Young Year ATX-50A - poor 12v regulation

[Th3_uN1Qu3]

The Young Year PSU here. It works... but at startup 12v is at 6.8v. I put a rail of molexes on it and it'll run a hard drive, with 12v settling at 10.85v. I also noticed that 12v varies quite a bit as the HDD starts up. 5v stays solid.

Even though it's full of fukyoos, it isn't bad caps. Ripple is very low, <30mV on all rails. At first i was reading the scope wrong, my hand was picking up noise and transmitting it to the probe. If the scope is hooked up without me holding the probe, or if i touch the case of the scope (grounded) while holding the probe, the noise goes away, and ripple shows low. Replacing the caps didn't make any difference in output voltage.

Here's the weird part: If a heavy load is put on 5v or 3.3v while the hard drive is connected, 12v will go in spec. If there is no load on 12v, it'll go to over 13v. This basically says that 12v is left out of the feedback loop! Could it really be true?

Edit: Found something cool. Here's how the Chinese load test their supplies... Yes the rail currents are mixed up, but you can figure out which one went where. If you click "Next" there's also temperatures of the various parts... along with estimated life times. Hmm. Those guys actually do their homework, but probably by the time the power supply leaves the factory, there's like half the parts removed.

[Th3_uN1Qu3]

Re: Young Year ATX-50A - poor 12v regulation

Man, am i having a productive day today. I got it.

The 12v rail was in the feedback loop. However, feedback is always taken before the pi filter, right? Well, this supply had a pi filter coil and a cap at the end of the pi filter. But... there was no main output cap for the 12v rail!!! So the 12v feedback was taken from the main inductor, before any cap, and what reached the 494 was the peak value of the inductor output instead of the actual DC voltage!

What i did was cut the trace for the 12v cap and wired it before the pi filter inductor. Now i'll have to figure out where i can fit an extra cap for the actual pi filter! Most likely a recap with slimmer, taller caps is in order.

The 12v rail still sits at ~7v with no load. However once the hard drive is connected, it goes to 12.10v, and it also behaves itself with high 3.3/5v loads now. Lame feedback loop is lame.

[kaboom]

Re: Young Year ATX-50A - poor 12v regulation

Like this? Ignore the inductor values, they were just defaults in this quick diagram.

[Th3_uN1Qu3]

Re: Young Year ATX-50A - poor 12v regulation

Yup, it was exactly like that. I found a place to fit the extra capacitor too.

[PCBONEZ]

Re: Young Year ATX-50A - poor 12v regulation

A lot of PSUs don't have full pi filters. Especially older ones.
Moving the FB point to the other side of L5 [in kaboom's drawing] would be enough I think.
[That and put in some real caps.]
.

[Th3_uN1Qu3]

Re: Young Year ATX-50A - poor 12v regulation

A lot of PSUs don't have full pi filters. Especially older ones.

I know, but in this case the pi filter coil was basically doing nothing except dropping voltage.

Moving the FB point to the other side of L5 [in kaboom's drawing] would be enough I think.

If i were to not install a proper pi filter that would be the case. However what i did is moved the existing cap (well, moved the trace) to before the pi filter coil, so the feedback point was left where it was and now gets the correct signal. Then i added an extra cap to form a full pi filter. I just used a Su'scon i got from another of those junkers, looks good, tests good, works good.

[That and put in some real caps.]

You know what's funny? This PSU has standard, general purpose, 85C fukyoos. SK series. All of them are good...

[PCBONEZ]

Re: Young Year ATX-50A - poor 12v regulation

I know, but in this case the pi filter coil was basically doing nothing except dropping voltage.

It wasn't a pi filter it was an LC filter and the coil doesn't 'do nothing'.
I though you rummaged through junker PSUs all the time scavenging parts?
You should have seen a bazillion of them by now.
.

[Th3_uN1Qu3]

Re: Young Year ATX-50A - poor 12v regulation

It wasn't a pi filter it was an LC filter and the coil doesn't 'do nothing'.

It was an LC filter with two coils in series. Pointless if you ask me, because the coil destined for a pi filter always has quite low inductance.

I though you rummaged through junker PSUs all the time scavenging parts?
You should have seen a bazillion of them by now.

I've seen missing coils altogether but never a missing (actually not even designed to be there) cap with the coil in place.

[PCBONEZ]

Re: Young Year ATX-50A - poor 12v regulation

LC filters are probably the most common topology used in cheap PSUs.
Very common.

The toroidal coils are not considered part of the filter although I've never really understood what good they do.

.

[Th3_uN1Qu3]

Re: Young Year ATX-50A - poor 12v regulation

LC filters are probably the most common topology used in cheap PSUs.
Very common.

Um, yes.

The toroidal coils are not considered part of the filter although I've never really understood what good they do.

The toroidal coils ARE the "L" in the LC filter. Any extra coils without caps both before and after them, are pretty much pointless.

[PCBONEZ]

Re: Young Year ATX-50A - poor 12v regulation

Um, yes.



The toroidal coils ARE the "L" in the LC filter. Any extra coils without caps both before and after them, are pretty much pointless.

The toroidal coils are often toroidal TRANSFORMERS [which is why they are referred to as COUPLED Inductors in the drawing] and it makes no sense to [to me] use coupled coils in transformers as coils in LC filters.

Further, if they are there for filtering, even with a cap between the coils you still don't have a pi filter, you would have two LC filters in series.
That would mean the caps uF value should be based on the uH of the coupled coil, not the later coil as would be if it were a pi filter.

This time I'm arguing with you because - I don't get it.
Never understood why they couple the coils there.
Teach me!
.

[Th3_uN1Qu3]

Re: Young Year ATX-50A - poor 12v regulation

if they are there for filtering, even with a cap between the coils you still don't have a pi filter, you would have two LC filters in series.
That would mean the caps uF value should be based on the uH of the coupled coil, not the later coil as would be if it were a pi filter.

It's just a naming thing, they're all cascaded filters.

The toroidal coils are often toroidal TRANSFORMERS [which is why they are referred to as COUPLED Inductors in the drawing] and it makes no sense to [to me] use coupled coils in transformers as coils in LC filters.

Well, you got some terms mixed up here. A TRANSFORMER by definition has a PRIMARY winding and one or more SECONDARY windings. Now, i'm sure you know what a transformer is and what it does so i won't stress this aspect more.

Now, a COUPLED INDUCTOR is a totally different animal. You will see that on the toroid you have several windings for the various rails. Since the very first computers the power supply needed to deliver 5v and 12v. (Now we have those fancy dc-dc converter thingies for 5v and 3.3v, and the 3.3v rail has always been dc-dc'd from the 5v, but let's focus on the older supplies).

Again, the power supply needs to output 5v and 12v. The 5v and 12v outputs will have different and sometimes wildly varying loads. Yet, both 5v and 12v need to be regulated against those loads, using the same pair of switching transistors! Sounds impossible, doesn't it? Well, here's how it's done. First of all, both rails are derived from the same transformer windings. The 12v winding is a little tricky and we'll talk about it some other time. Now, since the rails share the same wire, a load put on one of them will produce a voltage drop that will affect both, correct? Now the rails are already cross regulated to some degree but that isn't enough.

Here's where the coupled inductor comes into play. By winding all inductors on a common core, with a ratio dictated by the power requirements of each rail, here's what happens. When current changes thru an inductor, the magnetic field of the inductor changes, correct? Well, now that all the windings are on the same core, a field change in one of the windings forces the currents in all other windings to change as well. So when 5v is loaded, 12v will track it, and vice versa, allowing the same pair of transistors to keep both outputs in check. This is called group regulation.

Now, with the advent of small, efficient DC-DC converters group regulation has started to become less common, but it's still found in all older power supplies.

[kaboom]

Re: Young Year ATX-50A - poor 12v regulation

I'll start with a single output example.

L1 is the output inductor, L2 is the PI choke.

The output inductor limits the reflected dI/dT reflected back to the primary. More importantly, it integrates the constant amplitude, variable pulse-width from the rectifier. A nice easy way to have the 'main' output regulated and adjustable while having an 'aux' supply of fixed voltage. "494" supply in this case.

Those "fridges" that use a Peltier junction use this trick. When there is no demand for cooling, duty cycle is way down. The LED inside is also fed from the 494 supply, before the output inductor. About 22V here, regardless of where the main output sits. That all works out. Even though the pulse width is narrow at this point, the amplitude is as high as at full output. So it IS possible to have two different voltages.

So the 494 supply point's voltage is determined only by the input voltage and transformer ratio. It's the duty cycle that make it possible for the output voltage to be less.

A typical AT supply, with the '494 powered as in the diagram, will have the 494-supply sitting between 20 and 28V. Then how is it possible for a 12V output to exist? Our hero, the output inductor makes it possible! By integrating the duty cycle. And it saves the primary switches, as the cap-charging-current when C1 recharges doesn't happen in only a few degrees, thanks again to L1.

Bonez, the reason those output inductors are coupled in a computer PS has to do with cross regulation. It helps 'equalize' the current and balance voltages when the rails have different loads. Having individual output inductors would waste space and the balancing would not happen, as individual toroids would not be coupled.

It's really bad news if there are ever shorted turns in those coupled inductors. Think of shorted turns in an autotransformer, or better yet, closing the 'bypass' cutout across a pole-mounted voltage regulator before setting it to 'neutral.' The number of turns on that output inductor varies by the ratio of the different output voltages. So if the 5V output has 10 turns, the 12V output must have 24 turns! Otherwise, the transformer action of the inductor, at this point, will try to 'force' the voltages up or down. In other words, if the 12V winding had 30 turns and the 5V winding had 8, the output inductor itself would try to buck the 5V output down to 3.2! There are pulses here and some coupling occurs, even with the DC bias. And it would get much hotter as well. Even more if there are shorted turns.

Have a look at my "15 houses and a supermarket" distribution network. In the first, all is well and normal. Then, we lose a source phase. In the third, the neutral point of the bank with the delta secondary has been grounded. While the line voltages now 'appear' normal, the poor delta bank is smoking and shaking away, vainly 'refeeding' the lines. Hence the 'fourth cutout' on these dinosaurs, if/when any are even built anymore. A relic from the past, with 480Y/277 and 208Y/120 services ubiquitous now.

Not too different, in principal at least, from the coupled output inductor with the incorrect ratios- from shorted turns or bad design.

Hope the diagrams help,
-Paul

[kaboom]

Re: Young Year ATX-50A - poor 12v regulation

a field change in one of the windings forces the currents in all other windings to change as well. So when 5v is loaded, 12v will track it, and vice versa

Now, with the advent of small, efficient DC-DC converters group regulation has started to become less common, but it's still found in all older power supplies.

You got it! I hear ya on the DC-DC converters. Those huge power supplies are primarily 12V output units, capable of 30-70A total. Some of that capacity is used directly for the 12V output, through one or more current sense shunts. The buck converters for 3.3 and 5 get 12V in from the main supply.

-Paul

[PCBONEZ]

Re: Young Year ATX-50A - poor 12v regulation

Thanks guys. I get it now.

I understood most of it already but I've never walked through it with someone else before so my confidence level was low.

One thing that was tripping me up was that those "toroidals", while actually the part is a toroidal transformer in different applications, they aren't wired as transformers in this particular circuit and so they don't work like transformers.
- Once I got that through my thick head everything fell into place.

THANKS!
.

[PCBONEZ]

Re: Young Year ATX-50A - poor 12v regulation

kaboom
Your drawings are really clean and sharp.
What are you drawing them with?
.

[Th3_uN1Qu3]

Re: Young Year ATX-50A - poor 12v regulation

This is what makes building a split output SMPS way easier than i thought - same transformer winding and coupled inductor. If one rail is loaded, the other goes down with it - since they come from the same place, duh.

@ PCBONEZ: I think that's Multisim. I'm more partial to LTSpice.

[PCBONEZ]

Re: Young Year ATX-50A - poor 12v regulation

If I could find a free one I'd probably learn it.
Cost isn't justified because I'd use it for play, not work.
.

[PCBONEZ]

Re: Young Year ATX-50A - poor 12v regulation

Don't laugh too hard okay...
Has anyone ever tried something like this?
[Yeah I didn't finish but there's enough to see what I'm asking.]
-