Radiator Thermal Test Bench Progress

[Martinm210]

Spent alot of time working on this little testing facility project.

Idea being:

CrystalFontz to control and log Dallas DS18B20 digital sensors, .0625C resolution, 32 channels

Air IN = 2 sensors per 120mm fan
Air OUT= 1 sensor per 120mm fan
Water IN = 1 sensor
Water OUT = 1 sensor

I got all but the water sensors done, here are a few pics. I'm thinking I'll mount some heavy guage wires for the air in sensors so I can adjust their locations just right for different thickness setups.

I figured with the sensors all permanently mounted it would both eliminate the hassel of mounting sensors every time and also provide better consistency between tests for the sensor locations are fixed.

http://img381.imageshack.us/img381/6069/radiatorbench2gd5.jpg

http://img381.imageshack.us/img381/7481/radiatorbench3rf0.jpg

http://img381.imageshack.us/img381/5946/radiatorbench4oe5.jpg

I don't have my air flow meter yet, so for the time being I thought I'd just get the stand built and maybe start tinkering with a fan only type of test.

Any comments on the sensor placement, etc?:confused:

The only thing I was thinking is the air out sensors will also be measuring heat from the fans, but I figured that's ok since those are more informational than anything.

UPDATE 4-18-08
Ran a few preliminary tests today.

Here is my string of sensor ready for logging, 4 of them are in my computer that I'll simply disregard, but 14 of them are specifically for my radiator test bench:
http://img212.imageshack.us/img212/5452/radiatorcrystalfontzqv8.png

My very first test with a target heat load of 100watts, 15watts from the pump, 2 watts lost in my variac and 88 watts from the 300 watt water heater turned down until my "Kill-a-watt" registerd 88.
http://img212.imageshack.us/img212/3349/radiatorpa1203prelim1kl0.png
Takes a while longer to stabilize than I thought, I probably need to run this one again, but it's giving me an indication that at a bare minimum 20minutes of warm up is needed, but 30minutes is even better.

Then I tried a 200 watt run and found my first little glitch. The thermostat on the water heater kicked on mid way through the test...whoops..:)
http://img212.imageshack.us/img212/1996/radiatorpa1203prelim2ks2.png

Working out the bugs, but the logging is working nicely. My C/W results will be much lower because of heat lost through the tubing and reservoir, but I figured that's ok since a regular system will also. You just won't be able to compare c/w except other tests on the same test bed.

The first keeper:
http://img292.imageshack.us/img292/4069/radiatorpa1203prelim3je5.png

[nikhsub1]

The only thing I was thinking is the air out sensors will also be measuring heat from the fans, but I figured that's ok since those are more informational than anything.

Thanks!
Martin
Moving along nicely. Yes, you are correct but it won't matter so long as you use the same fans on each rad...

[Vapor]

In fact, using the same fans on each rad might be more helpful than using the same airflow.

End users can control which fan they use, but only maybe a handful know the kind of airflow they're moving.

Great looking testbed btw....the work and a tests you do are just awesome.

[disruptfam]

nice martin!

[Martinm210]

Thanks!

Yeah, I figured I'm going to stick more in line with what previous testing was done using specific fans and undervolt them, etc. At the same time when I get my anemometer, I'm going to see about building some sort of outlet shroud so I can also just record the actual airflow produced by those fans, then I would sort of have both.

Not sure what I'm going to do with the air out or water in, but it'll be information to look at and could be numbers to check some things.

[systemviper]

Hi Martin.
Thanks for all your effort, all this extra info helps us spend our money more efficiently, and that is Xtreme savings
from even more Xtreme info
coming from even more Xtreme test set-up's

You work has already helped me make some buying decisions .

So I want to thank you for all your time and effort! :up:

If there ever was a way I could help, I would.

I think we need a Martin fund, to help the cause......

Cause you need more test equiptment! :clap:

[biftek.]

You've got to be the Sheldon Brown of water cooling. ;)
Thanks for everything you do.

The test set up looks really nice.
Is the Neoprene along the edges of the wood? Looks pretty thick.

[skinnee]

This looks great!

Can't wait for the testing to begin!!! :eleph:

[DavidNJ]

You are one very industrious dude. Fortunately, my wife isn't expecting me to make cabinets! She just doesn't want me to spend money on manometers, temperature probes, rotational speed gauges, etc.

The testing setup could be better with a shroud; that would test the whole radiator and not just the space under the fan blades.. And the ability to vary up to a static load of .4-.6" with 40-50cfm/fan space flow to test the most restrictive radiators. A two-stage or some push-pull setups could get there with aggressive fans. The most aggressive here is rated at 51dba, in a two stage it would pull .7" with over 40cfm. With a shroud and some sound deadening would be in a range Ira would use.

I have an uninstall ed GTX480 here, and can easily imagine it improving right through .6". While I haven't tested it, my lowly MCR220 with single stage shrouded Panaflo 120x38 Ms could easily use more.

Some testing is needed, however I imagine the fans work best on the high flow side of knee in their curve. Then adding stages to get the pressure drop across the radiator to get there. However, without the shroud, a significant portion of the radiator surface may not be fully used.

Sidewinder has the Delta, and I'm sure he would make you a deal.

[Martinm210]

You are one very industrious dude. Fortunately, my wife isn't expecting me to make cabinets! She just doesn't want me to spend money on manometers, temperature probes, rotational speed gauges, etc.

The testing setup could be better with a shroud; that would test the whole radiator and not just the space under the fan blades.. And the ability to vary up to a static load of .4-.6" with 40-50cfm/fan space flow to test the most restrictive radiators. A two-stage or some push-pull setups could get there with aggressive fans. The most aggressive here is rated at 51dba, in a two stage it would pull .7" with over 40cfm. With a shroud and some sound deadening would be in a range Ira would use.

I have an uninstall ed GTX480 here, and can easily imagine it improving right through .6". While I haven't tested it, my lowly MCR220 with single stage shrouded Panaflo 120x38 Ms could easily use more.

Some testing is needed, however I imagine the fans work best on the high flow side of knee in their curve. Then adding stages to get the pressure drop across the radiator to get there. However, without the shroud, a significant portion of the radiator surface may not be fully used.

Sidewinder has the Delta, and I'm sure he would make you a deal.

I'll see what I can capture with this first. I figure a vast majority of users are using radiators without a shroud, probably 90% I would guess. In the end I want to capture performance that is closest to the conditions most people use them. The only thing artificial will be my heat load, everything else should be normal radiator operations and conditions. The fan hub definately affects performance, but if that's the way most people run them, then I would like to capture it like that.

Also with the open top, I purposefully allowed extra room with my air out sensor lengths so I can experiement with the effects of a shroud. It's pretty flexible like this, and I can try alot of different things, the only thing absolutely fixed is my water and air inlet sensors.

I also want to be able to display the results in a way that more people understand it. If I plotted CFM vs performance, I bet most folks would think that just means how many fans x their rated CFM which is not correct. But if I post a curve with a specific fan and fan voltage, that something people can relate to pretty well.

I will build a shroud I can slip over the fans and try to measure spot CFM when I get my anemometer. With a certain fan, I should be able to slip this shroud on and measure the actual CFM. If I do this with each type of fan, I should have captured enough to also do a CFM Vs Heat dissipated type of look. It would also indicate how restrictive the fins are (Not quite pressure drop, but an order of restriction). I might also try to measure pressure drop of the radiator fins in time, but that's going to be a bit tricky considering every manufacturer is a bit different and trying to get a one size fits all perfectly sealed shroud is going to be a challenge.

Just getting started, I'm sure I'll be tweaking something before I'm set in how I'm going to do the testing, and that will probably change over time as I learn more about it.:D

[plato333]

looks really cool man i was thinking about doing something similar.

[SiGfever]

Martin,

You do some fantastic work and we all appreciate your time and expenditures for these tests. :clap: :clap:

I noticed that you have the tubing tie wrapped to your Dwyer, that tubing is thick and tie wrapping was a good idea for a complete seal. :up: I recently purchased a used Panduit tie wrap tool (GS2B) and it sure does a great job.

[DavidNJ]

I'll see what I can capture with this first. I figure a vast majority of users are using radiators without a shroud, probably 90% I would guess. In the end I want to capture performance that is closest to the conditions most people use them. The only thing artificial will be my heat load, everything else should be normal radiator operations and conditions. The fan hub definitely affects performance, but if that's the way most people run them, then I would like to capture it like that.

Also with the open top, I purposefully allowed extra room with my air out sensor lengths so I can experiment with the effects of a shroud. It's pretty flexible like this, and I can try a lot of different things, the only thing absolutely fixed is my water and air inlet sensors.

I also want to be able to display the results in a way that more people understand it. If I plotted CFM vs performance, I bet most folks would think that just means how many fans x their rated CFM which is not correct. But if I post a curve with a specific fan and fan voltage, that something people can relate to pretty well.

I will build a shroud I can slip over the fans and try to measure spot CFM when I get my anemometer. With a certain fan, I should be able to slip this shroud on and measure the actual CFM. If I do this with each type of fan, I should have captured enough to also do a CFM Vs Heat dissipated type of look. It would also indicate how restrictive the fins are (Not quite pressure drop, but an order of restriction). I might also try to measure pressure drop of the radiator fins in time, but that's going to be a bit tricky considering every manufacturer is a bit different and trying to get a one size fits all perfectly sealed shroud is going to be a challenge.

Just getting started, I'm sure I'll be tweaking something before I'm set in how I'm going to do the testing, and that will probably change over time as I learn more about it.:D

Yes, most don't use shrouds. And it seems most have their radiators in push, which if anything is worse than not having a shroud. Especially if it is push without a shroud! :D However, designing to those standards reminds me of a story I tell why NYC subway seats or so narrow: because NYC has the only trains designed to crash into each other (true story). Codifying a ridiculous design standard. My guess is your tests would show that someone running 25mm fans in push on a GTX would get better performance with 38mm fans with a shroud in pull on a BIX.

However, I was hoping you would develop PQ curves for radiators and fans, and make a spread sheet for that. Add in your thermal testing, and you would have a complete analytical picture. Plug in a heat load, and the PQ curve would determine your temps. Of course, your spreadsheet may have to change to using a macro call to goal seek or solver.

P.S.
It looks like your frame is blocking 5 sq in of the radiator. Are the cross braces necessary? Additionally, it may be too close to the surface, blocking flow in stronger setups or less restrictive radiators. No by much, however, why not vertical?

[twwen2]

This is gonna be some of the most comprehensive and complete testing we've seen in a while. Reminds me of Cathar and Vapor's fan testing, all excellent stuff. Man i miss Cathar...:shrug:

Looking forward to the results Martin:up:

[Yeknom]

Thank you so much for putting this together; it really helps everyone on XS.

Looking forward to the results!

[Martinm210]

Thanks guys,
I'm still going to try and see if I can also develop P/Q curves for fans and the radiators, but the fan spec format will get me started. I'm still waiting for an anemometer I ordered, that I'll try to use just for that.

Anyhow, I got the water sensors made up on the lathe, hook it all up and none of my sensors came on line:eek: , so I had to do a little troubleshooting and after cutting my string of sensors 3 times I finally found the little sucker. I wired one of the sensors wrong, now I get to do a whole bunch of splicing back together again... A little panic set in though, at about $5 each, loosing a whole string of 14 from a mistake can be a little scary...all is well though.:up:

If all goes well and I'm not stuck out of town all week on business, I should get a chance to run a few preliminary tests by next weekend.:) First up is the PA120.3 to do some check ins and see how well I can or can't calculate c/w curves from this..

[Digz]

You guys are to complex all we need to know is temps at 12v , 7v and 5v :D :D :D

[DavidNJ]

You guys are to complex all we need to know is temps at 12v , 7v and 5v :D :D :D

With the fans correctly mounted or incorrectly mounted. In this test by Ira-k (http://www.overclock.net/water-cooling/204770-part-1-temp-s-out-shrouds.html) the difference between unshrouded and push and shroud pull was 3.25C in CPU temps and cooled Water to ambient Air temps went from 3.3C to 4.25C, an increase of 28%.. From measured input air went from 8C to 5C, 60% higher!

Comparisons......

Push-Pull with shroud average below ambient.....23.375C...3.625C cooler
Push-Pull with-out shroud average below ambient........19.75C
--------------------------------------------------------------------------------------------------------------

Pull only with shroud average below ambient.........21.125C......3C cooler
Pull only with-out shroud average below ambient..........18.125C
---------------------------------------------------------------------------------------------------------------

Push only with shroud average below ambient......20.25C.....2.375C cooler
Push only with-out shroud average below ambient..........17.875C
----------------------------------------------------------------------------------------------------------------

Water temp drop across rad with push-pull and shrouds.........1.025C
Water temp drop across rad with push-pull and no shrouds.....0.9C
----------------------------------------------------------------------------------------------------------------

Water temp drop across rad with pull and shroud..............1.275C
Water temp drop across rad with pull and no shrouds........0.95C
----------------------------------------------------------------------------------------------------------------

Water temp drop across rad with push and shroud...........0.975C
Water temp drop across rad with push and no shroud......0.875C
-----------------------------------------------------------------------------------------------------------------

Air temp drop across rad with push-pull and shrouds.........4.5C
Air temp drop across rad with push-pull and no shrouds....2.875C
-------------------------------------------------------------------------------------------------------------------

Air temp drop across rad with pull and shroud...............5.5C
Air temp drop across rad with pull and no shroud ........5.25C
-------------------------------------------------------------------------------------------------------------------

Air temp drop across rad with push and shroud..........6C
Air temp drop across rad with push and no shroud.....7.5C
------------------------------------------------------------------------------------------------------------------.

[voigts]

How you find the time to do all this is beyond me. Your work really is a great asset to the watercooling community as a whole. Just your flow estimator alone is so terribly helpful.

[septim]

another nice good long read from MartinM when this is over...

[Martinm210]

Yeah Ira has done some great work with the shroud effects. I'm thinking I might gut some old 120mm fans and run an experiement or two with shrouds as well. That's something anyone can make easy enough as a shround, take some old fans you don't use anymore and snip off the motor/blade, you've got a shroud..:D

[Xilikon]

That's a nice test bench :) As for the idea of holding the temp sensors wires in place, just use some steel wire ;)

[DavidNJ]

Yeah Ira has done some great work with the shroud effects. I'm thinking I might gut some old 120mm fans and run an experiement or two with shrouds as well. That's something anyone can make easy enough as a shround, take some old fans you don't use anymore and snip off the motor/blade, you've got a shroud..:D


Just get some plastic spacers, or cut some wood or aluminum tubing...even pvc tubing and slip it over the screws between the rad and the fan. Then wrap it with duct tape.

The fan shell as shroud uncovers the hub but not the frame. Over 20% of the area lost is shrouding from the frame.

[gojirasan]

Just want to add a +1 to appreciation for your testing. I am really looking forward to this radiator test and want to thank you for the time, effort, and money you are putting into this.

The imminent release of the Cryo-Z (along with its noise/weight/wattage and CPU-only limitations) only demonstrates further to me that this "hobby" will not be dying out any time soon and your testing will continue to further our knowledge.

@DavidNJ: This shrouding thing seems interesting. The premade ones are expensive and hard to find. Seems tempting for a nice DIY project. Although I might wait for the results of Martin's tests first. Do you have any pics of a homemade shroud you could post to give a better idea of what it should look like?

[XS_RoB]

Too funny - Everything I have purchased for WaterCooling - after my live and learn EXOS system, has been upon Martin's Test Results or Reviews of the actual products!
If a company was smart they would hire him / you Exclusively for this type of thing!

@Martin - you should link your site in your sig, if allowed by staff. Also throw a PayPal (Donate button on your site.) I know I can't afford too much right now (Mrs. Laid Off last week) But I can squeeze in a 10 or 20 here and there for your efforts.

I actually feel bad for using so much of your info... teasing. :)

Have a great one and thanks yet AGAIN - for all you do and have in process!

[NaeKuh]

NOOOOOOOOOOOOOOOOO Martin if you dont test the PA as being the king. Im gonna hate you in making me buy more h2o gear!

>:[

[Martinm210]

Thanks guys, that's what keeps me going....:D

[Martinm210]

Ran a few preliminary tests today.

Here is my string of sensor ready for logging, 4 of them are in my computer that I'll simply disregard, but 14 of them are specifically for my radiator test bench:
http://img212.imageshack.us/img212/5452/radiatorcrystalfontzqv8.png

My very first test with a target heat load of 100watts, 15watts from the pump, 2 watts lost in my variac and 88 watts from the 300 watt water heater turned down until my "Kill-a-watt" registerd 88.
http://img212.imageshack.us/img212/3349/radiatorpa1203prelim1kl0.png
Takes a while longer to stabilize than I thought, I probably need to run this one again, but it's giving me an indication that at a bare minimum 20minutes of warm up is needed, but 30minutes is even better.

Then I tried a 200 watt run and found my first little glitch. The thermostat on the water heater kicked on mid way through the test...whoops..:)
http://img212.imageshack.us/img212/1996/radiatorpa1203prelim2ks2.png

Working out the bugs, but the logging is working nicely. My C/W results will be much lower because of heat lost through the tubing and reservoir, but I figured that's ok since a regular system will also. You just won't be able to compare c/w except other tests on the same test bed.

[NaeKuh]

WOW! seriously great work.

BTW sorry martin, i flashed your tec blocks in lc forum. Your pm box is gonna get nuked soon. :wasntme:

[Martinm210]

Hehe.

I was just going to start my 4th run a bit ago by throwing the second 300 watt water heater in there and the darn thing just doesn't thermostat high enough.

So now, I've got the water heater in my lap pulled apart, going to solder a bypass wire to get around the thermostat, must have MORE heat!!...:D

I think I might bypass all of my heater thermostats, I just want solid power anyhow, just don't want to forget one on, it would probably boil water or start a fire..:eek:

[Martinm210]

Finally got a keeper, this is how they should all look:
http://img292.imageshack.us/img292/4069/radiatorpa1203prelim3je5.png

[systemviper]

Wow, Some great stuff, Now got to kick the brain into OC and try to undertstand the results :rofl:


As always, thanks for your effort, top notch work! :clap:

[Martinm210]

Thanks!

It's not going to mean much until I capture the trends. Right now it's looking like a 10C water over air temperature with the medium speed yates at 1.5GPM is going to fall in the 460-470 range.

Each of the above tests was more for just feeling out how long I really need to let the test run before logging temps and also how long I should log temperatures.

Also adjusting the heat load and plotting out these points will show what sort of curve or line develops. This will tell me how many different heat load points I should test for. So far it's looking pretty linear so testing at 100 watts and 200watts may not really be necessary.

In the end I'm planning to create a tool out of the information where you can estimate your water temperature with some simple pull downs.

[Vapor]

No kidding mcoffey....this is pretty kickass stuff you've got going on here Martin :toast:

[rockqc]

So much talent and skill. I'm just amazed at the stuff your able to do and info you provide. I've learned so much.

I think most of the members who post in this section (including myself) would agree.

Great work as always Martin. Looking forward to more results. :up:

[Jedda]

Hope you're enjoying doing it, Martin!
We appreciate your effort.

[Martinm210]

Thanks guys, this is alot of fun.

Always hated just doing pressure drop on wateblocks and not providing thermal stuff in addition, so I'm really happy that now I can do both on the radiators.

Looks like the key is letting the whole system stabilize for around half an hour before making a run.

Here is a 600 watt run, looking even better:
http://img525.imageshack.us/img525/5512/radiatorpa1203prelim4tv0.png

And even though the 100 and 200 watt runs were iffy at best, I plotted a trendline:

http://img525.imageshack.us/img525/6759/radiatorpa1203prelim5ag1.png

I can tell by this that I didn't wait long enough for stabilization for my 100 watt run. It looks to be low by about .2 to .3C.

Anyhow, I kind of like looking at this sort of curve better than normal c/w and heat dissipated relative to flow rate. I'll probably make a series of tests relative to flow rate just to see what the effects are, but since flow rate has less of an impact, I thought it would be nice to see water delat relative to heat dissipated.

The medium speed yates at 1.5GPM and 12V are dissipating really close to 500 watts for a 10C delta.

[Crankyhobo]

Seriosly Badass please keep us all updated!

[DavidNJ]

VERY nice graph. Kudos!!!

Now for a shroud and pressure measurement. Without the shroud the whole radiator isn't being used.

The pressure measurement would let us know where it stacks up vs. other fans.

One other measurement would be nice. Ambient air temp other than air in. From Ira's tests it seemed that Air In may be lower then Ambient.

Great work!

Updated:
Looking at the graph again, it is interesting how linear the temperature is for increasing heat load.

Do you have any Panaflo 120x38 M or Ls? The 120x25 SM Yates Loons fall between them in noise (if the published Yates Loon numbers are accurate), yet I would image either would draw more air through the Thermochill rad.

Second Update:

If the temperature differential is that linear with heat load, the interesting graph may be pressure differential (or air flow, on a pq curve either should define a point) on a shrouded radiator with the fan in pull (to take fan dimensions and vortex dynamics out of the measurement) vs. temperature differential. If the temperature differential changes linearly with heat load (less than a 10% variation over a 3x range in your tests), then the air flow to get a specific temperature differential at a heat load could be calculated.

Using a setup like your spreadsheet, and using goal seek as the calculating mechanism, the pressure differential/air flow for a temperature differential could be determined. If pq curves for fans also were in the database, a list of fan/voltage pairs for that pq could be listed, possibly with the sound level for that fan/voltage pair.

[Martinm210]

Yeah, it's going to be even more linear after I get my variac consumption figured out.

I might just pull the variac out of the picture all together, and pick up a 150watt water heater. Then just plan on doing 150 + the pump whatever it works out to be increments. I'm thinking a single point near 150, 300, 600 watts should be enough for triple rads and smaller rads should be fine with 150,300, 450.

I was assuming that the variac only consumed 2 watts, but that's at low voltages, it's closer to 5 or 6 up higher. My second round of 100 watts was also higher coming it at 1.90C.

http://img180.imageshack.us/img180/3756/radiatorpa1203prelim6nw8.png

http://img180.imageshack.us/img180/6480/radiatorpa1203prelim7nz3.png

I'll do some CFM type stuff too, just wanted to get this heat load thing started first.


The fan stuff is coming too. I've only got the medium speed panaflos though. But I've also got some slow speed yates and some of the 38mm 136cfm Scythe Ultra Kazes too that have really good pressure too.

It's all possible, just going to take some time and work. It could easily be a weekend's worth of work just to test one radiator with 4 sets of fan scenarios. As long as a measure actual fan CFM at the same time for now, I can work some other stuff in later.

Take a look at Vapor's fan charts for some good noise vs CFM data (http://www.xtremesystems.org/forums/showpost.php?p=2668922&postcount=5).

[DavidNJ]

I think you are making the radiator test way too hard by making it a fan test also. You just need a box you hook to the radiator with enough of a blower that can be create a pressure drop from .01" to maybe .5" across the radiator. I don't know what we have in practice now, however based on the fans I imagine it is in the .05-.1" range.

That would be quick, and neat. Then you can just select the fan with a shroud in pull that will duplicate the pressure drop you want.

BTW, what was your room ambient? Did it differ from air in?

[F1ZZY]

Hi

Long time lurker and 1st post for me here.

Loving your work Martin. It's this kind of effort that finally puts end to (some of) the bickering we see on forums - (like Cathar and his tubing results etc)

There are so many variables available for radiator heatload testing and you are doing a fine job trying to make sense of it. I am particularly loving the dT vs Watts dissipated graphs, I think they are revolutionary. Thermochill have always plotted a fixed 10degC dT.

You do realise that your efforts are going to result in the first full w/c simulator (now that you are adding the thermal assessments too) - Soon your spreadsheet will have to include components and fans used & ambient temps in order to provide speculative temps, ha ha.

Keep busy :)

[Martinm210]

I think you are making the radiator test way too hard by making it a fan test also. You just need a box you hook to the radiator with enough of a blower that can be create a pressure drop from .01" to maybe .5" across the radiator. I don't know what we have in practice now, however based on the fans I imagine it is in the .05-.1" range.

That would be quick, and neat. Then you can just select the fan with a shroud in pull that will duplicate the pressure drop you want.

BTW, what was your room ambient? Did it differ from air in?

Yes, but that wouldn't capture real world fan hub effects for folks that don't use a shroud, which I would esimtate is the larger majority of users.

I'm more interested in capturing something that is useful and easier to understand for people.

A CFM base flow chamber would have to be a large facility too, I just don't have the space where this little bench I'm using is very simple and small.

It'll work out..:D


I don't understand what the ambient vs air in question really is. air temperature is extremely sensetive and if I measure room ambient 10' away it really has very little correlation, measure it on the floor and it's probably cooler than air in, measure 6' up it's higher. Air temps are really really sensetive and different everywhere, that's why I've included 8 individual air in sensors to average out the mess of differences.

[nikhsub1]

Very nice Martin! This is exactly the sort of testing i begged Bill Adams to do, but he went with the 10C water/air differential instead which is much harder to decipher. You realize the can of worms you've opened though now right? Each test should include various air and water flow rates...

[bobo5195]

Care to plot your results as Time vs the LMTD

http://en.wikipedia.org/wiki/LMTD (the countercurrent one).
Cold stream = Air
Hot stream = Water.

The excel formula for it would be
with Wi being water in, Wo being water out Ai being air in and Ao being air out.

=((Wi-Ao)-(Wo-Ao))/Ln((Wo-Ao)/(Wi-Ai))

should be pretty much a straight line of results or a nice *slighty* smooth curve.

Edit
Oh and a scatter plot of (Wi-Wo)/(Ai - Ao). For a period of constant air and water flow rates.
The amount that this is a scattering of points vs a straight line (use a trend line intersecting at 0) is the value of "crapiness" for want of a better word of the experiment.

This should be varying with heat load.

Edit 2
Also that value call it K, or by its real name ratio of thermal heat capacities.

K=(Wi-Wo)/(Ai - Ao)

Would tell you fan CFM exactly (depending how good your apparatus is) if you varied Fan speed. Which would also tell you radiator flow resistance.

[DavidNJ]

Yes, but that wouldn't capture real world fan hub effects for folks that don't use a shroud, which I would esimtate is the larger majority of users.

I'm more interested in capturing something that is useful and easier to understand for people.

A CFM base flow chamber would have to be a large facility too, I just don't have the space where this little bench I'm using is very simple and small.

It'll work out..:D


I don't understand what the ambient vs air in question really is. air temperature is extremely sensetive and if I measure room ambient 10' away it really has very little correlation, measure it on the floor and it's probably cooler than air in, measure 6' up it's higher. Air temps are really really sensetive and different everywhere, that's why I've included 8 individual air in sensors to average out the mess of differences.

People who don't spend $10 and 30 minutes to build a shroud aren't really trying to leverage the numbers for the best performance: thermal or acoustic. Making tests that hamper radiator performance and don't provide numbers for people looking to tune thermal performance are equally lacking.

Isn't the goal for people to start designing their systems better rather than institutionalize bad design?

Ira's tests indicated that the higher speed air as it is sucked into the radiator has a lower temperature than ambient. That difference is a function of the effectiveness of the system in creating a pressure differential across the radiator design of the radiator. Failing to take it into account can make a more effective setup look less effective.

Taking the air temperature say 12-24" from the radiator should reflect the ambient temps around the radiator without any cooling affects from the fast moving air.

if you use a shroud (which Thermochill even sells for their units, which come with mounting holes) it may be very interesting to compare a PA120.3 to a BIX 360 with fans in the 20-30dbA range.

[gojirasan]

I think you are making the radiator test way too hard by making it a fan test also.
No. He is doing it right. Remember that test where the Thermochill was soundly beaten by the HWLabs GTX but only above 1200 RPM with a certain Yates fan? This is the first test in the English language where we might see such results duplicated, finally dethroning the PA120.3. It seems clear that nothing can beat the thermochill at very low CFMs/pressures. But the real question is how low exactly. If this can happen at a low enough sound level, then it will really be something very significant for most of us. This isn't SPCR. Not everyone here cares about utter silence. But most of us don't want a jet engine roar either.

It looks like you don't have the top rated fans though, Martin. Isn't there anyone who can contribute a few San Aces or some Scythe S-Flex Fs or Zalman ZM-F3s or Noctua NF-P12s? I do have a few of those San Aces you could borrow, but I only have one each of the Zalman and S-Flex. Actually I was thinking about scoring some of those medium speed Ultra Kazes. The problem with the Noctuas is I think they are too slow. They need to be overvolted or something. They would be great for this radiator showdown though.

[nikhsub1]

While I see your point of testing with shrouds, not all rad mfgrs have them... this would mean that Martin would have to MAKE a shroud for EVERY rad he tests... single fan, dual fan, tripple fan, quad fan ETC. Do you see the nightmare in this proposition? Plus, shrouds affect performance very differently on different rads, it is NOT LINEAR at all. The more densely packed the FPI, the more a shroud will help. A shroud on a PA rad is almost pointless.

[bobo5195]

Ira's tests indicated that the higher speed air as it is sucked into the radiator has a lower temperature than ambient.

I'm not sure that is possible from a thermodynamics POV given the conditions of the test.

Higher air speed would remove heat from the sensor faster if it was being warmed up though.

I have also never heard or seen it mentioned in textbooks or such. Good old Prof Kaynes would have brought it up.

[Martinm210]

I'm not sure that is possible from a thermodynamics POV given the conditions of the test.

Higher air speed would remove heat from the sensor faster if it was being warmed up though.

I have also never heard or seen it mentioned in textbooks or such. Good old Prof Kaynes would have brought it up.

Thanks and thanks for the equations, you're always making me dig my old reference manuals out and keep an eye on the equations.:)

First up I'm trying a few repeatability tests just to see that, then I'll play around with some other data concepts. The data is all there, I've been capturing with each test about 2,000 rows with 14 column of temperature data (logged every second) along with a timestamp and RPM), so there are plenty of numbers to play with.

Here is a sample of 40 or so rows of the 2000 or so I'm collecting in just one run:
http://img519.imageshack.us/img519/6862/radiatorpa1203prelim8qu2.png



Yeah the colder inlet vs ambient doesn't make sense to me either either, it was probably just error in the thermometers. If anything I've usually measured warmer temperatures at the inlet that's caused by recirculation of the exhaust. This is a big issue with alot of rear and top case mounted setups, my own system I was able to cut about 3C off by simply channeling air to exit and keep away from the inlet. In the end I don't find value in ambient, if the setup is done right ambient will equal Air In. Air in is what matters for heat exchange.

Even though I'm pretty impressed with these Dallas sensors, I still don't trust them for more than a few tenths of a degree and it's not uncommon for two sensors only a few inches apart to read as much as 1 degree difference. It's really amazing how much air temperature changes in such a small amount of space. After all air is considered and insulator, so it's not like measuring water and it's very high specific heat capabilities. Lots of sensors and logging for very long periods of time are the key to accurate air measurements is what I'm finding.

Also on the shroud vs. no shroud, I plan to do some controlled experiments to see if I can extract some good correction factors for a shroud vs. no shroud. I'm thinking on correction for 40mm hubs (25mm fans) and one for 55mm hubs (38mm fans) should be adequate, then you could adjust results for either shroud or no shroud.

If folks have the opportunity to see what those shroud performance results are, then they can decide for themselves if the effort is worth the hassle.

To keep the shroud design concept simple and easy for just about anyone, I'm thinking I'll try evaluating the effects of a shroud made from old 25mm fans with the motors snipped out of them. I could also play around with pull/pull, push/pull, etc. All of those difference scenarios where a flow chamber would only evaluate a shroud like condition, so it's not as flexible.

[DavidNJ]

I think the equation goes PV=kT or something like that. P goes down, V is constant.

You can trivially make a shroud with a spacer ($10 for a bag of 50 from McMasterCarr) and some duct tape. No harder than attaching the fan with a longer screw.

Just because some people who fasten the fan directly against the radiator in push is no reason to test that way. Those people are already giving up 50% of their cooling capacity...what difference does the radiator curve mean to them? They could improve their systems significantly by spending 30 minutes with $10 of hardware.

Update:
A better comparison would be: is it better to add a two stage fan-set to a 2-fan radiator in pull with a shroud than to go to a 3-fan radiator?

[Martinm210]

And here is a picture of an actual plumbed and running test in progress, sorry for the mess...
http://img519.imageshack.us/img519/5540/radiatorbench8ip1.jpg

[bobo5195]

I think the equation goes PV=kT or something like that. P goes down, V is constant.



yep its nR but thats about right the equation for an ideal gas law. There are some mods to account for compressibility given the complexness of air you better off using tables and charts as ideal gas is not a useful approximation.

too much to say on Martins post as im off down the pub to talk about Ammonia fuel cells. hint NH3 holds alot more hydrogen than methanol CH3OH

[DavidNJ]

The frame maybe two close to the inlet for the radiator, affecting (to a small degree) the numbers.

Question: why didn't you mount the radiator vertically? It would has worked with two vertical posts and some clamps.

[Martinm210]

The frame maybe two close to the inlet for the radiator, affecting (to a small degree) the numbers.

Question: why didn't you mount the radiator vertically? It would has worked with two vertical posts and some clamps.

Beacuse there is alot more differential between sensors that way, I get over 1 degree with my computer and the sensors are only aboug 6" apart, you'd probably have 3 degrees or more over a triple or quad rad mounted vertically.

The testing bench frame has 110mm opening, so there is no restriction. The radiator just barely sits on it's edges there, even on the PA which is wider than most. The only one I'll have to watch is the heatercore for that.

[DavidNJ]

Curious, why did the Air In numbers vary so much?

Input Average Std Dev
AirIn1 25.76 0.03
Airln6 25.98 0.09
Airln4 26.33 0.04
Airln3 27.01 0.07
AirIn8 27.31 0.04
Airln5 27.40 0.04
Airln2 29.58 0.03
AirIn7 34.35 0.04

[Martinm210]

Curious, why did the Air In numbers vary so much?

Input Average Std Dev
AirIn1 25.76 0.03
Airln6 25.98 0.09
Airln4 26.33 0.04
Airln3 27.01 0.07
AirIn8 27.31 0.04
Airln5 27.40 0.04
Airln2 29.58 0.03
AirIn7 34.35 0.04


I've got a pump that only about a foot away that's probably influencing some of the temp sensors, and sensor 7 isn't being used right now, it's sitting there idle along wiht #8 only for quad rads.

This is why I was saying you need alot of sensors. The sensors themselves have some error and the air temperature is different all over the place, that's why I put all the effort to include so many sensors and log temperatures for so long. If the temperature was the same everywhere I wouldn't have bothered with all the effort.

Take a look at coolingmasters radiator test, they also plastered the run with sensors, one is not enough, it varies alot more than you would expect. And they don't just sit there idle, they bounce around.

[Martinm210]

Ok, I've made the change in removing my variac all together and just reading the actual watts the Kill-a-watt registers plus 15 watts for my pump. This seems to have straightend the curve right out for a nice linear relationship. So I obviously had some error in the power consumption of the variac which is eliminated now.

With this I think I'll just plan on testing at around 300 and again at 600 watts and plot the linear trendline.:up:

http://img528.imageshack.us/img528/2797/radiatorpa1203prelim9xm3.png

Repeatability is really good too, I'd say within about .1C at 600 watts, so .1 of 12.12 is less than 1% error, so I'm happy with that. This should save a little time too.

Next up is a trial on measuring flow rate effects. I think I'll just do that at 600 watts at .5, 1.0, 2.0, Max Flow.

[DavidNJ]

I doubt fluid flow will have more than a 10% influence, if that. Much more interesting would be adding a shroud. Based on Ira's tests, that may drop your tests that may drop your 300-330 number to under 5C.

And maybe those Panaflos which probably have 50% more pressure at zero flow. Based on existing Thermochill stuff, the M, and virtually certainly the H, would drop you peak number under 10C.

When testing denser finned radiators like the GTX, I imagine they will be more sensitive to more pressure being developed (or less pressure in a pull :) )

[DavidNJ]

I've got a pump that only about a foot away that's probably influencing some of the temp sensors, and sensor 7 isn't being used right now, it's sitting there idle along wiht #8 only for quad rads.

This is why I was saying you need alot of sensors. The sensors themselves have some error and the air temperature is different all over the place, that's why I put all the effort to include so many sensors and log temperatures for so long. If the temperature was the same everywhere I wouldn't have bothered with all the effort.

Take a look at coolingmasters radiator test, they also plastered the run with sensors, one is not enough, it varies alot more than you would expect. And they don't just sit there idle, they bounce around.

Could some of the temperature anomolies also be the result of fan effects exaggerated by not using a shroud and my affects of the base (its supporting corners) and proximity to the table? Would standing the radiator upright (and using a shroud) eliminate the descrepencies?

[C'DaleRider]

First, let me thank you again, Martin, for all your testing and work you've done and are doing. It's great to have all this empirical data available to us "grunts" down here in the WC'ing trenches.

Second, David, let the shroud thing rest a bit. All I read in your posts is shroud, shroud, shroud. Honestly, you've become something of a broken record about that. Give it a rest for a bit.

I say that because of a few things. The vast majority using WC'ing don't use a shroud, for whatever reason. In my situation, a shroud is impratical, no matter how great one may be in actual use.....which I am ambivalent about, anyway. Space is my concern.....I've got my TC 120.3 top mounted in my case and with the rad and fans mounted, I've got less than a 1/2" of space between the rad's bottom and the motherboard's MOSFET heatsinks.....sinks that would definitely interfere with any attempt to mount a shroud.

I'd imagine that I'm not the only one with that sort of problem, too. While a shroud may be useful, it's not practical in my case and probably not in a lot of other situations, either. So, why test for something the majority choose not to use or cannot use, in any event.

Instead, more valuable information would be fan choice, radiator choice, and flow rates through the individual radiators. Those are factors that all WC'ing enthusiasts can use.....we all use rads, fans, and pumps with flow rates. So that info would be applicable to all of us.....


And just one more point.......my particular setup consists of a Q6600-G0 stepping, a Thermochill PA 120.3 radiator w/3 Scythe S-Flex "F" fans running at 7-8V, a Laing DDC2 pump (the original DDC2 18W pump, not the newer 3.2 version) with an XSPC top, a D-Tek cpu block with washer and quad nozzle, an EK S-Max NB block and all 3/8" tubing.

I've got an OC on the Q6600 of 3.6GHz at 1.35V.....and with the above setup, running OCCT hour after hour gives me max load temps of 56C on my "hotter" of pairs of cores in my cpu. (Ambient, or air in temps are at 24-25C.) Now, you're telling me I'm losing upwards of 50% of my cooling capacity because I'm running in push w/o a shroud?

I just really find that hard to believe. But then again, I'm just a 53 yo gent who is into WC'ing as a hobby.....and it's not my life. My temps are quite acceptable to me.....and wll below any I'd expect to see from decent air cooling......