Go Back   Pro/Forums > ProCooling Technical Discussions > Testing and Benchmarking
Password
Register FAQ Members List Calendar JavaChat Mark Forums Read

Testing and Benchmarking Discuss, design, and debate ways to evaluate the performace of he goods out there.

Reply
Thread Tools
Unread 10-24-2005, 04:26 PM   #51
Les
Cooling Savant
 
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
Default

OK,only need to throw out half the paper - where coolant velocity is varied.
The remainder stands.
That includes most of the above analysis.
Have edited analysis with treble ???
If the ??? is correct the consequences are a liitle difficult too digest.
It would suggest that if the spoke of 5x differential was used an error of 60%+ may be expected when comparing low fannage(0.5m/s say) with high fannage(3.5m/s).
Thought was easier to forget the whole paper than struggle with further analysis - have enough conundrums on plate.
Les is offline   Reply With Quote
Unread 10-24-2005, 04:32 PM   #52
bobo5195
Cooling Savant
 
Join Date: Aug 2005
Location: uk
Posts: 400
Default

Been thinking about his overnight and here is my executive summary. Using the paper previously mentioned and my uni notes.

First LMTD is not useful in the case of WC radiators this is because it takes no account of the load. As my uni notes say you have to use a geometry correction factor to analyse rads and such and need to know all 4 temps this is not useful. I remain unconvinced based on model arguments that LMTD has anything to do with actual rad performance. A rad is essentially 2D. Heat transfer occurs very quickly and depth is less than the length of a fan facing side. I am comfortable in saying that air basically gets heated and expelled from the rad. I believe work done earlier for the PA160 (which I’m badly remembering) shows that depth is not so useful. LMTD is also a pain to measure as it includes an outlet air temperature that is hard to quantify.

Any measure of rad performance should be a measure of the efficiency of the geometry to move heat to the air. This is obviously C/W or W/C from scaling arguments (heat in / DT). You could equally move to say based on my arguments above that it’s a function of frontal area so C/W per m^2. I am not happy with this as absolute performance is what we are after. C/W hence forth known as SD (aka W/C) is obviously a function of specific energy output of the system (in Watts or kack handed American units) and dT. This whole argument is about which dT is it. To my mind there are only two things that matter to that rad and that’s the input conditions. The output does not determine the performance. Hence SD = Q / (Tiw – Tia), which is exactly the same as in the paper.

Hence the argument of Bill that C = water in temp – ain in temp. Stews argument of usefulness is a good one but I am not sure it exactly quantifies the performance as well. If you know Q its is obvious to back calculate the results, but the form looks nasty as shown in image 1. I am half way to convincing myself that C/W for stews case is not independent of Q.

The problem with SD is it tells you nothing off the outputs without a little maths but it does make sense to your average joe on the street from a derivation point of view. Personally I prefer efficiency as it’s a nice quantitative number that tells you what you need to know and is equivalent to SD. It can also be calculated without going into the merky depths of what the cpu core power is. This is useful not only as a test but also because your never sure what heat a CPU is dumping to the WB. You will also notice that eff is proportional to the flow rate which is useful when considering small changes in flow rate and how they will impact your loop (from better tube routing etc.)

As Les pointed out you have changes in properties to worry about. I personally think that these changes are small in the case of material properties over temperature. Adding 1c to water isn’t going to change its properties. Air is slightly different as it has a larger rise and more likelihood of change but given the errors in measurement associated with measuring it I am happy to discount humidity and air Cp. There is obviously some change from fan geometry (and shroud) and fan speed so these should always be included but the eff formula allows small changes in fan speed to be modelled.

Eqs to flow once I’ve scanned them.
bobo5195 is offline   Reply With Quote
Unread 10-24-2005, 05:20 PM   #53
Les
Cooling Savant
 
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
Default

Quote:
Originally Posted by bobo5195
....As Les pointed out you have changes in properties to worry about. I personally think that these changes are small in the case of material properties over temperature. .....
I am not worried about changes in material properties.
My concern is
(1) The measurement of Heat DissipatedW) using Two-Twi without correction for dP*Q (~Coolant LPM*mH20/6)
(2) Whether C/W(or W/C) using either (Twi-Tai) or (Two-Tai) for C is (Twi -Tai) independent.My closer examination of the Delphi paper raises grave doubts
Les is offline   Reply With Quote
Unread 10-24-2005, 05:52 PM   #54
BillA
CoolingWorks Tech Guy
Formerly "Unregistered"
 
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
Default

did we not conclude that (1) was small ?

the Delphi paper speaks for itself I presume;
as the tests are normally run (in our world), Twi and Tai are both independent variables and setpoints
- is the C/W load dependant ?
in terms of efficiency my understanding is 'sure', the greater Twi - Tai the lower the C/W
but the changes, again, are not so large in our small rads at low velocities

EDIT
a very knowledgeable (thermo PHd) friend looked over the Delphi paper and made a dismissive comment about number manipulation and 'not science'. FWIW

Last edited by BillA; 10-24-2005 at 05:59 PM. Reason: add scurrilous comment
BillA is offline   Reply With Quote
Unread 10-24-2005, 06:02 PM   #55
Les
Cooling Savant
 
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
Default

"did we not conclude that (1) was small ?"


Not for this possible radiator "external rad .... with low Rwet at low water flow - high water velocity rad"(link)

"is the C/W load dependant ?"

Yes
Les is offline   Reply With Quote
Unread 10-24-2005, 06:05 PM   #56
bobo5195
Cooling Savant
 
Join Date: Aug 2005
Location: uk
Posts: 400
Default

i was about to say about small changes at low temps also the car water temp is exceptionally high so things are a bit iffy then.

I need to think about this correction for energy loss when water travels down a pipe. On one hand it makes alot of sense on the other the exteneed bernouli equation on which WC analysis is based is actually a load of bollocks. Either way W add in a block is small enough not to get me to worried saying that.

power = dP*Q

dP is proportional to mean velocity^2 which is roughly proportional to Q^2 (need to do maths to establish abit more)

Power = k * Q^3 so for high flow values i can see what your saying hmmm.

Last edited by bobo5195; 10-24-2005 at 06:06 PM. Reason: spelling
bobo5195 is offline   Reply With Quote
Unread 10-24-2005, 06:08 PM   #57
bobo5195
Cooling Savant
 
Join Date: Aug 2005
Location: uk
Posts: 400
Default

hmm i'll have to try and re read that delphi paper and try to find pot holes. Its not uncommon espeically as its practical design not to be spot on science wise.
bobo5195 is offline   Reply With Quote
Unread 10-25-2005, 03:40 AM   #58
Les
Cooling Savant
 
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
Default

Quote:
Originally Posted by Les
"is the C/W load dependant ?"

Yes

A drastically revised answer
I don't think C/W is load dependant.

Yes, I think you are correct "the greater Twi - Tai the lower the C/W" or at least I think C/W changes with Twi - Tai .
Yes, again, you are correct "the changes, again, are not so large in our small rads at low velocities " or the at least C/W changes with Twi - Tai vary with velocity.

Edit Thanks for helping crystallize my thoughts.
Upto now was wishing has thrown out "Delphi's problem child" with the bathwater

Last edited by Les; 10-25-2005 at 03:45 AM.
Les is offline   Reply With Quote
Unread 10-25-2005, 01:13 PM   #59
bobo5195
Cooling Savant
 
Join Date: Aug 2005
Location: uk
Posts: 400
Default

Had a look at some books (big books) and the delphi paper is on the right track but the rabbit hole goes deep.

There are reasons why the current approach does not work but they are beyond the level of this forum. They can losely be divided into 2nd law considerations and flow effects which all add up to something fairly significant.
bobo5195 is offline   Reply With Quote
Unread 10-25-2005, 01:24 PM   #60
BillA
CoolingWorks Tech Guy
Formerly "Unregistered"
 
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
Default

ah bobo
what do we make of your last sentence ?

and by "current approach" you mean C, or C/W, SD ?
more please

EDIT
I am aware of anomalous high flow data, it needed a correction
and the low flow data needed more resolution
the latter I,m dealing with, for the former I need understanding
BillA is offline   Reply With Quote
Unread 10-25-2005, 01:41 PM   #61
bobo5195
Cooling Savant
 
Join Date: Aug 2005
Location: uk
Posts: 400
Default

Yes current approach - SD and effectiveness. Alot of extra stuff maybe useful as well (the stuff on NTU in the notes i sent you certainly).

It could be extended to the p - NTU approach but that would not be more useful in this case. Where P is the non dimensional temprature.

My second sentence means that DIYers dont have the ability to measure it and you need masters grade engineering and physics to understand the second law considerations. Its new (1999) science and it is hard. The approach may or may not be correct.
bobo5195 is offline   Reply With Quote
Unread 10-25-2005, 01:49 PM   #62
bobo5195
Cooling Savant
 
Join Date: Aug 2005
Location: uk
Posts: 400
Default

If your really that interested put this on your christmass list

Fundamentals of Heat Exchanger Design (Hardcover)
by Ramesh K. Shah, Dusan P. Sekulic

http://www.amazon.com/exec/obidos/tg...X0DER&st=books
bobo5195 is offline   Reply With Quote
Unread 10-25-2005, 05:38 PM   #63
Les
Cooling Savant
 
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
Default

You can always present as "Fan/Radiator Effectiveness" without measuring Airflow.
Taken Papst 4412FGL@12v as 55.3cfm
Attached Images
File Type: jpg Monash4.jpg (15.1 KB, 12 views)
Les is offline   Reply With Quote
Unread 10-25-2005, 06:11 PM   #64
bobo5195
Cooling Savant
 
Join Date: Aug 2005
Location: uk
Posts: 400
Default

The point of measuring both is not so much as a better measure. The second approach has oviously more potential for error but as a check for correct results. Given that all apart from airflow are already used i thought that it would be worth doing. And knowing the airflow is quite a useful factiod as well.

I should congratulate thermochill on a good rad i never realised the eff. was that good. I could easily see the reaching a performance level of 60%. One book i was reading read that rad eff over 60% are not speced because it costs to much although high spec power station ones are approaching 99% with super technology.
bobo5195 is offline   Reply With Quote
Unread 10-25-2005, 08:15 PM   #65
Les
Cooling Savant
 
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
Default

Bobo
You are saying "the current approach does not work........( current approach - SD and effectiveness)"
You are suggesting that effectiveness is a check on "C/W(in)" and SD? Yet , to determine, it uses the same manipulation of the same data(+air-flow) .Doesn't make sense to me.
This means, to me, that you are concluding neither "C/W(in)" or SD can be used to characterize a radiator.
You have already expressed doubts over "C/W(out)" being independent of Q .
It does not leave much, if anything.
Yet , having dismissed all, your only suggestion is read a text on Heat Exchangers.
Surely this is a waste of time, in that it will surely confirm all measures "will not work".
Les is offline   Reply With Quote
Unread 10-26-2005, 07:39 AM   #66
bobo5195
Cooling Savant
 
Join Date: Aug 2005
Location: uk
Posts: 400
Default

Eff, SD and C/W (using input conditions are interchangeable), the equations are the same

Before I say anything going through the other method is a CHECK. It reduces the likelihood of error. Gives you two separate paths to the same answer. It tells you if your C/W gathered in the standard way is on the low or the high side. By gathering the results anyway you can tell loads of useful things. You can predict using fan PQ curves the flow through the fan and various other measures. More importantly it provides a check for Q. If your using a test bench setup you better make sure that your heating apparatus is right this does it. If your using a real loop then how on earth do you know Q to anything more accurate than 10%. Microprocessor heat for a given voltage is hardly exact. How do you know that a program loading a cpu will give X amounts of heat. The system is not well insulated, how much heat is being dissipated in the wires. Its not possible, repeat tests won’t eliminate it. Even you if got C/W exactly right and it was a perfect relationship, when it comes to someone else’s setup you can never be sure. Back pressure in the case will inhibit fan flow so they will not get the same results. As I’ve mentioned they are never go to be able to accurately guess the heat input to give them the right dT. What about radiators, because of manufacturing problems their performance is going to vary it might be within 5% of each other. These are engineering approximations, the fact that a result is given with in 10% is good enough. It tells you that this is better than this with a reasonable accuracy.

I’m saying the current approach works well enough. Its not exactly right but is good enough. The SD and eff approach is based on the 1st law of thermodynamics. Energy in = Energy out, but as you’ve already notice there are things like friction (which you’ve already calculated) which makes this approach less than absolutely correct. By the fact that SD is varying it is obvious that its not a geometric constant that is perfectly stable, independent of flow rate and heat input magnitudes. However the SD approach has been used for over half a century, engineers know its wrong but its easy to measure and is a good approximate catch all value. If people wanted more than SD they had to resort to graphs and correction factors normally.

As way of an example if you use first law to calculate gasoline engine performance you get a value about double what it actually is. Its in the right ball park and the 1st law will tell you loads of good ways to increase performance (it easily predicts that 2 strokes are better than 4 strokes, for example). An approximation using the second law is better still. After that your still not there and you basically need a myriad of gradually more accurate equations to make things right but the complexity goes up exponentially. A similar thing can happen with rads. SD is accurate enough, deriving SD then needing a page of maths to correct things is not the way to go.

As far as I can remember the 3, 2nd law errors that were identified in the book, were boundary layer effects, friction (water AND air) and losses form transferring heat over a finite temperature difference. There are others as well, flow mal-distribution, the fact that rads are thick, end effects from the sides dumping to the air. Change of thermal properties of the air (water content makes a near 5% difference in air Cp), errors in geometry, dust in the radiator, thermal properties of the air (you add 10% de-icer, things are going to change). I may post some of the results out of interest but the idea of applying and using them is a bit far fetched. For example removing heat addition from friction due to flow rate. Does trying to remove this effect actually make results more accurate? Does it tell you more about the radiator? Or does it just mean that you’ve got to recorrect for it to see how the rad actually performs.

In short SD gives you a nice approximate number it varies but its easy to understand and estimate. If don you honestly thing that Bills thermochill results are within 10% for an actual loop?
bobo5195 is offline   Reply With Quote
Unread 10-26-2005, 07:58 AM   #67
Les
Cooling Savant
 
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
Default

Quote:
Originally Posted by bobo5195
.
I’m saying the current approach works well enough. Its not exactly right but is good enough.
So you were not saying "the current approach does not work"
but rather the - the current approach is good enough.

Now you are exploring the limitations?

When you show the results of your exploration will they show the results are within 10%?
Les is offline   Reply With Quote
Unread 10-26-2005, 08:08 AM   #68
bobo5195
Cooling Savant
 
Join Date: Aug 2005
Location: uk
Posts: 400
Default

I'll go with the standard its a good approximation.

maybe, maybe not. I'll have to look at the equations. Either way the first eqs are directly from the first law so they are not going ot be massively out.

Consider that these effects have an impact on everything (WB, standard heatsinks) and tehn you'll being to see how far the rabbit hole goes.
bobo5195 is offline   Reply With Quote
Unread 10-26-2005, 09:44 AM   #69
BillA
CoolingWorks Tech Guy
Formerly "Unregistered"
 
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
Default

a trip through obscurantism ?
how has this exercise illuminated the basis for C ?

bobo
the phrase 'good enough' has a considerable history in WCing, your usage here is taken as literal
we know and accept that we are dealing with applied technology, and not science as such
and we, generally speaking ~ a dozen or so, do have an active interest in making improvements in our understanding, testing, and analysis
yes, most here really do understand that in a rad everything interacts with everything else - we are aware of the literature
in my case I am the sole source of rad data for 'our' industry (for which the test conditions and methods are described), outside of larger companies not directly involved

I have no problem with criticism and correction, I call such progress
do you have any specific suggestions ?
test methods, instrumentation, data collection, data analysis and/or presentation ??

all topics are open (though here we started with "C")
but at some point the action items need to be identified, can you help us progress ?

I too recommend books to the untutored, was this a snide pay-back ?
you have a lot of knowledge, can we benefit from it ?

Last edited by BillA; 10-26-2005 at 09:58 AM. Reason: incomplete, posted in error
BillA is offline   Reply With Quote
Unread 10-26-2005, 10:48 AM   #70
bobo5195
Cooling Savant
 
Join Date: Aug 2005
Location: uk
Posts: 400
Default

Bill:
I think this has illuminated the basis of C, the conditions are know and if you look at the SD to eff equation it is quite obvious how C is related to other inputs at least according to the first law of thermodynamics. The eff rules tell you what needs to be measured and for what case. The NTU relationship i sent you shows the relation with area and how to convert U about the place.

My worry is that there is a significant wall of knowledge coming up. I’ve spent 4 years getting my knowledge to this level (at what is supposedly the 3rd best mech eng course in the world) and a lot of things are already causing me to stop and think. The level is hitting the literature wall out there and becoming more on vague on what’s important and what is not. The 2nd law approximations established in the book i mentioned are some guys thinking. This guy admittedly is the former head of research at delphi so i figure he knows rads but it is still unconfirmed.

I mentioned the book above because that book will allow anyone here to go out and design one damn rad if read properly. Its 1000 pages long and is complete head **** but its everything that’s needed. It even contains notes on thermosypons so ppl here can work out exactly how good they think they are and if they are worth it (i believe they are not), maybe start a flame war with JoeC if they wish (this forum seems to like flame wars and it think that’s a good thing if handled well). I could summerise a lot of the work in there and help with the difficult bits but it’s a lot of my time to do that. There’s another book by Keys and London called compact heat exchangers which basically goes into the level here. At the library there’s a set of reference notes on this that im figuring are about 5000 pages thick but I don’t really want to be touching those. I get the feeling that the entire area is imprecise science with a lot of the info being in big labs.

Its all a question of what you want to know.

In further to your edits:

I’ve never actually built a test rig before and to be honest some people here know more than me about the practical stuff. My one suggestion as I have mentioned in my posts is calculating C/W two different ways for rads. I’ll have a look at any suggestions but my practical engineering is not so good. The only thing I would caution on is keeping it simple. A vacuum test chamber is not simple for example. If your going to spend money on complicated stuff make sure it’s a good set of sensors and maybe labview.

I do enjoy the ramblingness of forums. I feel that we have answered the C question here.

I recommend that book for reasons mentioned above, I am without doubt that it will answer all of your questions that can be answered and it will save me a lot of hassle writing information up and ME getting it wrong. Its not payback; its just as you can well accept answering a lot of simple and not so questions and checking that your answering right can be boring and repetitive.
bobo5195 is offline   Reply With Quote
Unread 10-26-2005, 11:08 AM   #71
Les
Cooling Savant
 
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
Default

Maybe some of limitations are using "C/W"or "W/C" values obtained at set Tai and Twi for operation at other Tai and Twi
Am crudely exploring these limitations using my old friend Kryotherm(must be good, it is Russian)
Possibly the most common situation is Tai is "comfort controlled at 25c(say)" and Twi varies.

Initial play suggests only a small influence on Rwetside for a 120x120x30
Possibly 1% for change from Twi-Tai=5 to Twi-Tai=25 for Rad with C/W=0.06

If anyone is interested may post any further numbers, if not will not.(if continue ferreting)
Attached Images
File Type: jpg Monash5.jpg (24.3 KB, 5 views)
File Type: jpg Monash6.jpg (18.7 KB, 3 views)

Last edited by Les; 10-26-2005 at 11:15 AM.
Les is offline   Reply With Quote
Unread 10-26-2005, 01:13 PM   #72
BillA
CoolingWorks Tech Guy
Formerly "Unregistered"
 
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
Default

I used to have Keys and London, informative but a bit dated
I do understand what you are saying, but something which cannot be conveyed cannot be used - and we do like the useful here
I am an experimentalist (or so described) so fret little about theory -> untill its time to interpret the data at which point I'm right back in the soup

vacuum is (hopeful) Ben, I do insulation

Les
I had data sets showing 1% (and less) differences, but |absolute| accuracy at this level is questionable - though the results were repeatable

should we attempt to list 'effects' considered 'too small' to justify climbing through our ear backwards ?
BillA is offline   Reply With Quote
Unread 10-26-2005, 03:15 PM   #73
Les
Cooling Savant
 
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
Default

Quote:
Originally Posted by unregistered
..climbing through our ear backwards ?
Finding determination of a suitable surface T of a flux-block(again for C) as painful(link) . Playing with "limitations" of rad testing is a convenient diversionary/delay tactic
Are you going to consider C for wb testing?
Les is offline   Reply With Quote
Unread 10-26-2005, 03:40 PM   #74
BillA
CoolingWorks Tech Guy
Formerly "Unregistered"
 
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
Default

oh lord Les, you're probably not referring to the inlet but to the dT basis ?
as in silicon (die ??) vs. IHS (case) vs. bp (device) ?

now we're in it
(for the rads C is Twi-Tai, end of story)
BillA is offline   Reply With Quote
Unread 10-30-2005, 02:11 AM   #75
Les
Cooling Savant
 
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
Default

Quote:
Originally Posted by Les
I don't think C/W is load dependant
Since no-one seems keen to move-on will dwell
Although not saying C/W(in) varies with load, I am far from happy it does not.
Example (using True MTD calculator,only valid for "TEMA" shell types but...)
Set Points(terminology?) Wi=40c, Ai=30c
At 100w Wi-Wo =1 (and Ao-Ai=9)
TMD=F(LMTD)=3.0, (LMTD= 3.6), C/W(in)= 0.1 C/W(TMD)=UA =0.03
At 200w Wi-Wo =2 (and Ao-Ai=8)
TMD=F(LMTD)=3.5, (LMTD=4.3), C/W(in)=0.05 C/W(TMD)=UA=0.0175

Think "UA" is the true thermal description and "(C/W)in" contains a few extra bits

Thoughts ?

Last edited by Les; 10-30-2005 at 02:41 AM.
Les is offline   Reply With Quote
Reply


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


All times are GMT -5. The time now is 02:34 AM.


Powered by vBulletin® Version 3.7.4
Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
(C) 2005 ProCooling.com
If we in some way offend you, insult you or your people, screw your mom, beat up your dad, or poop on your porch... we're sorry... we were probably really drunk...
Oh and dont steal our content bitches! Don't give us a reason to pee in your open car window this summer...