14V@ 20 Amps PSU: Smoked PCB after successful testing

[Razor_Edge]

Hi Everyone,

I have been working on to build a 14V, 20 Amps (230VAC +/- 10% input) from quite sometime now.

I recently rigged up complete PSU and loaded it to 13.8V @ 14 Amps. everything seemed fined. Since the Fets (IRF BC40) were a bit too hot on continuous run, i went for a new PCB design with more space for heatsinking.

After rigging up the whole unit, i checked and nothing seems to work now.

Driver buffer is getting hot, gate drive is getting load like hell, Fets have blown up 2 times and the output is rising from 0Volts to 13V in about 5~6 secs.

I am attaching the schematics for reference. Any suggestions on areas of improvement are most welcome.


NOTE: C37 is shorted and their are some typo's in input bulk caps, R23 and R24 open, R21 and R20 are 22K each.

Regards
Chaturvedi

 

[Razor_Edge]

snapshots of gate drive

There horrible gate drives were very clean on the 1st test iteration

 

[Eva]

Sorry but your gate drive seems to be completely wrong. I doubt it could have ever worked. Use the search engine, there are several threads on gate drive techniques.

 

[Razor_Edge]

Sorry but your gate drive seems to be completely wrong. I doubt it could have ever worked. Use the search engine, there are several threads on gate drive techniques.

Hello Eva,

Thanks for a prompt reply.
Yes, you are right. It is not working now. I am sure you have had a good look before making a comment.

My idea in this gate drive scheme was to drive the primary side Fets using a push pull drive method. It did work in the last iteration. I am using 3 parallel buffer of ULN2003 for each "push and pull".

I have read many threads on gate drive scheme and in most cases read the comments made by you. Problem is, i have not been able to find a fast enough transistor from a local vendor which can switch 1 amp current for the primary side of gate drv Xfmr.

Do you think SG3524 can drive BC40's directly?

Please comment.

Regards
Chaturvedi

 

[darw82]

I recently build 12V @60Amp PSU (Half bridge).
At such power level, gate drive circuit are really important thing.

The PWM controller is SG3524.
For the gate drive buffer, I use MC33152, and the gate drive xformer arrangment is exactly the same in slup169.pdf at page 36 (Fig 39)

The only thing I add from the picture is :
1 . Schottky diode from VCC to OUT_A and OUT_A to ground
2. Schottky diode from VCC to OUT_B and OUT_B to ground
3. Back to back 15V zener at each of the xformer output.

 

[darw82]

Eva,
When I load my 12V PSU to 60Amp, the output voltage drop to 9 - 10V. I have not try to fix it yet. Could you please tell me, what may caused it?

1. Is it caused by unstable feedback loop?
2. Do you think, insufficient of output capacitor may cause voltage drop?


The transformer winding are completely interleaved. The PWM controller are located in secondary side, so there is no delay in feedback loop.

 

[Eva]

Hello Eva,

Thanks for a prompt reply.
Yes, you are right. It is not working now. I am sure you have had a good look before making a comment.

My idea in this gate drive scheme was to drive the primary side Fets using a push pull drive method. It did work in the last iteration. I am using 3 parallel buffer of ULN2003 for each "push and pull".

I have read many threads on gate drive scheme and in most cases read the comments made by you. Problem is, i have not been able to find a fast enough transistor from a local vendor which can switch 1 amp current for the primary side of gate drv Xfmr.

Do you think SG3524 can drive BC40's directly?

Please comment.

Regards
Chaturvedi

Don't you have BD139 and BD140 bipolar transistors there?

To drive MOSFETs, you have to drive the transformer in a totem pole fashion rather than push-pull, the circuit should be able to sink and source current from the primary. You can use several paralleled buffers from a CD4050 or CD4049 to get rail to rail low-impedance logic gate-drive signals out of the SG3524 (the first thing that is going wrong in your circuit). Also, proper dead time is easily implemented with the help of the CD40xx and RCD networks because CMOS 4000 series switch at Vcc/2. Then you can drive some emitter-follower buffer transistors with these signals and then the transformer (through a 1uF DC blocking capacitor). Also, you can move the turn-off buffers (PNP) to the gate drive side ( transformer secondaries) and leave the 4000 buffers driving the transformer directly for turn off.

 

[TechGuy]

Eva,
When I load my 12V PSU to 60Amp, the output voltage drop to 9 - 10V. I have not try to fix it yet. Could you please tell me, what may caused it?

1. Is it caused by unstable feedback loop?
2. Do you think, insufficient of output capacitor may cause voltage drop?


The transformer winding are completely interleaved. The PWM controller are located in secondary side, so there is no delay in feedback loop.

Are your Half-Bridge Caps being saturated (ie too small for your output load)? What is the measured ESR of the Half-Bridge caps at the PWM operating frequency?

 

[luka]

Hi darw82

What is your main trafo ratio? It could be that you didn't wind enough turns for secondary, or I^2R losses.
Is it regulated? Probably it is.
what input caps did you use? They should be at least 1000uF each

 

[luka]

Hi Razor_Edge

Would it be easyer to redo PC psu, I did that and I have output current 12.5v@25A and can go over, if rectifier can handle it

 

[Razor_Edge]

Thanks Eva,

Yes BD139 and BD140 are both available but the cost is about 0.75$ each.

Design requirement is to complete the design for less than 22$...!

But, as you say. There is a big difference between working design and a good design... 🙂

I will check it out ASAP and post results.

Thank you once again.

Hi Luka,

i could not get you properly. You asked me about redoing PC psu. Is it possible to convert that 180 watts PSU to 280 watts just by redoing xfmr and output inductor?

I mean, there is hardly any space for big components.

Could you please elaborate?...

Best regards
Deepak Chaturvedi

 

[luka]

Hi

You can redo it. You would need to rewind secondary of trafo, remove all others(sec.) and connect the new winding to where 5v line was. Then you would have to work on feedback. There are two thing that you would need to change. One is overvoltage for 5v line from before would need to be removed or lifted up to above your voltage. And other thing is set feedback to work at yout voltage, ei. instead of 5v before to 13v now, or whatever you want.

Now 180w supply can be driven to 180w, probaby more, but not too much. My is/was 230w I think, but because of big transistors on primary side(they are original from the start), I got it over 300w for few seconds.

The thing you have to watch out is original 5v line current rating, as your new line would run on same track, you can't get much more then it was rated. (5v@25a can be 14v@25a, not 8v@40a)

Aslo output inductor and filter caps will have to be changed...

 

[Razor_Edge]

redoing PC PSU

Hi Luka,

Thanks, I will check out ASAP.

You are really fast in replying.

An out of the box question. Are you self employed?

 

[luka]

Hi

You could say self employed, but to tell the truth, I am still in school... so no luck of making big €€ 😀

Try to get Everyday Practical Electronics Magazine, 7.2006, inside is article where you can read this, I done my based on thet article.
If you won't able to get it, I will try to send you via email...

 

[Razor-Edge]

Don't you have BD139 and BD140 bipolar transistors there?

To drive MOSFETs, you have to drive the transformer in a totem pole fashion rather than push-pull, the circuit should be able to sink and source current from the primary. You can use several paralleled buffers from a CD4050 or CD4049 to get rail to rail low-impedance logic gate-drive signals out of the SG3524 (the first thing that is going wrong in your circuit). Also, proper dead time is easily implemented with the help of the CD40xx and RCD networks because CMOS 4000 series switch at Vcc/2. Then you can drive some emitter-follower buffer transistors with these signals and then the transformer (through a 1uF DC blocking capacitor). Also, you can move the turn-off buffers (PNP) to the gate drive side ( transformer secondaries) and leave the 4000 buffers driving the transformer directly for turn off.

Hello Eva,

Attached is the corrected gate drive. Please comment. Also, i have been using EE25 for gate driving, how about EE16 or may be EE13.5.

You also suggested CD4000 series for implementing dead time how about a diode resistor network (for each outut of SG3524), similar to the ones used at the mosfet gate? 🙄

 

[luka]

Hi

Gate drive looks good now, you can use smaller cap, since freq. is high, 470nF should be still big enough.

And be sure that you connect gate trafo right

 

[Razor-Edge]

Hi Luka,

Thanks for the comments, how about dead time?

I never had any problems of shoot through in the earlier design. But, Eva raised my suspicions on the subject. Especially when i the PSU will be loaded to short circuit conditions.

Any suggestion?

 

[luka]

Hi

I think that dead time that you set by SG is good for most of the time, but you should listen to Eva. Get is that you try and see if you need anything more then "stock"

Will you have short circuit conditions??

 

[Razor-Edge]

Dead time on PSU

Hi Luka,

Yes, design requires short circuit conditions to be taken care off.

I have used a CT so sense more than 2 Amps on primary and inhibit drive pulses.

But havent checked it yet. I want to first load the PSU with 14V@20 Amps.

Thanks for your comments. Just waiting for Eva to suggest any word of caution.

Eva, please help.

 

[luka]

Hi

Do you want for supply to shutdown or limit output current as in charger?

I have lost my protection in proces of redoing it