I am not using any drive circuit into my test. I used a capacitor in series with GDT primary.
But the final design should include a driver circuit.
Gate Drive Transformer Issues
- Thread starter twenglish1
- Start date
I am not using any drive circuit into my test. I used a capacitor in series with GDT primary.
But the final design should include a driver circuit.
That's amazing!..I will follow the way you test a toroid GDT using only series capacitor in the primary..but i'm still looking for some magnet wires that is suitable to my toroid GDT..then sooner or later, I'll post the pictures here if everything is completely set-up and ready..
Thanks again for your support!
Ok..I will also find UTP cable..sorry if I can't do it right away because I was given an extra job this month that preoccupies me with so much time..
But hoping that I can make sooner or later..
Thanks MicrosiM for the reply.
MicrosiM, have you got a part number for those formers or are they salvaged? I didn't have much luck finding any for sale with the usual distributors except direct from the manufacturer with large minimum orders.
Also out of curiosity, what were the conditions on the GDT waveform you posted; i.e. is that an open secondary or hooked up to a gate; if it was hooked to a gate, was there a bulk voltage on the drain/collector and if there was, was there a load on the output. I'm just curious because it looks so clean
Also out of curiosity, what were the conditions on the GDT waveform you posted; i.e. is that an open secondary or hooked up to a gate; if it was hooked to a gate, was there a bulk voltage on the drain/collector and if there was, was there a load on the output. I'm just curious because it looks so clean
MicrosiM, have you got a part number for those formers or are they salvaged? I didn't have much luck finding any for sale with the usual distributors except direct from the manufacturer with large minimum orders.
Also out of curiosity, what were the conditions on the GDT waveform you posted; i.e. is that an open secondary or hooked up to a gate; if it was hooked to a gate, was there a bulk voltage on the drain/collector and if there was, was there a load on the output. I'm just curious because it looks so clean
Hi.
there was no mosfet or igbt connected.
its open secondary.
When I connect mosfer. Wave will never look the same
That never means you cannot obtain robust results with GDT.
Waves after connecting mosfet or igbt is totally not the same.
but i have just built new 1kw smps with almost basic parts.
i buy my transformer based on my specifications. I learned a lot. I can sell you some.
@KX36
I would like to add, that if I use a transistor at the output of the GDT, wave will look much better, but that's going to be a trade-off between performance, switching losses, and many other critical factors.
I might be able to show a few pictures of my latest SMPS. I built it becasue of this thread, I conducted a test with my new GDTs to show the wave previously shown, and I liked the results, I found my self building a rugged SMPS with a few parts!
Regards
I would like to add, that if I use a transistor at the output of the GDT, wave will look much better, but that's going to be a trade-off between performance, switching losses, and many other critical factors.
I might be able to show a few pictures of my latest SMPS. I built it becasue of this thread, I conducted a test with my new GDTs to show the wave previously shown, and I liked the results, I found my self building a rugged SMPS with a few parts!
Regards
I'm back again in this thread..I tested already the toroid I posted here (the big one)..it really got almost a PERFECT square wave
and the amazing thing about the testing is it is connected to IGBT's gate and source..so it means the secondary is loaded..
and the surprising thing is It can burn 10 ohms 1w resistor without distorting the waveshape..see the pics below..
Thanks a lot MicrosiM...I have still some questions..Is it ok that the peak voltage of wave without 10 ohms load at secondary is 15v.
then with the resistor connected the peak becomes 10v only?
there is drop of 5v but it can burn the resistor coz it is 1W only..
Regards,
demykiko
you can adjust the turns ration of your GDT, and 15V is normal, and you have to try best voltage for your Mosfet/IGBT
Enjoy!
Thanks MicrosiM!..til the next test updates if time is enough!
A bit late to post this, but all my electronics stuff has been literally under 4 months worth of paperwork etc.
Here's the gate drive transformer I did for my last project. It's an Epcos B64290L44X830 (13.6mm diameter N30 material) ferrite ring with 20:40:40 turns on it primary inductance 1mH. It was designed for around 100kHz operation on this 50W 2 switch forward converter which actually runs at 125kHz. The enamelled copper wire is 0.315mm diameter, that's not too important, I just used what I had used in the main transformer (incidentally, that's an Epcos EF25 in N87). It's wound trifilar/bifilar as recommended in the PDF I have posted around before about GDTs. I know ECW straight into the board is not ideal but this is a prototype only, I don't have any formers for small vertically mounted ferrite rings and I intend at some point to gunk it down.
There's no reason to be salvaging wires from cat 5 cables and ferrite rings from CFLs etc not knowing how well they'll perform, this one costs under £0.20 and has a good datasheet.
The input is 380V dc from a PFC if you were wondering why you can't see a rectifier etc. It's one module in a bigger supply, but so far it's the only one I've gotten around to building up since it provides all the aux rails for primary and seconary. Always too busy...
Here's the gate drive transformer I did for my last project. It's an Epcos B64290L44X830 (13.6mm diameter N30 material) ferrite ring with 20:40:40 turns on it primary inductance 1mH. It was designed for around 100kHz operation on this 50W 2 switch forward converter which actually runs at 125kHz. The enamelled copper wire is 0.315mm diameter, that's not too important, I just used what I had used in the main transformer (incidentally, that's an Epcos EF25 in N87). It's wound trifilar/bifilar as recommended in the PDF I have posted around before about GDTs. I know ECW straight into the board is not ideal but this is a prototype only, I don't have any formers for small vertically mounted ferrite rings and I intend at some point to gunk it down.
There's no reason to be salvaging wires from cat 5 cables and ferrite rings from CFLs etc not knowing how well they'll perform, this one costs under £0.20 and has a good datasheet.
The input is 380V dc from a PFC if you were wondering why you can't see a rectifier etc. It's one module in a bigger supply, but so far it's the only one I've gotten around to building up since it provides all the aux rails for primary and seconary. Always too busy...
lynxlynx
A rumbling soft motor
full bridge shoot-through
I've had similar issue: when I connected GDT directly to SG3525 the signal was completely distorted. Then I added a BD139/BD140 driver.
Now I'm getting a massive shoot-through despite normal gate waves. I get 20mA at 350V with 4xIRFP450 without any load attached to output of full bridge.
I get this waveform (1.jpg), I use this circuit (2.jpg) and this is a ferrite core currently only I have (3.jpg). I guess if this little spike at end of each wave is to blame. Is my gate waveform is normal or damaged somehow? I also simulated large dead time with a huge cap from 8 pin to ground of SG3525 and indeed got a big dead time, but shoot-through condition starts to appear when IC starts oscillating with any duty cycle.
Ferrite core is from a input filter but it nicely burns 10R when I test it on separate 12V full bridge.
I understand that shoot-through can be a result of leakage inductance (according to http://thedatastream.4hv.org/gdt_crosscond.htm), but the photographed waveform is actually from 150R connected to outputs of driver, and without any load this spike persists on SG3525 pins. I used the layout in past and did not had any issues with it, but only on push-pull circuits.
Can someone help?
I've had similar issue: when I connected GDT directly to SG3525 the signal was completely distorted. Then I added a BD139/BD140 driver.
Now I'm getting a massive shoot-through despite normal gate waves. I get 20mA at 350V with 4xIRFP450 without any load attached to output of full bridge.
I get this waveform (1.jpg), I use this circuit (2.jpg) and this is a ferrite core currently only I have (3.jpg). I guess if this little spike at end of each wave is to blame. Is my gate waveform is normal or damaged somehow? I also simulated large dead time with a huge cap from 8 pin to ground of SG3525 and indeed got a big dead time, but shoot-through condition starts to appear when IC starts oscillating with any duty cycle.
Ferrite core is from a input filter but it nicely burns 10R when I test it on separate 12V full bridge.
I understand that shoot-through can be a result of leakage inductance (according to http://thedatastream.4hv.org/gdt_crosscond.htm), but the photographed waveform is actually from 150R connected to outputs of driver, and without any load this spike persists on SG3525 pins. I used the layout in past and did not had any issues with it, but only on push-pull circuits.
Can someone help?
Attachments
-
1.jpg -
2.jpg -
3.jpg
Jagd.Panther
New member
Hey Lynxlynx,
What did you probe to get this pic? Does your scope have a grid?
Anyway let's try to understand if you should see some quiescent current even if there is no load.
So you have four mosfets, each has Coss (output capacitance) about 330pf according to the data sheet. During one complete switching cycle all mosfets are switched on and off, so four Coss capacitor are charged to 350v and discharged. The charge of a cap is C U² / 2, or 4*330e-12*350*350/2 = 80e-6 = 80 uJ. If you have switching frequency set to 50 kHz (oscillator of sg3525 should be running at 100 kHz), the bridge would draw 80uJ 50000 times a second, or 80 uJ*50 kHz = 80e-6*50e3= 4.0 Watts. 4 watts at 350w is about 10 mA (the current will be actually higher if we take into account stray capacitance).
In order to be 100% sure that your quiescent current is OK you should measure rise/fall times at the gate of mosfet & check if your have enough dead time.
What did you probe to get this pic? Does your scope have a grid?
Anyway let's try to understand if you should see some quiescent current even if there is no load.
So you have four mosfets, each has Coss (output capacitance) about 330pf according to the data sheet. During one complete switching cycle all mosfets are switched on and off, so four Coss capacitor are charged to 350v and discharged. The charge of a cap is C U² / 2, or 4*330e-12*350*350/2 = 80e-6 = 80 uJ. If you have switching frequency set to 50 kHz (oscillator of sg3525 should be running at 100 kHz), the bridge would draw 80uJ 50000 times a second, or 80 uJ*50 kHz = 80e-6*50e3= 4.0 Watts. 4 watts at 350w is about 10 mA (the current will be actually higher if we take into account stray capacitance).
In order to be 100% sure that your quiescent current is OK you should measure rise/fall times at the gate of mosfet & check if your have enough dead time.
Jagd.Panther
New member
With regard to GDT drive, IIRC SG3525 can source/sink around 500 mA (edit: indeed, ±500 mA according to Fairchild's data sheet), so you can easily reduce resistors between SG3525 and emitter follower to 24-50 ohms. That would give you better drive. With your current set up and 330 ohm Resistors the maximum output current of the emitter follower is about (12Vcc - 1V rfor Vbe)/330 * Beta = ca. 3A for a Beta of 100. That's a peak drive current when the output is shorted. If we consider gate threshold voltage the current will be lower (about 2A). Since you drive four transistors, the effective peak drive current per transistor would be one quarter of that. If it's enough or not depends on the switching freq and desirable di/dt and switching losses.
Last edited:
Jagd.Panther
New member
A cap on 8th pin is for soft start, not for deadtime control! You should increase "discharge" resistor to increase dead time (NB doing that will lower oscillator frequency). Try higher value, you should get something like on the attached pic (probe on one of the winding of the GDT, the dead timetime can be seen on the centerline of thewave form).
Attachments
-
gdt-deadtime.jpg
lynxlynx
A rumbling soft motor
Thanks for your analysis Jagd.Panther! I indeed forgot about parasitic elements inside mosfets and around them (maybe parasitic inductances, but I've seen no ringing)
I also built a simulator for it in ltspice and indeed I see shoot-through even under ideal virtual conditions. So parasitic elements play a role. Good news I have enough heatsinks to play with if it will become a problem.
The scope is very old Soviet, and inside it's just a tube and lamps, it even has no grid indeed. I recently got another Soviet one with grid, but it needs VERY huge cleanup and it weights 25kg. Oh. So measuring rise/fall now is a problem.
I indeed calculated everything under 50kHz. So 20mA-10mA leaves me parasitic 10mA or about that (if your mentioned stray capacitances are involved too) which I need to eliminate. Good news I've eliminated spikes with simple RC filter (47Ohm-10nF) placed on outputs of IC now I need to rewind transformer.
I did another tests: I've replaced IRFP450 with IRF730 just as precaution not to burn 450's. Then I driven transformer with external circuit: http://lynxlynx.tk/log/files/schema.png (with C7, C8 = 510pF R2 set to 25k and full duty cycle) and tried shoot-through. And now I've got only 4mA. I will try the same with IRFP450 when I will wind the transformer again.
As for driver resistors then I need 2W ones and dissipate additional 4W+. That's currently a problem... but I will try.
I also built a simulator for it in ltspice and indeed I see shoot-through even under ideal virtual conditions. So parasitic elements play a role. Good news I have enough heatsinks to play with if it will become a problem.
The scope is very old Soviet, and inside it's just a tube and lamps, it even has no grid indeed. I recently got another Soviet one with grid, but it needs VERY huge cleanup and it weights 25kg. Oh. So measuring rise/fall now is a problem.
I indeed calculated everything under 50kHz. So 20mA-10mA leaves me parasitic 10mA or about that (if your mentioned stray capacitances are involved too) which I need to eliminate. Good news I've eliminated spikes with simple RC filter (47Ohm-10nF) placed on outputs of IC now I need to rewind transformer.
I did another tests: I've replaced IRFP450 with IRF730 just as precaution not to burn 450's. Then I driven transformer with external circuit: http://lynxlynx.tk/log/files/schema.png (with C7, C8 = 510pF R2 set to 25k and full duty cycle) and tried shoot-through. And now I've got only 4mA. I will try the same with IRFP450 when I will wind the transformer again.
As for driver resistors then I need 2W ones and dissipate additional 4W+. That's currently a problem... but I will try.
lynxlynx
A rumbling soft motor
Oh that's interesting. Datasheets often give a graphic where you see max. of 500Ohm and that's does not give a full control over dead time but I ever never experimented with that. I will try with trimmer. I need to build another one full bridge just as replacement for that 2xIR2153 thing