Problems with PC PSU (noise, ripple?)

[TCMSLP]

Hi All,

I have an old 450W ATX PSU.  I've snipped off all the standard cables and soldered on a standard amiga power cable.  I've connected the 3.3V sense cable (originally connected to the PC motherboard) directly to the 3.3V output;  apparently this is needed for good regulation.  I've also connected the 'power on' wire directly to ground (so PSU powers up when the mains is switched on).

Output voltages (+12V, -12V, 5V) all look good.  However, connect to the Amiga and I experience regular crashes.  Since fitting an Indivision AGA MkIICR I also now notice a lot of noise on the analogue Indivision outputs.

Switching back to an original Amiga PSU removes the noise and cures the crashes.

The video output noise leads me to believe there's either SMPS noise and/or ripple present on the DC outputs.   I've fitted small value decoupling capacitors on the outputs but these have had no effect.

I'm wondeing if the A1200 (68030, CF HD) perhaps isn't drawing enough current for proper regulation?   Or, are some PC SMPSs simply too noisy for Amigas?   I notice some of the EMC components have been omitted but I think these are more to do with switched mode interference making its way to the mains, rather than the output side.

I haven't measured ripple using a scope;  Is 12V ok to connect directly to a scope input or should I make a divider network?


Thanks,


Steve

[Delta]

Have you confirmed that earth ground and case ground are good?

When I get noise on anything I check this first.

[Tenacious]

Have you confirmed that earth ground and case ground are good?

When I get noise on anything I check this first.

Good point.  These PSUs are normally very well grounded to the motherboard ground.  OTOH, you want to be careful about running multiple GNDs to different places.  It could create noisy ground loops.  Maybe that is the place to start.

An Amiga will draw only a fraction of the 450 watts available.  I guess it's possible that it is not fully loading the PSU to the point of being stable.

Most scopes have very high impedance inputs, so as not to load the circuit under test.  I don't see why you could not directly monitor the 12VDC rail with it.

[TCMSLP]

I bonded both GND and cable shield to ground within the PSU.  However, you're right - there's several other paths to earth from audio, video cables etc.  Perhaps I should have bonded the shield to earth rather than gnd/0v?

Edit: Just checked and there's 0 ohm difference between GND and earth in the PSU.

[Tenacious]

I bonded both GND and cable shield to ground within the PSU.  However, you're right - there's several other paths to earth from audio, video cables etc.  Perhaps I should have bonded the shield to earth rather than gnd/0v?

Edit: Just checked and there's 0 ohm difference between GND and earth in the PSU.

Hardware system designers are VERY carefully to choose only one point to tie all of the various grounds (analog, digital, shield, power GND) together, usually on the motherboard.  If you tie them together somewhere else as well, that is a ground loop.  

Your 0 ohm reading does NOT confirm that you have no ground loop problem.

[TCMSLP]

I've bonded the cable shield to earth, I also tried leaving it floating.  No detectable difference on video output.

I'm now typing this using the original A1200 PSU without problems.

Perhaps it is ripple and/or SMPS noise after all.  Time to get the scope out...


Steve

[Zac67]

Possibly the PSU has problems with 3.3 V running idle - or it's broken.

You can measure the ripple portion by just switching your multimeter to AC if the frequency isn't too high.