You’ve probably noticed that everywhere you go — the doctor’s office, hotels, or retail shops, there are tiny PCs everywhere. These small PCs often show up on the surplus market for a very good price, but they aren’t quite full-blown PCs. They usually have little option for expansion and are made to be cheap and small. That means many of them have custom and anemic power supplies. We aren’t sure if [bm_00] needed a regular power supply to handle a graphics card or if the original power supply died, but either way, the HP small-form-factor box needed a new power supply. It took some clever work to be able to use a normal power supply in the little box.
At first, we thought this wouldn’t be much of a story. The motherboard surely took all the regular pins, so it would just be a matter of making an adapter, right? Apparently not. The computers run totally on 12V and the motherboard handles things like turning the computer on and off. The computer also was trying to run the power supply’s fan which needed some work arounds.
Granted, you could just wire the power supply to be on all the time, but it is nice to be able to turn everything off. The plan was to use the always-on 5V standby rail to drive a pair of relays. One relay senses the computer’s on/off switch and triggers the ATX power supply to turn on.
The problem is the computer wants to draw a little 12V power all the time. So, in an odd turn of events, a small boost converter changes the 5V standby voltage to enough current to drive the PC in the “off” mode. When the power supply’s 5V rails turn on, they throw the other relay to disconnect the boost converter and supply the real 12V supply.
There’s only one problem with that. The motherboard sees a power glitch when the switch occurs. So, there’s a hefty capacitor to smooth out the transient. Well, there’s another problem. In some cases, though, the boost converter couldn’t provide enough power for the motherboard before the boot process.
Honestly, we think we would just put a switch or a power strip in the supply’s AC cord and have been done with it. But we admire the tenacity and ingenuity.
Then again, you could just put the PC in the power supply. Around here, old power supplies usually get benched.
I’m old enough to remember when an “HP” logo on a device mean you had purchased the very best–or at least one of the very best– of that piece of electronics gear.
These days, the only ‘hack” better than replacing proprietary HP computer subassemblies with generic ones is to throw out the HP junk entirely and vow never buy another.
Those were the years “B.C.” (Before Carly)
+1
Their printers in the last 20 years are hot garbage too.
The build quality, both hardware and software-wise is shocking and they run Windows CE!
(The ones at work have a natty little sticker on the back that says “Powered by Windows CE” :P
“Toner low” when the b****y thing has half a cartridge left too!
I’ve got a secondhand Laserjet 1100 from the last century that came from a fleamarket 15-odd years ago and it is still working absolutely fine. It was from a veterinary practice originally so I was told and is just ticking over on the last of the toner carts it came with. Replacements aren’t too expensive either so I might just keep hold of it for a while longer.
HP used to be a trusted name. The only HP I’ll trust these days is the sauce! XD
LaserJet 5 here. Got it (broken fuser and feed gears) for free and rebuilt it myself over COVID wfh. Has been sitting on the network, printing when needed since then. Could not be happier with it.
1200 with RPi Zero W running CUPS here. New toner is like $2. I pay more for having it shipped.
Bought LaserJet 6mp used on eBay in 2005. Running strong 30k pages later, replaced toner three times. Only gripe is electricity usage.
Back when HP started going down the toilet, the Test Equipment division broke off and became Agilent.
Agilent now focuses on clinical lab equipment, and general test equipment split again into Keysight.
Keysight make some nice gear. HP? Not so much.
I know Intel shortly waffled on about a power supply standard they propped up that eliminated any 3.3v, 5v and negative voltage rails.
Knowing HP, they just copied Intels homework, changed a few things (pin-out and physical dimensions) and then rolled with it.
It’s the same with Dell Optiplexes, Lenovo Thinkcentres, and even later Acer Veritons.
There’s actually prebuilt 3rd-party PSU adapter cables (with 5v -> 12v boost converter) for HPs and Dells at least.
Is this really a proprietary power supply in these SFF systems? Or is it just the ATX12VO standard with a proprietary connector?
This type of PSU predates the ATX12VO standard though. The HP Elitedesk 800 G1 (Haswell/Intel 4th gen) from over a decade ago already had it, just like the Dell Optiplex 9020 (same generation). If anything, ATX12VO merely standardized something the OEMs were already doing.
I recently replaced a power supply in a Lenovo office computer that seemed to be just like a standard SFF ATX12VO PSU but with a very slightly nonstandard connector…to be fair it looks like the ATX12VO standard was still being ironed out when this computer hit the market. If I wasn’t able to find an exact replacement, swapping the connector on a standard one would’ve been an option.
This seems like a lot of work for something that should be simple. If it just expects a basic 12 V power supply then just give it a basic 12 V power supply.
If it’s all proprietary and designed to be as small and low cost as possible then it probably doesn’t have any normal connectors like PCIe either so it’s unlikely you would plug in anything else that would need an ATX power supply.
It’s not a basic 12V power supply, but more like a regular ATX PSU but with notable changes:
standby voltage is now 12V (up from 5V)
5V and 3.3V voltage rails removed, motherboard supplies those voltages.
I can think of three interconnected reasons why this approach was made:
simplify PSU design (lower costs)
improve conversion efficiency (older SFF PSUs weren’t even 80 Plus compliant)
increase rated output
Description of “standby voltage is now 12V (up from 5V)
5V and 3.3V voltage rails removed” describes ATX12VO
Or, after connecting the necessary wires to the motherboard connectors, just install another switch at the front panel to bridge Green+GND and start the PSU. Then press the second ( the original ) button to start the computer.
Or putting a 555 timer into service to activate the power button for you.
Use a latching pushbutton for the Power Switch if you like, or be fancy and use a momentary button and a latching relay that will release when the power LED goes out.
im still a fan of mains switches. that way you can make sure the damn thing isnt spying on you when its “off”.
Ah, the big awesome lever switches on some old AT PCs – nice clunk to them.
Or the ones on the AT PSU but inside the case – mechanically extended to the front. ;-)
I vaguely remember needing a key to turn those on. There was a “key” pin on the power supply connector, another actual cylinder key to lock the case, the power, and the keyboard,
I remember some whip’s on first conversations where there were 3 possible meanings for the word key , and several for the word lock all related to hardware.
Maybe I’m confusing it with PDP8s and PDP 11s
I vaguely remember needing a key to turn those on. There was a “key” pin on the power supply connector, another actual cylinder key to lock the case, the power, and the keyboard,
I remember some whip’s on first conversations where there were 3 possible meanings for the word key , and several for the word lock all related to hardware.
Maybe I’m confusing it with PDP8s and PDP 11s
Concern for ‘spying’, sure. But even just spontaneous self-starts for no reason (or, not MY reason). And phantom loads. And leakage current. And the threat of a power surge frying your supply even when the equipment is turn ‘off’.
Soft switches even (usually) have the fundamental bad design of having the same action (push the same button) to start and stop the machine. Even worse, sometimes it’s a long-press or a double-tap, and a loooong-press is ‘reset’. Arrgh.
Hard switches are good.
I recently ran into this same problem except I went with a larger case and a standard 1200 watt. Power supply with a $10 adapter from Amazon the power supply fan error message can be easily corrected by jumping a couple of pins
At first glance I thought this hack was for a HP Mini PC, but it is for a HP Elitedesk SFF (small form factor). Cos those minis need a genuine HP laptop supply to even POST.
I think the original hacker may have been barking up 2 wrong trees:
1. Have the power button circuit engage the PSU directly, then after a slight delay signal the motherboard to come on. (This may be a little tricky due to the power switch being on the PCB in these types of systems. But it’s an excellent excuse to try a custom round metal lit power button.)
The power changeover situation (which shouldn’t happen with solution #1 above) can be solved with a small FET board called an ‘ideal diode’. Most systems are rated for 1-2Amps of sleep current, so it shouldn’t require a very expensive board. The purpose is to fail over to the highest voltage, possibly with some configuration to delay switchover to prevent oscillation.
On the subject of what power supply to use, I have been getting FlexATX FSP500-30AS power supplies originally meant for NUC Extreme computers for $20 each with shipping on eBay (they accept offers, might as well get 2). I have one powering a 12700KF and RTX3080 right now after adapting it to 24-pin ATX. It has niceties like 2x 8pin PCIe power, and an ATX 12v 8-pin. You will have to solder onto a circuit board internally to bring out 5V, but since this project doesn’t even need 5V you should be fine.
Looking into the 10pin from the connector side with the clip on top:
_________Latch__________
3.3v Gnd PWOn Gnd Gnd
3.3v Gnd 5Vsb 12v 12v
A bonus with the FlexATX is you can position it so more airflow can reach the video card fans instead of being blocked by the power supply.
That looks a lot like the HP power supply I picked up dirt cheap at a local thrift store. After figuring out that it needed a PWM signal to keep the fan from screaming at full speed, it made a good power supply for some 12V LED strips. The one I got was built pretty well and RF quiet.