Auto shutdown idea??
 

Hey all, I was just sitting here bored out of my mind and came up with a possible idea to automatically shutdown a PC in the event of a w/c system failure....

What about a system similar to the auto-shutdown used by uninterupted power supplies?? The idea would be to take an electronic flowmeter and temp sensor and place it in the w/c loop then connect both to some sort of microcontroller which could be connected to the PCs serial port for monitoring. Then just write a simple monitoring app where the user an set the limits could monitor the data on the serial port and shutdown the computer if these are exceeded.

If anyone has any interest in something like this please let me know, especially if they know something about microcontrollers... I could easily write the software part.

Bill

Sounds way more expensive than need be.

A microcontroller? Why not just kill the PSU through a simple flowswitch?

I know there's been an article on just that on Bit-Tech

Think his idea is to have the software shut down the system rather than simply unplugging it... Generaly a shutdown is better than just pulling the plug for avoiding data corruption.

If something went wrong with the cooling CPU temps go up at a high rate. Seems you could monitor CPU temp via software and then do the shutdown? I would think it would be easy to do that. You could check for rate of change and initiate the shutdown when it exceeded a limit.

Since that is obvious, I'm thinking you may be looking to 'pre-sense' catastrophic failure, but I'm wondering if monitoring the CPU temp would still not accomplish that? Auto shutdown at 55C for instance ought to catch it so long as there was liquid still in the loop. If the failure was catastrophic due to lost coolant and the WB went dry, I'm not so sure you could pick it up fast enough to beat the bios shutdown?

Right now I am using MBM to monitor temps with it set to shutdown the PC if they get too high. Of course the real problem with this is it is using the winbond measurements which are not very accurate. I also thought that by monitoring via a combination of flow rate and water temp you would get some more options (maybe warning messages for decreases in flow). Plus I really don't want to have to wait for the temps to start rising before the computer starts shutting down (especially if there is a big problem like a dead pump), if the flow rate decreases below a certain level then start shutting down even if the temps havn't gone up yet.

This is all just food for thought though, until someone who can actually design a working circuit for it comes along then maybe I could actually build and test it!!!

Bill

pretty much the same question:
http://www.ocforums.com/showthread.p...84#post3177284


ive always just used mbm5. Having a tube like fall off a barb is worst case scenario. Do Opterons throttle? I dont think so, cant find any info on google about it.

The inaccuracy with the winbond chip isnt a big deal. The temp sensors are constantly inaccurate. So it will always show 45c when your temps are really 50c for example. The problem with it is there is a good delay between changes in mbm. If you set mbm to do a force shutdown, not a normal shutdown, at a temp thats within 5c of your max load Id imagine you would be ok for most situations.

We discussed this extensively, in the past.

The shut down cannot be entirely software based, because you can't guarantee that the PC hasn't crashed. Ideally, it'd have a software shutdown included, because there's no clean way of turning off the PC otherwise.

i've always theought a good way to measure a major failure in the system would be a water pressure sensor. Set it up so if it drops Xpsi/5mins shutdown, Xpsi/30 mins shutdown, Xpsi 60 mins shutdown, psi Below X Shutdown. X is used as a variable there because im not sure on the exact numbers to be used. the idea is that if the water pressure drops by a certiant ammount in a given time frame there is a leak in the system or the pump has failed, and if the system drops below a certiant pressure it is either low on fluid.

I am using a Gems Rotorflow with pulse output fed into a fixed I/O PLC. It gives me a rough idea of flow rate. (The PLC is for my automated startup sequence and monitoring).

These paddlewheel units are also available with setpoint switches installed.
edit: The paddlewheel is also a neat visual indicator...

I don't know how to initialize a software shutdown. But I can automatically hold the power button for 4s, etc. to shut down with the BIOS ...or take more drastic meaures... :rolleyes:

I'm almost positive there are commands to initiate a shutdown with any OS really. I would believe there is some part of the windows API that would allow you to do it. It's just a matter of digging it up.
:)

There have been a few of these posts and I don't get it. Am I missing something? I use mbm and one of the free shutdown proggies. Temps hit "X" and execute the shutdown.exe and instant shut off. I know my temps on the worst of days won't go over 40 so I set the threshhold for 48c and have tested it and it works. Am I missing something? Do I have too much faith in the software?

Seems like you'd want to check coolant level. Plenty of ways to do it...

The biggest problem I see in relying on a software solution is that 99% of the time people use the CPU diode temps which have proven unreliable time and time again. They are also not as accurate as some of us would like. Plus, like Bigben2K stated, if you are solely depending on software what happens if the system has already crashed.... you're sol and the software solution is going to provide you with 0% of protection.

Windows shutdown is possible through API calls. Specifically the ExitWindowsEx function located is the user32.dll for all you VB guys out there
Declare Function ExitWindowsEX& Lib "user32" (ByVal uFlags&, ByVal dwReserved&)
uFlags: (Combine flags using the OR function)
0 = EWX_LOGOFF
1 = EWX_SHUTDOWN (Powering off if possible)
2 = EWX_REBOOT
3 = EWX_SHUTDOWN (No Power off)
4 = EWX_FORCE (Force termination of processes that don't respond
dwReserved = 0
Returns non-zero on success, zero on error


Bill

Thank You for the information. I will put this to use. I have some I/O issues to sort out still -but again, thanks.

i don't think temps are a reliable way to do it because it could be hours after a leak starts that your temps raise enough to activate your shutdown and by that time your computer is fried

I may someday add a level switch to my tee hose. I think that I will rebuild it at some point anyway...

As Ben stated there are threads dealing with this. I might do some research and revive one of them. I am getting more interesred in this -if for no other reason than to achieve further automation...I thrive on weird things. :rolleyes:

Hey CoolROD, can you post a pic of your flowmeter setup? I made one that is quite similar to your design, in fact I even know the exact flow meter you're using, just too bad they were sold only in packs of 4...

Here's a picture of my flow meter that right now is connected to a buzzer that goes off once the flow drops below a certain point. I eventually want to implement some kind of shutdown function to complement the buzzer alarm.
http://img.photobucket.com/albums/v1.../FlowMeter.jpg

That is a sweet sensor you have there. I bought (1) directly from gems sensors online. Unfortunately I live on the road and don't have a picture ...I will gladly post one next week. My old web page is outdated by about a year now -so that is no help.

How are you generating your tach signal?

If you have a buzzer wired in, why not wire in a relay on the buzzer signal to cut the power on the PSU? Use the ATX power on signal wire maybe?

IIRC, the powersupply is designed to provide power, so long as the PS_ON (usually pin 11?) is grounded. If the voltage is high, it has to cut out. You'd probably need a way to take it off of ground. I could be pretty off base though, as I was only looking into a way to activate a PSU without the motherboard.

Here's a good link on something that would work pretty well. A delayed signal that would allow time for a shutdown via something like Shutdown Now, and if there is a crash and the computer can't shutdown, it would hard power down the comptuer.

http://linuxcar.sone.jp/pic.en.html

on a different note - a cheap "low coolant" alarm