GDT questions
..., Mon Aug 14 2006, 04:33AMI have started building a sstc (got first light 2 nights ago running off tlc555 going into a single irfp460 chopper into a 4 turn primary at about 100v in) and plan to use a full bridge--and thus need a gdt. Likewise I have a few questions I can't find decent answers
to. I should point out that I will be running a full bridge of irfp250's or prif460's, or possibly some hgtg40n60's or hgtg27n120's at 1mhz. I am feeding it with a tc4421/tc4422 in the to-220 package on the same heat sink as the rest of the transistors (cos they were getting pretty hot on the little heatsink I had them on driving just 2 fets).
1. How do you figure out how many turns you need on the core? I tried anywhere from 1 to 25 turns feeding a ifrp250 load with a 10r gate drive resistor and a pair of 10v zeners back to back on it and I never saw any ringing and the rise/fall times stayed constant.
2. Is there a resonable limit to how much power a core can handle? I have been using a small emi supression core (about .5" in diameter, .13" thick and .75" long) wound with 25 turns of 32awg tri-filar wire (very tightly twisted) into 2 10nf dummy loads (ceramic caps that got pretty hot at 1mhz) and the core got pretty hot (me thinks about 130ish)... I worry that it can't handle all 4 transistors.
3. How important are the rise/fall times on the output? Feeding an irfp250 they have a rise/fall time of about .25us to 10v, which seems a little long for 1mhz... What can I do to make it smaller? I tried shorting the gate resistor, but that only dropped it down to .2us...
4. Will magnet wire handle the voltage? I only plan to run off 170v into the bridge, so it seems like things should be ok there... would it be safe to try for 350v, 1kv?
Thanks!
Re: GDT questions
Steve Ward, Mon Aug 14 2006, 06:53AM
The same as any other transformer, using the material specs, core cross sectional area, and you can calculate the V*s/turn it will handle before saturating. But in general, 10-18 turns is about right for your average 1" diameter ferrite running at 100khz or upward and 30v p-p on the primary. You need less turns as the frequency increases.
If it gets hot, its likely either saturating, or you have a short that you arent aware of. Ive had this happen with magnet wire usually, where it shorts to the core, and the core gets pretty hot. 25 turns was likely far too many turns at 1mhz. You could probably go as low as 3-5 turns at that frequency, though a little more would be playing it safe. Its possible that the 32 awg wire was getting hot, and not necessarily the core? I usually use something around 24awg. Cat-5 networking wire works pretty well, you can untwist and twist it with a drill.
Thats pretty slow for that fet... i'd imagine 50nS is quite easy to achieve. Remember, your gate drivers have some resistance too. I think your 32awg wire and using so many turns might have been limiting things somewhat as well. Its also nice to know if you have the right core material too as this could potentially reduce the performance of your GDT.
I dont trust magnet wire anymore. Its too easy to scrape the insulation off on the ferrite core as you are winding it.
Steve Ward, Mon Aug 14 2006, 06:53AM
1. How do you figure out how many turns you need on the core? I tried anywhere from 1 to 25 turns
The same as any other transformer, using the material specs, core cross sectional area, and you can calculate the V*s/turn it will handle before saturating. But in general, 10-18 turns is about right for your average 1" diameter ferrite running at 100khz or upward and 30v p-p on the primary. You need less turns as the frequency increases.
2. Is there a resonable limit to how much power a core can handle? I have been using a small emi supression core (about .5" in diameter, .13" thick and .75" long) wound with 25 turns of 32awg tri-filar wire (very tightly twisted) into 2 10nf dummy loads (ceramic caps that got pretty hot at 1mhz) and the core got pretty hot (me thinks about 130ish)... I worry that it can't handle all 4 transistors.
If it gets hot, its likely either saturating, or you have a short that you arent aware of. Ive had this happen with magnet wire usually, where it shorts to the core, and the core gets pretty hot. 25 turns was likely far too many turns at 1mhz. You could probably go as low as 3-5 turns at that frequency, though a little more would be playing it safe. Its possible that the 32 awg wire was getting hot, and not necessarily the core? I usually use something around 24awg. Cat-5 networking wire works pretty well, you can untwist and twist it with a drill.
How important are the rise/fall times on the output? Feeding an irfp250 they have a rise/fall time of about .25us to 10v, which seems a little long for 1mhz... What can I do to make it smaller? I tried shorting the gate resistor, but that only dropped it down to .2us...
Thats pretty slow for that fet... i'd imagine 50nS is quite easy to achieve. Remember, your gate drivers have some resistance too. I think your 32awg wire and using so many turns might have been limiting things somewhat as well. Its also nice to know if you have the right core material too as this could potentially reduce the performance of your GDT.
4. Will magnet wire handle the voltage? I only plan to run off 170v into the bridge, so it seems like things should be ok there... would it be safe to try for 350v, 1kv?
I dont trust magnet wire anymore. Its too easy to scrape the insulation off on the ferrite core as you are winding it.
Re: GDT questions
EDY19, Mon Aug 14 2006, 02:14PM
Even though this doesn't have too much to do with the problem (maybe the slow rise times? )Have you tried the UCC versions of the chips? The seem to be favorable when you get into higher frequencies.
EDY19, Mon Aug 14 2006, 02:14PM
Even though this doesn't have too much to do with the problem (maybe the slow rise times? )Have you tried the UCC versions of the chips? The seem to be favorable when you get into higher frequencies.
Re: GDT questions
Steve Ward, Mon Aug 14 2006, 03:27PM
Well, they are both 9A pk drivers, but the UCC drivers do use a better technology, which they claim is sorta like having a 12A driver. Id expect performance to be pretty similar if it was limited by the driver chips, though.
Steve Ward, Mon Aug 14 2006, 03:27PM
Well, they are both 9A pk drivers, but the UCC drivers do use a better technology, which they claim is sorta like having a 12A driver. Id expect performance to be pretty similar if it was limited by the driver chips, though.
Re: GDT questions
..., Mon Aug 14 2006, 04:01PM
I have the ucc versions, but as of now I am using way too much current to use them... A full bridge of 250's at 1mhz isn't an easy load (me calculates about 3w just from discharging/charging the caps, then a few extra for the gate resistor and zeners which are dissipating several watts per fet...)
1. But how would you tell when it is saturating? I never saw any difference in the waveforms, or any ringing whatsoever, but the no load current of the driver did seem to level off around 5 turns, the number of turns and the corresponding no load current from the driver is listed here. No gdt connected 50ma; 0 turn 1800ma; 1 turn 780ma; 2 turn 400ma; 3 turn 270ma; 4 turn 220ma; 5 tun 210ma; and it sorta levels off around there. So 5 turns should be ok?
2. I know it was the core getting hot, and the wire was not shorted. I suppose it was the cure saturating, but will having less turns help solve that? I was avoiding cat5 because the window on the core is pretty small, but if I only need 5 turns I should ok...
3. The resistance of the windings is 0.8ohms (measured), and reducing the number of turns down as low as 1 didn't speed it up... The rise time at the input of the gdt is about 50ns from -12v up to 0v, but about 200ns to get all of the way up to 12v (the signal goes strait up to 0v, then at about 5 it arks over). The fall times are the opposite; it drops like a rock (about 10ns) to 0v, then slows down to get to -12v, about 200ns total.
4. I will try for cat5...
..., Mon Aug 14 2006, 04:01PM
I have the ucc versions, but as of now I am using way too much current to use them... A full bridge of 250's at 1mhz isn't an easy load (me calculates about 3w just from discharging/charging the caps, then a few extra for the gate resistor and zeners which are dissipating several watts per fet...)
1. But how would you tell when it is saturating? I never saw any difference in the waveforms, or any ringing whatsoever, but the no load current of the driver did seem to level off around 5 turns, the number of turns and the corresponding no load current from the driver is listed here. No gdt connected 50ma; 0 turn 1800ma; 1 turn 780ma; 2 turn 400ma; 3 turn 270ma; 4 turn 220ma; 5 tun 210ma; and it sorta levels off around there. So 5 turns should be ok?
2. I know it was the core getting hot, and the wire was not shorted. I suppose it was the cure saturating, but will having less turns help solve that? I was avoiding cat5 because the window on the core is pretty small, but if I only need 5 turns I should ok...
3. The resistance of the windings is 0.8ohms (measured), and reducing the number of turns down as low as 1 didn't speed it up... The rise time at the input of the gdt is about 50ns from -12v up to 0v, but about 200ns to get all of the way up to 12v (the signal goes strait up to 0v, then at about 5 it arks over). The fall times are the opposite; it drops like a rock (about 10ns) to 0v, then slows down to get to -12v, about 200ns total.
4. I will try for cat5...
Re: GDT questions
williamn, Tue Aug 15 2006, 01:14AM
I use cat5 for all of my gate transformers and use 30AWG Kynar wire for my current transformers. Whilst experimenting with my 750khz audio modulated coil I found that the UCC drivers can handle higher frerquencies than the Tc4422/21 drivers. The tc4422/21 drivers got hot much faster than the UCC37322/21 drivers, and both are suposed to be rated at 9A.
williamn, Tue Aug 15 2006, 01:14AM
I use cat5 for all of my gate transformers and use 30AWG Kynar wire for my current transformers. Whilst experimenting with my 750khz audio modulated coil I found that the UCC drivers can handle higher frerquencies than the Tc4422/21 drivers. The tc4422/21 drivers got hot much faster than the UCC37322/21 drivers, and both are suposed to be rated at 9A.
Print this page