Radiator CW vs Air/Coolant Delta-T
 

Okay, so I'm updating my approximator with all the excellent work BillA did at Swiftnets: A manufacturer who publishes PQ and CW charts.
Woot!

Anyhooowwww, I am looking at the MCR220QP page, and there's a so-BillA-looking chart for C/W vs flow rate here:

http://www.swiftnets.com/assets/imag...cw-600x440.gif

(I'd link to the page, but witht the menu system, I can't. Oh heck, at least not properly, but here's the page using properties: http://www.swiftnets.com/products/MCR220-QP.asp

Anyway, the test says there is a kept-constant 10 degree difference between the water temp and the ambient air temp.

Now, naturally in my simulator as per real life, the more efficient a radiator is the closer to ambient the water temp runs. Its often, for the rigs I see and comment on, less than 10 degrees.

So, how linear is it? How do I approximate it? Do I bother?

Typically I don't see in my approximator the water temp getting under 2 degrees above ambient. However, I imagine at that temperature, the C/W is not going to be as good as when the delta-T is 10 degrees.

I'd love a chart showing, for just one or two setups (perhaps just vary the airflow once) showing the impact of air to coolant delta-T in the range of 2 to 20 degrees.

Heck, given all the hacks in my approximator, I'd settle for someone having an approximation of the shape of the graph.
Is it linear?
1/X?
X^2

My brain says that at a zero delta-T you'd get zero heat flow(regardless of anything else) so that's one point on the graph. The chart referenced above shows another single point on my proposed chart. The question is, does the relationship between the delta-T and the C/W continue linearly, or does it improve (so increasing the delta-T to 20 degrees does more than double the performance of the rad) or does it taper off (law of diminishing returns and all that).

My money's on the latter.

I just don't want to have my approximator mega-inaccurate as it assumes the same effectiveness at delta-T = 2 as at delta-T = 10 and delta-T = 1000 (for that matter).

I think it's linear with delta T. However there would be an element of diminishing returns, due to secondary losses, the higher the dT. Also, of little account perhaps, that the lowering of the viscosity with higher temperatures will have an impact.

Edit: Silly me, misread. Heat transfer is linear. C/W is constant, but subject to effects of viscosity .

We are in LMTD territory.
The overall heat transfer coefficient can be considered constant.
Can of worms.

Edit. Some Heat Exchanger calculators Link.
Unfortunately "Crossflow Compact Heat Exchangers" are not directly covered

Thanks Les - I'll just leave it constant and not worry about it ever again. The facts and figures "sourced" from the web are varied enough in their production (and influence of marketing depts) to always have my numbers as approximations only. Just thought the C/W might be linear or heavily influenced by the delta-T leading to massive errors.

BTW: I was too scared to follow the link: your engineering skills and knowledge are on another plane of existance to mine. I've downloaded the "heat transfer physics textbook" pdf thing, and that scared me off doing anything remotely scientific around heat transfer. Can of Worms? No, entire oceans of worms each plucked from Dune.

The C/W measurements are another matter.
They are the worms
None of the radiator testing measures the "Overall Heat Transfer Coefficient".

Hmmm.
What I am using the CW for is pretty straight forward.
I start with the water temp at zero above ambient, and then I creep it up until the radiator can dump the same amount of watts that the blocks and pump are adding in. The C/W is sourced from BillA's thermal-management-testing article for the HE rads, and I will soon add the Swiftech range.
I'd just calculate it straight out, but my aim is to be able to cater for circular-relationships where the heat of the coolant impacts the heat able to be dumped out. Probably wasted but writing loops in java is easy.

Seems to roughly correlate with reality.

What is better for a quick approximation done by an idiot? Not interested in equations that involve star signs etc. Simple maths for quick approximations.

I know the worms I am in with CW. I am contacting a certain manufacturer now as their block's CW values are an order of magnitude better than the Cascade. Err, don't think so Tim.

For simulation of cpu and radiator combo would probably use Phaestus(cpu)+Billa(rad) with a pinch of Roscal
Its worms all the way.

Edit Corrected omission of Billa