I need to decrease the leak. inductance (remove some turns from the series coil or to increase the intrefier ) ,and to design some snuber for the secondary diode (got 400v at 70v output) maybe another secondary in the oposite direction with a small rectifier diode to clamp the spike
Hi Dexter,
good job.
I have made something similar (24v/100A with PFC) and I have some questions for you.
1) which controller are you using and how is your control strategy (voltage mode or current mode)?
2) how are you measuring the primary current? (shunt, current transformer)
3) mosfets: the irfp460 are very old mosfets not at all suitables for phase shift zvs. These mosfets needs to have a fast low Qrr body diode otherwise they will blow at light load or generate a nightmare of EMI. On top of that you are using 2 of them in parallel so the drain to source capacitance is doubled. You need a very large leakage inductance to make the system go in zvs with such big mosfets capacitance.
Very large leakage creates nasty spikes on the seondary side which are hard to clamp.
I suggest you to change the mosfets: infineon has some good ones in the CoolMOS CFD series that can do the job.
For my SMPS I have used Ixys IXTQ480P2; its a beast but very hard to find.
4) output diodes: what is the rated voltage? If there is too much ringing you can calm it down with a parallel RC snubber (one for each diode) a starting point can be 1nF and 20...100ohm; the bigger the capacitor the better the ringing is attenuated at the expense of power loss in the resistor.
Overshoot is another story: a paraller RC will be not effective on overshoot unless you can burn a lots of watts in the resistor. I fixed this issue with a partially regenerative RCD clamp. Note that in my design I lose around 15W only for snubbing the diodes.
5) output inductor: where is it? How it is made.
6) 97% efficiency? I hope that it is only the phase shift and not including the PFC.... Seriously forget it with the mosfets you are using now. The most difficult part is the transformer,
you can easily lose 4...10 times the losses you have calculated because of skin and proximity effects. In my design I use a ETD49 core wound with litz wire and using a current doubler synchronous rectifier and it gives me around 96% efficiency at 100A output.
Thats good job but making a 5k smps which does not explode every two minutes will be an hard job...
Ciao
-marco
good job.
I have made something similar (24v/100A with PFC) and I have some questions for you.
1) which controller are you using and how is your control strategy (voltage mode or current mode)?
2) how are you measuring the primary current? (shunt, current transformer)
3) mosfets: the irfp460 are very old mosfets not at all suitables for phase shift zvs. These mosfets needs to have a fast low Qrr body diode otherwise they will blow at light load or generate a nightmare of EMI. On top of that you are using 2 of them in parallel so the drain to source capacitance is doubled. You need a very large leakage inductance to make the system go in zvs with such big mosfets capacitance.
Very large leakage creates nasty spikes on the seondary side which are hard to clamp.
I suggest you to change the mosfets: infineon has some good ones in the CoolMOS CFD series that can do the job.
For my SMPS I have used Ixys IXTQ480P2; its a beast but very hard to find.
4) output diodes: what is the rated voltage? If there is too much ringing you can calm it down with a parallel RC snubber (one for each diode) a starting point can be 1nF and 20...100ohm; the bigger the capacitor the better the ringing is attenuated at the expense of power loss in the resistor.
Overshoot is another story: a paraller RC will be not effective on overshoot unless you can burn a lots of watts in the resistor. I fixed this issue with a partially regenerative RCD clamp. Note that in my design I lose around 15W only for snubbing the diodes.
5) output inductor: where is it? How it is made.
6) 97% efficiency? I hope that it is only the phase shift and not including the PFC.... Seriously forget it with the mosfets you are using now. The most difficult part is the transformer,
you can easily lose 4...10 times the losses you have calculated because of skin and proximity effects. In my design I use a ETD49 core wound with litz wire and using a current doubler synchronous rectifier and it gives me around 96% efficiency at 100A output.
Thats good job but making a 5k smps which does not explode every two minutes will be an hard job...
Ciao
-marco
Hy,
1) controller uc3895 voltage mode
2) shunt and if i remember by ina138
3)I'm not using irfp460 , somthing like spw20nk50 , i don't remember exactly , tomorrow i can tell you , they are 20A at 500w and about 0.2 ohm each .I know about Infineon C6 but they are expensive , and i'm a student without a job.
4) I,m using 2 inductors each having 15 turns on an e40 core and 3paper for the gap .
5) My bridge diode is rated at 600V 30A , yes ringing is there , i don;t like the dissipative snubber, maybe another winding in opposite direction with a slightly more turns to clamp the voltage .
6) eff 97 % calculated , the core is dissipating 22w and i used cooper foil 0.1mm thick and 40 mm wide (2 in paralel) .(current is 17A at 5kw 400v simulated ).
Tomorrow i will test it at 400W and 100v input limited by may transformer , next test will be with an auto transformer . Will post picture .
Hi Dexter,
if you are powering at 80V input you should get something around +/-43V at the output with quite 100% duty cycle.
Your output voltage is +/-75V (if I remeber correctly) so when it is powered from a PFC at 400V the controller will run at around 35% duty cycle. This is
way too low for a phase shift converter. You will have a lot of circulating current just for nothing that heats up the mosfets and the transformer.
I think that your transformer turn ratio is too high, you need decrease the secondary turns to make the system run at >80% duty cycle at 400V input.
A phase shift converter works well only with a limited range of input voltage
(300...400V will be a good choice), the wider the input range, the higher the circulating current, the lower the efficiency.
You need to test your system at 400V to check for proper ZVS transitions.
Shunt: where are you sensing the current? can you post a schematic?
Transformer connection: your system is configured in voltage mode; did you place a coupling capacitor in series with you transformer primary? Without this capacitor it will never work due to transformer staircase saturation.
Inductors: how many uH on your E40 cores
mosfets: STW20NK50Z (from ST) has slow and high Qrr body diode. For phase shift it is mandatory to use mosfets with fast, low Qrr diode.
Some possibilities:
-Infineon: Cool MOS CFD/CFD2 (not the C3/CP/C6 series the body diode is
very slow)
-ST: all models with *ND at the end of the part number (Infieneon are far
better anyway and cost less than ST)
-Fairchild: all models with *NF at the end of the part number (never tried
them but seems fine on paper)
-IXYS: all models that starts with IXFxxx or the Polar P2/P3 series.
(in my opinion the best mosfets on the market for this application also
cheaper than ST)
This body diode issue is very important a phase shift converter will never work with mosfets with slow body diode.
@savu:
I have used UCC3895 configured in voltage mode @ 100kHz (200kHz osc).
I have never used LTC1922 but it seems to be a good choice since it provides dedicated outputs for synchronous rectifier drive. With UCC3895 it becomes quite tricky to make the SR work properly. There is a recent improvement of UCC3895 which is called UCC28950 that seems similar to LTC1922 it worth to take a look at it.
if you are powering at 80V input you should get something around +/-43V at the output with quite 100% duty cycle.
Your output voltage is +/-75V (if I remeber correctly) so when it is powered from a PFC at 400V the controller will run at around 35% duty cycle. This is
way too low for a phase shift converter. You will have a lot of circulating current just for nothing that heats up the mosfets and the transformer.
I think that your transformer turn ratio is too high, you need decrease the secondary turns to make the system run at >80% duty cycle at 400V input.
A phase shift converter works well only with a limited range of input voltage
(300...400V will be a good choice), the wider the input range, the higher the circulating current, the lower the efficiency.
You need to test your system at 400V to check for proper ZVS transitions.
Shunt: where are you sensing the current? can you post a schematic?
Transformer connection: your system is configured in voltage mode; did you place a coupling capacitor in series with you transformer primary? Without this capacitor it will never work due to transformer staircase saturation.
Inductors: how many uH on your E40 cores
mosfets: STW20NK50Z (from ST) has slow and high Qrr body diode. For phase shift it is mandatory to use mosfets with fast, low Qrr diode.
Some possibilities:
-Infineon: Cool MOS CFD/CFD2 (not the C3/CP/C6 series the body diode is
very slow)
-ST: all models with *ND at the end of the part number (Infieneon are far
better anyway and cost less than ST)
-Fairchild: all models with *NF at the end of the part number (never tried
them but seems fine on paper)
-IXYS: all models that starts with IXFxxx or the Polar P2/P3 series.
(in my opinion the best mosfets on the market for this application also
cheaper than ST)
This body diode issue is very important a phase shift converter will never work with mosfets with slow body diode.
@savu:
I have used UCC3895 configured in voltage mode @ 100kHz (200kHz osc).
I have never used LTC1922 but it seems to be a good choice since it provides dedicated outputs for synchronous rectifier drive. With UCC3895 it becomes quite tricky to make the SR work properly. There is a recent improvement of UCC3895 which is called UCC28950 that seems similar to LTC1922 it worth to take a look at it.
The power supply is designed to work at 400v input and 2*160v output .Now in order to resonate i used a small inductor in series with the power transformer (16uH) .When i tested today with an lower output rezistor the power was the same , because the series inductance is way to big , and not enough voltage is applied to the power transformer .In order to decrease it's value i removed a half of the core (just for test ) .the voltage jumped from 50V @8ohm to 65v@8 ohm , but the transistor are heating . For the final test i will use a 6uH series inductance to lower the circulating currents (transformer has 2u ) and will resonate at @2.5kw from may calculations .
Now the mosfets , they were cheap , and this means poor results , maybe i will use some fast rectifiers (mur460) .The ina138 is in the lower side of the buss after the filtering caps but before the h bridge .
I use GDT and a tc4422 buffer and a pnp turn off buffer for each mosfet to minimase ringing .
I didn't use dc blocking cap , because i think if the current increases the voltage applied to the transformer will decrease (pulse by pulse limiting ina138 connected to the cs pin of uc3879 ) .That's why i didn't use current sens transformers in order to prevent them from saturating and not detecting the increasing current or DC.
The inductance of the output filter is big 90uH for each bobin .If i remember when i was simulating it in pSim at 85%DOD the voltage was 160 with this configuration of transformer ,but the series inductance needed to be changed to almost 7 uH ........ all these calculations and simulations were made in april-may 2011 , maybe the value posted here are slightly wrong , i don't remember everything.
Hi Dexter,
don't use those fets. It is not a matter or Rdson.
All these mosfets has slow body diode you absolutely need something with fast body diode for phase shift zvt.
The STW/STY have very good rdson specs but they are very hard to drive properly (high Qg and Ciss). Their body diode is extremely slow 26uC of reverse charge and 73A Irrm....
You need something with no more than 2..3uC of Qrrm to work properly on phase shift supply.
Anyway you don't need extremely low rdson at this power level, something with around 0.1R is ok.
ciao
-marco
don't use those fets. It is not a matter or Rdson.
All these mosfets has slow body diode you absolutely need something with fast body diode for phase shift zvt.
The STW/STY have very good rdson specs but they are very hard to drive properly (high Qg and Ciss). Their body diode is extremely slow 26uC of reverse charge and 73A Irrm....
You need something with no more than 2..3uC of Qrrm to work properly on phase shift supply.
Anyway you don't need extremely low rdson at this power level, something with around 0.1R is ok.
ciao
-marco
Update
I did find this FCH76N60NF , rds on 0.03ohm Qrr ~ 2nC , the problem is the Qg , but with only one , it is still biger than my 2 transistors in paralel , maybe I will change the drivers 9A peak with some capable of 12A , or use totem pole bipolar transistors , but first , i will try to blow the SW..
I did find this FCH76N60NF , rds on 0.03ohm Qrr ~ 2nC , the problem is the Qg , but with only one , it is still biger than my 2 transistors in paralel , maybe I will change the drivers 9A peak with some capable of 12A , or use totem pole bipolar transistors , but first , i will try to blow the SW..
Hi you don't need such big fets; you will have many troubles in driving them properly and you risk also worst performances compared to a smaller fet.
Big fet means slower transition times and thus higher switching losses.
On top of that big fet means big Coss. In phase shift you need to resonate out the the Coss with your leakage inductance to achieve zvs. The leakage inductance will be bigger reducing the available secondary duty cycle and thus inreasing the primary current to compensate.
At those power level stay on something with around 0.1...0.15R that will be ok and much less expensive than those beasts from Fairchild.
The smaller the fet the faster the body diode (for a given fet technology) and this is good for phase shift topology.
I suppose that at your power level you will have something around 10Arms in your fets at full load so with 0.1R rdson you will have around 10w per fet of conduction loss that are easily managed by a to247 package.
Avoid paralleling ultrafast diodes with fets without adding a series schottky to the fet. It is much more cheap, easy and performant a fet with fast body diode.
Big fet means slower transition times and thus higher switching losses.
On top of that big fet means big Coss. In phase shift you need to resonate out the the Coss with your leakage inductance to achieve zvs. The leakage inductance will be bigger reducing the available secondary duty cycle and thus inreasing the primary current to compensate.
At those power level stay on something with around 0.1...0.15R that will be ok and much less expensive than those beasts from Fairchild.
The smaller the fet the faster the body diode (for a given fet technology) and this is good for phase shift topology.
I suppose that at your power level you will have something around 10Arms in your fets at full load so with 0.1R rdson you will have around 10w per fet of conduction loss that are easily managed by a to247 package.
Avoid paralleling ultrafast diodes with fets without adding a series schottky to the fet. It is much more cheap, easy and performant a fet with fast body diode.
I know about the Coss , from my simulations the primary current at maximum value is 25A , 5kw and 400V in , leak ind is 6uH at this power level with my original transistors it will resonate at about 3kw , and those too have big Coss . I don't wont to resonate below 3kw , because at maximum power without a saturable inductor in series ,the circulating currents introduce new losses .
Sorry but it is 2.6kw or 5kw?
Fch47n60 can be ok it is similar to the onesI proposed some time ago.
Maybe for 5kw you will need 2 in parallel or a single fch76n60. For 2.6kw
this is more than enough.
I have made a 2.4kw phase shift using 4xIXTQ480P2 with very good results. I have also tried with spw35n60cfd with similar results.
For 5kw I don't know, I have never tried something that powerful.
Btw: what is your switching frequency? You can consider lowering it a bit and using IGBT + parallel fred instead of mosfets. In a phase shift smps the bridge is soft switched and fast 600v type IGBTs can do the job better than mosfets. They are also much more easy to drive (lower Qg).
Fch47n60 can be ok it is similar to the onesI proposed some time ago.
Maybe for 5kw you will need 2 in parallel or a single fch76n60. For 2.6kw
this is more than enough.
I have made a 2.4kw phase shift using 4xIXTQ480P2 with very good results. I have also tried with spw35n60cfd with similar results.
For 5kw I don't know, I have never tried something that powerful.
Btw: what is your switching frequency? You can consider lowering it a bit and using IGBT + parallel fred instead of mosfets. In a phase shift smps the bridge is soft switched and fast 600v type IGBTs can do the job better than mosfets. They are also much more easy to drive (lower Qg).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.