High Voltage Forum
Tesla coils => Solid State Tesla Coils (SSTC) => Topic started by: Thunderstruck on March 16, 2019, 06:32:36 AM
-
While I'm waiting for some parts to arrive I decided to try and master the dark art of GDT winding. Got some advice on a correct core selection, watched a few videos on the topic and decided to have a go at it.
Well, I wound 5 different cores, and I did not get what I wanted.
I used cores with high AL, Cat5 Network Cable Wire and on one occasion Magnet Wire and got very similar results every time.
I am trying to achieve a good waveform at about 65 - 70kHz
Image 1 - Core specification ( AL5400, T35, 27mm OD )
Image 2 - First attempt
Image 3 - Waveform achieved
Image 4 - Richie Burnett's advice on corrective action for that waveform
Image 5 - He said more turns - More turns it is then ;D
Image 6 - Resulting waveform - Still sloping, and as you can see there is no more room for wire turns
Other attempts had pretty much the same results
I am quite sure that I am not doing something right here - or RS Components sent me wrong cores.
-
You did also choose cores with almost identical properties, its not even a 1000 permeability in difference, so you will only see the differences from number of turns and how tight coupled the wires are to eachother/the core.
What kind of test jig did you use? A driver IC --> GDT --> MOSFET gates?
Did you see this practical design guide / experiment? http://thedatastream.4hv.org/gdt_practical.htm
-
The one thing missing from Burnett's advice: reduce the driver impedance.
Was that testing with a function generator (50 ohms)? Looks about right for it.
Tim
-
You did also choose cores with almost identical properties, its not even a 1000 permeability in difference, so you will only see the differences from number of turns and how tight coupled the wires are to eachother/the core.
What kind of test jig did you use? A driver IC --> GDT --> MOSFET gates?
Did you see this practical design guide / experiment? http://thedatastream.4hv.org/gdt_practical.htm
Mads, in both cases it was the same core. I did say 5 different cores in my original post, what I wanted to say is 5 different ways ( twisted, straight, trifillar, etc. ). I was expecting to see at least a half decent waveform because it is a high permeability core, especially after watching a few videos.
Test signal was supplied from the FG directly, i do not have a driver circuit to test the GDT with.
I suspected that my FG is not really capable of driving the GDT...
I read that guide a few times, only thing missing is screenshots of waveforms after each attempt.
The one thing missing from Burnett's advice: reduce the driver impedance.
Was that testing with a function generator (50 ohms)? Looks about right for it.
Tim
Tim, yes I tested the transformer with my FG which is a little bit ancient :)
-
I dont know if this will help but thare a cuple of articals
http://ferroxcube.home.pl/news/gate%20drive%20trafo.pdf
https://www.coilcraft.com/edu/gate_drive_transformer.cfm
http://seansoleyman.com/solid-state-tesla-coil/
your outputs have good rise time but it looks like they are saturating because the waveform is saging after 7us. usaly your signal generator has a 50 ohm inpedance and may not be able to drive it enough. build a gate driver circuit on a proto board using a mosfet diriving ic or the output of a universal driver. a biger core might help and or using more stacked cores if you had some
-
If you apply a voltage step to an unloaded (and perfect) transformer, the primary will look like an inductance. It will then respond with a linear current rise, which will induce the desired voltage step in the secondary. In a non perfect situation, you will have a primary winding resistance and also a voltage source, which might sag under the loading current. Both will cause a decrease of the linear current slope and then to a drop of secondary voltage. If the core saturates, the primary current will shoot up, causing more voltage loss without increasing the magnetic field, which would keep the output voltage stable. So this will aggravate the resistance effects.
So first, you have to make sure, that the applied primary voltage doesn't drop. Then you have to check, whether the magnetizing current, i.e. the current supplied to the unloaded transformer isn't excessively large. If so, increase the number of turns or look for a higher u core. A too large magnetizing current can also be caused by a saturating core. Also keep in mind, that the above mentioned resistive losses will increase as you load the secondary.
-
Thank you for all the comments, a lot of useful information there. It is definitely not a case of just winding some wire on a core.
Clinical definition of insanity is tepeating the same thing expecting a differnet outcome. :o
So I decided to have another attempt ;D
This time I got 3 different cores:
125-3320 Al 13500 T38
212-0982 Al 5750 N30
212-0976 Al 7000 N30
I managed to wind 11 turns through the T38 core and got exactly the same result as before.
Logical conclusion is that the way I’m driving the GDT’s is inadequate ( somebody already mentioned this )
I will have to build a test rig, fortunately I got a few PCB’s for profdc9’s SSTC, I reckon that will do just fine.
Strangely enough, yesterday I wound a GDT ( not really caring about how it will perform ) just to test my UD board and got a perfect waveform, go figure....
-
I built Profdc9's Half Bridge SSTC driver and connected the GDT's I built. This time I'm getting half decent results, waveform actually looks ok.
As you are probably aware, this is all quite new to me, so I am quite excited achieving things that many of you achieved years ago and consider them ordinary. But, it's all about fun, right ?
Screenshots below show output of two different GDT's I made.
First image shows waveform from a GDT I previously tried running from a FG only with terrible results
Second image is a waveform from a GDT made with 11 turns of CAT5 cable - all four pairs of wires. Signal from 2 only shown.
That waveform confused me a bit until I realized that there is "deadtime" in the signal. Only recently I watched a video on YouTube on that topic - quite glad I recognized it in the waveform. ;D
-
Yep thats it, nice work, do you have a link for Profdc9 scematic and witch core did you use if I may ask. If you add a little load resistance in paralell with your transformer output you might be able to get rid of those spikes at each transition
-
Yep thats it, nice work, do you have a link for Profdc9 scematic and witch core did you use if I may ask. If you add a little load resistance in paralell with your transformer output you might be able to get rid of those spikes at each transition
You can find Prof’s schematic, and a lot of other useful stuff here:
https://github.com/profdc9/DRSSTC-PCB-Pack
It is under Half Bridge SSTC
Core i used in the first screenshot is EPCOS, Al 5400 T35 material, 27mm OD ( ball of wire on the photo below ;D )
Second core is also EPCOS, Al 13500 T38 36mm OD ( connected to the driver on the photo below )
-
That waveform confused me a bit until I realized that there is "deadtime" in the signal. Only recently I watched a video on YouTube on that topic - quite glad I recognized it in the waveform. ;D
You have about 1.8us of deadtime in both shoots and i can't figure it out how...
The Half Bridge SSTC schematic doesn't have any built in deadtime and the TC4420 have delay times less than 100ns and matched Rise and Fall times of 25ns or are you using different IC's?
-
That waveform confused me a bit until I realized that there is "deadtime" in the signal. Only recently I watched a video on YouTube on that topic - quite glad I recognized it in the waveform. ;D
You have about 1.8us of deadtime in both shoots and i can't figure it out how...
The Half Bridge SSTC schematic doesn't have any built in deadtime and the TC4420 have delay times less than 100ns and matched Rise and Fall times of 25ns or are you using different IC's?
I am using UCC37322 instead of TC4420, so that might be the reason.
I also wondered what causes the dead time ( well to be honest I would not know what does it even if there was a dedicated component )
What controls the dead time was going to be my next question.
I already had spare UCC’s, so I did not want to spend more on TC’s since they are interchangeable in this case.
-
I am using UCC37322 instead of TC4420, so that might be the reason.
I also wondered what causes the dead time ( well to be honest I would not know what does it even if there was a dedicated component )
What controls the dead time was going to be my next question.
UCC37322 is faster than TC4420...
Given that you got the same deadtime in both cases at different frequencies i'd say the propagation delay from an extra schmitt inverter on one of the UCC's is what produces the deadtime.
What is the part number of U6 (from schematic) that you are using?
-
Part number from Digikey is 296-3503-5-ND, CD40106BE.
From what I gather, deadtime is not a bad thing, right ? Gives time to the devices to switch off and on completely without running into each other ( going linear )
-
I don't think it should be a problem. The body diode of the one MOSFETs should go into conduction after the other MOSFET is turned off. It is a longer dead time than I have observed but it should probably be ok.
Note that the SSTC should be driven using the antenna feedback. The built-in oscillator is mostly to "jump start" the oscillation of the Tesla coil.
-
It might be a problem, that the middle voltage in your square wave can cause a situation in which the FET is not completely blocking. That will happen for both FETs at the same time.
-
your UCC37322 may have built in deadtimet to prevent lachup but I am surprised to see that it makes it through the core. 1.8us is a bit much but I don't think it is an issue. check it with your scope if you have dead time going into the transformer. If you were using bricks 1.8us would be fine. If you have fead back is it a SSTC or a DRSSTC
-
I looked on the output of my board. I get a 880 ns dead time. I am not sure why you are getting twice as much. Even this is kind of high for the switching time of the CD40106 with 15 volt supply. I should probably have a board revision and use a CD4027 and use both the Q and not Q outputs which would change at the same time. This would also ensure 50% duty cycle, but would add an extra part.
I don't think it's a problem to have zero voltage in the gate drive transformer. The overcurrent shutoff in fact does exactly this, where it forces both gate drivers low to turn off the MOSFETs.
I have a plot of the gate drive signal when the antenna port is drive by 200 kHz with a signal generator. The dead time is long which might lower power output however.
-
I have a solution to greatly reduce the dead time.
Replace D3 and D4 with 1N4148. They are 1N5819 in the design. I will update the schematic.
It should work much better, here's a screenshot. The dead time is now 320 ns.
Dan
-
Well ! This certainly took a different turn, I wanted to make a test circuit for GDT’s, and it resulted in a circuit change !
Dan, thank you very much for your help with this. More than likely I will use this driver in a TC build in the future so thanks for troubleshooting and making changes.
One thing that I am not clear with is jumper settings, so far I figured out AC/DC input jumper, but others are a bit harder to figure out from the schematic ( for me at least ).
I’d like to do a test with FG connected to the antenna input next.
-
@profdc9:
I don't think it's a problem to have zero voltage in the gate drive transformer.
I should have checked IGBT specs. They indeed turn off at 0 gate voltage. But I still have a little doubt: Think of a hard switching turn off situation. And suppose, there is a little internal inductance between emitter of die and package outside. The dropping collector-emitter current will cause a negative voltage jump at the emitter, which fakes a positive gate voltage. Therefore it might be an advantage to drive the gate to negative voltages as fast as possible. Just a guess.
-
Well you have to realize that there's a lot there in the PCB Pack and it could always use some improvement. But I wanted to try to create something so that all the circuits one needs could be in one place like Mads with his series of guides. One of the hardest parts about building a Tesla Coil is trying to figure out where the accurate information is, and this was a struggle for me getting started, so feedback like yours is a big help to try to make the situation better.
As for the dead time issue, here is my understanding of it. When driving an inductive load like a tesla coil primary, one achieves ZVS (zero voltage switching) for a half/full bridge by having a dead time between the turning off of the upper/lower transistor and turning on the lower/upper transistor. When both transistors are shut off, there is still current flowing from the inductor. The current is diverted from the transistor that is now shut off to the the body diode or reverse pack diode of the other transistor. Since the other transistor is now in forward conduction, its voltage drop is near zero. So you do get some energy loss from excessive dead time which is current X diode drop voltage X dead time duration. Now the other transistor is turned on at nearly zero voltage which results in a lower switching loss. The disadvantage of excessive dead time is you get a little more loss in the reverse diodes but the benefit of ZVS switching for the on condition and a low probability of shoot-through.
Dan
Well ! This certainly took a different turn, I wanted to make a test circuit for GDT’s, and it resulted in a circuit change !
Dan, thank you very much for your help with this. More than likely I will use this driver in a TC build in the future so thanks for troubleshooting and making changes.
One thing that I am not clear with is jumper settings, so far I figured out AC/DC input jumper, but others are a bit harder to figure out from the schematic ( for me at least ).
I’d like to do a test with FG connected to the antenna input next.
-
Well you have to realize that there's a lot there in the PCB Pack and it could always use some improvement. But I wanted to try to create something so that all the circuits one needs could be in one place like Mads with his series of guides. One of the hardest parts about building a Tesla Coil is trying to figure out where the accurate information is, and this was a struggle for me getting started, so feedback like yours is a big help to try to make the situation better.
As for the dead time issue, here is my understanding of it. When driving an inductive load like a tesla coil primary, one achieves ZVS (zero voltage switching) for a half/full bridge by having a dead time between the turning off of the upper/lower transistor and turning on the lower/upper transistor. When both transistors are shut off, there is still current flowing from the inductor. The current is diverted from the transistor that is now shut off to the the body diode or reverse pack diode of the other transistor. Since the other transistor is now in forward conduction, its voltage drop is near zero. So you do get some energy loss from excessive dead time which is current X diode drop voltage X dead time duration. Now the other transistor is turned on at nearly zero voltage which results in a lower switching loss. The disadvantage of excessive dead time is you get a little more loss in the reverse diodes but the benefit of ZVS switching for the on condition and a low probability of shoot-through.
Dan
[/quote]
To be honest, I stumbled on that issue, I was not even aware that is an issue, but I am glad that it lead to an improvement.
So far I built 2 circuits out of your github pack, interrupter and driver, both working well ( yet to program Attiny85 and try Midi function on the interrupter )
I also had Pcbway manufacture your half bridge and full bridge PCB's, so I might try to put one of them together as well.
Anyway, I did more tests on my driver board.
Image 1 shows waveform with dead time of 1.16us
I swapped UCC37322's with TC4420's to see what is going to happen, and as Dexter pointed out, TC's are slower therefore the problem got worse, as you can see in Image 2.
Not only did the dead time increase to 2.2us, the waveform itself has a weird spike in it.
Then I put the UCC's back and replaced D3 and D4 with 1N4148's, dead time reduced to 230ns - Image 3
I also added 270 Ohm resistor to get rid of the voltage spikes.
Image 4 shows waveform at 200kHz with FG connected to the antenna input
Image 5 shows all 4 GDT outputs running, only CH 1 has 270 Ohm resistor connected, rest show voltage spikes.
-
I have to confess I had an ulterior motive for designing the SSTC circuit this way. I hope to try to use to make an induction heater using the SSTC board.
Basically, use a feedback coil rather than antenna near the workpiece to drive it into resonance. So I designed it with some stuff that may not have been necessary. I did not really worry about the dead time but I thought I should look into it after you brought it up.
If you need even less dead time, you can swap R5 and R11 for 4.7k resistors. The limit is that the overcurrent circuit can only sink 6 mA and so if the resistors are too low resistance, the overcurrent protection does not work.
Dan
-
I have to confess I had an ulterior motive for designing the SSTC circuit this way. I hope to try to use to make an induction heater using the SSTC board.
Basically, use a feedback coil rather than antenna near the workpiece to drive it into resonance. So I designed it with some stuff that may not have been necessary. I did not really worry about the dead time but I thought I should look into it after you brought it up.
If you need even less dead time, you can swap R5 and R11 for 4.7k resistors. The limit is that the overcurrent circuit can only sink 6 mA and so if the resistors are too low resistance, the overcurrent protection does not work.
Dan
I have no idea how much dead time is enough for the correct operation of a SSTC, I suppose I’ll just have to build one and see what happens, then adjust accordingly.
Since you are looking for feedback, there is one more thing that I noticed when assembling the driver.
Two pots are very close to each other, which is not a problem if you are using 16mm pots or panel mounting them, but for some reason in Australia I can only get 17mm pots which are just a little bit too big if you mount them directly onto the PCB.
I had to remove some material from each of them to make them fit.
I am not sure if anyone else had a similar problem with slightly larger pots, but moving each pot only by 1mm in opposite directions would help.
Also, one of the pots will come very close to touching the opto-isolator if opto-isolator is mounted into the ic socket. Again, this is not a problem if pots are mounted onto a panel.
-
I moved the OC LED backwards so the potentiometers may be better separated, there should be plenty of space now.
Dan
-
profdc9 wrote:
As for the dead time issue, here is my understanding of it. When driving an inductive load like a tesla coil primary, one achieves ZVS (zero voltage switching) for a half/full bridge by having a dead time between the turning off of the upper/lower transistor and turning on the lower/upper transistor. When both transistors are shut off, there is still current flowing from the inductor. The current is diverted from the transistor that is now shut off to the the body diode or reverse pack diode of the other transistor. Since the other transistor is now in forward conduction, its voltage drop is near zero. So you do get some energy loss from excessive dead time which is current X diode drop voltage X dead time duration. Now the other transistor is turned on at nearly zero voltage which results in a lower switching loss. The disadvantage of excessive dead time is you get a little more loss in the reverse diodes but the benefit of ZVS switching for the on condition and a low probability of shoot-through.
Certainly a nice way to avoid shoot through. And, as you write, turn on is at 0 voltage for inductive loads. In a DRSSTC, though, the current might change polarity during dead time, so you won't switch on at 0 voltage. That will happen at comparatively low currents, so it probably doesn't matter much.
-
what is the input vs output waveform look like on the UCC's