12v line problem

PSUs: The source of DC power for all components in the PC & often a big noise source.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee, Devonavar

Post Reply
ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

12v line problem

Post by ryboto » Thu Oct 16, 2008 7:47 pm

So, I've got a Dell Y2515(DA-2) power brick driving my system. My issue is that the 12v line seems kinda low. When idling, using a DMM the 12v measures 11.92v, PC Probe shows 11.8v. When I fold with my GPU(HD 3870), the 12v drops to 11.46v(DMM) and 11.31v(PC Probe). If I also engage the SMP folding for the CPU, PC Probe shows the 12v dropping further, to 11.2v, sometimes 11.14v. I haven't checked what the DMM says, I imagine it will be ~0.1v more than PC probe.

Are these safe voltages? I've been GPU folding for a while with this setup, and never had any stability issues. Power consumption is ~180W when both folding clients are running, that's AC draw. The power brick is supposedly capable of 220W DC output, and I'm probably only stressing it with ~150-160W.

Should I be looking for an alternative brick? One user suggested shortening the length of the power cable from the brick to the PC, which by default is ~4-5ft. Any insight/suggestions/anything would be very much appreciated.

dukla2000
*Lifetime Patron*
Posts: 1465
Joined: Sun Mar 09, 2003 12:27 pm
Location: Reading.England.EU

Post by dukla2000 » Thu Oct 16, 2008 11:53 pm

You are right to be concerned: not sure what the latest spec is but ATX12v v1.3 said +-5% which would be 11.4V minimum.

There are a couple of debating points, not least which spec your config needs/is trying to adhere to. As things are apparently stable there is no immediate need to 'fix' anything. But I would guess the PSU is operating out of spec and so more likely to fail or cause other side effects.

And I would trust the DMM rather than any mobo/BIOS reported value, but even on the DMM you have a problem.

Although the PSU is rated for 220W that is a combined total of specific maxima on each of the output rails - could be you are exceeding the 12V max (and if that brick has any similarity to an ATX supply under the 5V & 3.3V maxes).

Then again, I presume you have a laptop? In which case all the spec/logic above is irrelevant as it is superseded by the (probably unpublished) spec of the laptop input DC.

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Fri Oct 17, 2008 3:08 am

dukla2000 wrote:Although the PSU is rated for 220W that is a combined total of specific maxima on each of the output rails - could be you are exceeding the 12V max (and if that brick has any similarity to an ATX supply under the 5V & 3.3V maxes).
The PSU is a 12V DC brick. I'm not using it in a laptop, this is my main PC. The brick is a 220W rebranded by Dell for their slimline PC's. It only has 1 rail, and it's all 12v.

Thanks for the info, but I'm still not sure what I should do.

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Sat Oct 18, 2008 5:23 pm

Well, I did two things. First I checked the potential from the brick with 0 load. I jumpered the on switch, and measured 12.4v DC. So, it's losing ~1v under load. I plugged my 400W FSP Zen into the system, and the 12v is monitored by PC Probe, and it hasn't budged from 11.98v, idle or load, though I don't know the accuracy of the program.

What should I do here? I really want to use the power brick. Could it be an issue with the brick itself, or how it's wired to the PC?

ntavlas
Posts: 811
Joined: Mon Jul 16, 2007 2:35 pm
Location: Greece
Contact:

Post by ntavlas » Sun Oct 19, 2008 6:24 am

In your other thread you mention that in a previous test the fsp also gave a low voltage, 11,5 in speedfan compared to 11.98 dmm in your latest test. Was the previous test done with the custom wiring?

If wiring was the only difference between the 2 tests it`s quite possibly the issue, maybe a badly soldered cable or a low quality molex splitter?

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Sun Oct 19, 2008 6:50 am

ntavlas wrote:In your other thread you mention that in a previous test the fsp also gave a low voltage, 11,5 in speedfan compared to 11.98 dmm in your latest test. Was the previous test done with the custom wiring?

If wiring was the only difference between the 2 tests it`s quite possibly the issue, maybe a badly soldered cable or a low quality molex splitter?
Yes, nice find. The FSP was showing low voltage. I was using only the PCIE power connector to power the whole system, and I was using custom wiring. Now I'm powering the system with the FSP directly, but since there's no room for the PSU in the case, it just sits on top and looks like a mess.

What pointers could you give me regarding the soldering? I've got 3 incoming power lines going to a section of a 24 pin motherboard connector, as that is what the DA-2 fits. From there, the three 12v, and 3 ground wires become one 12v, and one ground. To the 12v and ground lines I then need to make a total of 9 solders. Two are for the 12v and ground to the Pico, I've got a 12v and 2 grounds to the molex-PCIE adapter, and 2-12v+2-ground for the 4pin ATX 12v connector. Should I make individual solder points for each? minimizing in solder usage?

Arvo
Posts: 294
Joined: Sat Jun 10, 2006 1:30 pm
Location: Estonia, EU :)
Contact:

Re: 12v line problem

Post by Arvo » Sun Oct 19, 2008 7:06 am

ryboto wrote:Are these safe voltages? I've been GPU folding for a while with this setup, and never had any stability issues.
Easy - if the system is running stable, then voltages are safe :)
You just can't kill your components with little (10%) undervoltage, no need to worry about.

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Re: 12v line problem

Post by ryboto » Sun Oct 19, 2008 7:52 am

Arvo wrote:
ryboto wrote:Are these safe voltages? I've been GPU folding for a while with this setup, and never had any stability issues.
Easy - if the system is running stable, then voltages are safe :)
You just can't kill your components with little (10%) undervoltage, no need to worry about.
That's what I'd always thought, issue here is that it's stable with GPU folding, but when I load all 4 cores+GPU, even though I'm within the capabilities of the power supply, the system crashes, or reboots. It seems stable with Prime95 indefinitely, with 3d apps, but once I load all 4 cores and the GPU, the system craps out.

Arvo
Posts: 294
Joined: Sat Jun 10, 2006 1:30 pm
Location: Estonia, EU :)
Contact:

Post by Arvo » Mon Oct 20, 2008 8:13 am

Sorry, I was under impression that your system is always stable and you are only worried about voltage dropping below standards.
--
You need to measure brick output voltages with DMM at brick side, this way you can see, where voltage drops - before power cable (brick output stabilization is weak) or after (cable is too long/thin). If brick can not give enough current, then you're out of luck - but if voltage drops along the power cable, you can just make it shorter.
(Same for internal power wires - they are probably shorter anyway, but you need to measure voltage drop on these too.)
--
To avoid crashes you may experiment with CPU (and maybe video card) voltage too. Well, there can be two opposite cases, I can't say, which one will actually help (if at all):
1. You lower CPU voltage, overall power consumption will decrease, input voltage will rise and entire system becomes more stable;
2. You increase CPU voltage (a very little bit!) and it becomes just more stable despite input voltage falling somewhat (this may help only if other components on motherboard are less sensitive to undervoltage than CPU power converter).

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Mon Oct 20, 2008 9:53 am

Arvo wrote:Sorry, I was under impression that your system is always stable and you are only worried about voltage dropping below standards.
--
You need to measure brick output voltages with DMM at brick side, this way you can see, where voltage drops - before power cable (brick output stabilization is weak) or after (cable is too long/thin). If brick can not give enough current, then you're out of luck - but if voltage drops along the power cable, you can just make it shorter.
(Same for internal power wires - they are probably shorter anyway, but you need to measure voltage drop on these too.)
The only problem with measuring the voltage at the brick, is that I'll either have to open it up, and have it exposed, or cut into the wire. I guess it couldn't hurt anything . If the brick output is weak, I'd hope it was limited to my unit. These things were meant specifically for a Dell SFF system, and those require 12v like mine. I wouldn't have a problem buying a replacement if I knew I just had a lemon.

As for the CPU voltage, that never budges, it's always reported as constant. I'm undervolting, this chip doesn't need much in the way of vcore. The video card is also undervolted at idle. It has a modified BIOS, though it really only saves 1-3W.

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Mon Oct 20, 2008 1:21 pm

So, here's what I tried,

Measured the resistance of the DA-2 12v and ground cables. The meter was going haywire, readings were going up and down. Couldn't get a stable reading.

Measured no-load voltage at brick, 12.38-9v. This is the same voltage measured at the end of the 12v/ground cable, so no voltage loss at no-load condition.

measured voltage at the molex of the Pico with only a 12v 120mm fan powered on, 12.38v at the molex, and 12.38v at the brick.

Measured the resistance between the Pico 12v/ground connector(the one I made) and the terminals on the Pico. Both the 12v and the ground measure .4-5 ohm.

Arvo
Posts: 294
Joined: Sat Jun 10, 2006 1:30 pm
Location: Estonia, EU :)
Contact:

Post by Arvo » Tue Oct 21, 2008 7:54 am

No-load voltage measurements are no indication. Voltage drops under heavy load (=big current), not while idling.

0.4 ohm is way too much (and must be wrong OR I can't exactly imagine your circuitry or both :)). When your system consumes say 120W, then DC current is 10 amps, on 0.4 ohm resistance voltage drop should be 0.4ohm*10A=4V, what is unreasonable.

Below 1 ohm resistance measurements are usually inexact anyway, unless you have very good multimeter.

If possible, measure +12V on load at all important points - brick output, Pico input and Pico output/motherboard connector (be careful not to short something!).

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Tue Oct 21, 2008 8:49 am

Arvo wrote: If possible, measure +12V on load at all important points - brick output, Pico input and Pico output/motherboard connector (be careful not to short something!).
I'll try, but the terminals on the Pico are right next to each other...I don't think it would be a problem to check the brick without the cover on, but the pico might be trouble.

Arvo
Posts: 294
Joined: Sat Jun 10, 2006 1:30 pm
Location: Estonia, EU :)
Contact:

Post by Arvo » Tue Oct 21, 2008 1:11 pm

ryboto wrote:As for the CPU voltage, that never budges, it's always reported as constant. I'm undervolting, this chip doesn't need much in the way of vcore.
You cannot know about short voltage changes on CPU socket. It is quite possible that feeding CPU power converter with lovered voltage makes CPU Vcore a bit unstable.
--
What about measurements on pico terminals - you can use common ground (not hard to find - chassis will do) and one almost completely insulated thin DMM probe (use wire insulation to cover it) to measure both pico ground and +12V (you need to subtract values afterwards). At least I hope that groud is common - you'll discover that shortly :)

I don't know exact english terms for many things, I hope you can understand me.

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Sun Oct 26, 2008 2:42 pm

Alright, well, finally I was able to test the 12v line again. Under CPU load using Prime95, which causes the system to draw ~120W AC, the 12v line was measured at a few points:
at the brick: 12.38-9v
at the Pico: 11.84v
At the PCIE connector: 11.84v
At the second molex from the pico: 11.79
Reported by PC-Probe: 11.7v

at idle, the power measured at the brick is 12.38-9v. I'd test the voltage where the brick meets the connector, but I'm too scared I'll short something. I'm considering it right now, and if I find the courage to do it, I'll post back.

So, this implies that the brick is good, but either it's wire is too long, the soldering is poor, or the connection between the brick wire and my diy connector is poor, or it's a combination of all of the issues.

edit: measured power consumption at the connection, 11.91v under load.

Arvo
Posts: 294
Joined: Sat Jun 10, 2006 1:30 pm
Location: Estonia, EU :)
Contact:

Post by Arvo » Mon Oct 27, 2008 9:10 am

If I understand correctly (wiring is brick-connection-pico-mobo), then the largest voltage drop (about 0.4-0.5V) occurs in long cable from brick to your connection? If you increase power up to 180W (1.5x), then drop would be about 0.6-0.8V (corresponds to your other mesurements nicely).

Can you take a picture of your DIY cabling (connector)? With all components and wires and soldering points visible? I'm sorry but I just can't clearly imagine your setup and exact measuring points. Mark this 11.91V point on picture too.

Why I'm asking about so many details? Because I wish not give some bad advice. I could say - cut your brick output cable to third of its length and you're done - but maybe I just misunderstood some little point and this won't help. Or I could say - resolder your wiring with thicker wires - but maybe problems lie in brick cable then :)

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Mon Oct 27, 2008 10:35 am

Arvo wrote:If I understand correctly (wiring is brick-connection-pico-mobo), then the largest voltage drop (about 0.4-0.5V) occurs in long cable from brick to your connection? If you increase power up to 180W (1.5x), then drop would be about 0.6-0.8V (corresponds to your other mesurements nicely).

Can you take a picture of your DIY cabling (connector)? With all components and wires and soldering points visible? I'm sorry but I just can't clearly imagine your setup and exact measuring points. Mark this 11.91V point on picture too.
I understand why you're asking, and it all makes sense. I'll take a picture when I find the time, the connection is fairly poor. I have made two, the first one I made was from a 20-24pin adapter, and had cables as leads coming from the connector. The one I made for the Pico was made by removing a 24pin adapter directly from a motherboard, and because of that, the leads are small and weak. I'll take a picture of the Pico connector, and the other one I made to better explain this.

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Mon Oct 27, 2008 8:04 pm

pictures will have to wait until tomorrow, I don't have my camera tonight.

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Wed Oct 29, 2008 6:03 pm

haven't had any time, pictures tomorrow night.

edit: well, that didn't work, but I did take pictures today, but I don't have a data cable for my camera! great. Probably wont get uploaded until tomorrow morning.

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Sun Nov 02, 2008 10:18 am

My soldering is terrible here, not sure if that's the issue, or if it's the thinness of the connector metal.

Image
So, here's the connector that plugs into the brick cable. The upper 3 solder points are for 12v, the lower 4 consist of 3 ground, and 1 pwr-on connector. The pieces of metal coming from the molex connector are cheap and thin.

Image
The Green and black wires coming from the right are from the connector, the other green and black continue on to be soldered to the PicoPSU lead wires and the PCIE power connector. The two yellow and two ground are for a 4-pin atx power connector that plugs directly to the motherboard.

Image
This is the final solder joint, the white and black wires are from the Pico, the Yellow and Black are from the PCIE power connector.

Image
Here's a wider view of the connector and the first solder joint.

Image
just a close up.

Image
another angle.

Image

Here's the other connector I could use, it's soldered to a PW-200-M. The molex comes from a 20-24pin adapter, so it has lead wires, unlike the one I cannibalized from a motherboard.

Arvo
Posts: 294
Joined: Sat Jun 10, 2006 1:30 pm
Location: Estonia, EU :)
Contact:

Post by Arvo » Sun Nov 02, 2008 1:27 pm

On the first image is connector where you measured 11.91V? Then the biggest problem seems brick cable itself (like you already suspected in very first post). Soldering may be terrible :), but looks electrically safe.

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Sun Nov 02, 2008 3:30 pm

Could thin leads from the connector cause the drop? While the soldering might be electrically safe, the leads are small and thin, and break with very little force. The cable was obviously designed for the brick, would Delta really make them knowing such a voltage drop would occur?

Arvo
Posts: 294
Joined: Sat Jun 10, 2006 1:30 pm
Location: Estonia, EU :)
Contact:

Post by Arvo » Mon Nov 03, 2008 9:44 am

Brick output (at cable connector) stays nicely within limits (5%, from 11.4V to 12.6V). Your DYI parts an Pico add just about 0.25V drop at full load, what shifts output voltage off the limits. Seems that motherboard and other components are designed without any reserve too.

Thin connector leads should not add noticeable resistance - they are very short anyway.

What I myself would do:
* cut brick 12V cable to about 50cm/2ft (or less, if possible)
* buy some cheap terminal block for screw connection (for 4-6 sq mm)
* make all joints (from brick cable to PCIE, 4pin 12V and Pico) in same terminal block, using enough thick wire (2.5 sq mm should be enough). You can use two terminal blocks - one for ground, another for 12V. Feed all cables/wires from one side and all bridges at other side; connect blocks together somehow.

Both terminals and (solid) wire (if needed) you can obtain in local electric shop. Maybe there are more interesting terminal blocks available, avoid spring-loaded ones though.
To help fastening stranded wires you should solder them or better buy some wire ferrules (in same electric shop) - you get safer contact.

This way you should s(h)ave about 0.5V, IMHO.
--
But it can be possible that all this doesn't help. Maybe brick output is just little unstable at full load - that you can't measure with DMM, but power instability affects system behavior significantly.

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Mon Nov 03, 2008 11:19 am

Arvo wrote:Brick output (at cable connector) stays nicely within limits (5%, from 11.4V to 12.6V). Your DYI parts an Pico add just about 0.25V drop at full load, what shifts output voltage off the limits. Seems that motherboard and other components are designed without any reserve too.

Thin connector leads should not add noticeable resistance - they are very short anyway.

What I myself would do:
* cut brick 12V cable to about 50cm/2ft (or less, if possible)
* buy some cheap terminal block for screw connection (for 4-6 sq mm)
* make all joints (from brick cable to PCIE, 4pin 12V and Pico) in same terminal block, using enough thick wire (2.5 sq mm should be enough). You can use two terminal blocks - one for ground, another for 12V. Feed all cables/wires from one side and all bridges at other side; connect blocks together somehow.

Both terminals and (solid) wire (if needed) you can obtain in local electric shop. Maybe there are more interesting terminal blocks available, avoid spring-loaded ones though.
To help fastening stranded wires you should solder them or better buy some wire ferrules (in same electric shop) - you get safer contact.

This way you should s(h)ave about 0.5V, IMHO.
--
But it can be possible that all this doesn't help. Maybe brick output is just little unstable at full load - that you can't measure with DMM, but power instability affects system behavior significantly.
Based on the lack of voltage drop at the brick under load, it is implied that the brick is somewhat stable, correct? I kept the leads on the brick for 10-15 seconds, and saw no fluctuation. The 12v goes from 12.4v(idle and load) at the brick, to 11.91v(load) at the first connector. So, there's loss at the connection or in the wire.

I can easily shorten the brick wire. My only issue with using a screw terminal, is that I can't easily disconnect the brick from the system once it's setup. It's a very good suggestion though, one I will consider. I will have to stop by a hardware shop soon and see what supplies they have in regards to electrical wiring. Maybe there are some molex connectors with leads available somewhere online?

If none of this helps, it's no huge issue, the connectors should be cheap. Thanks again for all of the suggestions. I will post back here if I find parts online, or at the hardware store.

Arvo
Posts: 294
Joined: Sat Jun 10, 2006 1:30 pm
Location: Estonia, EU :)
Contact:

Post by Arvo » Tue Nov 04, 2008 8:14 am

ryboto wrote:Based on the lack of voltage drop at the brick under load, it is implied that the brick is somewhat stable, correct?
Not exactly - without oscilloscope you cannot see short fluctuations.
I can easily shorten the brick wire. My only issue with using a screw terminal, is that I can't easily disconnect the brick from the system once it's setup.
You can add some molexes or other power connectors, like push-on terminals.

This seems to be educating thread for me - I've learnt many english terms for electrical equipment :)

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Sun Nov 23, 2008 7:47 am

OK, just an update. I finally had time to cut the bricks wire, and splice/solder it back together. It's 20'' now including the 8 pin connector at the end. Now PCProbe shows 11.93v under full system load. At idle, it's 11.98-12.04v. I'm quite content with this. When I was using the PicoPSU fed by the FSP Zen, I was only seeing 11.8v from PC-Probe, although, the benefit with the FSP is that PC-Probe never measured any fluctuations.

I could probably clean up the solder joints, and use the better connector that's on the PW-200, and solder the PCIE power directly to the 12v line, but considering how long it took me to find time to get to this point, that might be pushing it. Thanks for all the suggestions!

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Fri Dec 05, 2008 8:35 pm

Well, the problem was semi-fixed, at least in regards to the CPU. I just used rthdribl to stress the GPU only, and the 12v, monitored by PCProbe showed the voltage to drop from 11.98v idle to 11.54v under load. So, something is up with the wiring for the GPU, or the Pico, since the card gets power through the board and through the PCIE power adapter.

Aris
Posts: 2299
Joined: Mon Dec 15, 2003 10:29 am
Location: Bellevue, Nebraska
Contact:

Post by Aris » Fri Dec 05, 2008 11:42 pm

ryboto wrote:Well, the problem was semi-fixed, at least in regards to the CPU. I just used rthdribl to stress the GPU only, and the 12v, monitored by PCProbe showed the voltage to drop from 11.98v idle to 11.54v under load. So, something is up with the wiring for the GPU, or the Pico, since the card gets power through the board and through the PCIE power adapter.
So just to sum up, all you did was shorten the power line between the brick and your PC?

Did you replace the original line at all with thicker wire?

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Sat Dec 06, 2008 7:55 pm

Aris wrote: So just to sum up, all you did was shorten the power line between the brick and your PC?

Did you replace the original line at all with thicker wire?
To sum it up, all I did was shorten the brick's power wire, between the brick and the PC. This solved the issue with low 12v line when the CPU was under load. The issue still persists with the video card under load though, implying that the solder joints/wire gauges for the PCIE connector and/or the PicoPSU are not up to the job. The system is 3D stable, but I was really hoping to get a mild OC going, which is going to still be put on hold until I can find some time to redo the connections.

Post Reply