Suggestion: Macron MPT-xxx and MPT-xxx2

[LongRunner]

Considering that momaka's MPT-301 had a seized Superred fan, I indeed don't exactly see fit to let it slide with no deductions. Pentium's comment that it "seems to run cool for an old half bridge unit" (a few surprisingly efficient half-bridge designs were tested in the previous el-cheapo round-ups), along with momaka's comment that it has good regulation for a group-regulated design (bettered only by Hipro IIRC) suggests that it could indeed be a worthy unit (for use in old PCs) after recapping (and a new fan if Superred), but that's something we won't know for sure until this year's round-up is complete.

[LongRunner]

In the earlier 80mm-fan version, which had a mains pass-through outlet, the wiring arrangement placed that outlet between the two stages of the EMI filter.
[…] if too much current is drawn through that outlet (especially with the unit in standby mode), the filter choke could get rather hot.

Confirmed; up to 4.17A (1000W @ 240V) is fairly safe (≈55K rise on the neutral winding, assuming 105°C as the maximum allowable), but 5.83A (1400W @ 240V) is pushing it and the full 10A would surely toast it. (The switch is rated 6A, so testing to about that seems fair.) However, the filter PCB does have two through-holes besides the inlet pins, so you can re-arrange the connections to put the outlet before the filter (and switch).

Another MPT-xxx nit I found is the use of a single 220V MOV (ZNR1) from one AC terminal to the mid-point of the bulk capacitors; on 115V this works fine, but on 230V it will just shunt the surges across one bulk cap and into the other (which one depending on the polarity of the surge). So I would recommend installing a 470V MOV in the position ZNR2 (just after the fuse); since these (at least those sold in 230V markets) usually lack the voltage selector, you can then remove ZNR1. Later PCB revisions also support the normal two MOVs across bulk caps (ZNR3 and ZNR4), and the MPT-xxx2 finally used them.

[LongRunner]

Here's what to do:

• Especially if altering the voltage feedback, remove the original output cables (or at least connectors), so no-one plugs in normal computer hardware and damages it. However, make sure to desolder the +5VSB wire properly; if you just melt-and-pull (as I did initially), you may solder-bridge it to the voltage feedback node (linked to pin 1 of the TL494) and then the main supply won't run at all .
• Access the TL494's unused second error amplifier at pins 15 and 16. (At pin 16 you can just cut the trace to remove it from original circuit, pin 15 is harder as it's solidly connected to the adjacent pins; I desoldered and removed the TL494, severed and lifted-off the pin 15 pad, and then reinstalled it.)
  You can then connect the feedback and reference components of your choice (I duplicated the original 4.7kΩ feed and 47kΩ + 10nF compensation) to pin 15, and shunt of your choice (in my case 10mΩ made from a coiled Australian 7/0.5 earth wire, which is nominally 1.5mm² but actually 1.37mm²; although I really ought to use something with a lower temperature coefficient than plain copper ) to pin 16 (although if you obtain a suitably DC-precise opamp, amplifying the shunt voltage would be good for improved accuracy). You can arrange a voltage divider (using a potentiometer if continuous adjustment is desired) fed from either +5VSB or the TL494's internal reference, to configure the current feedback threshold.
• Disable UVP by removing J16 (adjacent to the LM339).
• Disable OVP by removing D37 (adjacent to J17, between the LM339 and TL494), if you modify the voltage feedback (I removed R54, which is for +12V feedback) to allow more (up to almost +10V/+24V is possible, varying somewhat with the mains input and primary ripple).
  Of course, if you do so, you should also uprate the output capacitors to 10V and 25V (I used Rubycon ZLJ 2200/10 and Panasonic FS 3300/25 which fit nicely, along with Nichicon PW 100/25 on the negative outputs; Rubycon YXJ is suitable for the latter if you want to keep full capacitance, but many similar designs used 100μF there so they should be good enough).
• If those don't let the unit run, you can disable the protection circuit altogether by removing the diode (adjacent to C35, fourth part in from the PCB edge after three resistors) between pin 2 of the LM339 and pin 4 of the TL494.
• If you want the fan to run properly, you would need to either connect a 5V type to the standby supply, or provide an alternative 12V regulator; I rectified the 12V winding output (you can use either a ≥80V common-cathode rectifier, in my case a salvaged F06C20C, or a ≥40V diode from the +12V rectifier cathode) and regulated it down in a 7812.
• Connect the output wires of your choice (I'd recommend 4mm² or 10AWG for +5V and ground, at least 1.5mm² or 14AWG for +12V) and have fun.

I can take pictures soon if you're ready.

UPDATE: I've obtained and mounted a proper 10mΩ resistor (25W aluminium-bodied, crudely mounted onto the secondary heatsink fins with a liberal amount of thermal paste; hopefully that'll do well enough for a dissipation of 10.24W at 32A).