DRSSTC: choosing the "right" surge impedance
Luca, Tue Mar 30 2010, 02:27PMHi guys,
I am planning my (first) DRSSTC.
Some specs:
Secondary: diameter=14cm (5.5 in) length 60 cm (23.5 in) 1600 turns AWG 28
topload 10x50 cm toroid (4x20 in)
res. freq. 110kHz
Full bridge of IXGN60N60C2D1 at 320VDC
I am going to set the max peak current to about 600A, in order to avoid excessive stress on the IGBTs
Now, I have to design primary circuit...
I understand that less impedance means higher current and faster secondary voltage build-up, but, in terms of performance (spark length) which value of surge impedance should I choose?
Now I am considering something between 150-200nF primary cap which gives 10-7.5 ohnm of impedance.
Thanks for any help
Regards,
Luca
Re: DRSSTC: choosing the "right" surge impedance
Goodchild, Tue Mar 30 2010, 03:28PM
umm.. I would go with a lower OCD value for the 60N60s I would go no more than 500A with thoses.
Goodchild, Tue Mar 30 2010, 03:28PM
umm.. I would go with a lower OCD value for the 60N60s I would go no more than 500A with thoses.
Re: DRSSTC: choosing the "right" surge impedance
Steve Conner, Tue Mar 30 2010, 06:04PM
I choose the surge impedance to be about 10 times the inverter's output impedance, but Steve Ward's newer coils have a lower impedance and seem to work great, so now I'd say 5 times. Anything between 5 and 10 seems to work, though.
The formula I use for the inverter output impedance is (4/pi)*(DC bus voltage/current limit) but since it's just a rough estimate, you can leave out the 4/pi.
And if it were a half bridge inverter, you'd need to divide the impedance by 2.
Steve Conner, Tue Mar 30 2010, 06:04PM
I choose the surge impedance to be about 10 times the inverter's output impedance, but Steve Ward's newer coils have a lower impedance and seem to work great, so now I'd say 5 times. Anything between 5 and 10 seems to work, though.
The formula I use for the inverter output impedance is (4/pi)*(DC bus voltage/current limit) but since it's just a rough estimate, you can leave out the 4/pi.
And if it were a half bridge inverter, you'd need to divide the impedance by 2.
Re: DRSSTC: choosing the "right" surge impedance
Mads Barnkob, Tue Mar 30 2010, 07:02PM
Steve Ward wrote this on his site about the 40N60 IGBTs
Mads Barnkob, Tue Mar 30 2010, 07:02PM
Goodchild wrote ...
umm.. I would go with a lower OCD value for the 60N60s I would go no more than 500A with thoses.
umm.. I would go with a lower OCD value for the 60N60s I would go no more than 500A with thoses.
Steve Ward wrote this on his site about the 40N60 IGBTs
3/16/05:
A lot has changed. New secondary (6.5"x21" 30awg) and now running on primary feedback. Also a new MMC, 5 strings of 2 caps in series for .375uF. A new primary was needed to tune (requires 7 turns now). The primary is tuned to 55khz, the secondary Fr is about 75khz. Sparks easily hit 6' with measured 650A+ in the primary circuit. I tested these IGBTs to over 1000A at 95khz so this should be no problem at all. Everything runs nice and cool at 55khz and the sparks are pretty darn hot! Enjoy some pictures from today (notice the red LED on the base of the coil, indicating that the overcurrent circuit tripped). The coil was running 120bps and 150uS on period.
Re: DRSSTC: choosing the "right" surge impedance
Goodchild, Tue Mar 30 2010, 11:59PM
Well ya those are 40N60s there a hell of a lot better than 60N60s ive personaly tested 60N60s and they fail at about 600A to 700A
Goodchild, Tue Mar 30 2010, 11:59PM
Mads Barnkob wrote ...
Steve Ward wrote this on his site about the 40N60 IGBTs
Goodchild wrote ...
umm.. I would go with a lower OCD value for the 60N60s I would go no more than 500A with thoses.
umm.. I would go with a lower OCD value for the 60N60s I would go no more than 500A with thoses.
Steve Ward wrote this on his site about the 40N60 IGBTs
3/16/05:
A lot has changed. New secondary (6.5"x21" 30awg) and now running on primary feedback. Also a new MMC, 5 strings of 2 caps in series for .375uF. A new primary was needed to tune (requires 7 turns now). The primary is tuned to 55khz, the secondary Fr is about 75khz. Sparks easily hit 6' with measured 650A+ in the primary circuit. I tested these IGBTs to over 1000A at 95khz so this should be no problem at all. Everything runs nice and cool at 55khz and the sparks are pretty darn hot! Enjoy some pictures from today (notice the red LED on the base of the coil, indicating that the overcurrent circuit tripped). The coil was running 120bps and 150uS on period.
Well ya those are 40N60s there a hell of a lot better than 60N60s ive personaly tested 60N60s and they fail at about 600A to 700A
Re: DRSSTC: choosing the "right" surge impedance
Luca, Wed Mar 31 2010, 08:30AM
Well, in my case that would be around 6.5 ohm which means about 230nF...
Thanks for the info
Luca
Luca, Wed Mar 31 2010, 08:30AM
Steve McConner wrote ...
I choose the surge impedance to be about 10 times the inverter's output impedance, but Steve Ward's newer coils have a lower impedance and seem to work great, so now I'd say 5 times. Anything between 5 and 10 seems to work, though.
The formula I use for the inverter output impedance is (4/pi)*(DC bus voltage/current limit) but since it's just a rough estimate, you can leave out the 4/pi.
And if it were a half bridge inverter, you'd need to divide the impedance by 2.
I choose the surge impedance to be about 10 times the inverter's output impedance, but Steve Ward's newer coils have a lower impedance and seem to work great, so now I'd say 5 times. Anything between 5 and 10 seems to work, though.
The formula I use for the inverter output impedance is (4/pi)*(DC bus voltage/current limit) but since it's just a rough estimate, you can leave out the 4/pi.
And if it were a half bridge inverter, you'd need to divide the impedance by 2.
Well, in my case that would be around 6.5 ohm which means about 230nF...
Thanks for the info
Luca
Re: DRSSTC: choosing the "right" surge impedance
Mads Barnkob, Wed Mar 31 2010, 09:05AM
One wouldn't know from looking in the datasheets, the 60N60 are superior in almost any aspect.
Mads Barnkob, Wed Mar 31 2010, 09:05AM
Goodchild wrote ...
Well ya those are 40N60s there a hell of a lot better than 60N60s ive personaly tested 60N60s and they fail at about 600A to 700A
Mads Barnkob wrote ...
Steve Ward wrote this on his site about the 40N60 IGBTs
Goodchild wrote ...
umm.. I would go with a lower OCD value for the 60N60s I would go no more than 500A with thoses.
umm.. I would go with a lower OCD value for the 60N60s I would go no more than 500A with thoses.
Steve Ward wrote this on his site about the 40N60 IGBTs
3/16/05:
A lot has changed. New secondary (6.5"x21" 30awg) and now running on primary feedback. Also a new MMC, 5 strings of 2 caps in series for .375uF. A new primary was needed to tune (requires 7 turns now). The primary is tuned to 55khz, the secondary Fr is about 75khz. Sparks easily hit 6' with measured 650A+ in the primary circuit. I tested these IGBTs to over 1000A at 95khz so this should be no problem at all. Everything runs nice and cool at 55khz and the sparks are pretty darn hot! Enjoy some pictures from today (notice the red LED on the base of the coil, indicating that the overcurrent circuit tripped). The coil was running 120bps and 150uS on period.
Well ya those are 40N60s there a hell of a lot better than 60N60s ive personaly tested 60N60s and they fail at about 600A to 700A
One wouldn't know from looking in the datasheets, the 60N60 are superior in almost any aspect.
Re: DRSSTC: choosing the "right" surge impedance
Luca, Wed Mar 31 2010, 09:14AM
I made the same remark looking at the datasheets. Both the nominal and peak currents are the same for the two devices...
I will try 600A and let's see what happen
Luca
Luca, Wed Mar 31 2010, 09:14AM
Mads Barnkob wrote ...
One wouldn't know from looking in the datasheets, the 60N60 are superior in almost any aspect.
Goodchild wrote ...
Well ya those are 40N60s there a hell of a lot better than 60N60s ive personaly tested 60N60s and they fail at about 600A to 700A
Mads Barnkob wrote ...
Steve Ward wrote this on his site about the 40N60 IGBTs
Goodchild wrote ...
umm.. I would go with a lower OCD value for the 60N60s I would go no more than 500A with thoses.
umm.. I would go with a lower OCD value for the 60N60s I would go no more than 500A with thoses.
Steve Ward wrote this on his site about the 40N60 IGBTs
3/16/05:
A lot has changed. New secondary (6.5"x21" 30awg) and now running on primary feedback. Also a new MMC, 5 strings of 2 caps in series for .375uF. A new primary was needed to tune (requires 7 turns now). The primary is tuned to 55khz, the secondary Fr is about 75khz. Sparks easily hit 6' with measured 650A+ in the primary circuit. I tested these IGBTs to over 1000A at 95khz so this should be no problem at all. Everything runs nice and cool at 55khz and the sparks are pretty darn hot! Enjoy some pictures from today (notice the red LED on the base of the coil, indicating that the overcurrent circuit tripped). The coil was running 120bps and 150uS on period.
Well ya those are 40N60s there a hell of a lot better than 60N60s ive personaly tested 60N60s and they fail at about 600A to 700A
One wouldn't know from looking in the datasheets, the 60N60 are superior in almost any aspect.
I made the same remark looking at the datasheets. Both the nominal and peak currents are the same for the two devices...
I will try 600A and let's see what happen
Luca
Re: DRSSTC: choosing the "right" surge impedance
Steve Ward, Wed Mar 31 2010, 09:53PM
Ive had questionable results when pushing the IXYS 60N60 parts, they seemed to explode when the older fairchild parts did not. I suspect the die size is actually smaller for the IXYS compared to the fairchild.
You can always give it a try and see how it works.
But as ive said before, im no longer into pushing parts way beyond their specifications, it doesnt get you anywhere in the long run. Unless you really go through the thermal transient modeling (which unfortunately the 60N60 datasheet is screwed up) and prove that the delta Temp on the die is quite safe, then you guys are mostly just pulling numbers out of your ass.
My thoughts on surge impedance: faster rise time seems to be more efficient at producing the longest sparks for a given power consumption. You are ultimately limited by the silicon and other resistive losses. I think the guideline Conner gave is pretty good if you dont have any other idea of where to start. If you are seeking to minimize peak current, then you likely want a higher surge impedance, but too high means that you cant dump as much energy into the sparks (the system is held back). I have found that i can achieve a moderate spark length with a high Z primary at some low current, and when using a low Z primary it takes more peak current to get the same spark. The only difference is that the low Z primary still has the ability to go much beyond and produce even bigger sparks, at a cost of even higher primary current.
Going further, it seemed to me that small coils could never be optimized for a really fast energy transfer in the same way that big coils could be. So small coils are often detuned so that they only excite 1 pole frequency which allows the secondary voltage to grow un-bounded before spark breakdown (it essentially removes the notch). But, on big coils i found that i didnt need to detune the primary to get good performance, and tuning it for a "notched" waveform yields a better net energy transfer to the secondary, making operating more efficient than the detuned method.
Its difficult to explain this in a short forum post. For anyone interested, i suggest you read up on Antonio's theoretical description of the DRSSTC
Its sort of a pre-requisite to really understanding the subtleties of DRSSTCs, they can really be quite complex if you want them to be. Or, just tune for maximum smoke if you dont want to be bothered with all the thinking.
I also recommend Antonio's DRSSTCD program, just be mindful that with any simulation: garbage in = garbage out. It takes a bit of interpretation and working with a real system to know whether or not the results are realistic when you have a spark shooting off the top of your coil (which is, afterall the hardest part of all of this analysis).
Steve Ward, Wed Mar 31 2010, 09:53PM
Ive had questionable results when pushing the IXYS 60N60 parts, they seemed to explode when the older fairchild parts did not. I suspect the die size is actually smaller for the IXYS compared to the fairchild.
You can always give it a try and see how it works.
But as ive said before, im no longer into pushing parts way beyond their specifications, it doesnt get you anywhere in the long run. Unless you really go through the thermal transient modeling (which unfortunately the 60N60 datasheet is screwed up) and prove that the delta Temp on the die is quite safe, then you guys are mostly just pulling numbers out of your ass.
My thoughts on surge impedance: faster rise time seems to be more efficient at producing the longest sparks for a given power consumption. You are ultimately limited by the silicon and other resistive losses. I think the guideline Conner gave is pretty good if you dont have any other idea of where to start. If you are seeking to minimize peak current, then you likely want a higher surge impedance, but too high means that you cant dump as much energy into the sparks (the system is held back). I have found that i can achieve a moderate spark length with a high Z primary at some low current, and when using a low Z primary it takes more peak current to get the same spark. The only difference is that the low Z primary still has the ability to go much beyond and produce even bigger sparks, at a cost of even higher primary current.
Going further, it seemed to me that small coils could never be optimized for a really fast energy transfer in the same way that big coils could be. So small coils are often detuned so that they only excite 1 pole frequency which allows the secondary voltage to grow un-bounded before spark breakdown (it essentially removes the notch). But, on big coils i found that i didnt need to detune the primary to get good performance, and tuning it for a "notched" waveform yields a better net energy transfer to the secondary, making operating more efficient than the detuned method.
Its difficult to explain this in a short forum post. For anyone interested, i suggest you read up on Antonio's theoretical description of the DRSSTC
Its sort of a pre-requisite to really understanding the subtleties of DRSSTCs, they can really be quite complex if you want them to be. Or, just tune for maximum smoke if you dont want to be bothered with all the thinking.
I also recommend Antonio's DRSSTCD program, just be mindful that with any simulation: garbage in = garbage out. It takes a bit of interpretation and working with a real system to know whether or not the results are realistic when you have a spark shooting off the top of your coil (which is, afterall the hardest part of all of this analysis).
Re: DRSSTC: choosing the "right" surge impedance
Luca, Thu Apr 01 2010, 09:14AM
Steve, I have found two different 60N60 parts from IXYS...
IXGN60N60: datasheet dated 2000. Datasheet quite poor (no graphs), max peak current is 200A, 1ms @ 25°C
IXGN60N60C2: datasheet dated 2009. Two times faster than the older part (half the gate charge) and max peak current is 300A, 1ms @ 25°C
By extrapolation of the output characteristic, I assume that with a collectror current of 300A Vce is about 4.5V (Vge=15V). So, a peak current of 300 A with a period of 1ms means a pulse with energy of 1.35J.
Considering a pulse of 700A with a Vce of 10V (rough estimate) and a duration of 100us the dissipated energy is 0.7J... In theory, no problem of junction over heating...
Luca
Luca, Thu Apr 01 2010, 09:14AM
Steve Ward wrote ...
Ive had questionable results when pushing the IXYS 60N60 parts, they seemed to explode when the older fairchild parts did not. I suspect the die size is actually smaller for the IXYS compared to the fairchild.
Ive had questionable results when pushing the IXYS 60N60 parts, they seemed to explode when the older fairchild parts did not. I suspect the die size is actually smaller for the IXYS compared to the fairchild.
Steve, I have found two different 60N60 parts from IXYS...
IXGN60N60: datasheet dated 2000. Datasheet quite poor (no graphs), max peak current is 200A, 1ms @ 25°C
IXGN60N60C2: datasheet dated 2009. Two times faster than the older part (half the gate charge) and max peak current is 300A, 1ms @ 25°C
By extrapolation of the output characteristic, I assume that with a collectror current of 300A Vce is about 4.5V (Vge=15V). So, a peak current of 300 A with a period of 1ms means a pulse with energy of 1.35J.
Considering a pulse of 700A with a Vce of 10V (rough estimate) and a duration of 100us the dissipated energy is 0.7J... In theory, no problem of junction over heating...
Luca
Re: DRSSTC: choosing the "right" surge impedance
Steve Conner, Thu Apr 01 2010, 12:42PM
Richie and I have always recommended going easy on the peak current, too. When I first built my Mjollnir coil, Steve laughed at me because I put four miniblocks in it and ran them at only 450A. But 5 years later I'm still using the original set of IGBTs. (Touch wood.)
This approach to reliability is summed up by a TV program I once saw on tuning cars. "Don't waste money on tuning, just stick a bigger engine in it."
Steve and I used to have different methods for designing DRSSTCs, but his make somewhat bigger sparks than mine, and I like his method of tuning large coils that excites both poles at once for fast energy transfer. So nowadays I just agree with him.
Steve Conner, Thu Apr 01 2010, 12:42PM
Steve Ward wrote ...
Unless you really go through the thermal transient modeling (which unfortunately the 60N60 datasheet is screwed up) and prove that the delta Temp on the die is quite safe, then you guys are mostly just pulling numbers out of your ass.
Of course we are pulling numbers out of our ass! This is the Internet after all Unless you really go through the thermal transient modeling (which unfortunately the 60N60 datasheet is screwed up) and prove that the delta Temp on the die is quite safe, then you guys are mostly just pulling numbers out of your ass.
Richie and I have always recommended going easy on the peak current, too. When I first built my Mjollnir coil, Steve laughed at me because I put four miniblocks in it and ran them at only 450A. But 5 years later I'm still using the original set of IGBTs. (Touch wood.)
This approach to reliability is summed up by a TV program I once saw on tuning cars. "Don't waste money on tuning, just stick a bigger engine in it."
Steve and I used to have different methods for designing DRSSTCs, but his make somewhat bigger sparks than mine, and I like his method of tuning large coils that excites both poles at once for fast energy transfer. So nowadays I just agree with him.
Re: DRSSTC: choosing the "right" surge impedance
Luca, Thu Apr 01 2010, 01:08PM
Regarding this item, I'm performing a lot of PSpice simulation in order to try to better understand DRSSTC behavior...
The point is that one should have a detailed electrical equivalent model of the streamers otherwise simulations are quite meaningless... The main difficulty is that streamer load is something "dinamic", very complicated to model...
At the moment, as streamer load, I am using a 220k resistor with a series capacitor that increases during streamer growth.
Regards,
Luca
Luca, Thu Apr 01 2010, 01:08PM
Steve Ward wrote ...
Its sort of a pre-requisite to really understanding the subtleties of DRSSTCs, they can really be quite complex if you want them to be. Or, just tune for maximum smoke if you dont want to be bothered with all the thinking.
I also recommend Antonio's DRSSTCD program, just be mindful that with any simulation: garbage in = garbage out. It takes a bit of interpretation and working with a real system to know whether or not the results are realistic when you have a spark shooting off the top of your coil (which is, afterall the hardest part of all of this analysis).
Its sort of a pre-requisite to really understanding the subtleties of DRSSTCs, they can really be quite complex if you want them to be. Or, just tune for maximum smoke if you dont want to be bothered with all the thinking.
I also recommend Antonio's DRSSTCD program, just be mindful that with any simulation: garbage in = garbage out. It takes a bit of interpretation and working with a real system to know whether or not the results are realistic when you have a spark shooting off the top of your coil (which is, afterall the hardest part of all of this analysis).
Regarding this item, I'm performing a lot of PSpice simulation in order to try to better understand DRSSTC behavior...
The point is that one should have a detailed electrical equivalent model of the streamers otherwise simulations are quite meaningless... The main difficulty is that streamer load is something "dinamic", very complicated to model...
At the moment, as streamer load, I am using a 220k resistor with a series capacitor that increases during streamer growth.
Regards,
Luca
Print this page