Hello all,
My name is Dave an I'm an electronics design engineer. Brett's test setup was unfortunately flawed. He used very good equipment, EXCEPT for something extremely important: he used a Variac to set his AC voltage. A variac is essentially a HUGE power transformer with a variable secondary winding. The voltage you set with a variac will VARY DEPENDING UPON THE LOAD. It will vary slightly-- but I've seen some poorly made ones that vary significantly---remember there is no storage capacitor on a variac--- the more you load it down, the lower it's output voltage gets. In order to use one you need to reset it FOR EVERY SINGLE TEST.
what do I use at work? A Kikusui or HP powersupply. those suckers deliver PERFECT POWER, VERY CLOSE to 1.00 power factor, and a nice clean sine wave all the way up to 1kHz. Of course they costs tens of thousands of dollars.
Also left out of Brett's calculations was heating loss and power factor loss. those will add to your AC current draw--heating loss will further reduce your output voltages.
Trust me on this, component tolerances have little to nothing to do with voltage variations--The PSU is a "Flyback" type, where a highspeed, very low RdsOn MOSFET switches on and off a power transformer--which has multiple secondaries. 1 secondary is for 5v, one for 12v. All the 3.3V and lower rails are done by the regulators located on the motherboard. 5.1V is a special rail--I forget what its for, but ATX supplies must have it. Off of these 5, 12v rails there are 2ndary switching regulators which use LC filters that filter out all high frequency MOSFET switching noise and are tuned to limit Capacitor ripple to VERY LOW levels (VERY LOW ESR/ESL caps are used.) You may notice in your power supplies that there appears to be 2 mini transformers on the board: 1 is the L (inductor) for the +5V line, it's secondary is the -5V line (simply a power inverter,) the other is the L (inductor) for the +12V line, it's secondary is the -12V line (simply a power inverter too.)
The 5.1V line is special, I'm not quite sure what it's for--though I get the impression that it's ALWAYS ON--it's what lights your motherboard LEDs when the power is off--like on an Asus board--I believe it was added per the ACPI spec--but I'm just guessing on that.
I have worked on countless switching products, both cheap and expensive--when all else fails, component tolerances are not the answer--if they were, you can do an experiment--go to radio shack or wherever and buy 1%, or mil-spec 0.1% --0.001% parts and swap them in. If your results aren't PERFECT then you know there is more to it than component tolerance. Tolerance is mainly an issue with analog only or audio products--switching power supplies require a VERY TIGHTLY CLOSED CONTROL LOOP--otherwise they explode quite easily or produce AWFUL results.
Fixed voltage regulators like computer power supplies are comparatively easy to work on--try working on multi-output VARIABLE ones!
Brett if you are reading this, I commend you on your excellent testing procedures, you used the best equipment you had available and drew a hypothesis--like a good technician should. Just a little more research on the Fly-Back switchmode PSU topology would give you that extra information you need to adjust your hypothesis.
--Bruce Leeroy
My name is Dave an I'm an electronics design engineer. Brett's test setup was unfortunately flawed. He used very good equipment, EXCEPT for something extremely important: he used a Variac to set his AC voltage. A variac is essentially a HUGE power transformer with a variable secondary winding. The voltage you set with a variac will VARY DEPENDING UPON THE LOAD. It will vary slightly-- but I've seen some poorly made ones that vary significantly---remember there is no storage capacitor on a variac--- the more you load it down, the lower it's output voltage gets. In order to use one you need to reset it FOR EVERY SINGLE TEST.
what do I use at work? A Kikusui or HP powersupply. those suckers deliver PERFECT POWER, VERY CLOSE to 1.00 power factor, and a nice clean sine wave all the way up to 1kHz. Of course they costs tens of thousands of dollars.
Also left out of Brett's calculations was heating loss and power factor loss. those will add to your AC current draw--heating loss will further reduce your output voltages.
Trust me on this, component tolerances have little to nothing to do with voltage variations--The PSU is a "Flyback" type, where a highspeed, very low RdsOn MOSFET switches on and off a power transformer--which has multiple secondaries. 1 secondary is for 5v, one for 12v. All the 3.3V and lower rails are done by the regulators located on the motherboard. 5.1V is a special rail--I forget what its for, but ATX supplies must have it. Off of these 5, 12v rails there are 2ndary switching regulators which use LC filters that filter out all high frequency MOSFET switching noise and are tuned to limit Capacitor ripple to VERY LOW levels (VERY LOW ESR/ESL caps are used.) You may notice in your power supplies that there appears to be 2 mini transformers on the board: 1 is the L (inductor) for the +5V line, it's secondary is the -5V line (simply a power inverter,) the other is the L (inductor) for the +12V line, it's secondary is the -12V line (simply a power inverter too.)
The 5.1V line is special, I'm not quite sure what it's for--though I get the impression that it's ALWAYS ON--it's what lights your motherboard LEDs when the power is off--like on an Asus board--I believe it was added per the ACPI spec--but I'm just guessing on that.
I have worked on countless switching products, both cheap and expensive--when all else fails, component tolerances are not the answer--if they were, you can do an experiment--go to radio shack or wherever and buy 1%, or mil-spec 0.1% --0.001% parts and swap them in. If your results aren't PERFECT then you know there is more to it than component tolerance. Tolerance is mainly an issue with analog only or audio products--switching power supplies require a VERY TIGHTLY CLOSED CONTROL LOOP--otherwise they explode quite easily or produce AWFUL results.
Fixed voltage regulators like computer power supplies are comparatively easy to work on--try working on multi-output VARIABLE ones!
Brett if you are reading this, I commend you on your excellent testing procedures, you used the best equipment you had available and drew a hypothesis--like a good technician should. Just a little more research on the Fly-Back switchmode PSU topology would give you that extra information you need to adjust your hypothesis.
--Bruce Leeroy
