EC ATX-350W mod

[Behemot]

Continue from here.

I have indentified positive voltage feedback from this scheme (the input on pin 1 of 7500B) below. It is not mine PSU but I can find similarities so I will refer to it here and there.


If I wanted to decrease voltage on +3,3 V, do I decrease the R76?

[Behemot]

Re: EC ATX-350W mod

Changing the -12V coil for small, resistor-type inductor had brought the rail almost to ATX specification with arround 130 mV peak-to-peak. Squeezing 680uF cap on +12V rail brought the rail down to arround 120 mV peak-to-peak. It is still out of specs, but definitelly better than 180 mV, right This was aprox. 80 % of rated power. This thing could definitelly use coils on both +5 and +12 V. The +3,3V rail has one and it's ripple is well within specs, the other rails are hitting or exceeding it. Unfortunatelly, there is no room for any coils inside.

Efficiency wise, it's POS. Efficiency at maximum power (calculated ass aprox. 330 W from actual - not nominal - output power and voltage) before OPP intervened was aprox. 67 %. I think that it is easier to just mark it as 300 W than push it to 350 W for the cost of being out of specs anyway and having terrible efficiency. I don't want to know what efficiency it possibly had before I have changed all that silicon?

So I will just try to bring the +3,3 V rail down from the terrible 3,6 V maximum load and leave it be, instead of pushing the OPP further. I think this was good lesson for me so far, don't need to push my luck

[PeteS in CA]

Re: EC ATX-350W mod

Can't really make out the reference designators, but I see the PWM feedback voltage sums the +5V, +3.3V and +12 V. So you can't really adjust those O/Ps independently. I'm sure you already know that. I'm not sure whether increasing or decreasing the resistor from the +3.3V will decrease the +3.3V. You could find out by paralleling that resistor with a ~100K - make a small change - and see which way it changes. Another thought occurs to me. What happens to the +3.3V if you add a small pre-load, maybe a 100 ohm, .25W or a 47 ohm, .5W?

[Behemot]

Re: EC ATX-350W mod

There are some small load resistors, I didn't noticed them sooner. Not that they do their work properly enough, voltages are fine, but ripple is high until I increase load on all raisl incl. +5 V SB. But the major problem is, due to changing the rectifiers for shottkies, Vdrop decreased. Now the +3,3 V is over spec even with no load and it increases all the time up to 3,6 V maximum load as the supply pushes more power to whole secondary.

I know it's group regulated, but since other voltages are very well within spec, more on the higher side too, there is not problem in decreasing all of them. The +5V rail may drop slightly below 5 V. I am gonna learn from this next time when I will just add second superfast to +3,3 V (average drop will be than between single shottky and single superfast) and two schottkies for +5 V to make as low drop as possible.

I can try to add some pot, good idea, there is 270k resistor in mine supply so I will have to add something bigger. Or I will just replace the resistor with ~270k pot and change it directly.

[Behemot]

Re: EC ATX-350W mod

I've tried messing with resistors on both +5 V and +3,3 V in there, trying both +50 and -50 % resistance and it still is arround 3,5-3,6 V which is still high. I guess the Vdrop reduction is too significant. Do you think it may harm something, going up to arround 3,7 V?

[PeteS in CA]

Re: EC ATX-350W mod

I'm one of the resident analog dinosaurs, so I don't really know what the digital electronics will tolerate. IIRC from my days at Delta Products, the spec tolerance for the +3.3V is +/- 3%. 3.6-3.7V is about 10% high. Of course, the Delta PC P/S designs I saw used mag-amp regulation on the +3.3V. The +5V and +12V were summed for the main loop.

For the design you are looking at, I'm thinking reduction in the rectifier forward drops for all the high current outputs with some degree of balance might have improved your results. You mentioned earlier that all the outputs are on the high side, with the +3.3V being the worst case. Maybe changing the resistor from PWM pin 1 to ground slightly (higher value?) would reduce all three high current outputs' voltages, bringing your +3.3V in-spec high, with the +5V and +12V being in-spec low?

[Behemot]

Re: EC ATX-350W mod

UAAAAAAAAAAAAAAAAAA THE BLOODY THING IS IMMORTAL, KILL IT WITH FIRE!!!!

Using 1 megaohm resistor I drove the resistance from the pack to ground to 4k2 from 4k1. Cause even with these high values there is something accross them making the average resistance go down. Maybe small internal resistance of the 7500, I don't know.

Anyway, now 5 V is under 5 V as well as 12 V is under 12 V. Not remember exact numbers cause it is not important. The +3,3 V is still 3,47 V no load! Uaaaaaa

[Th3_uN1Qu3]

Re: EC ATX-350W mod

Are you 100% sure the 3.3v isn't regulated by a TL431 somewhere? I've yet to see a PSU which has 3.3v in the feedback loop of the TL494.

[Behemot]

Re: EC ATX-350W mod

All those three voltages are really being summed and than conencted to the 7500/494 with two resistors between this and ground, the same as in the Seventeam.

And, well, if I load ONLY the +3,3 V with some minor load, than the voltage drops to arround 3,3 V. But if I load any other rail too, it jumps back cause there is just more power being pushed to secondary.

I really think they have beefed the wirings at main transformer counting with the huge voltage drop on superfasts. Now when I swapped it for schottkies, the voltage jumped as much as 0,65 V (schottkies usually have half the Vdrop as superfasts), it is just too much cause it is almost 20 % of that rail nominal voltage. So no matter what I do, there is more power being pushed to the wirings than anything is consuming in there. That's my theory anyway

[Behemot]

Re: EC ATX-350W mod

One great idea came to my mind as a last try to bring the +3,3 V down. It uses mag-amp, right? So can I affect it somehow by changing the number of turns on the toroid before rectifier, or is it much more dependant on it's core, so changing number fo turns won't help it? Cause I am thinking, if I used more turns, than it should saturate sooner and give lower voltage on output…?