-12 volt problem

[everell]

Got a "bad" HiPro HP-P3527F3 from a computer shop. On power up the 5vsb, 12 volts, 5 volts, and 3.3 volts all looked good. Power good working OK. But my power supply tester (bought on Ebay) beeped and for -12 volts indicated LL (low voltage).

So I pulled the power supply apart. No bulging capacitors, no burned areas or heat damage. So I began tracing the -12 volt circuit. Output filter caps were OK, then a turn or two on the group regulation coil, then a diode from source to ground. But the diode to the main switching transformer was never installed. Also, there was NO WINDING connected to the transformer pin which went to the missing diode. What???? So where does the -12 volt line get its power?

Take a good look at the schematic of the -12 volt circuit. The ONLY source of energy coming to this circuit is from the group regulation coil. Now it makes sense. The more load you put on the power supply 12 volt and 5 volt circuits, the more voltage that will be induced and transferred to the -12 volt circuit. That means that the -12 volt output will be low unless you load the power supply properly.

Looks like this power supply fooled a computer repair tech who most likely tested the psu with a cheap tester, saw that the -12 volts was low, pulled the psu as bad and replaced it with a "bargain basement 500 watt psu" that could do the job. Customer was probably happy to get computer repaired along with a more powerful psu!

[R_J]

Re: -12 volt problem

Whats the chances that "computer tech" has the same tester off ebay that you used ?

[kaboom]

Re: -12 volt problem

I've got a similar HiPro in a "DOS game computer" I assembled last spring.

After recapping while testing with only the internal fan running, -12 reads around -9.5. Put some load on it and -12 came right up.

They're deriving the -12 output from the +12 winding thru the coupled inductor.

Whats the chances that "computer tech" has the same tester off ebay that you used ?

Leave it to a "computer tech" (present company excluded, of course ) to completely miss and disregard the limitations of said "tester"- for one thing, those don't even load the device under test!

Yet the "repair" is always the same: deer season, as usual...

[Wester547]

Re: -12 volt problem

The HP-P3527F3 is a model built for eMachines/Gateway. Think cheap. The -5V linear regulator doesn't even have a heatsink in this model and the ventilation leaves much to be desired. In fact, it's built much less spectacularly than the older Dell 250W beasts. This model is in fact based on a much older (2001) model from Hipro known as the HP-P302GF3, which is built much better IMHO (much more overspec'd output filter and secondary side, as well as the primary, bigger input electrolytics, better ventilation, linear voltage regulators for -12V and -5V on a small heatsink, etc... and IIRC, the 2001 unit does have a diode and winding for -12V). I believe that PSUs utilizing forward topology, that properly implement -12V, use the coupled inductor (magamp) as a flyback converter of the sort, with one end of the -12V winding grounded and the other connected to the output diode. That still may result in poor -12V regulation unless you use a linear regulator.

So what you're saying about this unit is that it has no rectifier diode for -12V (and thus steals current from the +12V rectifier, along with the linear regulated -5V rail)?

[PeteS in CA]

Re: -12 volt problem

With some 7V across it and no heatsink, I hope that uA7905 isn't supplying much current. If it is it could overheat and shut down (I'm pretty sure it has internal protection against overheating or excess power dissipation).

[everell]

Re: -12 volt problem

The next "bad" power supply from this computer shop is a Bestec (HP part number 667893-003) with BAD written on the case in big bold letters. I like Bestec power supplies, so I figured this one deserved a chance. The 5vsb, 12 volts, 5 volts, and 3.3 volts all came up fine. The power good signal was good. But on my Ebay tester, the -12 volts read LL (Low). The label on the case says -12 volts @ 0.3 Amp.

So I opened the power supply for a look see, and discovered there were NO components installed around the -12V marking on the pc board. In fact, there was no blue wire connected between the pc board and the 24 pin connector. Also no white wire for -5 volts between the pc board and the 24 pin connector. I conclude that the Ebay tester is correct, the -12 volts is low (actually, nonexistent). Looks like this is another bad computer power supply because the -12 volts has failed.

Sure is a nice Bestec power supply.....except no -12 volts available!

[kaboom]

Re: -12 volt problem

The one I have is an HP 210PP, which is an AT unit. IIRC, the 7905 is heatsinked.

It also doesn't have the PG comparator. This PS was made for a late-90's compaq PC, where the PG comparator was on the mobo. In order to use this with a normal AT board, I made my own PG comp on a small board and put it inside the PSU case, and ran a lead/pin to the connector.


Now I have seen those small 90W supplies for those K6 aptivas not heatsink the 7905. Or maybe they had a "finned flag" type of thing, as they were rated for less than 0.3A on -12.

The HP-P3527F3 is a model built for eMachines/Gateway. Think cheap.

The "eMachining" of Gateway was one of the biggest fails, up there with HP/compaq, perhaps even IBM/lenovo. Ironic and sad how Gateway merged with practically their own antithesis. But that's another thread, possibly another forum.

The -5V linear regulator doesn't even have a heatsink in this model and the ventilation leaves much to be desired.

-5 is likely present for spec-only, and those are the same heatsinks the "top fan" PSUs of the last real Gateways. EvilMachines may have suggested "recycling" the old design, just as they were known to use obsolete video card designs while at the same time claiming "it's got an AGP card." Sure, it was an AGP card, but less capable than most PCI VGA cards. All marketing.

This model is in fact based on a much older (2001) model from Hipro known as the HP-P302GF3, which is built much better IMHO (much more overspec'd output filter and secondary side, as well as the primary, bigger input electrolytics, better ventilation, linear voltage regulators for -12V and -5V on a small heatsink, etc... and IIRC, the older unit does have a diode and winding for -12V).

The ones up to around mid-2004 (Gateway) had something like 15-18V into a 7912, which then gave regulated -12. The unreg in came from the coupled inductor, into a 470u 25v cap, then the 7912. Of course, the output of the 7912 fed the 7905, if present. Most of those units did not have -5, and used the white wire for fan speed control. There's an opamp ckt in these which reads the internal thermistors and drives a TO-92L transistor that controls the fan. A pullup resistor on the mobo connects to this white wire (the "reserved" -5V pin). An NPN collector also connects to this node. When the mobo/system is cool, this transistor is biased on, pulling "fan-ctl" down, and running the PSU fan "normal to slow." As the mobo heats up, that NPN is biased off and fan-ctl goes higher, making the PSU speed up its fan.

It's possible to almos stop that fan by connecting that pin to a load, such as a normal mobo where that pin is a -5v input. When doing so, remove that pin from the ATX header! Better yet, tie fan-ctl to +5 thru a 56k-100k resistor. Lower values=faster fan.

I believe that PSUs utilizing forward topology, that properly implement -12V, use the coupled inductor (magamp) as a flyback converter of the sort, with one end of the -12V winding grounded and the other connected to the output diode. That still may result in poor -12V regulation unless you use a linear regulator.

In the one from the compaq, they've got the winding for -12 reversed WRT to +12, just doing 1:1 coupling. The poor regulation at no-low load is a result of low duty cycle- not enough total energy coming from the transformer. Even +12 runs a bit low at no load. Even just a few amps on +5 balances things out; at this point, 0.5-1A on +12 only improves things.

As AT supplies didn't have aux supplies to feed the PWM controller, the former had to have higher min-loading than ATX ones, simply because dropping the duty cycle too much would destabilize the PWM. AT supplies were generally "kickstarted" by biasing the two transistors on slightly; one conducts more than the other and takes over via the driver/oscillator/coupling transformer, which saturates and then drives the other switcher. This happens over a few cycles and eventually, the PWM (TL494, KA7500, etc) comes out of UVLO and takes over, driving the switchers thru the coupling transformer with some 2SC1815s/1845s/945s. The PWM is fed from about +18V, taken from the +12 rectifier, but before the output inductor. Since it's straight from the rectifier, the full amplitude is present regardless of duty cycle. It's the output inductor (in combination with the filter caps) that differentiate the constant-amplitude, variable duty waveform into a constant amplitude (we hope) DC voltage, within the current capacity of the unit.

So what you're saying about this unit is that it has no rectifier diode for -12V (and thus steals current from the +12V rectifier, along with the linear regulated -5V rail)?

Yes, the +12 rectifier ultimately supplies +12, -12, and (via a 7905), -5. Neat huh?


Such a unit that can do 8A on +12 might have a 12A or 15A +12 "master rectifier."
IOW, that single rectifier would see 8A from +12, 2A total from internal -12, 1A from external -12, and 1A from -5 (via 7905). This is worst case- not all AT PSUs did 1A on -5/-12.


OTOH, if that same rectifier only supplied +12, a 10A unit would generally be enough for 8A out.

[Wester547]

Re: -12 volt problem

Now I have seen those small 90W supplies for those K6 aptivas not heatsink the 7905. Or maybe they had a "finned flag" type of thing, as they were rated for less than 0.3A on -12.

Well, those linear regulators do have a metal tab to conduct heat away from the three terminal regulator. I don't think it makes much of a difference, though. I've seen some sinkless 7912 regulators in a number of Deltas, but they are pretty close to the fan and the label purports no more than 0.2A-0.3A.

-5 is likely present for spec-only, and those are the same heatsinks the "top fan" PSUs of the last real Gateways. EvilMachines may have suggested "recycling" the old design, just as they were known to use obsolete video card designs while at the same time claiming "it's got an AGP card." Sure, it was an AGP card, but less capable than most PCI VGA cards. All marketing.

The mobos those PSUs were fitted to (Socket 754 and 939 MSI) certainly didn't need -5V.

The ones up to around mid-2004 (Gateway) had something like 15-18V into a 7912, which then gave regulated -12. The unreg in came from the coupled inductor, into a 470u 25v cap, then the 7912. Of course, the output of the 7912 fed the 7905, if present. Most of those units did not have -5, and used the white wire for fan speed control. There's an opamp ckt in these which reads the internal thermistors and drives a TO-92L transistor that controls the fan. A pullup resistor on the mobo connects to this white wire (the "reserved" -5V pin). An NPN collector also connects to this node. When the mobo/system is cool, this transistor is biased on, pulling "fan-ctl" down, and running the PSU fan "normal to slow." As the mobo heats up, that NPN is biased off and fan-ctl goes higher, making the PSU speed up its fan.

It's possible to almos stop that fan by connecting that pin to a load, such as a normal mobo where that pin is a -5v input. When doing so, remove that pin from the ATX header! Better yet, tie fan-ctl to +5 thru a 56k-100k resistor. Lower values=faster fan.

Ah yes. You are referencing the special Newton which controls the fan speed via the white wire and the pin on the mobo. 15-18V sounds about right, the unregulated input of -12V (before the regulator) might be a bit higher in some PSUs (16V-20V).

In the one from the compaq, they've got the winding for -12 reversed WRT to +12, just doing 1:1 coupling. The poor regulation at no-low load is a result of low duty cycle- not enough total energy coming from the transformer. Even +12 runs a bit low at no load. Even just a few amps on +5 balances things out; at this point, 0.5-1A on +12 only improves things.

The regulation would be good with some load on it, but a linear regulator would likely improve upon it further at the expense of efficiency.

As AT supplies didn't have aux supplies to feed the PWM controller, the former had to have higher min-loading than ATX ones, simply because dropping the duty cycle too much would destabilize the PWM. AT supplies were generally "kickstarted" by biasing the two transistors on slightly; one conducts more than the other and takes over via the driver/oscillator/coupling transformer, which saturates and then drives the other switcher. This happens over a few cycles and eventually, the PWM (TL494, KA7500, etc) comes out of UVLO and takes over, driving the switchers thru the coupling transformer with some 2SC1815s/1845s/945s. The PWM is fed from about +18V, taken from the +12 rectifier, but before the output inductor. Since it's straight from the rectifier, the full amplitude is present regardless of duty cycle. It's the output inductor (in combination with the filter caps) that differentiate the constant-amplitude, variable duty waveform into a constant amplitude (we hope) DC voltage, within the current capacity of the unit.

Ah yes, ATX PSUs, especially half-bridge units, do have an auxiliary +12V rail which supplies power to the secondary PWM controllers.

Yes, the +12 rectifier ultimately supplies +12, -12, and (via a 7905), -5. Neat huh?

And they steal "energy" from the +12V winding too, from the sound of it. Not only that, but IIRC, the +12V rectifier in the HP-P3527F3, like in the Dell 250W Hipros, drives the +3.3V linear regulator. In the "300W" HP-P3527F3, this is all done via a single MBR20100CT or equivalent, on the output of a forward converter.

Such a unit that can do 8A on +12 might have a 12A or 15A +12 "master rectifier."
IOW, that single rectifier would see 8A from +12, 2A total from internal -12, 1A from external -12, and 1A from -5 (via 7905). This is worst case- not all AT PSUs did 1A on -5/-12.

OTOH, if that same rectifier only supplied +12, a 10A unit would generally be enough for 8A out.

Well, if the topology is half-bridge, it could probably do 10A and then some even with -12V and -5V drawing current, being heatsinked and getting airflow (if the ultrafast is rated for 10A). If it was forward topology, I wouldn't risk more than 7A (again with a 10A ultrafast).

Sure is a nice Bestec power supply.....except no -12 volts available!

Nice. Bestec went cheaper than Hipro.

[stj]

Re: -12 volt problem

-5v is only usually used as a reference for the audio section on old stuff,
the -12v is also used to drive the negative bias of the RS232 serial port drivers.
so if your not using the serial ports you probably wouldnt notice it missing.

[ChaosLegionnaire]

Re: -12 volt problem

yea i noticed recent mobos from a few years ago dont even have serial port headers anymore. guess those mobos wont miss not having any -12v rail.

man that -12v rail tapping its power from the 12v and -5v rail is really cost cutting at its finest! here's another solution the cheapskate manufacturers can try: nag, nag, nag really hard at intel to drop the -12v rail from the next version of the atx specs since recent mobos dont even bother having a serial port header anymore! so now no one will decry their cheap asses for cost cutting the -12v rail anymore!! of course, where would the fun be in that?

[Longbow]

Re: -12 volt problem

Thanks for documenting this type of circuit. The -5 and -12 outputs are obviously used for very low load circuits. Still, eliminating the secondary winding is very sloppy and I must say typical of production practices in you know where. As far as the tester, hanging a load resistor on the -12 volt output might improve the operation.

[mariushm]

Re: -12 volt problem

Officially, RS-232 considers a "1" bit as anything between -3v and -25v.

Most equipment out there considers a "1" anything from -8v to -30v, so as long as the voltage is higher than -8v you're fine.

Even the ancient MAX232 chips which are powered from 5v accept up to +/-30v and use voltage doublers and inverters to produce +8.5v and -8.5v from 5v, for "0" and "1" bits... so if that's acceptable, you really don't have to worry about low voltages on -12v rail.