Log 3LVIS: (3) (L)itre (V)ery small case with (I)nternal psu and (S)creen

Valantar

Shrink Way Wielder
Jan 20, 2018
1,857
1,793
No clue either...

To sum up:
- There is no problem on the 'computer' hardware parts as you are able to start the computer with an ATX PSU
- You are able to start the computer when the PDCB is connected to an external brick, so the PDCB is also fine
-> then the only remaining part that can be faulty is the PSU

But
- The PDCB output values when attached to the PSU give good results on the tester
-> so the PSU seems to be working as expected, at least when there is no power draw or very few

Don't you have another setup that requires less power to test your PSU with? Like a motherboard with an embedded CPU drawing 20W from the wall or something.
Or having the PSU to take care of the motherboard only, and the ATX PSU for the CPU?
If that's not risky for your hardware that is, just thinking aloud here...

Well maybe it's time to summon our 12V chief magician @Thehack !
That's the problem with those PSU testers - they don't really test much, as they place zero load on the PSU. And a PSU needs to be pretty broken to not show correct voltages at zero load.

There are cheap ways to do some simple load testing (even if it is by no means equivalent to proper PSU testing): Buy a set of powerful 12V lamps - you should be able to find up to 60W relatively easily (car headlights, for example). Connect these to the PSU, ideally in parallel to the PSU tester. As this is a pure test setup and at low voltage, there's no need for fancy wiring - just get it hooked up, ideally in a way that lets you relatively easily connect and disconnect the loads. See how the PSU reacts to scenarios such as the sudden addition of loads from zero power draw by monitoring voltages on the PSU tester. (Worth noting: incandescent lamps are purely resistive loads, while a PC is a complex combination different loads. Using pure resistive loads for PSU testing is thus not really realistic - but it can give some indication of PSU behaviour.) If the 12V voltage drops significantly, then that's your answer - which would likely be due to an insufficient bulk capacitor in the PSU (likely IMO for a compact, low output LED driver), coupled with too slow transient response (if the PSU has no voltage monitoring, which is likely the case for an LED driver, then it has no way of compensating for sudden load spikes and the accompanying voltage drops), either causing the cap to drain and the motherboard to no longer see a sufficient voltage for operation, or it might even trigger some form of protection in the PSU if the inrush current from the PC is sufficient. Either way, that would result in the PC simply not booting, as from it's point of view it's not actually receiving power.

Another highly useful tool in a scenario like this would be a clamp style multimeter/ampmeter, to measure the current draw from the PSU when turning on the PC. The filter caps on a large CPU VRM can cause quite high inrush currents when connected to power, even if nothing in the system is actually drawing significant power - the filter caps want to fill up as quickly as possible. This could be enough to trigger OCP on your PSU if it's sensitive.
 

infoberg

Average Stuffer
Original poster
Mar 23, 2021
73
49
That's the problem with those PSU testers - they don't really test much, as they place zero load on the PSU. And a PSU needs to be pretty broken to not show correct voltages at zero load.

There are cheap ways to do some simple load testing (even if it is by no means equivalent to proper PSU testing): Buy a set of powerful 12V lamps - you should be able to find up to 60W relatively easily (car headlights, for example). Connect these to the PSU, ideally in parallel to the PSU tester. As this is a pure test setup and at low voltage, there's no need for fancy wiring - just get it hooked up, ideally in a way that lets you relatively easily connect and disconnect the loads. See how the PSU reacts to scenarios such as the sudden addition of loads from zero power draw by monitoring voltages on the PSU tester. (Worth noting: incandescent lamps are purely resistive loads, while a PC is a complex combination different loads. Using pure resistive loads for PSU testing is thus not really realistic - but it can give some indication of PSU behaviour.) If the 12V voltage drops significantly, then that's your answer - which would likely be due to an insufficient bulk capacitor in the PSU (likely IMO for a compact, low output LED driver), coupled with too slow transient response (if the PSU has no voltage monitoring, which is likely the case for an LED driver, then it has no way of compensating for sudden load spikes and the accompanying voltage drops), either causing the cap to drain and the motherboard to no longer see a sufficient voltage for operation, or it might even trigger some form of protection in the PSU if the inrush current from the PC is sufficient. Either way, that would result in the PC simply not booting, as from it's point of view it's not actually receiving power.

Another highly useful tool in a scenario like this would be a clamp style multimeter/ampmeter, to measure the current draw from the PSU when turning on the PC. The filter caps on a large CPU VRM can cause quite high inrush currents when connected to power, even if nothing in the system is actually drawing significant power - the filter caps want to fill up as quickly as possible. This could be enough to trigger OCP on your PSU if it's sensitive.
Thank you so much for these hints, @Valantar !

I pondered getting a fwe lamps to put load on the PSU, but decided to test it with components from another build first. Cross-testing the PDCB/PSU combo against that build showed exactly the same results. I then checked the current during the turn-on sequence with the multimeter to check the output voltage of the PSU and a power meter to check how many watts are pulled from the outlet. It turns out that after switching on the PC, the output voltage of the PSU falls from 12.6V down to about 2V for a second, before rising again to 12.6V, probably after the mainboard stops drawing any current due to the low voltage. I think there's no sense in investigating any further as this PSU would only be usable with modifications, which I don't trust myself on performing. I guess it also means that the 50ms rise time spcified by the PSU makers is nonsense, so I will try to return it to the seller, let's see how that goes...

And of course, it's also time to think about alternatives.
 
  • Like
Reactions: Valantar

Valantar

Shrink Way Wielder
Jan 20, 2018
1,857
1,793
Thank you so much for these hints, @Valantar !

I pondered getting a fwe lamps to put load on the PSU, but decided to test it with components from another build first. Cross-testing the PDCB/PSU combo against that build showed exactly the same results. I then checked the current during the turn-on sequence with the multimeter to check the output voltage of the PSU and a power meter to check how many watts are pulled from the outlet. It turns out that after switching on the PC, the output voltage of the PSU falls from 12.6V down to about 2V for a second, before rising again to 12.6V, probably after the mainboard stops drawing any current due to the low voltage. I think there's no sense in investigating any further as this PSU would only be usable with modifications, which I don't trust myself on performing. I guess it also means that the 50ms rise time spcified by the PSU makers is nonsense, so I will try to return it to the seller, let's see how that goes...

And of course, it's also time to think about alternatives.
Yep, that sounds like there's a current spike when turning on the system (likely when charging the various caps around the motherboard) which overpowers the PSU and causes it to reset or just drop to a very low output voltage.

The rise time is likely true, but the problem is that you're not looking at a PSU off-to-PSU on scenario, but a PSU on but zero load-to-PSU on with significant load scenario. Even if on a different scale, it's similar to how high end GPUs this generation (3080s and 3090s particularly) have been shutting down PCs with seemingly sufficiently rated PSUs due to high transient load spikes draining their (small for the wattage rating) bulk caps and causing undervoltage scenarios. Rise time is from PSU off to PSU on after all (and assumes the load is present when the PSU is switched on), and doesn't really tell you about the transient handling of the PSU, or how its bulk capacitance relates to its overall rated output power.

One question: have you tried setting the BIOS to power on with AC power detected, and switching the PSU off and then on? I have no idea if this would change anything, but it would be interesting to see if it affected the behaviour in any way. (My thought: is there a difference between a steady-state, zero-load PSU that's on vs. a ramping, starting up, recently-off, charging its caps PSU in how they handle high transient loads?)
 
  • Like
Reactions: infoberg

BaK

King of Cable Management
Bronze Supporter
May 17, 2016
666
641
I then checked the current during the turn-on sequence with the multimeter to check the output voltage of the PSU and a power meter to check how many watts are pulled from the outlet. It turns out that after switching on the PC, the output voltage of the PSU falls from 12.6V down to about 2V for a second, before rising again to 12.6V, probably after the mainboard stops drawing any current due to the low voltage.
Thanks for the feedback, sad but interesting to know such voltage drops can happen.

One question: have you tried setting the BIOS to power on with AC power detected, and switching the PSU off and then on? I have no idea if this would change anything, but it would be interesting to see if it affected the behaviour in any way. (My thought: is there a difference between a steady-state, zero-load PSU that's on vs. a ramping, starting up, recently-off, charging its caps PSU in how they handle high transient loads?)
Good idea! Worth a try!
 

infoberg

Average Stuffer
Original poster
Mar 23, 2021
73
49
One question: have you tried setting the BIOS to power on with AC power detected, and switching the PSU off and then on? I have no idea if this would change anything, but it would be interesting to see if it affected the behaviour in any way. (My thought: is there a difference between a steady-state, zero-load PSU that's on vs. a ramping, starting up, recently-off, charging its caps PSU in how they handle high transient loads?)
Interesting suggestion, I had not thought about that. Indeed the AC power loss setting was set to remain off. I switched it to power on after power loss and as expected the PC tried to boot, flashed the LEDs, then tried to boot again, (because power from PSU was lost) and over and over again. The voltage output of the PSU kept going from 12V before switching the PC on, then down to 2V when the PC tried to boot and up again to 12V when booting failed, then down to 2V again when the next boot was attempted and so on.

But the most interesting thing happened when I switched the PSU off. At that moment (and reliably every timed I repeated it), the CPU fan started to spin for a tiny bit, but stopped after turning about 1cm...

But other than that, no change. It was kind of an experiment to use this LED PSU, so at least now I know why they can't be used. It's a shame that there are no other options available for PSUs that will fit inside this case without modification. Guess I need to go for the Meanwell and mod the case...
 
  • Like
Reactions: BaK and Valantar

Choidebu

"Banned"
Aug 16, 2017
1,182
1,174
Cheers for having the audacity and bravery to actually try led psus for PCs. Folks 'round here 'ave been asking whether they can - with naysayers (me included) saying it won't work so they shouldn't even try.

Basically voltage drop to 2V means that 3% load regulation is a load (pun intended) of b*llcrap. More like 88% lol.

For sub 200W PC load, maybe - and that's with a big asterisk - just maybe, 500W++ led psu can boot it.

Running it reliably is another story. PCs are remarkably noisy beasts as power consumer. Cpu switching cores on and off, ram refreshes, etc. Led psus are not made for these type of loads. I'm curious whether transient response time/graph report for one even exists.

But don't get me wrong I'm not trying to be that 'I told you so' guy - I watched your thread since day one with a little sigh at first but some anticipation that you might pull it off somehow. So really, thanks for putting the hammer down on this convo. Respect.
 

infoberg

Average Stuffer
Original poster
Mar 23, 2021
73
49
Cheers for having the audacity and bravery to actually try led psus for PCs. Folks 'round here 'ave been asking whether they can - with naysayers (me included) saying it won't work so they shouldn't even try.

Basically voltage drop to 2V means that 3% load regulation is a load (pun intended) of b*llcrap. More like 88% lol.

For sub 200W PC load, maybe - and that's with a big asterisk - just maybe, 500W++ led psu can boot it.

Running it reliably is another story. PCs are remarkably noisy beasts as power consumer. Cpu switching cores on and off, ram refreshes, etc. Led psus are not made for these type of loads. I'm curious whether transient response time/graph report for one even exists.

But don't get me wrong I'm not trying to be that 'I told you so' guy - I watched your thread since day one with a little sigh at first but some anticipation that you might pull it off somehow. So really, thanks for putting the hammer down on this convo. Respect.
Thanks, I attribute my going ahead to a mix of stubborness and stupidity 😅
As the available explanations didn't satisfy my curiosity, I just needed to try it out for myself. I guess now we know for sure...

But at least not all is lost with this build. For the time being, I will keep using an external power brick, just need to enlarge one of the WiFi antenna holes to fit the original inlet of the PDCB. The C6 inlet and switch will just keep sitting there unused, not wasting any space that would be needed for something else.

Been quite busy at work lately, but it's now time to think about the LCD screen. Maybe I should have used a Thin Minit-ITX board to make more space for the screen. But then again, that would mean more space but less cooling if going with one of the low-height cooling solutions...
So for now I am sticking with the build components, trying to come up with a solution for a hidden hinge.
 

BaK

King of Cable Management
Bronze Supporter
May 17, 2016
666
641
So for now I am sticking with the build components, trying to come up with a solution for a hidden hinge.
Just had a shot at it, trying to keep it simple.


The idea is to have the little top front panel to be movable, next to the main top panel when the latter is down, and offset towards the front when the main top panel is raised up.
Now let's find out how to make it slide like that... not that simple after all I'm afraid 🙄
 

infoberg

Average Stuffer
Original poster
Mar 23, 2021
73
49
Wow, that's a great mockup! It goes in a direction which I hadn't thought of before - having the bulge of the hinge (whatever it is called in proper English 😁 ) mounted facing the bottom of the case. This enables flush mounting on top without using a bulky frame. I think the top would be made out of 1,5mm aluminium, this means about 2mm of space will be needed to the smaller top front panel when lifting the top at an angle. I might even just leave a 2mm gap between the two parts. In an ideal world the screen would lift up by the push of a button, but I don't know if any programmable miniature hydraulic components for such a task exist - or how to google for them.

Sadly I didn't have enough time to continue the project, too busy at work lately, which is a good sign after more than 18 months working at home and only for a few hours per week :)
 

BaK

King of Cable Management
Bronze Supporter
May 17, 2016
666
641
Wow, that's a great mockup! It goes in a direction which I hadn't thought of before - having the bulge of the hinge (whatever it is called in proper English 😁 ) mounted facing the bottom of the case. This enables flush mounting on top without using a bulky frame.
Thx, glad to have brought up a new option with the bulge of the hinge downward!

I think the top would be made out of 1,5mm aluminium, this means about 2mm of space will be needed to the smaller top front panel when lifting the top at an angle. I might even just leave a 2mm gap between the two parts.
You could fill the 2mm gap with a magnetic strip to prevent unwanted objects to get inside the case.
That could also be an option to detach/move the top front panel, if you can deal with some extra thickness.

Sadly I didn't have enough time to continue the project, too busy at work lately, which is a good sign after more than 18 months working at home and only for a few hours per week :)
No hurry, hope you are not giving up the hinge idea.
Maybe putting the project aside for a while will let you find the proper solution!
I anyway still have this in a corner of my mind , will let you know if I think about something else.