Project: RPM Flow Sensor for Liquid Cooling

[zittware]

Seeing the interest in this forum; I figured I'd start my contribution by keeping you up-to-date on a mini-project for my Aluminum Cube ARs worklog ( http://episteme.arstechnica.com/eve/...1#549007923731) ...

I applogize for those who are regulars to both forum; but wanted to make a contribution here since there seems to be a history here regarding these sensors.

This Design is Copyright 2005 by John Zitterkopf. ALL RIGHTS RESERVED!
Terms of use:
Free for individuals for PERSONAL use.
Business MAY NOT build, include, or sell this product for any reason without prior WRITTEN consent from John Zitterkopf

The idea of this sensor is to report the flow rate of the liquid in a water cooled pc. Ideally; this sensor would plug into the RPM fan header of your motherboard. I've been tossing ideas around in my head for about a year.

I was going to try some of the Innovatek FlowMeter PRO Rev 3.6 devices; but they are impossible to find here in the USA. No one keeps a stock; and rarely gets shipments. Too expensive too.

So I took matters into my own hands and began to think of alternate ways of doing what I wanted.

Start with some mini Flow indicators found on the web. I'm not going to post a direct link here... The BestLabDeals.com boneheads are not worthy of my endorsement. If you really have to know where to get them; see my ARs thread. Instead I'll post some other "vendors" who might be better at delivering these parts:
http://service.belart.com/cat/199370002.html
http://www.omegabiotek.com/supplyweb/flowmeters.htm

These indicators are the perfect size for a small case - About the size of a US Quarter:


The idea:
Mount an IR (infrared) side-emitting diode on one side of the indicator... and an IR side-detecting Transistor on the opposite side.
Then with support circuitry detect the make/break of the IR light beam as indicated by the paddle wheel. Given a faster paddle wheel; the faster the transistor will switch.

So; first step... how to mount the transistor & diode? At first I thought about drilling a hole into the case of the indicator... but quickly dismissed that... What kind of bone-head drills into a water tight device? Too much risk of springing a leak.

In comes the smoke grey plexi I've used through my mod... Cut a circle with a drill press... drill the hole in it... mount it to the side. Epoxy and clamp the side once the holes are lined up "across" the indicator.
Epoxy the transistor and diode into the small holes and whala: Instant Mockup

Here's a top view of the mockup:


And here's a side view of the mockup:

I applogize for the grainyness of the above picture. I need a "macro" lens for my DigiCam. :(

Now that I have the mechanicals taken care off;
I'll let the Epoxy cure overnight.

I'll begin designing the support electronics; such as diode and transistor biasing... as well as possible hysterisis to detect the paddle pulses. I have some ideas; but will need to experiment on what it'll take to make this flow rate sensor work. :wink:

I'll post progress as I make it... if there is interest.

[Pug]

Nifty.

I did a project at school based on the same principle; where we triggered an Exit sign to say goodbye when people walked toward the door.

Never thought of using it under these circumstances. Good call.

[fivecheebs]

Yeah! Great idea.

There is certainly some interest here. I take it you will be measuring the number of passes the vanes. How will you calculate the flow rate?

[Squelch]

I like it.

Non contact, minimal flow impact, and electrically isolated. All in one nifty little package.

I hope you get enough pulses from it to be accurate.

[scopEDog]

Definitely, thats a great idea. Are you looking to have it plug into a mobo header for rpm monitoring? Would be kinda neat to have its own 7 segment (a few) to show rpm.

[Starbuck3733T]

John, this is definitely the way to fly!

[dutchcedar]

zittware... I really appreciate what you're doing here... it seems like a great solution.

I stopped in on your worklog at ARS... and take my hat off to ya. Very, very nice work you're doing on the cube.

[zittware]

Originally Posted by fivecheebs

I take it you will be measuring the number of passes the vanes. How will you calculate the flow rate?

Correct.
I thinking of flow rate as a function of "RPM"...
I am toying with the idea of "calibrating" the RPM vs flow rate... but I'm not sure that is really necessary.

For my case; I'm just interested in "balancing" the flow through my parallel paths. IE relative to each other. I'm not really interested in an acutal flow rate measurement.

Are you looking to have it plug into a mobo header for rpm monitoring? Would be kinda neat to have its own 7 segment (a few) to show rpm.

Correct. Fan mobo header.
For me a 7seg is overkill... I have a 15" LCD moded into my Aluminum cube to do that already. ;)

[zittware]

Last night, I breadboarded up a circuit based upon
http://www.discovercircuits.com/PDF-...stphotonpn.pdf
fast response circuit in the upper right hand corner.

I'm happy to report that inital reports look good.
I was able to capture a square wave based upon the speed of the paddle wheels on my Oscope.

I noticed a "hump" during the low cycle of the squarewave. At this point; I'm not sure if this is an artifact of my WorldWar2 era Oscilloscope or a real issue.

I suspect it's a real issue because I noticed that the paddle wheel appears to be slightly transparent in the middle. The edges of the paddle wheel are reinforced with more material; as such it is more opaque... and would block more IR light. This would explain the "hump" in the cycle.

I suspect I can tune this out using the voltage divider given by 100k/4.7k.

[zittware]

No engr should be without a bread board:



I'll post a schematic later when I actually have the design in Eagle. Right now the circuit uses a standard LM385 dual opamp... I need to convert it to a quad opamp design (for the 4 sensors I'll be needing) and to surface mount. I'll also need to add a transistor and maybe a linear regulator to make it compatible with a mobo fan header.

I hooked my mockup to my koolance pumps:


And the result for your viewing pleasure; a short video:
RPM as the flow is restricted

I appologize for the low quality video; only motion camera I have is a camera cellphone.
Codec is DiVX
Soundtrack brought to you by CDMaster32 and He's Dead Jim.Mod.

The video shows that as I "pinch" the water flow the square waves get longer... ie less rpms... less flow.

For those Oscilloscope junkies; the time/div was set to 5ms/div and 0.5V/div.

Keep in mind the RPM is three times the acutal number because there are 3blades which break the IR light beam. This probably a good thing; because the rpm needs to be readable by the on-mobo fan sense chip. If I build a circuit to convert this to one transition per revolution; we'd likely have a problem with the sense chip measuring the rpm.

This won't be a problem for me because MBM5 allows me to /3 on the fan settings screen.