SG6105 IC analysisWith switching PSU's the duty cycle of the switching FET's (duty cycle = time on versus time off) is controlling the output voltage of the PSU, in order to make the 220V AC mains voltage regulable we have to convert it into DC voltage first before we can chop it on and off. The more output voltage you want, the longer the FET's have to be switched on. This is all not too important too know, it's just good basic knowledge you may need whenever working on ATX PSU's. The fact that only one IC regulates the entire voltage range is quite interesting for us, let's us have a closer look at the System General controller and find out how it does all that:
Via a feedback circuit existing out of resistors the controller is reading the high power +5V and +12V output voltages. Thus, for us, most interesting is the Pin 17 labeled "IN", since it's this part of the circuitry that we have to modify to change output voltage. If we alter the voltage at pin 17 the controller will think that the output voltage is changing, while in fact it was not. This will cause the controller to adjust the duty cycle of the high side Q1 and Q2 FET's and so the output voltage of the PSU will change. What I did was disconnect the pin 17 from the PSU PCB so that I could place a variable resistor in between, making it possible for me to adjust the voltage at the pin 17 and thus also the output voltage of the PSU. Yes, it has become a fancy rheobus now!
Downside here is that the old PSU design features only one power transformer for all the voltage levels, thus changing the duty cycle at the primary side will change all voltage levels, not only the 12V level for example. Another interesting fact is that the controller is able to read live voltage without any kind resistor feedback circuit. Pin 2, 3 and 7 are being use to check the 3,3V, 5V and 12V lines. These are however not feedback circuits like with pin 17, these are protections circuits used to check if the live voltages are in between certain boundaries, in between -10% and +10% for example. Very nice to have whenever the PSU is no longer able to maintain it output voltage, the PSU will sense too low voltage and thus shut itself down for safety reasons. This however makes our PSU-Fan controller only good for changing very few percentages and so we have to get rid of this protection circuit.
What I did was building a second low power circuit which transforms the 220V PSU supply voltage into 12V, 5V and 3,3V. Then I disconnected pin 2, 3 and 7 from the printed circuit board and hooked up my self made low power circuit, now the SG6105 would always see the ATX PSU voltage lines in between the boundaries and never shutdown. Never really means never though, if the PSU would go into error causing the output voltage to fluctuate heavily then the PSU will not shutdown either. That's why I decided to add a panel volt meter to the PSU, it's handy to adjust voltage this way but it is also handy to check if the PSU is still running good.
When I tested the PSU I noticed that at around 9V the PSU would shut itself down once again… I forgot to remove the negative voltage protection! Negative line voltages are being monitored at the NVP pin 6. Here the voltages are not read live, but again through a resistor network and an internal current source. A comparator will check if the voltage at pin 6 is lower then a 2,1V reference voltage, if not then the comparator will enable a PSU shutdown. Altering the voltage level at the pin 6 via an added resistor didn't work out too well, but since this is a low current source I took the chance of grounding the pin. This worked out pretty well, my Frankenstein was finally alive! Let's have a look at how the PSU now looks like internally ->