I'm designing 1kW multiphase forward converter.
From DC 12V to DC 60V. (1kW)
Since the input current can be very high, more than 80Amps.
I think it is better to be done in 4 phase.
In the attached picture.
There are two triangle wave form.
The first one is non inverter.
The other one is inverted.
Then the triangle fed to LM319 comparator. And set the duty cycle to 30% square wave.
Now I have 2 phase non overlap square wave.
My plan is fed this square wave to (OSC) pin 3 of SG3524.
Then the RT CT set to 90% of square wave frequency.
Each of SG3524 run in 180 degree out of phase.
So I will have 4 phase signal. But, the problem is. Im not sure, will this work?
From DC 12V to DC 60V. (1kW)
Since the input current can be very high, more than 80Amps.
I think it is better to be done in 4 phase.
In the attached picture.
There are two triangle wave form.
The first one is non inverter.
The other one is inverted.
Then the triangle fed to LM319 comparator. And set the duty cycle to 30% square wave.
Now I have 2 phase non overlap square wave.
My plan is fed this square wave to (OSC) pin 3 of SG3524.
Then the RT CT set to 90% of square wave frequency.
Each of SG3524 run in 180 degree out of phase.
So I will have 4 phase signal. But, the problem is. Im not sure, will this work?
Attachments
This 12VDC input.
I will not use full bridge.
Push pull is okay, but centre tapped transformer construction become difficult.
Beside 4 phase forward will have lower ripple current in input capacitor. Very fast transient load.
Single phase push pull will need a capacitor with 70A ripple current rating.
I will not use full bridge.
Push pull is okay, but centre tapped transformer construction become difficult.
Beside 4 phase forward will have lower ripple current in input capacitor. Very fast transient load.
Single phase push pull will need a capacitor with 70A ripple current rating.
Input capacitor ripple current is duty cycle dependent, and for high duty cycles there is not improvement at all despite how much phases are employed. The major improvement happens at 50% duty cycle only...
I wouldn't use forward topology nor multi-phase approaches for such a SMPS. A single push-pull toroid with a single output inductor will do for 1KW.
I wouldn't use forward topology nor multi-phase approaches for such a SMPS. A single push-pull toroid with a single output inductor will do for 1KW.
actually ive found that the ripple is best with just a little less than 50% DT..adding some dead time...i havent quite been able to solve the FT to determine why
Eva,
With single phase push pull, peak input current to the mosfet will be 110A, plus it must block 24V, twice input voltage.
Transformer peak current more than 100A.
And thats 70A ripple current is for 30% duty cycle.
I think thats not good.
Have you read "slup231.pdf" from TI?
Btw, what do you think about my idea?
Will this work?
With single phase push pull, peak input current to the mosfet will be 110A, plus it must block 24V, twice input voltage.
Transformer peak current more than 100A.
And thats 70A ripple current is for 30% duty cycle.
I think thats not good.
Have you read "slup231.pdf" from TI?
Btw, what do you think about my idea?
Will this work?
But why 30% duty cycle? Such a push-pull converter should be designed to run at 100% duty cycle for 9V input (low line limit) and thus 75% for 12V input.
This would yield approx. 113A peak primary current and 85A average input current, so input capacitors will have to sink 85A during 25% of the time and source 113-85=28A during 75% of the time. RMS capacitor current would be: sqr(85^2*.25+28^2*.75) = 49A
In a two phase push-pull approach with same turn ratios, each phase would have to deliver half the output current, that's 113/2=56A. But adding current consumption from both phases we get 113A during 50% of the time and 56A during 50% of the time. Ripple capacitor current would be 29A rms (this time a perfect square wave).
Conclusions: Now we have two sets of primary switches, two sets of gate drivers, a very complex layout requiring a double sided metallized PCB, an ultra-complex control circuit (forget about SG3525 and TL494 since two synchronized out of phase sawtooths are required), two transformers to wind, two sets of output diodes, two output inductors, more heatsink insulators and more surface required for placing power devices... and we haven't been even able to reduce to half the amount of required input capacitors!!! since ripple current is now 29A rms versus 49A.
And now we have introduced a potentially troublesome issue with current sharing between phases since nothing guarantees both phases to carry the same current except two independent inner current mode control loops driven by a single outer voltage mode loop.
My advice: Double the amount of input capacitors instead of doubling the entire circuit 😀😀😀 (and think twice and do your own numbers to check whether what these application notes propose is actually advantageous or not in practice).
This would yield approx. 113A peak primary current and 85A average input current, so input capacitors will have to sink 85A during 25% of the time and source 113-85=28A during 75% of the time. RMS capacitor current would be: sqr(85^2*.25+28^2*.75) = 49A
In a two phase push-pull approach with same turn ratios, each phase would have to deliver half the output current, that's 113/2=56A. But adding current consumption from both phases we get 113A during 50% of the time and 56A during 50% of the time. Ripple capacitor current would be 29A rms (this time a perfect square wave).
Conclusions: Now we have two sets of primary switches, two sets of gate drivers, a very complex layout requiring a double sided metallized PCB, an ultra-complex control circuit (forget about SG3525 and TL494 since two synchronized out of phase sawtooths are required), two transformers to wind, two sets of output diodes, two output inductors, more heatsink insulators and more surface required for placing power devices... and we haven't been even able to reduce to half the amount of required input capacitors!!! since ripple current is now 29A rms versus 49A.
And now we have introduced a potentially troublesome issue with current sharing between phases since nothing guarantees both phases to carry the same current except two independent inner current mode control loops driven by a single outer voltage mode loop.
My advice: Double the amount of input capacitors instead of doubling the entire circuit 😀😀😀 (and think twice and do your own numbers to check whether what these application notes propose is actually advantageous or not in practice).
Eva,
Thanks for your advice. You're so right about what you said.
Complex circuit will not be a problem for me. I used to build double side SMD circuit with only fax paper and laser jet printer.
And there a few thing I don't understand
Since I never build 2 phase push pull.
Do you mean,
1. I have to wind the transformer with 2 centre tapped primary
and 2 centre tapped secondary (All in the same toroid
core)?
2. Then connect the secondary output together, and connect it
to the diode, --to--> inductor --to--> capacitor --and-> output ?
Am I right?
Thanks for your advice. You're so right about what you said.
Complex circuit will not be a problem for me. I used to build double side SMD circuit with only fax paper and laser jet printer.
And there a few thing I don't understand
Since I never build 2 phase push pull.
Do you mean,
1. I have to wind the transformer with 2 centre tapped primary
and 2 centre tapped secondary (All in the same toroid
core)?
2. Then connect the secondary output together, and connect it
to the diode, --to--> inductor --to--> capacitor --and-> output ?
Am I right?
A two phase push-pull is made out of two independent push-pull converters (including independent output filters, but with its outputs joint in some way after them) whose oscillators are synchronized in such a way that one converter is half cycle out of phase respect to the other.
This is not accounting for current sharing... Altough making each converter produce 30V DC output and connecting both outputs in series to get 60V may do the trick... 😀
As a control circuit, you may try a master SG3525 and another slave SG3525 whose sync line is fed with a pulse anytime the sawtooth voltage of the master IC reaches a certain point (trimmed to be at half cycle ).
This is not accounting for current sharing... Altough making each converter produce 30V DC output and connecting both outputs in series to get 60V may do the trick... 😀
As a control circuit, you may try a master SG3525 and another slave SG3525 whose sync line is fed with a pulse anytime the sawtooth voltage of the master IC reaches a certain point (trimmed to be at half cycle ).
- Status
- Not open for further replies.