How do you calculate how a rad performs based on the data?

[bobo5195]

How do you calculate how a rad performs based on the data?

I.e I have one of bills nice graphs how do I use it to tell me how a rad performs for a set of conditions (heat dump, air temp etc)

I am prepping this into a spreadsheet (can all ready calculate root performance to beyond excels accuracy) but want to know how the WC community does it. I will probably implement the science way of doing things (efficiency - NTU, which have the advantage that you get do a lot of calc of relatively crappy data) but if there is another way I’ll do that.

[jaydee]

I think you need more than one graph of data. One graph for Pressure drop and flow rate and then one for heat dissipation and flow rate.

I usually get an idea what my flow rate is before adding the radiator. Then using their Pressure drop information for the rad calculate what my flow rate will be after adding the rad to the loop.

Once I got the flow rate then use the Flow rate and heat dissipation graph to get an idea how much heat the rad will dissipate at the flow rate I am running. If the rad fits the wattage I will be dumping then good to go. If not start over using a better radiator.

Also a 3rd graph with Thermal resistance and flow rate might be good to. Then you can get an idea what CFM fan is needed.

Might be a little off on that but pretty sure that is what I used to do.

[billbartuska]

Originally Posted by bobo5195

How do you calculate how a rad performs based on the data?

I.e I have one of bills nice graphs how do I use it to tell me how a rad performs for a set of conditions (heat dump, air temp etc)

I am prepping this into a spreadsheet (can all ready calculate root performance to beyond excels accuracy) but want to know how the WC community does it. I will probably implement the science way of doing things (efficiency - NTU, which have the advantage that you get do a lot of calc of relatively crappy data) but if there is another way I’ll do that.

Calculates flow rates, not rad performance, but may give you some ideas.
http://www.xtremesystems.org/forums/...d.php?t=151627

[bobo5195]

Originally Posted by billbartuska

Calculates flow rates, not rad performance, but may give you some ideas.
http://www.xtremesystems.org/forums/...d.php?t=151627

I know of martins spreadsheet and in fact rewrote it myself to include a split loop calculations. I then rewrote that one to the version I have now to prevent errors. The reason of the question is I am currently adding the thermal data at the moment and wanted to know “the enthutiast” as billA would have said way of doing things rather than the text book.

With regards fan vs flow rate the industry standard thing is to say air thermal capacity should be a little higher than water thermal capacity. i.e.

Flow rate water * 4200 J/kg/C = air flow rate * 1000 J/kg/C

This will tell you if you need to add more fans (ie a bigger rad) or if your rad is too small. It also suggests that the performance of a two fan rad with screaming deltas should be equal to a quiet rad with 3 fans. This is useful as you can easily test if your rad needs more thermal capacity by upping the flow rate with cheap high cfm fans.

The other thing with rads is its abit hard to estimate what exactly detla T is. Water temperature varies across the radiator surface and air temperature varies through the device. A pass rad has two delta Ts of whatever you specify as air goes through the colder side then through the warmer side. It then matters which way the air is flowing and you have to compute the effect of heat dump on the second pass.

Standard procedure is to use the LMTD (for counter current flow) as the temperature difference with a correction factor (either constant like 0.7 or based on thermal capacities0 cos the result you get out of that is wrong.

See wikipedia for the LMTD as writing it out is hard
http://en.wikipedia.org/wiki/Log_mea...ure_difference

There is another method called eff-NTU which is based on the how close the performance of the rad is too an ideal one of same size then using empirical correction factors.

The two above approaches have the advantage that rad performance is just one C/W number. Similarly it doesn’t matter what fan you use only the air flow rate. So you can type away CFM numbers to your hearts content or use fan curves to estimate how things perform on a rad. Both these approaches have the disadvantage that they are some what uninteligiable to an average reader involving a lot of voodoo magic and they involve long, boring, hard to check formulas that I wanted to avoid.

[Marci]

Basically same way as Jaydee...

The only way I've ever managed to translate the data into useable is giving folks a "with this fan, this heatload, this flowrate, you can get your coolant to within XX degrees of ambient" based purely on the deltas Bill used in the Thermal Dissipation Charts. Coolant temp is all I'll aim people for... can't predict final CPU Temp / GPU Temp etc due to the differences in everyone's Thermal Resistance testing etc...

The formula in the spreadsheet-source of Bill's PA Series graphs (at www.thermochill.com/PATesting) allows me to extrapolate different air > coolant deltas... standard graphs are at dT of 10... graphs HERE are extrapolated for a dT of 15... could recalc `em for an extrapolated dT of 5 too... but then the airflow required to cool the given heatload tends to be ridiculous when it comes to the noiselevels that airflow would present, hence not bothered with dT5 yet...

S'all very loose and inaccurate but it's as good as I can explain... and folks seem to understand the concept...

I summarised it into a 5-step procedure using combination of ExtremePSUCalc to (over-)estimate heatload, Martin's Flowrate calc to get the flowrate, Heat Dissipation graphs to roughly gauge the CFM required, then Cathar's Fan-on-Radiator testing as a loose guide to choose the fans... Step5 is simply fitting it all in the case and thus irrelevant.

One of the issues is you can't account for the backpressure imposed by the case, which has knock on effect on airflow thru the rad... so you have to go overboard on the CFM vs what the graph tells you just to ensure that chassis backpressure isn't an issue... (ie: We can only guage performance based on the kit in free air. Soon as it's in a case, all the figures for thermal dissipation at blah airflow vs blah liqflow go out the window, unless we assume that whatever fans and quantity thereof that are attached to radiator pulling air into case are matched with the same exhausting air from the case... then it remains vaguely valid)

See http://www.over-clock.com/ivb/index.php?showtopic=20277

Where in the UK are you Bobo??

[bobo5195]

That’s interesting and what I figured. I figured If I was to make something I would just do it and let the data catch up although TIM stuff is as always suck it and see. I think that at least something that calculates something gives you a feeling for what’s going on, even if the numbers are all balls. The n umbers can be changed easily anyway. Besides it’s a nice intellectual exercise.

I would add that the textbook (at least the mech eng / motorsports way) is useful that it just says that radiator performance is X regardless of flow etc. This is probably not the case in real life as flow rate does effect heat transfer performance. It never comes up in mech eng as watercooling as rather big relative changes in input characteristics for similar systems. As with the all engineering once you look at what the equations imply its all wrong anyway.

I work in central London (a few minutes from Piccadilly Circus). With regularish trips to Cardiff.