???
Now I must ask about your intended goal for this project.
a) Building a SMPS that works ?
b) Learning to brew your own SMPS?
For a) there are two common paths:
a1) Build a proven design, unchanged. Chance of success >70%.
a2) Roulette of circuits and components. Success rate of this method has massively improved (say >30%) thanks to internet forums, because there are many contributors. Consequently there is a chance that somebody has a proven design which is very similar and can be used as reference. Nevertheless it is likely to suffer from high component waste during gambling. In case of winning a game, still the reliability of the circuit is hard to predict.
For b) you must draw your personal satisfaction from learning electronic details in practice and theory. In case of power electronics unfortunately also approach b) can cause high component waste. BUT by the end of the day you will know why it works and most likely will have a rough idea about the circuit limitations. Time line and really usable hard ware results of the first prototype(s) should be less important.
Success rate of b) is similar like a2), because most people do not draw enough satisfaction from small steps and give up before reaching the break through.
Your last question is pointing to a2). Unfortunately I do not have a similar resonant circuit on my desk and lacking the time to calculate/simulate one for this thread.
But it looks like DJ Lecco has such a resonant circuit on the desk...
*curious*
Now I must ask about your intended goal for this project.
a) Building a SMPS that works ?
b) Learning to brew your own SMPS?
For a) there are two common paths:
a1) Build a proven design, unchanged. Chance of success >70%.
a2) Roulette of circuits and components. Success rate of this method has massively improved (say >30%) thanks to internet forums, because there are many contributors. Consequently there is a chance that somebody has a proven design which is very similar and can be used as reference. Nevertheless it is likely to suffer from high component waste during gambling. In case of winning a game, still the reliability of the circuit is hard to predict.
For b) you must draw your personal satisfaction from learning electronic details in practice and theory. In case of power electronics unfortunately also approach b) can cause high component waste. BUT by the end of the day you will know why it works and most likely will have a rough idea about the circuit limitations. Time line and really usable hard ware results of the first prototype(s) should be less important.
Success rate of b) is similar like a2), because most people do not draw enough satisfaction from small steps and give up before reaching the break through.
Your last question is pointing to a2). Unfortunately I do not have a similar resonant circuit on my desk and lacking the time to calculate/simulate one for this thread.
But it looks like DJ Lecco has such a resonant circuit on the desk...
*curious*
Paralleling IGBTs is tricky.
Just hard wiring in parallel is not suitfull. Switching losses and conduction losses as well would not be shared in a symmetric manner between both devices.
Most likely it is more resonable to go for a full bridge, fortunately this also reduces the currents on the primary side.
Hm, I am not convinced that your half bridge is generally overloaded, your IGBTs should be strong enough. It still appears to me that there is some hidden issue somewhere. But searching this issue can be pretty annoying.
In fact there is a chance that things turn better in a full bridge.
But this means also a lot of building efforts including a new power transformer.
P.S.
Heavy winter... yes, I understand. That's tough!
Just hard wiring in parallel is not suitfull. Switching losses and conduction losses as well would not be shared in a symmetric manner between both devices.
Most likely it is more resonable to go for a full bridge, fortunately this also reduces the currents on the primary side.
Hm, I am not convinced that your half bridge is generally overloaded, your IGBTs should be strong enough. It still appears to me that there is some hidden issue somewhere. But searching this issue can be pretty annoying.
In fact there is a chance that things turn better in a full bridge.
But this means also a lot of building efforts including a new power transformer.
P.S.
Heavy winter... yes, I understand. That's tough!
...not sure if I got your request right.
Are you still facing defects, or do you just wish get a better efficacy?
If you just wish to get a better efficacy, then the simple step towards a full bridge will not change much.
To improve efficacy you will have to tackle all the losses that are distributed over the entire circuit.
Here the typical strong contributors
-Switches
-Power transformer
-Power choke
-Secondary rectifier
-Auxiliary supply
Are you still facing defects, or do you just wish get a better efficacy?
If you just wish to get a better efficacy, then the simple step towards a full bridge will not change much.
To improve efficacy you will have to tackle all the losses that are distributed over the entire circuit.
Here the typical strong contributors
-Switches
-Power transformer
-Power choke
-Secondary rectifier
-Auxiliary supply
Before you switch to resonant operation, you will need to determine the voltage transfer function over the frequency range you intend to operate. You have two options there, You either move the Bus voltage up/down in response to load transients (while operating at resonance all the time) of you move the frequency so that the primary tank impedance matches your output impedance. From the frequency you selected and the resonant inductance (I suspect you are using the leakage) I suspect you have a very low Q resonant tank (Cr will be large compared to Lk) and that you will have trouble operating the IGBT's over that wide a frequency range (result of low Q). If you intend to move the bus up and down, you will need an intermediate buck/boost stage since the PFC front end will have a low loop bandwidth and will not be able to skew fast enough to handle load transients. If you do not have a PFC front end then you will need a boost or buck to do the voltage movement for you.
In any event, dead time control becomes critical for the resonant topology.
I hope my comments helped.
Tony
THX dtproff and ChocoHolic.
Winter is still strong, and my igbt is not arrived.
I replace igbt whit irgp4063dpfb (of recovery/ second hand), I set the frequency at 25khz and i test it at 14.4v / 130A loud.
In 10 minutes the temperature of heatsink incrise at 58C(start from 24C) and heatsink is cooling with a small fan.
waveform primar (frequency is wrong) http://oi41.tinypic.com/ieil2p.jpg
What power can be obtained with irgp4063dpfb in half bridge configuration, at ~230V, frequency at 25khz, and temperature of heatsink at 80C ? (in datasheet i dont see Load Current vs frequency vs temp)
Winter is still strong, and my igbt is not arrived.
I replace igbt whit irgp4063dpfb (of recovery/ second hand), I set the frequency at 25khz and i test it at 14.4v / 130A loud.
In 10 minutes the temperature of heatsink incrise at 58C(start from 24C) and heatsink is cooling with a small fan.
waveform primar (frequency is wrong) http://oi41.tinypic.com/ieil2p.jpg
What power can be obtained with irgp4063dpfb in half bridge configuration, at ~230V, frequency at 25khz, and temperature of heatsink at 80C ? (in datasheet i dont see Load Current vs frequency vs temp)
I am afraid, I can't tell you exactly what power level you can hit for the IGBT's I usually use MOSFET's (efficiency) but I can tell you as a practical matter most engineers would run a half bridge to about 750W. Above that I would be using a full bridge. This is just a rule of thumb... you can push them way higher than that just as you can push a flyback to the KW range.
Tony
You can simply reduce the frequency from 40kHz down to 25kHz without running into saturation?
Which number of turns do you have on the primary and which core?
The possible power with certain switches always depends on the very detail of switching, means the happenings +/- few hundreds of ns around transition.
That's why it is usually not stated in data sheets.
Which number of turns do you have on the primary and which core?
The possible power with certain switches always depends on the very detail of switching, means the happenings +/- few hundreds of ns around transition.
That's why it is usually not stated in data sheets.
...but let's assume you are close to the conditions which was used for the switching energies in the data sheet.
Then you would have 80W switching losses in each device... and let's add another 20W for conduction losses (20A x 2V close 50% duty cycle ). Means 100W.
Rth=0.45K/W.
Consequently the junction will be approx 45 Kelvin above case temp.
Coming from Tjmax=175C, we have to keep the case below 130 C.
If the heat sink is at 80 C, then we can drop max. 50K between case and heat sink. Means even if you can ensure a low Rth between case and heat sink of 0.5K/W, you are at the limit....
Better keep the heat sink below 60C.
Then you would have 80W switching losses in each device... and let's add another 20W for conduction losses (20A x 2V close 50% duty cycle ). Means 100W.
Rth=0.45K/W.
Consequently the junction will be approx 45 Kelvin above case temp.
Coming from Tjmax=175C, we have to keep the case below 130 C.
If the heat sink is at 80 C, then we can drop max. 50K between case and heat sink. Means even if you can ensure a low Rth between case and heat sink of 0.5K/W, you are at the limit....
Better keep the heat sink below 60C.
Hello ChocoHolic . THX for replay.
I set the frequency at 25khz becose igbt drown more current at this frequency.
The ferrit kaschke R63 material k2004. Trafo windings 13/2+2
I test smps whit different loud, start from 50A, 80, 130, and 150A at 25khz.
The final test when a have 150A loud and add anader loud 30A ...I heard a strange sound from trafo or choke coil (zzzz) and low side igbt blow.(it is posible to have a big peak current when conect a second loud 30A, ).
The waveform on gate igbt is very dirty and depend of output loud. (150A= very dirty waveform gate, but amplitude is mentain at 15V)
Next configuration, full bridge igbt
I decided to try optodrivere HCLP3120 instend of GDT.
I wait for 2 weeks my componet but are still under red code.
Excuse me English!
I set the frequency at 25khz becose igbt drown more current at this frequency.
The ferrit kaschke R63 material k2004. Trafo windings 13/2+2
I test smps whit different loud, start from 50A, 80, 130, and 150A at 25khz.
The final test when a have 150A loud and add anader loud 30A ...I heard a strange sound from trafo or choke coil (zzzz) and low side igbt blow.(it is posible to have a big peak current when conect a second loud 30A, ).
The waveform on gate igbt is very dirty and depend of output loud. (150A= very dirty waveform gate, but amplitude is mentain at 15V)
Next configuration, full bridge igbt
I decided to try optodrivere HCLP3120 instend of GDT.
I wait for 2 weeks my componet but are still under red code.
Excuse me English!
Last edited:
I think it is possible to saturate choke coil, it is built with an unknown core (yellow iron core R39)
I have calculated the AL of this unknown core : L(12000nH)= Al*10 (turns) x 10 (turns) => Al =120
and have simulated the current of the choke coil : Halbbrückengegentaktwandler
The peak current is 156A for 130Aloud , Current*turns=1000 A (in most cases)
I have 7 turns on choke coil => Crurrent x 7 = 1000A => Current = 142A
Are my calculations correct ? Is it possible for this core to saturate an cause the IGBT to blow ?
I have calculated the AL of this unknown core : L(12000nH)= Al*10 (turns) x 10 (turns) => Al =120
and have simulated the current of the choke coil : Halbbrückengegentaktwandler
The peak current is 156A for 130Aloud , Current*turns=1000 A (in most cases)
I have 7 turns on choke coil => Crurrent x 7 = 1000A => Current = 142A
Are my calculations correct ? Is it possible for this core to saturate an cause the IGBT to blow ?
Hi mgm2000,
yes if you have reduced the frequency you have increased the peak current in the output choke.
156A peak at 130A load means 52A pk-pk ripple which is acceptable at this current level, it is 40% of the output current. You can even reduce a bit the inductance value and increase the ripple. Doing this will reduce the B on the core possibly avoiding the core saturation.
Check also your output capacitors ripple ratings! Eating 52A pk-pk is not a joke for the capacitors.
Btw. If you use unknown cores you are looking for problems. From your descriprion it seems an iron powder core. These core lose inductance progrssively as the current increase and don't have a sharp saturation characteristic as plain ferrite cores.
I don't think that your smps blows because of saturation; check your gate driver. Dirty waveforms on gates are likely to kill everything.
Good luck for your project
-marco
yes if you have reduced the frequency you have increased the peak current in the output choke.
156A peak at 130A load means 52A pk-pk ripple which is acceptable at this current level, it is 40% of the output current. You can even reduce a bit the inductance value and increase the ripple. Doing this will reduce the B on the core possibly avoiding the core saturation.
Check also your output capacitors ripple ratings! Eating 52A pk-pk is not a joke for the capacitors.
Btw. If you use unknown cores you are looking for problems. From your descriprion it seems an iron powder core. These core lose inductance progrssively as the current increase and don't have a sharp saturation characteristic as plain ferrite cores.
I don't think that your smps blows because of saturation; check your gate driver. Dirty waveforms on gates are likely to kill everything.
Good luck for your project
-marco
...urghs, besides issues with the gate drive...
From the given data core saturation of the output choke is likely.
You say R39... I am guessing an Ae around 100mm² (Ok?)
You say AL=120 and 7 turns ==> approx 6uH (correct?)
With 156A this would theoretically translate to a flux density around 1.3T.
Core will saturate earlier, inductance will drop.
Be sure to have saturation issues in the output choke, also previously at 40kHz.
It is not evident whether this was the reason for your defects or not, but it is possible. It would cause higher max currents in the IGBTs.
What has happened with your gate drive? The older screen shots were not catastrophic.
From the given data core saturation of the output choke is likely.
You say R39... I am guessing an Ae around 100mm² (Ok?)
You say AL=120 and 7 turns ==> approx 6uH (correct?)
With 156A this would theoretically translate to a flux density around 1.3T.
Core will saturate earlier, inductance will drop.
Be sure to have saturation issues in the output choke, also previously at 40kHz.
It is not evident whether this was the reason for your defects or not, but it is possible. It would cause higher max currents in the IGBTs.
What has happened with your gate drive? The older screen shots were not catastrophic.
THX ChocoHolic and mag
I wonder if there may be problems from GDT.
GDT is made from kaschke R16. It is to small core for drive 2 igbt?
http://oi40.tinypic.com/292766u.jpg
And another problem is, I have no decoupling capacitor on 310V network. It is possible to influence switching GDT ?
I wonder if there may be problems from GDT.
GDT is made from kaschke R16. It is to small core for drive 2 igbt?
http://oi40.tinypic.com/292766u.jpg
And another problem is, I have no decoupling capacitor on 310V network. It is possible to influence switching GDT ?
Last edited:
I was left with no igbt and use all that falls in my hand.
Today I used a mosfet fda59n30 (yes I know 300v)
Test 130A waveform from gate drive (input capacitans 4670pF)
http://oi43.tinypic.com/2003alv.jpg
Today I used a mosfet fda59n30 (yes I know 300v)
Test 130A waveform from gate drive (input capacitans 4670pF)
http://oi43.tinypic.com/2003alv.jpg
R16 should be large enough. Easily.
How many turns do you have? Looks like more than 20 turns...
Just calculate the voltage time product and the resulting flux density.
For rectangular signal shape:
U*Ton=N*deltaB*Ae
No rail cap?
Means your 310V drops to zero?
In any case rail decoupling does have influence.
Starting to wonder... appearing to me that your SMPS worked better than one should expect (no rail cap and saturating output choke and since today dirty gate drive, but still delivering 2kW without immediately blowing.
).
How many turns do you have? Looks like more than 20 turns...
Just calculate the voltage time product and the resulting flux density.
For rectangular signal shape:
U*Ton=N*deltaB*Ae
No rail cap?
Means your 310V drops to zero?
In any case rail decoupling does have influence.
Starting to wonder... appearing to me that your SMPS worked better than one should expect (no rail cap and saturating output choke and since today dirty gate drive, but still delivering 2kW without immediately blowing.
).If we are speaking of 1.3T in the choke it will saturate, no matter of which material it is made. If it is an iron powder core at 130A you will not have 6uH but probably much less. Your peak current will increase a lot if the choke start to loose significant inductance under load.
You need for sure a core with greater Ae.
GDT: if you have decreased the frequency you need to be sure that the GDT can handle the pulse without saturating.
How many turns you have on the gdt? What is the core Ae?
Check your B in the GDT at 25kHz and if needed increase the number of turns.
Bpeak=Vdrive*Ton/N/Ae, try to keep Bmax<250mT for the GDT.
From the picture it seems that you GDT is wound really bad. Wind it trifilar to reduce the leakage inductance as much as possible.
You need for sure a core with greater Ae.
GDT: if you have decreased the frequency you need to be sure that the GDT can handle the pulse without saturating.
How many turns you have on the gdt? What is the core Ae?
Check your B in the GDT at 25kHz and if needed increase the number of turns.
Bpeak=Vdrive*Ton/N/Ae, try to keep Bmax<250mT for the GDT.
From the picture it seems that you GDT is wound really bad. Wind it trifilar to reduce the leakage inductance as much as possible.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.