k. 12.6mH at 220kHz is 17.4k reactance, so phase shift is arctan(1k/17.4k) = 3.3 degrees. That's the limit at infinite power. So, the phase lead is dropping from almost 90 towards 3.3 at the start of each burst.
R1 in my schematic is just further isolation of the HC14 input from CT feedback voltages below 0 or above 5V. The 1N4148 diodes clamp the voltage to one diode drop below 0 and above 5V. That's the same as the HC14 input, so the input and the 1N4148 diodes share the current. R1 forces most of the current to stay in the 1N4148 clamp diodes. Modern HC14 inputs can handle input current fairly well without disrupting operation, so R1 isn't all that important. Old parts were more sensitive. (The original schematic had a note about changing the clamp diodes to schottky parts, which is another way to accomplish the same goal, as they have lower forward voltage drop than the HC14 input The suggested 1N5818 part is rather high capacitance, however, so I'd not recommend that particular schottky diode.)
I'm a bit puzzled as to why the scope signal amplitude ramps up at the start of the burst so much faster than it ramps down at the end. Is this normal for SSTCs? If so, will someone please explain why?
Passing the Tesla secondary ground lead once through the CT core as you pictured is effectively 1 turn. The "turn" is just quite large, counting the path to the ground/counterpoise and back through the air to the top-load. 2 turns would be one more loop of wire, so the ground wire goes through the core twice. As long as one turn produces a signal large enough for reliable starting, more just adds to heating in the 1k resistor.
Post by: babass on November 29, 2019, 09:40:00 PM
Anyway, if I replace R2 by a potentiometer, maybe I can adjust the phase shift at high power to get optimal results, right ? and them replace it with normal resistors, I think it won't survive for long. Or should I keep R2=1k and change the capacitor value ?
Post by: davekni on December 01, 2019, 04:30:10 AM
Does your scope have a separate trigger input? If not, measuring phase with a single channel will be difficult. If it does have a trigger input, connect that to a fixed place, ideally the enable input. Trigger on the lead edge of enable. Then use the single channel to measure points around the loop: CT output, 4069 input, 4069 output, gate drive transformer inputs (each side separately), and H-Bridge outputs (if your probe can handle 200V). Set the scope for say 5us/division to see the startup phasing. This exercise should show phasing and hopefully give clues on the surprisingly-rapid HV output start-up. Scope your antenna pick-up as well with the same trigger and horizontal scale. Images of all those should be enough information to figure out the puzzles.
Of course, your call on whether it's worth all that measurement and analysis. I have fun figuring out puzzles, so tend to ask for lots of information.
I'm still hoping someone else can comment on the fast rise and slow fall of your antenna waveform.
Post by: babass on December 10, 2019, 11:24:50 PM
Anyway I connected 1 channel at the output of the 4069, and 1 channel acting like an antenna.
I tried to adjust the phase by replacing R1 by a potentiometer, it won't change anything, even with a 100k pot, and it ended up smoking. Phase won't change and emited sound isn't louder at all. I'll try changing the capacitor value maybe, if I choose a capacitor with similar impedance of the current transformer one maybe it will work better.
Post by: davekni on December 11, 2019, 05:38:17 AM
I'm loosing track of which resistor is R1 and which is R2. Do you have a second resistor between clamp diodes and 4069 input? Or is there just a single resistor (with series capacitor) from CT output to the clamp diodes? Is it that single resistor you are calling R1? (As I mentioned previously, the second clamp-diode to 4069 input resistor isn't critical.)
I'd mistakenly thought your scope had only one channel. With two, the external trigger isn't needed. Keep one probe on the 4069 output and trigger on that channel. Move the other probe from one point to the next, taking images at each place. (Adjust vertical as needed.) For these measurements, I'd suggest going back to the normal 1k for the CT output resistor. Start with the antenna for the second channel (as in your last image, but repeat just to make sure the setup is stable, no stray POT leads for the resistor now). Then move to the CT output, then after the 0.1uF blocking capacitor, then to after the 1k resistor (scope at the clamp diode junction), then to UCC27425 pin 2 or 4, then to UCC27425 outputs (pins pins 5 and 7, each separately, to make sure there's no issues with the chip), then to the node between the blocking capacitor and gate-drive transformer primary. If you are using your 120V supply (and not line power directly), make sure the negative supply output is grounded and move to the lower two H-Bridge gates (one at a time), then to the H-Bridge outputs (should be 0 to 120V mostly-square waves), and finally to the node between that output blocking cap and the Tesla primary coil. That's a lot of images, but it will almost certainly explain the entire phase-shift situation, where the delays are unexpectedly large (or blocking caps too small or whatever). Again, your call on whether it's worth making all those measurements.
Post by: babass on January 28, 2020, 10:02:54 PM
Anyway, since the beginning as my resonnant frequency was pretty low (220kHz) I didn't put a capacitor on the top of my tesla coil. But I tried to add one "just too see" and... it works much MUCH better. I think the even lower resonnant frequency makes the phase between the current transformer and the driver circuit even lower.
Also I need to add a DC blocking capacitor to avoid excess heating from my IGBTs. I apparently need a MKP capacitor that would act as a dead short at the resonnant frequency. I found the one in the picture. Will this one do the trick ?
Post by: davekni on January 29, 2020, 06:16:56 AM
Great to hear that it's working better!
Post by: Mads Barnkob on January 30, 2020, 10:04:57 AM
Also I need to add a DC blocking capacitor to avoid excess heating from my IGBTs. I apparently need a MKP capacitor that would act as a dead short at the resonnant frequency. I found the one in the picture. Will this one do the trick ?
Scroll down to DC blocking capacitor and I have shown how to calculate if its suitable: http://kaizerpowerelectronics.dk/tesla-coils/sstc-design-guide/
Post by: babass on January 31, 2020, 01:45:27 AM
Also, I'm now using PCBs for my circuits. And it seems to work even better now. I stuck a little heatsink on the driver ICs. I probed the signal between gate and emitter of an IGBT, and the signal is nearly perfect. No oscillation.
For the freewheeling diodes I'm using 4 DSEI30-06A, rated 600V, 37A, 35ns. I think that's fast enough. But they are heating a little (not a lot, just getting warm a little) is it normal ?
Post by: davekni on January 31, 2020, 06:04:43 AM
Being an SSTC, there's usually some H-Bridge current at the switching time. The diodes across the IGBTs that are turning-on will conduct for a short time until the current reverses. If the phasing is off, that could increase diode power, either by higher forward current for longer, or by H-Bridge switching after the current zero-crossing (leading to diode reverse-recovery losses). As long as the IGBTs aren't overheating, I'd be inclined not to worry about a little diode warming.
Of course, scope plots can reveal more of the actual situation. Is that one scope image of the gate-drive signal on the driver board or at one of the IGBT gates? 12V is a bit on the low side for driving IGBT gates. Current ratings are given at 15Vge. However, the typical spec's for your IXGH40N60A parts appear to work fine at 12V.
Post by: Mads Barnkob on January 31, 2020, 09:42:22 AM
So some heating on a much smaller external TO-247 device should be expected. You could add a small passive heat sink to it.
Post by: prabhatkumar on March 21, 2020, 04:23:05 PM