Super 5vsb mod for Bestec ATX-250 12E

[mockingbird]

Re: Super 5vsb mod for Bestec ATX-250 12E

Very interesting and informative posts.

[PeteS in CA]

Re: Super 5vsb mod for Bestec ATX-250 12E

I was reading about that very thing in the data sheet. So I removed my junk box 20N60 and replaced it with a brand new Digikey Fairchild 5N60 FET. Fired it up, and the Vdd on the NCP1216 was very stable across all load resistors at 11.12 volts. That's up a little from the previous 20N60 which had NCP1216 Vdd at 6.5 volts. Data sheet says it normally runs at about 10 volts. So I will leave the new 5N60 FET in the circuit.

At a previous employer, the engineer had to optimize his design to improve the efficiency by a couple of percent. One of the things that helped was finding some MOSFETs with only slightly higher R(D-S On), but significantly lower gate capacitance. Decreasing the the power necessary to drive the gates more than balanced the extra power due to the increased R(D-S On).

The modern Fairchild Semiconductor came from National Semiconductor (now part of TI) spinning off discretes and some low profit jellybean parts they no longer wanted to make. Fairchild also bought several other companies' unwanted lines. So Fairchild markets a couple of MOSFET lines with different part numbering systems. When I was laid off from my previous employer, who makes motor controllers, Fairchild was their got-to vendor for new MOSFET requirements. Their controllers stressed MOSFETs, and at that time, Fairchild's products were bullet proof.

[PeteS in CA]

Re: Super 5vsb mod for Bestec ATX-250 12E

BTW, because of the duty cycle, when the schottky rectifier is conducting, it may be conducting a higher current than the current delivered to the load. So the V(f) when it is conducting will be higher than the V(f) at the load current.

[budm]

Re: Super 5vsb mod for Bestec ATX-250 12E

I worked for Fairchild Discrete division in the 80's, the last dept, was in the Power MOSFET fab in the last 4 years before National bought us out and shutdown the plant in 86'. I really do not miss the bunny suit, but missed a lot of good people I used to work with.
I am glad it is doing well on its own now.

[c_hegge]

Re: Super 5vsb mod for Bestec ATX-250 12E

Looks like it can do just about 5A before the 5VSB voltage goes out of spec on the lower limit (4.75V). Very impressive mod!

Not quite. It was actually 4.74A. an impressive mod nonetheless, though.

[everell]

Re: Super 5vsb mod for Bestec ATX-250 12E

Up to now, I have been focusing on the 5vsb winding of the transformer and the feedback winding that provides the Vdd supply voltage to the 5vsb pwm chip. There is one other winding on the transformer. It provides the Vdd supply voltage to the main pwm chip, a UC3842. This chip has a spec of 30 volts max. So measuring this feedback voltage rail with the different load resistors, the voltage varied from 17.43 volts to 39.04 volts. This is too high. Other brands of power supplies would probably have a lower voltage....the Bestec 12E 5vsb transformer is unique. This would also explain why the 220 uF/25 volt capacitor on this rail tends to explode!

Also, I am now not using the feedback winding to provide Vdd supply voltage to the 5vsb pwm chip. So I decided to tackle this problem first. Why not use it to power a fan? With the capability of a 5 amp load, some extra cooling wouldn't hurt. Using the existing diode, a 4.7 uF filter capacitor, and a 100K bleeder resistor, I tested using two different fans. One was a Jamicon from a Bestec rated at 12 volts and 0.19A. The other was a Delta (source unknown) rated at 12 volts and 0.34A. Power up and test using my assortment of load resistors, results were great. The voltage differences between the two fans was so slight as to be neglegible. Also in lookiing at the other feedback rail, the one providing Vdd supply voltage to the main pwm chip UC3842, the extra loading provided by the fan brought this voltage down to an acceptable level, measuring 17.68 volts to 24.81 volts. Two birds killed with one stone!

Next I want to look at the current limiting resistor. For now I have included a spreadsheet showing the results when adding a fan.

[everell]

Re: Super 5vsb mod for Bestec ATX-250 12E

The current sensing resistor which is used to shut off the 5vsb circuit for overcurrent condition is located between the source connection of the FET power transistor and ground. Up to now all of my testing was done using a 1.0 ohm resistor. I obtained an assortment of resistors measuring 1.0, 1.2, 1.3, 1.5, 1.7, and 2.0 ohms. Increasing the value of this resistor does indeed cause the 5vsb to shut off sooner. At 2.0 ohms the circuit didn't want to come up at all. The values 1.0, 1.2, and 1.3 worked best with little difference. The other two values shut the 5vsb down before it could reach the goal of 5 amps. So I reinstalled the 1.0 ohm resistor. Here is a spreadsheet showing the data from this test.

[everell]

Re: Super 5vsb mod for Bestec ATX-250 12E

I put the circuit board back in the case, and the 5vsb did not power up at all! What now!!!! I pulled the circuit board out slightly to get some readings, and they were all over the place. I thought I had blown something. So finally I took the circuit board completely out of the case. Seems I played cat and mouse for a few weeks. Didn't know what was happening. I could not get consistent results. I tried changing the 431 regulator chip. That didn't help. I decided to replace the optoisolator, so cherry picked one from my junk box and installed it. Still inconsistent results. Put it back in the case and still wouldn't start up. I was about ready to completely change the pwm add on circuit board, when I began to wonder if I had a ground loop. Took the circuit board out of the case. Added a wire about 3/4 inch in length between the circuit board ground and the screw holding the circuit board to the heat sink. The heat sink is soldered to AC common, so the ground path should be correct. That did it. Now I am getting stable readings again, in or out of the case. What a headache!!!

While I was experimenting, I thought that adding a 0.1 uF capacitor across the input filter capacitor of the 5vsb pi filter would help filter high frequency trash and make for a better output. Wrong, it only acted as additional loading and caused the output voltage to drop faster than normal. My thanks to ldsishere for his experiments with the snubber circuit. I think I will try that.

I removed the SB540 diode and installed a 10 amp TO220 shottky to the heat sink. Before removing the ground loop problem, it worked worse. After adding the additional ground to the pwm board, the 10 amp diode worked better. I am including a picture of the how I installed the diode in the 5vsb output circuit.

When cherry picking the optoisolator, I learned some new things about optoisolators. I think I will start a new thread to open a discussion on optoisolators.

Now to do some torture testing. With a one ohm load to get about 5 amps, the power supply only lasted about two minutes, then stopped. Increasing the load resistance slightly, it ran for about 20 minutes. Another slight increase and it ran for about 45 minutes. The only component getting hot was the transformer. When it would start to fail, the pwm would sense overcurrent and shut the circuit down. So far it has not totally blown the transformer. With a 2 ohm load the circuit seems to be very stable over time. That means that the transformer rated for 10 watts was stable at 12.5 watts, but when asking it for 25 watts......just couldn't handle it.

Final conclusion - designing this circuit for a continuous 5 amp output was a failure. BUT...the design was a definite improvement over the original two transistor circuit. The feedback windings on the original transformer required some thinking and circuit redesign.....the added fan (or some other load) was needed to bring those feedback voltages down. The higher voltage available in the feedback winding is probably the cause of so many 12E power supplies failing with a "blown up" 220 uF capacitor on the feedback rail.

[LDSisHere]

Re: Super 5vsb mod for Bestec ATX-250 12E

Your solution to the heat sink mounted diode is very creative. I was lucky with the SL-400 as making a TO-220 diode work only required me to slightly bend the legs.

I would be interested in you trying a NCP1200 and comparing the results with what you achieved with the NCP1216. Since the NCP1216 is supposed to be a newer version of the NCP1200 you would expect it to perform better however in actuality that may not be the case. I did a little bit of experimenting with the NCP1216 and it did not perform as good as the NCP1200 so I am wondering if you would have a similar experience.

[ben7]

Re: Super 5vsb mod for Bestec ATX-250 12E

Your solution to the heat sink mounted diode is very creative. I was lucky with the SL-400 as making a TO-220 diode work only required me to slightly bend the legs.

I would be interested in you trying a NCP1200 and comparing the results with what you achieved with the NCP1216. Since the NCP1216 is supposed to be a newer version of the NCP1200 you would expect it to perform better however in actuality that may not be the case. I did a little bit of experimenting with the NCP1216 and it did not perform as good as the NCP1200 so I am wondering if you would have a similar experience.

The newer one probably has different/better features, and it likely operates differently than the old one - that could be why, especially if you just swapped them without re-designing the circuit.

[everell]

Re: Super 5vsb mod for Bestec ATX-250 12E

I did a little bit of experimenting with the NCP1216 and it did not perform as good as the NCP1200 so I am wondering if you would have a similar experience.

In what way did you find the performance not as good? In our earlier experiments we were using the feedback winding from the 5vsb transformer to provide power to the NCP1200/1216. This should not be done. That is where some of our problems came from. Also something needs to be done with that feedback winding to keep the other feedback winding from going too high in voltage. And finally, the transformer is a limiting factor in overall performance. The transformer was originally designed to handle only 2 amps max in most of these power supplies.

[ben7]

Re: Super 5vsb mod for Bestec ATX-250 12E

And finally, the transformer is a limiting factor in overall performance. The transformer was originally designed to handle only 2 amps max in most of these power supplies.

Have you tried making your own transformer?

[everell]

Re: Super 5vsb mod for Bestec ATX-250 12E

Whereas I don't need a 5vsb that will handle more than 2 amps, I think making a transformer would get into more than I can handle. I learned a lot from this project. I think psu manufacturers could improve on the 5vsb to provide the added power. But economics being what they are, I expect to continue seeing lots of 2 amp 5vsb two transistor circuits.....and of course critical capacitors gone high esr and

[LDSisHere]

Re: Super 5vsb mod for Bestec ATX-250 12E

In what way did you find the performance not as good? In our earlier experiments we were using the feedback winding from the 5vsb transformer to provide power to the NCP1200/1216. This should not be done. That is where some of our problems came from. Also something needs to be done with that feedback winding to keep the other feedback winding from going too high in voltage. And finally, the transformer is a limiting factor in overall performance. The transformer was originally designed to handle only 2 amps max in most of these power supplies.

It was several months ago when I tried the NCP1216 so I do not remember specifics. I did this testing on the SL-400 and the only difference was the chip. On this supply, the 5vsb voltage would sag noticeably more using the NCP1216 as the load was increased. Your experiments with the optoisolator makes me wonder if it may cause the difference in the results. In other words maybe the NCP1216 did not play well with the same optoisolator that worked just fine with the NCP1200. I still have a "cheapo" supply setup for the NCP12XX boards that I may do some experimenting with to just compare the two chips. (When and if I get the time.)