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 03-04-2006, 11:21 AM   #1
BillA
CoolingWorks Tech Guy
Formerly "Unregistered"
 
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
Default assumptions require examination

all done and said is predicated on assumptions, they are unavoidable
but it is instructive to be aware of such, and their potential for error when ignored

wrt heat transfer, 'ideal' contact areas are assumed
and even more significantly from an experimental viewpoint:
1) the 'source' is assumed to be perfect - which is never true wrt flatness
1A) while it is merely constant if only one source is used
1B) AND if no wear or degradation is occurring - which is known not to be true
2) the DUT is assumed to be representative of all such items - which is 'true' only if statistically validated
3) AND the source/sink thermal coupling is 'correct' - again, 'true' only if statistically validated

the above does not mean that testing cannot yield conclusions,
just that extreme prudence is needed in the interpretation of test results (a sample of one is always questionable);
all of the above listed 'variables' are interrelated and indeed interactive,
change one slightly and the apparent results are quite different

for illustration an enhanced bp image is attached made with an optical flat in a reflex viewer using monochromatic light from a filtered mercury bulb,
each 'fringe' (light/dark band) represents a height (difference) of ~0.000273mm (0.000010749in.).
- the example shown has a huge valley running across it, the square outline in the center is 10x10mm
Attached Images
File Type: jpg (non) flatness.jpg (234.0 KB, 30 views)
BillA is offline   Reply With Quote
Unread 03-04-2006, 12:12 PM   #2
BGP Spook
Cooling Savant
 
BGP Spook's Avatar
 
Join Date: Sep 2005
Location: Virginia
Posts: 153
Default Re: assumptions require examination

Quote:
Originally Posted by BillA
all done and said is predicated on assumptions, they are unavoidable
but it is instructive to be aware of such, and their potential for error when ignored

wrt heat transfer, 'ideal' contact areas are assumed
and even more significantly from an experimental viewpoint:
1) the 'source' is assumed to be perfect - which is never true wrt flatness
1A) while it is merely constant if only one source is used
1B) AND if no wear or degradation is occurring - which is known not to be true
2) the DUT is assumed to be representative of all such items - which is 'true' only if statistically validated
3) AND the source/sink thermal coupling is 'correct' - again, 'true' only if statistically validated

the above does not mean that testing cannot yield conclusions,
just that extreme prudence is needed in the interpretation of test results (a sample of one is always questionable);
all of the above listed 'variables' are interrelated and indeed interactive,
change one slightly and the apparent results are quite different

for illustration an enhanced bp image is attached made with an optical flat in a reflex viewer using monochromatic light from a filtered mercury bulb,
each 'fringe' (light/dark band) represents a height (difference) of ~0.000273mm (0.000010749in.).
- the example shown has a huge valley running across it, the square outline in the center is 10x10mm
Would you care to give an absolute(or relative) minimum vs. respective maximum?

Personally, I find it difficult to accuratetly interpret multi-banded monochromatic images.

By my interpratation, guessing the number of bands to be no more than 30 (probably a high estimate, from center of "valley" to edge of "valley" and assuming that every band indicates a depression rather than any change), there is ~8 micrometers (0.00819 mm) variation between peak and depression.

Do not claim to be an expert, so maybe it is my naivety, but I would not clasify 8 micrometers "huge."
Granted ANY variation does have a negative effect.

Again, could you give absolute or relative min/max.

Edit: Changed "interpreting" to "interpret," "a change" to "any change," "And I do realise" to "Granted"
__________________
I can't spell, but I am working on it.

Last edited by BGP Spook; 03-04-2006 at 12:29 PM.
BGP Spook is offline   Reply With Quote
Unread 03-04-2006, 12:55 PM   #3
BillA
CoolingWorks Tech Guy
Formerly "Unregistered"
 
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
Default Re: assumptions require examination

along the bottom ~30, and about the same vertically on the left

counting from lower left to upper right is what will define the 'depth' and the fringes become too dense to count,
their curvature can be seen going from upper left to lower right, the upper right corner is quite 'high'
- will be very apparent from the grease (assuming the die is flat)

added a detail image with more contrast
compare to this http://www.thermal-management-testing.com/PF3fringe.jpg
(a 10mm sq is not much larger than that central area)

of course this is only half of the equation, the die/CPU relative flatness is an issue of equal significance
- consider the case of one convex and the other concave ?
- what is the more typical case is both convex (certain if hand lapped - other than by N8 or Steve at PolarFlo)

the effect on the wb's apparent C/W is what is huge, more so than the actual dimensions might suggest (w/o testing)

any appraisal of wb C/Ws is incomplete w/o understanding the contribution of flatness,
could be due to this or that or whatever, assumptions are a tricky business
Attached Images
File Type: jpg detail.jpg (179.6 KB, 11 views)
BillA is offline   Reply With Quote
Unread 03-04-2006, 01:08 PM   #4
BillA
CoolingWorks Tech Guy
Formerly "Unregistered"
 
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
Default Re: assumptions require examination

and restudying the original makes me think my initial assumption was wrong, i.e. backwards

the band across the die area is probably high, not low
I suspect that the 'low' areas are out of focal length
- no way to verify now, have a viewer but not that wb

what was I saying about assumptions ?, lol
same effect though re heat transfer
BillA is offline   Reply With Quote
Unread 03-04-2006, 01:49 PM   #5
BGP Spook
Cooling Savant
 
BGP Spook's Avatar
 
Join Date: Sep 2005
Location: Virginia
Posts: 153
Default Re: assumptions require examination

Quote:
Originally Posted by BillA
along the bottom ~30, and about the same vertically on the left

counting from lower left to upper right is what will define the 'depth' and the fringes become too dense to count,
their curvature can be seen going from upper left to lower right, the upper right corner is quite 'high'
- will be very apparent from the grease (assuming the die is flat)

added a detail image with more contrast
compare to this http://www.thermal-management-testing.com/PF3fringe.jpg
(a 10mm sq is not much larger than that central area)

of course this is only half of the equation, the die/CPU relative flatness is an issue of equal significance
- consider the case of one convex and the other concave ?
- what is the more typical case is both convex (certain if hand lapped - other than by N8 or Steve at PolarFlo)

the effect on the wb's apparent C/W is what is huge, more so than the actual dimensions might suggest (w/o testing)

any appraisal of wb C/Ws is incomplete w/o understanding the contribution of flatness,
could be due to this or that or whatever, assumptions are a tricky business
I count ~23-25 but will call it 30 for easy and to give benefit of the doubt.

Can see how convex to convex or concave to concave would magnify thermal resistance and how it would need to be accounted for.

Suppose <10 microns variation for contact surfaces.(more later) Worst we could see would be about ~20 microns metal to metal gape.

How does claimed thermal resistance of AS5 factor in then?
http://www.arcticsilver.com/as5.htm
http://www.arcticsilver.com/ceramique.htm

Claimed thermal resistance of AS5
<0.0045°C-in^2/Watt (0.001 inch layer)
and ASC
<0.007°C-in^2/Watt (0.001 inch layer)

0.001 in = 25.4 microns


Quote:
Originally Posted by BillA
and restudying the original makes me think my initial assumption was wrong, i.e. backwards

the band across the die area is probably high, not low
I suspect that the 'low' areas are out of focal length
- no way to verify now, have a viewer but not that wb

what was I saying about assumptions ?, lol
same effect though re heat transfer
If we could guarentee (in a reasonable manner) <10 microns variation on a contact surface. Couldn't we apply the claimed thermal resistance across an area as worst case to better derive total thremal resistance?

If claimed thermal resistance is bogus. (Opinion: at least worth testing throughly.) Couldn't we start by defining what the actual thermal resistance (as per worst case thickness) of something like AS5 is?

As a first step anyway.

Or am I makeing too big an assumption?
__________________
I can't spell, but I am working on it.
BGP Spook is offline   Reply With Quote
Unread 03-04-2006, 02:06 PM   #6
BillA
CoolingWorks Tech Guy
Formerly "Unregistered"
 
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
Default Re: assumptions require examination

ask AS about AS stuff

heat transfer is primarily through the actual contact points as I understand the literature,
the flatter the surfaces the greater number of points and/or area of contact
a uniform film of 0.001" w/o any substrate contact would be a pretty good insulator, easily demonstrated by removing the film
"uniform film" is an (imaginary) assumption requiring 2 absolutely flat surfaces

TIM testing is a specialty of its own, lots of info on how to do it;
see Inco's testing for some TIM values, I defer to him

EDIT
you said "need to be accounted for"
I'm not suggesting a calculation methodology which is what you seem to be driving at

rather an awareness that the number of variables not explicitly quantifiable necessitate temperance in the making of assumptions/conclusions

Last edited by BillA; 03-04-2006 at 02:22 PM.
BillA is offline   Reply With Quote
Unread 03-04-2006, 03:01 PM   #7
BGP Spook
Cooling Savant
 
BGP Spook's Avatar
 
Join Date: Sep 2005
Location: Virginia
Posts: 153
Default Re: assumptions require examination

Quote:
Originally Posted by BillA
ask AS about AS stuff

heat transfer is primarily through the actual contact points as I understand the literature,
the flatter the surfaces the greater number of points and/or area of contact
a uniform film of 0.001" w/o any substrate contact would be a pretty good insulator, easily demonstrated by removing the film
"uniform film" is an (imaginary) assumption requiring 2 absolutely flat surfaces

TIM testing is a specialty of its own, lots of info on how to do it;
see Inco's testing for some TIM values, I defer to him
If "a uniform film of 0.001" w/o any substrate contact [is] a pretty good insulator" then how can AS claim (low?) thermal resistance?

I must be missing something obvious. Else, why bother with thermal interface material at all?


"'uniform film' is an (imaginary) assumption requiring 2 absolutely flat surfaces"

Maybe a restatement would clarify:
1. If we knew the theoretical thermal resistance at 25 microns with perfectly flat metal -"uniform film"- perfectly flat metal or could establish real-life worst case thermal resistance at 25 microns
2. Then if could gurantee <10 microns variation per mating surface with thermal resistance filling the gap.(by reasonable means as yet unknown/discussed)

Would we then be better able to test anything? Assuming point 1 turns out to be arbitrarily small under most scenarios.

Edit:
Quote:
Originally Posted by BillA
EDIT
you said "need to be accounted for"
I'm not suggesting a calculation methodology which is what you seem to be driving at

rather an awareness that the number of variables not explicitly quantifiable necessitate temperance in the making of assumptions/conclusions
Indeed.
Can we quantify a worst case variance and find it to be either negliable or account for a given (reasonable) size gap and make sure we do better than such a gap?


Will have to get back to this, need time to track down and digest Incoherent's posts and other relevant information. Maybe talk to AS. (HA!)

Real life commetments are also catching up with me, though.
__________________
I can't spell, but I am working on it.

Last edited by BGP Spook; 03-04-2006 at 03:08 PM.
BGP Spook is offline   Reply With Quote
Unread 03-04-2006, 03:18 PM   #8
BillA
CoolingWorks Tech Guy
Formerly "Unregistered"
 
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
Default Re: assumptions require examination

both Colin and Nevin will be happy to speak with you, good people

don't understand the difficulty, 'anything' can be tested now
I'm commenting on assumptions, who (other than I) has looked at flatness wrt wb testing ?

before you get too far into flatness characterization and subsequent calcs best see what data you are going to have to work with
??

BTW, laser interferometers are better for the quantification of flatness
(I'm not clever enough to build one, don't have the bucks to buy one)

BTW, it is unnecessary to reproduce each post; only you and I are posting, I can follow what you are saying

EDIT
no, one does not 'quantify' a worst case by making an assumption re its magnitude
"to make sure we do better" is a mfgr activity, the tester is concerned with the DUT and its characterization
I'm talking about testing, not the analysis of process variation and defining a 'lower' process limit for mfgn
the tester does not have a clue about such (assuming that it was defined by the mfgr) w/o a very large set of samples over a number of production lots
no 'outside' tester will ever have access to a sufficient quantity of pieces (or the time !) to define the process variation, a test is a window

Last edited by BillA; 03-04-2006 at 03:51 PM.
BillA is offline   Reply With Quote
Unread 03-04-2006, 06:44 PM   #9
Brians256
Pro/Staff
 
Brians256's Avatar
 
Join Date: Oct 2001
Location: Klamath Falls, OR
Posts: 1,439
Default Re: assumptions require examination

Quote:
Originally Posted by BillA
no 'outside' tester will ever have access to a sufficient quantity of pieces (or the time !) to define the process variation, a test is a window
That is something I've been worried about since day one. We "test" waterblocks and make assumptions about the product line. I presume that the variation is small enough to make our testing valid. Otherwise, testing is worthless.
Brians256 is offline   Reply With Quote
Unread 03-04-2006, 06:53 PM   #10
BillA
CoolingWorks Tech Guy
Formerly "Unregistered"
 
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
Default Re: assumptions require examination

these things are machine made, generally ok - but not identical (looked at the 6000)
not enough to try to select a 'good one' for review
I attribute the variation largely to bp flatness (avgd mountings)
less with a TTV, larger base - more stability [lol]
BillA 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 07:37 PM.


Powered by vBulletin® Version 3.7.4
Copyright ©2000 - 2017, 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...