Replace ultra fast recovery diode with schottky?

[dsctech]

I am working on improving an old half bridge ATX power supply. Since most computers use mostly 12V, I want to upgrade the rectifier on the 12V rail. Right now it has a 20A Ultra fast recovery diode, with 200V peak repetitive reverse voltage. I want to add at least a 20A schottky, or maybe 30A (I have some lying around) mainly to increase the efficiency. It also has a spot on the PCB to add another rectifier in parallel, which I'm hoping would also increase the efficiency. My main question is, the schottky's don't have as high of peak reverse voltage ratings, and I'm wondering if using a schottky good for 45V or 60V would be good enough. I appreciate any help/insight.

I'm also going to replace all of the chinese electrolytics with high quality japanese lytics

 

[WimRFP]

When the peak reverse voltage exceeds the rating for the schottky, I would not try it. Mayby you can measure the voltage waveform across or at the input of the rectifier under varying load. Make sure that you don't create a short circuit via a probe lead or other ground connection. Note that reverse spikes can be very narrow due to the recovery characteristic of the rectifier. If your oscilloscoope has low bandwidth, you may not see the actual spike or ringing. When using shottky rectifiers, the spike may reduce significantly.

If you don't have an oscilloscope, you can measure the negative (reverse) peak voltage with a seperate small signal fast diode (for example some 1N4148 in series with a 47 Ohm resistor) together with a multimeter with high impedance. Maybe you need to add a 1nF capacitor parallel to your voltmeter to make sure you measure the real peak voltage.

 

[dick_freebird]

Schottkies above 60-100V or so, tend to have guardrings
which then start to store charge at higher Vf (If) operation.
That then kills the no-storage-time feature. You may have
to look real carefully at ratings / specs for turnoff.

But I really wonder how much you stand to gain efficiency-
wise. I'm thinking at 12V it's single-digit-%, tops - and you
could figure this by the Vf@If difference divided by 12V.

Figure the 12V rail means peak reverse real voltages about
25V on the bridge (?) so 2X margin puts you at 50V or so.
But there's probably some overshoot scenario that makes
more a good idea. And silicon diodes are all pretty similar
in price so 100V vs 200V may be a nit and they just stock
the higher voltage in the parts system, or something.

 

[schmitt trigger]

I'm afraid that you will really require an oscilloscope to view the ACTUAL reverse voltage, which includes any ringing.

Schottkies have excessive (as compared to a silicon junction diode) reverse currents as the reverse voltage approaches its PIV rating. These actually increase with temperature. Therefore, it is customary to provide more voltage derating on Schottkies than in junction diodes.

 

[dsctech]

Hey guys, thanks for the replies. I have a scope but I hardly know how to use it. I tried replacing the schottky with a different schottky with 100V reverse voltage and the PSU will not turn on. I think it is detecting it as a short because it makes a quiet tick when I try to turn it on, but just once. And it's still giving 5.05V on the 5VSB. It originally had a shorted NPN transistor on the primary, and I replaced that with a tested good one. I'm guessing this old design just doesn't like the schottky?

 

[WimRFP]

Do you have a circuit diagram of the supply? If not try to locate the control chip (such as TL494), number of primary switching transistors, etc. Maybe you can find a circuit that is almost yours. You can find some example schematics at: http://danyk.cz/s_atx_en.html of various ATX supplies.