the3dfxdude wrote on 2021-09-10, 22:53:
snufkin wrote on 2021-09-10, 21:49:
This might be daft, but would it need to have an actual regulator? Looking around, I think the current requirement for -5V was only around 300mA. Going from -12V to -5V would mean burning around 2W maximum, and you can get things like zener diodes in packages that can handle that
Yes... I guess if that's who you trust to do design 😀 let's burn power for something that is unused, when the ATX supply could have provided it if it was really needed.
The main point is, if the motherboard does it assuming ATX spec has changed, there is going to be a device you can point to on the board creating it. It would be interesting if it is true. Because like I just said above, having the ATX PS supply do it is a better answer than plopping any device on a board, that aint gonna get used anyway.
Well, it'd only burn power if current was being drawn (power proportional to current), the 2W (actually 2.1W) would only be burnt if the full 300mA was being drawn (7v * 0.3A = 2.1W). I was just speculating that they might not have used a normal regulator chip, so there could be device on the board to generate -5V, but it wouldn't be obvious.
The motherboard's from when? 2000ish? So -5V was still a requirement of ATX PSUs (pre ATX 1.3), but I expect there were already discussions about dropping it. So for the sake of adding one small component, a motherboard designer might have decided to generate it locally so it'd still be available if someone replaced their PSU in a couple of years time. Just in case someone has some obscure ISA card that still needed -5V. Don't think I've ever used any card that needed -5V.
But now we've got an actual measurement from Kahenraz, and there's definitely no -5V on the ISA slots. But the value in the BIOS does show changes, so it's almost certainly not just a value fixed in firmware/software. Maybe it was easier to reuse a voltage monitor chip from a previous design and add a couple of resistors between e.g. -12V and GND to fake the -5V input to that one chip, than to add one line to the BIOS code to fix the measured value at -5.
Definitely weird.