CrystalFontz thermal monitoring, small success.

[Martinm210]

So I decided before I do any more thermal testing that I was going to find a way to log water/air temperatures automatically to ease the pain a little. When you start splitting degrees on two blocks, gathering more and more data seems to be key in making any worthwhile conclusions....so...

In the end I decided to go with a Crystal Fontz CFA633 for the following reasons:

-Logging of up to 32 digital thermal sensors
-Digtial Sensors are very accurate with extremely good resolution. .5C accuracy with .0625C resolution and .01C relative accuracy.
-Digital Sensors are digital so there isn't error introduced by the thermal wire (Resistance) leading up to the coupler or transistor, etc.
-Logging exports to excel.

So here are the sensors I started with, just 4 sensors, two for air on the radiator and two for the CPU loop, one inlet and one outlet. I chose to do this so I can measure any differences in outlet temperatures, I thought this may be something that will change with higher restriction blocks/nozzles, etc. It would also provide a secondary sensor to confirm the inlet is operating correctly.



This is how I integrated the sensors as thermal probes for measuring the water temps, low restriction, but good sensor placement.


A standard Air or surface sensor construction, basic three wires and all wires can be connected in series which is nice.


The LCD with the four temperature probes turned on the display. There is alot you can do with the LCD, I only know enough to be dangerous..


And a short quick and dirty logging of both TAT and the CrystalFontz. As an added bonus they both log temperatures in the exact same time format so correlating the two along the X-axis is really easy.



Some observations:

Water Inlet/Outlet temperature differences in this high flow setup(3GPMish) are so small, I can't even really measure it. I'll see if this changes with a high restriction nozzle setup. I always figured you'd see a degree or two even at these flow rates.

I did some calculations on specific heat and figured it takes about 350+ watts to change a 1.5GPM flow rate just one degree, so it would make sense I suppose that at 3 GPM, it would take nearly 700watts to change one degree. The E6600 doesn't even make a mark, or at least not one strong enough to measure.

TAT produces some wacky ups and downs, check out the swap between core 0 and 1 where they sort of switch places. I guess this all the more emphasizes the need to log temperatures for a very very long time to minimize some of this. I'm thinking I'll start testing and run it for at least an hour, maybe more just to gather more results to average. It's too bad it fluctuates so much, but it is what it is..

Resolution on the CrystalFontz Dallas DS18B20 Digital sensors and relative accuracy seem to be very good. You can see by the two water sensors that they hug the same temperature within the .0625 resolution very well, you can't even hardly tell there is a blue line under the orange one, but there is and those are two completely separate sensors.

It's a bit suprising to see the difference between the to radiator intake sensors. Lower to the floor the cooler the air and that's more significant that you would think...

Anyhow, it's working and I think I'm finally ready to do a little thermal testing. Going to let the computer do the work..

The question is what should I use the other 28 channels for...

[NaeKuh]

OMG martin you gotta sell me a couple of those inline temp probe!!!!!



I want a couple with wires. :T now that i think about it... those would probably work on my senor board on my matrix orbital(it only takes 3 prong temp probes!!). OMG martin you definitely need to make me a few with wires!!!

Well the center piece would be good enough. :T

[septim]

cool. more toys... i like...

[MrToad]

Thanks Martin (again). Wish you had picked a different area of expertise though.

If you would have gone for pure physics, we'd have cold fusion by now, and sod off all those fossil fuel burning power plants.

Anyway, I like how you rigged those temp probes. Neat and tidy ^^

I may actually let you work inspire me (read: plagiarize) and do the same. I like it more than my obtrusive approach

The probe is the same one you are using, but it's sleeved and then attached through an EK G1/4 plug.

[Soulman84]

great job martin also in italy someone of the staff of "pctuner" did the same,measuring h2o temp before and after cpu wb, and also changing position of components in the loop.
Same results of your testing..no variation

[Xilikon]

I love it, especially the water thermal sensor Wonder if this will fit with any temp sensor for M-Cubed or Aquaero ? If yes, I'm in line for 1 myself.

[Jedda]

Pretty much the only water temp changes you'll see will match the changes in ambient. Could be different when the rad is really being pushed but I doubt it.
However, like myself, you have to see for yourself and that's a nice bit of quality kit to do it with. All accuracy, little bling, at a good price.

[Shocker003]

[QUOTE=MrToad;2759590]Thanks Martin (again). Wish you had picked a different area of expertise though.

If you would have gone for pure physics, we'd have cold fusion by now, and sod off all those fossil fuel burning power plants.


I totally agree with you Mr Toad
Nice work Martin and i hope you will start a small company that produces nice sensors for us. Keep the good work up

[Martinm210]

Thanks!

Not sure what exactly these DS18B20 sensors will work with, but I like them alot! Very simple little digital self contained units. I bought 10 of them on ebay for $4.50 each, so I have a few more left over to experiment around with. Several will be wired into a test rig wiring that I'll extend off my existing set, then I can run a string of sensors off to test external loops.

It still adds up to a little money, but not bad. I bought the unit used on ebay for $25, bought some cables and the hard drive bay for $40 from CrystalFontz, then another $45 for the sensors, so in the end I have about $110 or so into it.

I like that sleeved sensor, I was going to try out something similar if this doesn't work, but they seem ok right now. I just drilled a hole small enough to slip them through the acetal and encased it all in epoxy which seems to do a good enough job, but not as clean as that...very nice!

I guess the interesting part for me is how unprecise TAT really is. Not sure if everyone else sees the same thing, but even in this little 10 minute run, there is a good 3-5 degree variability in the TAT temps. All the more reason to run very long logged test runs to average that mess out some.

I'm just happy to have it all logged. Now rather than running 5 individual smaller 10minutes tests logging it all down manually, I can just let it rip for a few hours and come back with the work all done..

[Xilikon]

Just a funny thing to point out, in your LCD display, you should write H2O instead of H20

If you can log everything so 1 line contain all the temps at a given time, you should register ambient temps as well then when you have a good Excel log, you can weed off results if the ambient is not the goal (like 21C so delete lines with ambient not 21C). This will let you have a very good average with a precise ambient temp.

[coolmiester]

It’s a great job on the sensor Martin but I can’t help thinking even though the sensor itself is tiny it may interfere with the flow rate quite a bit?

Wouldn’t the D-Tek Sensor approach have been more beneficial in this instance?


@ MrToad – you could always drop yours inline into a DD T fitting but your probably onto that one already.

(not trying to take anything away from the great innovation)

[Martinm210]

Just a funny thing to point out, in your LCD display, you should write H2O instead of H20

If you can log everything so 1 line contain all the temps at a given time, you should register ambient temps as well then when you have a good Excel log, you can weed off results if the ambient is not the goal (like 21C so delete lines with ambient not 21C). This will let you have a very good average with a precise ambient temp.

You guys are using water!??...got liquid Hydrogen running here, the 20 atom variety...

Fixed...

Yeah, I could run another probe to the front of my case although the bottom radiator inlet sensor should be very very close, after I push my case under my desk, I set barriers against the sides of the radiators to minimize any recycling of air. But I had planned on doing something similar for my other loops anyhow, and my chipset loop has an MCR220 up front, so it would be perfect right there. Got 32 channels to burn, might as well make use of them...

It’s a great job on the sensor Martin but I can’t help thinking even though the sensor itself is tiny it may interfere with the flow rate quite a bit?

Wouldn’t the D-Tek Sensor approach have been more beneficial in this instance?


@ MrToad – you could always drop yours inline into a DD T fitting but your probably onto that one already.

(not trying to take anything away from the great innovation)

I could always test it for restriction, but it's not restrictive at all per my highly scientific blow through test , I really couldn't see a difference between a straight through T and one with a probe sticking in there. I've experiemented a bit with my old sensor in my T-line and while you can leave it up and out of the flow, it makes it less responsive. For the most accurate realtime reading I would recommend that at least a portion of the tip of the sensor is in the stream of flow.

In the end it's probably being anal about the accuracy of the CrystalFontz sensors when you start looking at how poor the Core Sensors are, but I'd rather make it as accurate as possible. Also the CrystalFontz can only use the Dallas sensors, it can run two types, a cheaper 2C accuracy version and these which are good to .5C.

[Xilikon]

Another reason for sticking the probe a bit further in the water stream is to avoid laminar flow. The water on the outer rim would be slighty warmer than the water in the middle. Just little details like that can make a difference

[Martinm210]

Thanks!!
I'll give those a shot, I'm sort of starting all over here, so I might as well experiement a little with the options..

[Jedda]

Those regular dips every 30 secs after 9.50:24 with the other core flat lined really are interesting. Would love to know whats going on.

[MrToad]

@ MrToad – you could always drop yours inline into a DD T fitting but your probably onto that one already.

(not trying to take anything away from the great innovation)

I've got it "plugged" into the EK Res ^^ Can't be more "out of the way".

Before that I was thinking on plugging it into a Norgren 1/4" BSPP Tee

I wouldn't be able to use the tee you linked even if I wanted. As you can see in the picture it protrudes too much from the plug. It's measured to sit right above the flow when using the fitting I linked above.

Still, thanks for the suggestion. I quite like the DD acrylic one

[SiGfever]

Those sensor hubs are fantastic. O-K now you gotta quit your regular job and start making pieces parts full time for us addicts.

[NaeKuh]

Those sensor hubs are fantastic. O-K now you gotta quit your regular job and start making pieces parts full time for us addicts.

i so want a couple of those ring temp probes.

use a coupler on one end mount it to the outlet of your radiator or inlet, and bam! you got barb'd temp probes mounted on the radiator directly

[Xilikon]

i so want a couple of those ring temp probes.

use a coupler on one end mount it to the outlet of your radiator or inlet, and bam! you got barb'd temp probes mounted on the radiator directly

Damn, why didn't we think of this !! With a couple of male G1/4" couplers, we can get the in and out temps as close as possible to the radiator. I might try that myself.

[Martinm210]

Yeah, you could do that with a DD delrin Tee, that's basically all I made except minus the one hole in the side. A G1/4 plug would work fine for that though...

BTW, I also found out I can log motherboard CPU, GPU, and System temps using Nvidia system monitor,
You just have to import the file space delimeted, and merge it with the rest.

That might be another good reference.