can we agree on the basis of "C" in C/W ?

[bobo5195]

U is the thermal resistance per unit area

TEMA types are not vaid here.

End effects and assorted problems like that are going to be proportional to Q.

[Les]

1) Used UA
2) Stated only valid for TEMA. Unsure whether invalid for Compact Xflow. Have you an example?
3) "End effects and assorted problems" = "extra bits and pieces"

Are we agreeing there are problems?

[BillA]

sure
some are sidestepped by the general similarity of form (six 120mm fan/rads can be externally very similar, six 2-core (dual row) rads can be compared to each other)
with enough data on different aspects then those can be characterized
depends on the type of/purpose for the data collection

many other 'problems' can be minimized by focusing on the actual conditions of use

10% is rather good for 'hand calcs' such as we are doing
I generally settle for an understanding of 'which was better' in testing, but such will involve smaller increments than I would presume to calculate with accuracy

[bobo5195]

We agreed that there were problems ages ago but that there were to small to be worried about.

To measure second law effects you have to take into acoount all 4 temps which is a bugger for testing especially as they are only useful in scaling calcs not overall performance.

TEMA is invlaid use xflow results. There are actually many types of crossflow as well, in our case counterflow might be valid as well.

Actual conditons can play a bit part. Attaching a rad to a warm case full of warm air is obviously going to cause an effect for examplem thats there is obviously no point focusing on.

Another way of including UA is given below. Eq 2 and 3 show relations for eq 1. NTU is a commonly used perfromance indicator for rads. Cmin is whatever masslfow * cp is lower. This is why on other areas i say that either fan or water flow rate is important not both. The effect is easily seen in bills graphs where performance bottoms out at a set point.

[Les]

Originally Posted by bobo5195

We agreed that there were problems ages ago but that there were to small to be worried about.
.

"We agreed that there were problems ages ago"
Yes
".....but that there were to small to be worried about."
with the exclusion of me.
You have conjured up 10% (or was it 11%) out of thin air. May well be 10% but have not been shown evidence.

Yes, use crossflow.Why I was asking you for "crossflow example"

The interrelations shown can be useful in presentation data. but do little to validate method. Unless you can demonstrate with an example - proof of pudding.

[Les]

Originally Posted by bobo5195

.... The effect is easily seen in bills graphs where performance bottoms out at a set point.

Attached plot
Not the plot you wished.
Old data and possibly errors(link to data)
However shows performance bottoming out.
This tells us nothing.
The values however suggest something may be amiss
Need examples


Edit1: Added link
Edit2 : Reply to

Originally Posted by bobo5195

The jet airliner where you entrust your life to a heat exchanger uses this method..

Not,I hope, only using C/W data obtained for a (hot - cold) differential possibly 10x that experienced in service

[bobo5195]

We agreed that they were too complicated to worry about. Proof of how right they are is impossible as you dont know what you are comparing them to.

The current method was accepted practice in the industry from the 40's to the presnt day. Various books use it as a good first order interpretation. The jet airliner where you entrust your life to a heat exchanger uses this method. Its wrong yes but not wrong enough.

[BillA]

hold it please
let me generate a set of accurate data for Les et all to work with (several weeks +)
(a shame Swiftech does not seem inclined to put up their rad comparison)
I am unsettled by references to 'small' values wrt my earlier testing, it was not that accurate

[Les]

No intention of creating work.
If "small" values you are referring to are the E=100%+ for Big Momma, feel these are caused by P/Q curve error . Fouling (rabbit hair) comes to mind.

[bobo5195]

anythign above 60% for a computer rad take with a pinch of salf. More than likely they are caused by other means.

A really good rad test would have all 4 temps, 2 mass flows and Qin. If we have all of those accurately you can quantify pretty much everything. I think that gives you 4 ways to get C/W so errors should be minimal.

[Les]

Originally Posted by bobo5195

anythign above 60% for a computer rad take with a pinch of salf. More than likely they are caused by other means..

Evidence please.
My modeling suggests a range of 20-100% is possible in a 120x120mm rad.

Edit1: Added more points to graph
Edit2 : Finished off sums, attached

[bobo5195]

I am convinced anything over at least 80% is completely wrong. Definitly. The only heat exchangers that achieve over 95% are the huge industrial type that cost more than your house. If there are results over 100% then they are quite obviously wrong. In this forum we obey the laws of thermodynamics. The other flow could be Cmin.

Car heat exchangers are in the 40% range. A paper i was reading expressed worries about using a heat exchanger of over 60% as it is "too good" because you end up paying more money for it.

[Les]

Again a source or evidence would be useful.
Use big book and calculate for some sample 120x120rads

Edit: Of course results over 100% are wrong. These are beermat calculations using rough, average Specific Heats and Convection Coeffs. Normalizing(terminology) max % of 101.9% to 100%(or less, 95% say) would be no biggy

[bobo5195]

Originally Posted by Les

Again a source or evidence would be useful.
Use big book and calculate for some sample 120x120rads

Edit: Of course results over 100% are wrong. These are beermat calculations using rough, average Specific Heats and Convection Coeffs. Normalizing(terminology) max % of 101.9% to 100%(or less, 95% say) would be no biggy

Source is the big book i linked to above. I dont have any data but i will see if i can whip up a spreadsheet. I need to have a look over how to do some proper error calculations though as it could get abit involved.

[BillA]

from a procedural perspective, if you have questions ask
I'm presently w/o data but should have some soon (?)
the air side is characterized by the fan inlet air temp only (as described earlier)

the data is not 'wrong', it is our assumptions regarding the data, etc.

EDIT
just took a look at "Radiator" Effectiveness v Air Flow and I see the units "Air Flow (83.min)
this was measured with a thermal anemometer and a traverse across the 6" duct used for a 'wind tunnel'
one thing is quite certain, the 'tunnel' - not being calibrated - IS going to be off at at least one end of its range, and likely both

iffy data to be beating up each other over

[bobo5195]

I had a read of my heat exchanger book on the train. Bit of a headfock. Alot of the 2nd law stuff is not applicable here as we do not know Air temp out

[Les]

Hear you Bill.
However my current dispute with Bobo does not involve data, it is the theoretical expectations of PC radiators.
Have finished current set of sums and will keep further thoughts to myself.

Final update is now attached to Post 86

[bobo5195]

Do you mean a full analytical model from first principals?
2nd law equations which mean not very much?
How to design a good 120mm rad?
Variation of c/w with various gemetric parameters?

2nd law stuff cannot be computed from current data as all 4 tempratures are need. Measurement of current data is fairly simple, although bills +/- 1 air temp resutls in huge errors, and there is the problem of no test for air water content

[bobo5195]

This is quick and rough but it looks right.

3 sheets, condensed is the one your after(tidier interface), designed for converting bills thermochill results (C/W) into eff, SD and "bobo" C/W (being the correct one of course).

Enjoy.

Results are not entered properly (not proper bill C/W), but i have attempted to calculate errors and show how things are affected. Shows limits of DIY

CAUTION
Bills C/W results read of the graph will effect the eff results alot, take with pinch of salt.

[Les]

Initial comments
1)Bill at that time was , I think, using a LMT for water T component of C in C/W
2) Your sums appear correct if using correct units in Cp(air)m calcs. I get ~10w/c to your ~600 .
Bloody units. Will check mine .Whichever, it explains all.
3) Hate magic in Excels - messes up my simple sums - I suppose I must learn sometime.

Edit: Checked mine, looks ok, but forever screwing-up unit conversions
Air @ 0.5m^3/min , Cp=1005.70 J/kg-C, Density=1.1774 Kg/m^3

m[Kg/hr]Cp= 35523.34 (J/hr)/C = 9.867593 W/C

Will check again in the morning.
Anyone?

This would make Effectiveness ~ 97% at all considered water flow-rates
Not in disagreement with Post 86's modeling predictions for a 24 FPI rad

[Les]

Morning
Checked again, my numbers are fine.
Your numbers are also fine.
However you are, I think, working at AirFlow=30m^3/min( 0.5m^3/s), which is 75X the AirFlow at which there is a "Measured Heat Dissipation" of ~50w ( Graph 12)
Looks GIGOish


Also not at all happy in your description of Bill's C/W.
His C/W, IMO, should be C/"Measured Heat Dissipated"
Using current practice this is, for these tests, 5/"Measured Heat Dissipated"
As can be seen from Graph 12 it decreases with increasing flow.

[bobo5195]

Originally Posted by Les

Morning
Checked again, my numbers are fine.
Your numbers are also fine.
However you are, I think, working at AirFlow=30m^3/min( 0.5m^3/s), which is 75X the AirFlow at which there is a "Measured Heat Dissipation" of ~50w ( Graph 12)
Looks GIGOish


Also not at all happy in your description of Bill's C/W.
His C/W, IMO, should be C/"Measured Heat Dissipated"
Using current practice this is, for these tests, 5/"Measured Heat Dissipated"
As can be seen from Graph 12 it decreases with increasing flow.

Yes the units are wrong, i'm fairly sure about the other stuff. The eff seems high and i'll have to recheck my notes. It could be this high in real life as the operationg conditions are close together so entropy loses could be quite high. Notice alos how large the error bars are so this eff value even if it is right is clearly not acceptable as a good test.

[Les]

Originally Posted by bobo5195

. Notice alos how large the error bars are so this eff value even if it is right is clearly not acceptable as a good test.

Think is partially covered by : : " was measured with a thermal anemometer and a traverse across the 6" duct used for a 'wind tunnel'
one thing is quite certain, the 'tunnel' - not being calibrated - IS going to be off at at least one end of its range, and likely both"

Plus the measurement error bars are some what arbitrarily set(case of ask Bill) and think your Cp and Density ranges are rather large(no numbers just ....)

(Fan+Rad)'s Eff. may well be a better option

Entropy - dunno but happy as long as don't bring in activation volume

[bobo5195]

consider psycromery (spelling?) science of water in air. Creates large density and heat capacity changes in air.

Not sure what bill means by tunnel and calibration. If it is way off then that could explain results. In fact it makes this entire argument null and void.

Water results were quite high but i was worried about additives.

[Les]

Originally Posted by bobo5195

(1) If it is way off then that could explain results.(2) In fact it makes this entire argument null and void.

(1) Yes. However results can be considered as (fan +rad) which gives info on rad :: Erad>E(rad+fan).
(2) No
My dispute is that you(with the Big Book backing) say that when E is above 60% the data should be taken with a pinch of salt, and when E is above 80% the data is completely,definitely wrong.
I say above 60% and 80% are acceptable values, as indicated by my modeling

You and Big Book would automatically discard this data in any case.
I would accept unless there are reasons to reject