Gate drivers ICS Dying( UCC37322 and UCC37321 )

[prabhatkumar]

Hey !! i tried making the sstc 3 from the kaizer electronic site
(http://kaizerpowerelectronics.dk/tesla-coils/kaizer-sstc-iii/). This is my first post here (and also the first post on any electronics forum, so please excuse my faults here ).
I made the pcb from the express pcb file you provided by conventional toner transfer method at home. The pcb is working and I verified with a scope without the bridge supply on ( Only 12 volts was fed for the oscillatory part ( with drivers ics, gdt and the mosfet in place), only thing i didn’t connect the primary wires to the tesla coil . I checked it by running a slayer exciter on the same tesla coil to be used with the half bridge later with antenna in close vicinity ( 3-4 cm from the coil ). The wave forms were as expected just with a slight damping which could be fixed by reducing the gate limiting resistor( i used 5.6 ohm originally ) . But as soon as i connect the supply to the bridge( just 12 volts to check) , bang !! my driver circuity went in the current limiting mode and immediately the top ucc became hotter than normal. ON checking everything i realized this is not the only victim of this crash but also my 74hc14 inverter went bad i.e I dont have any pulses at the base of the driver ics. Now believe me i tried to build a full bridge with ir2110 before and burnt like 10 ir2110 before with the same issue, as soon as i connect the supply i goes bang !!( at least the ir2110 worked till the supply voltage was raised till 25 volts dc).
So please help me out as i dont have the money to invest on lot of ucc ics they are very expensive and also not available where i live ( I have to order them from mouser).
PS: Is it because the the primary was not connected anywhere or somehting else which i am not aware of . Also the the entire electrncs drive circuitary is connected to a 12 volt lab bench power supply (with current limiting on . I restricted the cureent to maybe 500mA ). And also the circuity was consuming 200 mA with no supply on the bridge side . The drivers were oscillating as they should with feedback from the slayer exciter.

EDIT: For those of you all who dont have the time to scroll the 3 pages of this post, I will tell what fixed the issue. The thing was i powering the tesla coil( SLAYER EXCITER USED HERE FOR FEEDBACK PURPOSED ONLY FOR CHECKING THE DRIVER) and my driver circuit
from a single bench power supply with 3 independent(isolated outputs. the power supply was hy3005-3 or similar ones sold there). As noob I am with electronics, I thought that it would work out fine, but as pointed in this post below by ritaismyconscience, the high voltage noise got back into the supply though the wires and interfered with other channels also and there must be huge voltage spikes which could have blown the ucc chips and the schmitt trigger. But then  I the powered the slayer exciter and the driver circuit with simple old iron transformers laying around and then it worked out fine. Also go through this post as there are many other things suggested below like adding resistor to the outputs of the gdt which is better than no resistor of course etc. )

[ritaismyconscience]

Could you send a few pictures?

[prabhatkumar]

Yes sure !! The only change I made was that replaced the 74 hc14 with cd40106
The two hanging wires are the wires to be connected to the primary

[prabhatkumar]

I can't upload the waveforms at the gates of the mosfet since too bad , the mosfet drivers arent working anymore after the aforementioned problem I encountered.

[johnf]

What sort of GDT core have you used??
it might be a powdered iron type that will cause the drivers chips to supply too much current

[prabhatkumar]

I don't know the exact type but then I guess its a correct one because the waveforms looked pretty decent at the gate without the bridge supply connected . I also tested the gdt with a function generator on hand and observed the waveforms which also looked acceptable .
 And as I said that this is my second attempt on a building a bridge in a previous attempt I built a ir2110 full bridge without a gdt which also behaved in a similar fashion i.e as soon as I connected the supply voltage for the bridge the drivers failed . And surprisingly none of the mosfets failed ( maybe because I had current limiting on my bench power supply )
I am wondering if it has to do something with the current limiting of my bench power supply for gate driver circuitry because maybe when the gate driver need peak current for switching they just can't get it as my supply goes into current limiting mode.
And maybe also that the primary is not connected so the mosfets are basically operating in open circuit .Also I cannot verify this theories without a concrete solution given by the people replying here as I don't have a lot of ucc ICS to spare( only 1 one more pair to go . And thanks for replying !
 PS: LINK FOR THE GDT I USED https://m.aliexpress.com/item/32861649468.html?spm=a2g0n.orderlist-amp.item.32861649468&aff_trace_key=743794a95edc4ceb95f68af9d7a029e3-1584258913329-02359-_etFMgO&aff_platform=msite&m_page_id=6622amp-EQHr2YJdaiHjqhUjjScjvA1584555238490&browser_id=7dd3e04ea5b7412a86b5466a7ba2fc52&is_c=N

[johnf]

I do not see any gate protection on your mosfety gates ie 25 volt transil or zener diodes gate to source.
As you do not know your core parameters you might need to put a resistor in series with your white wire to limit current into your GDT your ucc chips will haqve a max current rating of a few amps select a resistor to limit this current to this max ie if the ucc chip can do 9 amps then a 1.2 ohm series resistor to your GDT you can then measure across this resistor with your scope as long as your circuit is floating away from earth to see how much current is going into your GDT. Ie if you are getting significant number of volt pulses across the resistor then your core has insufficient ui and you are allowing your driver chips to supply past their rating. The scoping of this point will also let you know if the driver chips are behaving ie not high frequency oscillating which does destroy them.
do you have the 1uf ceramic and 10uf electro decoupling caps as per the chip datasheet installed

[prabhatkumar]

Thanks for the reply !! I haven't placed the TV's diodes because am just checking if the circuit works for some extent so my voltage is only 12 volts.
I have 1uF ceramic caps installed just beside the ic but no 10uF as is it not there in the kaizer sstc 3 schematic . Though I will add them on my own now to check .
So should I do this test without the power supply to the bridge or with the supply connected to the bridge . Because I think putting no power supply on the bridge side makes no difference as long as the signals at the gates are correct ( basically I don't want to risk another pair of ucc chips ) .
And yes I have put tape around the earth wire of my scope ( I know it's not a good practice but it's only for temporary purpose ) so I will try to take a reading at the input of the gdt .I will upload pics of the waveforms once done.
Also I as said in the previous message this is not the first time I am encountering a similar problem . Because last time when I build a full bridge out of ir2110 in a failed attempt , it had exactly the same problem. But then I never used the gdt . The output of ir2110 was connected directly to the respective mosfets . So could it be something else which we are missing ? Btw what made you select a 1.2 ohm , anything specific ?

[ritaismyconscience]

You're using the same power for both the driver and the bridge right? I think HV spikes are getting into your 12V supply, which destroys all the chips.

Consider using two different power supplies. Also definitely add more decoupling capacitors as johnf mentioned.

[prabhatkumar]

I have dual bench power supply( actually triple with a fixed 5v output on the third ) and all the supplies are isolated from one another so I doubt that will be the cause but I am not sure I am still a newbie so no idea of RF stuff .
Also regarding the high voltage rf created a , I have Slayer exciter running beside the driver so that the antenna could pickup signals and the driver circuitry could begin oscillating which it does properly . It's only when I apply power to the bridge side , the drivers fail.
Also my ucc together take around 200 mA doing the switching job without the bridge supply connected. Is that normal or something wrong here ?

[johnf]

12 volts applied to driver chip through a 1.2 ohm resistor will give a max of 10 amps into a dead short negelcting losses in the switch outputs so this is very close to the 9 amps that the driver chip can provide you will need the resistor to be a few watts to allow a short time to do the measurement

[prabhatkumar]

Ok thanks I will try . But is there something else which you all can think of so that I try that also ?

[ritaismyconscience]

Also if you have an oscilloscope, you could check power rails to make sure no voltage spikes are occuring.

[prabhatkumar]

Thanks for the replies everyone !! But all that means is that I will have to sacrifice another pair of ucc chips now
And regarding the power rails spikes , I think I can't take measurement of that when taking the measurement across the resistor at the put of the ucc chips ( since then they both have different
grounds and I don't have a diffrential probe ) .

[prabhatkumar]

Could I just replace ucc37322 with tc4420 and ucc37321 with ds0026??
I know the pinouts are different but then I will take of that . So will it matter that the ICS are different from another ?

[John123]

I think it might need freewheeling diodes on the output, not sure if the IC includes them (you could measure them with a diode checker). Maybe they do have built in diodes but they're not enough for the freewheeling current of your gate drive transformer, who knows.

I've read online that the tc44xx series prefers schottky diodes here or else latchup could still occur, but for the ucc3732x series I've personally had no issues with the UF400x series (UF4004 and below have lower forward drop and less recovery time).

Not saying they can't work without them but its usually standard practice to include some kind of freewheeling path for this kind of transformer drive.


I think I used 10 ohm gate resistors for IRFP260s (that was for flyback frequencies of 40-50khz, the switching rise and fall times should be a small fraction of the total "on" period, gets more important at high frequencies and powers) with 18v zener diodes on the gate, the gate resistor needs to limit the peak currents to less than the max rated of the gate driver chip at the supply voltage. The back to back zeners should be chosen so the worst case they shouldn't allow the gates to reach 20v, but not too low that they start conducting and clipping under normal operation (in practice that's 15-18v imo).

Also the higher power zener diodes 1.3w+ have a lot of junction capacitance, which can slow down switching speeds for SSTC use. This is what worked for me when I built my half bridge resonant flyback driver.

Not saying Mads circuit is wrong, but every build is different and things like parasitic inductance and capacitance differ between what you've built and what Mads built.

[prabhatkumar]

  I guess thats a nice point to have !! I will first upload the waveforms and voltage across the 1.5 ohm resistor as told above and then I will proceed to the modifications which John123 mentioned. So basically I will not connect the supply to the bridge now just to do the above mentioned tests and then verify with you all . Also I am going to use 20 volt TVS diodes instead of the zener for clamping the voltage at the gates. Thanks !!

[John123]

If you've got another power supply spare/battery you can temporarily energize the bridge using it, use a dummy resistor in place of the tesla coil primary. Whilst that doesn't really stress the bridge it'll show you the effect millar capacitance is having on the gate drive waveforms.

Those nice square waves when the main bridge is powered off quickly start looking different when the MOSFETs have got voltage on their drains, if you've got any old switching transformers laying around then connect that in place of the tesla coil primary (load one secondary winding) and see how the inductive loading changes the waveforms (again low voltage like 12v to the bridge, no mains at this point!).

For the antenna just tuck it around the core once, that's usually enough for the oscillator to start doing its thing.

This way you can poke around with a scope and not have to worry about something going bang or high voltages referenced to the mains.

In my experience the gate drive waveforms never look like the perfect square waves when the fets are switching actual power, the trick is to make sure the fets aren't turning off when they shouldn't (going below 10v g-s) or turning on when they shouldn't (rising above the gate-source threshold voltage, usually about 2-3v for most power fets).

The dead time between the two gate drive signals and recovery times of the diodes mod I posted can sometimes cause the gate drive transformer to "wonder".

I'm trying to find the page where someone detailed these problems, I'll post back if I manage to find them for you.

Edit: Found the page, it was actually written about a different gate driver IC but the same could still apply "UC3710 Latch-Up Issue - Update" http://electronsbefree.blogspot.com/p/electronics.html

More edits: Not sure about your core without looking at waveforms, but it's sometimes luck of draw when it comes to random cores. I've had both bad luck and good luck when it comes to picking randoms (including the hf choke types they slip a few turns around in equipment, often between circuit earth and mains earth and other noise suppression).

What I know to actually work for definate are Fair-Rite 77 5977006401 for flyback frequencies of a few khz up to 100khz, and fair-rite 78 for tesla coil use between 200khz to 500khz. Not saying you have to get these exact cores but something with the same specs as the 78 version for SSTC use.

Random cores can sometimes work, but in my case I had to wind a lot more turns to stop them from making the gate drive chips get hot.

And to your question mixing different gate drive chips, possibly. If one's significantly slower or faster than the other then it could cause delay and phase shift of one of the drive waveforms, this is bad as it might lead to shoot through (a brief point where both fets are on at the same time).

Fiddling with the gate drive resistor on one of the fets could perhaps solve this, often putting a diode in parallel to the gate resistor to enhance the turn off is used to trim switching times.

But speeding up switching comes with its own risks, again it's one of those things other people have had success with but in my case I've always needed to put a resistor in series with this diode or else turning the fet off too fast without some resistance effectively excites an LC tank circuit made up of the fet gate-source capacitance and the inductance of the gate drive transformer = voltage spikes and ringing.

Could you get a picture of the underside of the board? Maybe something could still be accidentally shorted if only the one gate drive transformer is getting hot. Are they definitely UCC37321 and UCC37322 and not the UCC37323 version which is like both built into one package.

Sorry for the wall of text but I went through similar frustration with gate drive transformers back in 2012.

[prabhatkumar]

Thanks John123 for the amazing site and other information your provided. And I can definitely understand the frsutration as it seems to be building up with me also now. The only problem is that in that site they use a 10uF tantalum which i dont have , Is it ok to use a 10uF electrolytic ??
Ok so here I will post the waveform of the various points in the circuit. I am doing this without the power supply connected to the bridge. Also my setup for now is as follows. I have just used a old 9 volt transformer for powering the driver circuit instead of the bench power supply( the slayer exciter is powered by the bench power supply). The voltage at the input after connecting the driver circuit is 12.1 V. I have added the 18 volt TVS diodes(1.5KE18CA) (I still dont  know why the gate voltage reaches 20 volt peak to peak, maybe because no supply and load is there for the current to flow.)
I have also changed the bridge capacitor to a value closer to what i will use actually( I have used a 400 V 150uF) , which I will upgrade later.
Also I have used 1 ohm resistor in series with a 0.47 ohm resistor which should be give around 1.5 ohms total resistance. I also have added another film capacitor of 100nF at the power rails of the UCC , and now I have to say that the ics are colder than before. The film caps definitely perform superior to ceramic ones here.
I have attached the oscilloscope probes to various points to get the measurements , so hopefully you all could point out the problem going in there.I have also attached the pictures of the bottom side of the board and the top also.
The wave forms
1)interrupter output
2)gdt current ( voltage across 1.47 ohm resistor)
3)gate signals( without the tvs diodes here )
4)gate with gdt
5)power line (12volt ripple)
6) Bottom side
7) Top side
Current setup(Quite messy)

[John123]

Is it me or does the third image have less voltage amplitude on the yellow waveform than the green one? Are those both gates at the same time?

Tantalum for decoupling the gate drive ICs? You can substitute it with a low ESR electrolytic of a few hundred uF and above along with some kind of smaller film or ceramic capacitor in parallel, its job is to supply those fast peak currents, without it the supply lines parasitic inductance can cause voltage spikes across the supply pins of the chips.

I decoupled the hell out of my gate drive chips with a combination of spare capacitors I had laying around. I used ceramic, electrolytic (220uF low ESR ones from an old switched mode wall wart secondary side) and film, basically any I could find in the parts box lol. No need to go out and buy new parts, just use what you've got.

Also I remember a forum post on 4hv saying that certain caps don't often work well as the DC blocking capacitor in series with the gate drive transformer primary, something about the capacitance varying with frequency and voltage applied. I used a film capacitor for this, it was 2.2uF pulled from an ATX PSU, for tesla coil use the smaller value is ok but try adding one in parallel and see if it improves the gate switching speeds.

8.7v into 1.47 ohm = 5.918 Amps peak, what does this look like with the regular 12-15v or whatever you're using is supplied? As without these current sense resistors in place and more voltage this could get close to exceeding the 9 amps peak rating. And the MOSFETs need at least 10v to fully turn on or else their on state resistance will be higher and they'll dissipate more power when switching a real load.

The peak to peak gate voltage is ok as its swinging from + 10v to -10v, assuming the ground lead on the scope is connected to the source pin of the fet. It's one of the advantages of a gate drive transformer with bipolar drive, the negative turn off bias stops the miller capacitance from turning the device back on when switching high dv/dt stuff. If peak to peak exceeds 40v then you've got problems, I'd probably change that to display peak instead as it's more useful here.

Apart from that they don't look too bad, a bit slow but no ringing. What's the slayer exciter doing? Is that part of this circuit?