PDA

View Full Version : Water cooling system failure protection


xcr20
08-31-01, 10:14 PM
I was in another thread (Home Made Water Block) where N3D2000 was designing a type of waterblock that could handle a fan to provide extra cooling and cpu protection in case of a pump failure. So everyone who has a water cooling rig had (or forgot) to think about what would happen in the event of a pump failure. If you had the chance to see what it does (see example at http://www.overclockers.com/tips220/ ), you might wanted a solution to prevent this kind of accident. After thinking of it for a few days, I came with a possible (and cheap) solution.

A lot of ppl know about thermal sensors. How about using it in a protecting device ? The principle is simple, get the level of water temp then if it's too hot, power down psu. And this can be done with cheap IC's soldered on a little pre-made pcb that could be done quite simply.

I'm about to finish the prototype of a thermal sensors monitor (college project). So it will monitor 4 probes and 1 of those will be the water temp. If water temp raise above previously set level (comparator), a relay is set to send High level to power switch, shutting down computer. Maybe some system like this has been done or ppl did a commercial product that overclockers can use...So if I talk about something usual in overclocking please tell me. But if not, then tell me if it would be interesting to continue this thread and maybe post a design and make an easy HowTo.

Thanks !

xcr

dimmreaper
08-31-01, 10:54 PM
Why not just solder some wires on the MoBo to make the system temp thermistor a remote one. Then drill a place for it in the waterblocks base. Then set MBM5 to shutdown when the temp gets too high (BTW type "rundll.exe shell32.dll,SHExitWindowsEx 5" in the run program dialog box and see what happens). Seems like it would be a lot cheaper and easier to me.

ken257
08-31-01, 11:15 PM
If you buy a good pump like a Eheim it is not a problem. The fan on a normal cpu cooler(especially these high rpm fans) will have a shorter life span then a quality pump.

dimmreaper
08-31-01, 11:19 PM
Originally posted by ken257
If you buy a good pump like a Eheim it is not a problem. The fan on a normal cpu cooler(especially these high rpm fans) will have a shorter life span then a quality pump. Agreed!

touser
08-31-01, 11:26 PM
my question is what about the temp monitor in the bios? my bios has an auto shutdown feature to shut down at the temperature i tell it..now does this not work? i have never tried it..has anyone else? maybe i will try and see what happens :)

xcr20
08-31-01, 11:41 PM
Thanks for the advice guys ! I think i was getting things too complicated (Or being paranoid maybe...well i don't know if i could rely entirely on software to protect the computer). So good pump is a good way to start. Mine is not a Eheim :( Is there a way to know the time to live of a pump ? How can you visually know it's time to change it ? And how big is the probability of a failure. Is it happening often ?

xcr

IFMU
09-01-01, 01:55 AM
Well theres a small problem there... i cant solder... did i spell it right? lol.. hell i dont even think i can spell it... then as far as software... yea... i put my life in it... wtfe.... lol.... all i gotta say there is WINDOWS!!!

VashTheStampede
09-01-01, 02:22 AM
Couldn't you just somehow wire up a LED that will stay lit if the pump is operating normally? Bleh I don't know how to solder either, though I do have one around here...

~RT~

Froggy1
09-01-01, 02:39 AM
You could "T" in a Pressure switch to light an LED or shut the system down!

OddOne
09-01-01, 03:11 PM
I have my own approach to the problem...

http://www.ocmojo.com/guides/electronic_flowmeter/images/completed_flowmeter.jpg

I'm wrapping up a how-to article on building your own flow alarm. The prototype above makes an alarm sounder freak out when the flow drops below a preset level. It's all solid-state, no moving parts, with no electronic parts in direct contact with the coolant.

And, as the pic hints, it's an unholy simple circuit. :D

Since the article is being posted on a website I'm webmaster for and the forum has a no-links policy I'll not post the URL - you'll have to flash me a message or some E-mail for the location. (Some folks will figure out the URL on their own anyway.)

oO

dimmreaper
09-01-01, 03:31 PM
That looks cool OddOne.

But realistically you won't need pump failure protection anyway with many CPUs (unless your runnin a peltier). Coppermine and newer Intel CPUs have overheat protection built in. Palomino and Morgan AMD cores do as well I believe. If these chips get too hot they will shutdown to prevent damage.

OddOne
09-01-01, 04:00 PM
Actually, if you run water cooling you need a failed-pump alarm of some sort no matter what processor you're using. Sure, they do include thermal protection to some degree but in the real world a processor will go into meltdown if the water flow stops. This is true of all current P3, P4, and Athlon family processors, thermal protection notwithstanding.

The current record for frying a T-Bird is 6 seconds.

EDITED TO ADD: This is what a waterblock looks like after a pump dies on an Athlon 500:
http://www.overclockers.com/tips220/bad.jpg

oO

xcr20
09-01-01, 05:33 PM
I think we should'nt rely on existing cpu protection or software as they are not as reliable as hardware.

Try to imagine...Your sensor detects abnormal condition of system then let's say MBM gets it and does the super fast exit command (as dimmreaper suggested it) What if Windows hang when exitting ? Or if you messed up the exit command ? You're fried (and your components too)...I would not personnaly take that chance.

OddOne, that was exactly what I had in mind and it's really simple. Would there be a protection if you're away and the alarm's beeping ?

xcr20

OddOne
09-01-01, 11:54 PM
OddOne, that was exactly what I had in mind and it's really simple. Would there be a protection if you're away and the alarm's beeping ?

Not as shown, but nothing says you couldn't have it trip a time delay circuit that could operate a relay that in turn could cut power to the system after say 30 seconds.

oO

dimmreaper
09-02-01, 04:39 AM
Hey OddOne. How hard would it be to make a device that uses a thermister to keep an eye on CPU temps, and say when the CPU reaches 65C it cuts the power with a relay? Would that require any programmable logic ICs? Or could you do it with just cheap readily available parts?

Something along these lines could be usefull in aircooling aplications as well as watercooling aplications . . . .

SavageHenry
09-02-01, 09:37 AM
If we're going to all the trouble of auto shutdown, why not have the alarm communicate with the PC through the serial (or USB) port and issue a shutdown command that way? This works pretty reliably for UPSs when power dies.

I suppose there would still need to be something in place in case the system takes too long to shutdown, or windows doesn't cooperate. . .

YMAN
09-02-01, 09:59 AM
When A pump dies it may be time for a New Computer! :cool:

Why not have some type of backup pump system?
One fails - an electronic circuit turns on the other one!

OddOne
09-02-01, 01:33 PM
Hey OddOne. How hard would it be to make a device that uses a thermister to keep an eye on CPU temps, and say when the CPU reaches 65C it cuts the power with a relay? Would that require any programmable logic ICs? Or could you do it with just cheap readily available parts?

Something along these lines could be usefull in aircooling aplications as well as watercooling aplications . . .

Certainly doable with an op-amp being used as a voltage comparator, and a pair of resistive voltage dividers for the input circuit, one with a potentiometer in it to set the cut-off point, the other with the thermistor in it to act as the sensor. When the voltage drop from the thermistor decreases too much it'd swing the op-amp output, and that could be used to trip a relay, drive a pwoer MOSFET, etc.

I'd think it could be done with a 741 op-amp IC, a power MOSFET (say an IRF520) as a driver, the thermistor, a potentiometer, and a handful of resistors and capacitors.

The trick would be mounting the thermistor, as it'd need to be in direct contact with either front or back of the processor die. If it wasn't, the setting would have to be adjusted to compensate for the difference between detected and actual chip temps.



If we're going to all the trouble of auto shutdown, why not have the alarm communicate with the PC through the serial (or USB) port and issue a shutdown command that way? This works pretty reliably for UPSs when power dies.

I suppose there would still need to be something in place in case the system takes too long to shutdown, or windows doesn't cooperate. . .

Therein lies the problem. If temps have elevated to a critical level, odds are Windows has already crashed or will be in the process of crashing as the processor destabilizes. The best backup plan would not depend at all on the computer itself.

oO