|
|
Testing and Benchmarking Discuss, design, and debate ways to evaluate the performace of he goods out there. |
Thread Tools |
10-10-2005, 02:30 PM | #51 |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
10mm ID
0.8 lpm Re=1902 0.9 lpm Re=2139 Many would say ~0.85 LPM Give Mir if wish Ta for Excel, even tho I hate them- u use too much magic(formulas) which confuse my simple Excel operations - pasting to other work sheets Last edited by Les; 10-10-2005 at 02:37 PM. |
10-10-2005, 02:51 PM | #52 | |
Cooling Savant
Join Date: Sep 2003
Location: Vallentuna, Sweden
Posts: 410
|
Quote:
You should like that one, not a single formula in it. |
|
10-11-2005, 08:59 AM | #53 | |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Quote:
Getting uncorrected C/W above yours Probably done something stupid, but thought better check before looking again for silliness. Watts calculated from 5mm spacing( 0.127551c/w) Am simply adjusting reported dTs differences by 0.056888 c/w(2.23mm) for C/W(wb+TIM) and 0.102041c/w(4mm) for C/W(TIM) Edit Added TIM Inco Last edited by Les; 10-11-2005 at 12:53 PM. |
|
10-11-2005, 10:28 AM | #54 | |
Cooling Savant
Join Date: Sep 2003
Location: Vallentuna, Sweden
Posts: 410
|
Quote:
Not quite sure what you mean Les. I am preparing a Post with calculations so you can see how I generate the numbers in the logfile. |
|
10-11-2005, 11:05 AM | #55 |
Cooling Savant
Join Date: Sep 2003
Location: Vallentuna, Sweden
Posts: 410
|
1. @FLUX=393*0.00010*(@LOW4 + @TOP3*-1)/0.005077
2. @CU=(@LOW4 + @TOP3*-1)/5.077 3. @TI=((@CU*1.988 + @FB2)*-1 + (@TOP3 + @CU*-2.081))/@FLUX 4. @CW=(@FB2 + @CU*-2.017 + @WIN5*-1)/@FLUX + @TI*-1 5. @FLOW=60*(@FLUX/(4186*(@WOT6 +@WIN5*-1))) 6. @SYS=(@FB2 + @CU*-2.017 + @AIR1*-1)/@FLUX + @TI*-1 These are cut and pasted directly from my DAQ monitoring script. Translated into readable equations as follows: 1. Heat flux = k x A x (Tdie_low - Tdie_top)/L (heat in watts, Fourier, q=k.A.dT/L, k=393W/m°C, A=0.0001m^2, L=0.005077m, [nominally 5mm]) 2. Cu dT = (Tdie_low - Tdie_top)/L (Temperature gradient in °C/mm) 3. TIM C/W = ((Tdie_top - Cu dT x 2.081) - (Cu dT x 1.988 + Tfluxblock))/Heat flux or: (Extrapolated die top surface temp minus extrapolated FB lower surface temp, i.e Interface dT) divided by watts 4. C/W waterblock = (Tfluxblock - Cu dT* x 2.017 -Twater_in )/Heat flux -TIM C/W or: (Extrapolated FB top surface temp minus Water in temp) divided by watts minusTIM i.e Air to WB base C/W 5. Flowrate (lpm) = 60 x (watts/(Cp x (Twater_out - Twater_in)) 6. System Cw = (Extrapolated FB top surface temp minus Air temp) ...Not very useful, air is too variable I realise I have named these very badly, particularly Die Top, sounds like die top surface, actually Die upper (top!?) sensor and Die low, = lower die sensor. I will rename them for future logs. One thing I have noticed, different k here (393) than model (392). won't make any observable difference but shows I am making mistakes. Note the scripts inability to deal with the minus operation, I have to use + xx*-1. Edit added pic Last edited by Incoherent; 10-11-2005 at 11:23 AM. |
10-11-2005, 12:05 PM | #56 |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Was only inquiring whether "FLUXBLOCK 2", "DIE TOP 3 and "DIE LOW 4" where raw data values.
Using those as raw data getting the above graph.. Graph generated with 1. Heat flux = k x A x (Tdie_low - Tdie_top)/L (heat in watts, Fourier, q=k.A.dT/L, k=392W/m°C, A=0.0001m^2, L=0.005m) 2. Cu dT = (Tdie_low - Tdie_top)/L (Temperature gradient in °C/mm) 3. TIM C/W = ((Tdie_top - Cu dT x 2.) - (Cu dT x 2 + Tfluxblock))/Heat flux or: (Extrapolated die top surface temp minus extrapolated FB lower surface temp, i.e Interface dT) divided by watts 4. C/W waterblock = (Tfluxblock - Cu dT* x 2.23 -Twater_in )/Heat flux -TIM C/W or: (Extrapolated FB top surface temp minus Water in temp) divided by watts minusTIM i.e Air to WB base C/W My values should show uncorrected values.Designated as (C/W) rather than (C/W)". These (C/W)wb have previously been lower than (C/W)"wb. Here, mine is higher than yours. Was going to simply multiply (C/W)wb by 1.038069 to get (C/W)"wb. 1.038069 is my new number now I have(poached from you) the " positions in the Fluxbloc.However my values are higher. Something is amiss. Edit Like the picture. Yes, does help. Last edited by Les; 10-11-2005 at 12:18 PM. |
10-11-2005, 01:18 PM | #57 | |
Cooling Savant
Join Date: Sep 2003
Location: Vallentuna, Sweden
Posts: 410
|
Quote:
I understand. All the temperatures are raw or at least ADC counts converted to voltage converted to resistance then spun through the Steinhart-Hart equation with coefficients generated from the calibration. Not sure what's going on. Can you simply use a multiplier? I'll double check this. |
|
10-11-2005, 01:25 PM | #58 |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Dunno but was going to check against your sums.
Seems your W" is a tad higher than mine - only calculating now eg 61.67384 vs 61.51689896 The corresponding W(mine) is 62.4642592 These are line1 in above pic (Log Period 25470) Edit CudT" the same at 1.569308647(les) and 1.569309(Inco) The difference is the 393 and 392 Cu. Which is it? Obviously neither but.......... Last edited by Les; 10-11-2005 at 02:02 PM. |
10-11-2005, 02:03 PM | #59 |
Cooling Savant
Join Date: Sep 2003
Location: Vallentuna, Sweden
Posts: 410
|
Don't know Les, this is what I get.
The TIM has settled quite a bit. |
10-11-2005, 02:27 PM | #60 |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Your quick.
True positions = 2.23,2,2,and 5? - my L 2.017,.........5.077 are my L" Will leisurely do some(depending on how much data u post) L" and post overnight/tomorrow Last edited by Les; 10-11-2005 at 02:33 PM. |
10-11-2005, 04:29 PM | #61 | |
Cooling Savant
Join Date: Sep 2003
Location: Vallentuna, Sweden
Posts: 410
|
Quote:
Meantime, a rather clean log (some formulae) Look for more stable flowrate, (grounding problem fixed - no more flyers at high flowrates, plus a recalibration of water temperatures) |
|
10-11-2005, 04:50 PM | #62 |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
All looks sweet after all.
Made some wholesale cock-up some where. First graph crap. Not even know where offending curve came from. Only data on work sheet resembling is your "System C/W". Have recalculated nothing. . Ratio 1.038069429 is way out. Horse lost. Not been my day. Ta for extra Excel. With the amount of data you produce even I am considering some magic. Thinking of LMTD ? Morning Edit . Added Series3. Every thing behaving. Retaining "L=2.23......" in picture . Effect on predictions of moving "Shadow Model" zero to the, possibly more realistic, wb/Fluxbloc interface ? Having a look-see. Lunch update Not sure whether needs Fermi to have another crack at it. Not sure is better. Not sure ............ Though, is food for thought Manually getting Ts and Ls is a chore and my magic is very limited Last edited by Les; 10-12-2005 at 06:45 AM. |
10-12-2005, 11:33 AM | #63 | |
Cooling Savant
Join Date: Sep 2003
Location: Vallentuna, Sweden
Posts: 410
|
Quote:
I'll work some Excel magic, will post new L''s (L'''?) and result when done. Edit: Of course. The numbers for L''' are exactly the same. They should be, that's the whole premise. Last edited by Incoherent; 10-12-2005 at 11:57 AM. |
|
10-12-2005, 11:59 AM | #64 |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Dunno but need careful think after re-crunching with new parameters.
Am producing new L" but slowly - am useless against the clock. If not get your more reliable(my opinion) L", will then plug into Series2 and 3 - again slowly. Then we look-see? Edit "Of course. The numbers for L''' are exactly the same. They should be, that's the whole premise." Will find out - slowly Edit Yes,or at least Flovbloc 2.23 gives 1.65 gr at 2.016578788 Last edited by Les; 10-12-2005 at 12:12 PM. |
10-12-2005, 12:14 PM | #65 | |
Cooling Savant
Join Date: Sep 2003
Location: Vallentuna, Sweden
Posts: 410
|
Quote:
Meantime I'm ready to try a remount of the block. All this data has been the 2nd mount of the block for the sake of getting the testbed characterised. It's important that I be able to reproduce this mount for mount so that any TIM results I generate have meaning. With the previous 12x12mm fluxblock this was no problem, I was consistantly able to get very similar numbers mount for mount. Remains to be seen if it is possible with the 10x10. My feeling is that the TIM settling time is longer with this one, can't think of any reason for that except perhaps that the paste has aged six months or so. |
|
10-12-2005, 12:55 PM | #66 | |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Quote:
Yep Ls are the same . |
|
10-12-2005, 01:08 PM | #67 |
CoolingWorks Tech Guy Formerly "Unregistered"
Join Date: Dec 2000
Location: Posts: 2,371.493,106
Posts: 4,440
|
yes, get fresh paste
|
10-12-2005, 05:01 PM | #68 | |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Quote:
Cannot split The uncorrected values later/tomorrow. Update Added uncorrected . Also cannot split( the two have done) So no wiser Last edited by Les; 10-12-2005 at 07:27 PM. |
|
10-13-2005, 12:28 AM | #69 |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Happy the "Shadow Correction" method applies to the fb/die IF where IF is included in the model.
Happy the "Shadow Correction" values are correct for C/W(TIM). However have grave niggles about C/W(wb).values Here IF is not included in the model. Moving zero again to water/bp IF. This is closer to reality. However the gr=1.65 constraint does not apply.in bp section. In fact the constraint varies from wb to wb. The wb has to be modeled before any error correction can be applied. Ugh. |
10-13-2005, 02:36 AM | #70 | |
Cooling Savant
Join Date: Sep 2003
Location: Vallentuna, Sweden
Posts: 410
|
Quote:
Horrible isn't it? L''s for another model, see chart for details 0.000 1.976 3.953 3.953 6.030 11.105 13.456 |
|
10-13-2005, 02:54 AM | #71 |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Interesting
Dumbfounded you have come up with numbers so quick. Still digesting. May edit this post(rather than repost) as thoughts emerge |
10-13-2005, 03:14 AM | #72 |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Have I got this right: Tdev curves are about same for two examples.
Wattage L 5.075 wb+TIM L 1.976 TIM L 4.054 Like(iI think) the always + Tdev Reasoable guess for MCW6000 at *lpm Have you plugged in ? Last edited by Les; 10-13-2005 at 03:37 AM. |
10-13-2005, 03:59 AM | #73 | ||
Cooling Savant
Join Date: Sep 2003
Location: Vallentuna, Sweden
Posts: 410
|
Quote:
Quote:
|
||
10-13-2005, 04:21 AM | #74 | |
Cooling Savant
Join Date: Oct 2001
Location: Wigan UK
Posts: 929
|
Quote:
The MCW6000was rant To expand on rant The 20,000 h(eff) on 50x50x5mm bp = 77,519W/m^2C h(eff) on die-surface The 20,000 h(eff) on 50x50x5mm bp would be a fair guess of conditions in MCW6000 @ 0.25 lpm.However in tests there is TIM which reduces h(eff) to ~50,000(on fb-surface) My model 17,000 on 40x40 @ 0.25 lpm 75,000 on 40x40 @ 10 lpm Values are in Book1 Excel. No idea whether dimensions are correct. Consider much higher h(eff) in small patch thin bp wbs. For MP-05-SP with 1mm bp 10,000 on 14.4x14.4 @ 0.25 lpm 250,000 on 14.4x14.4 @ 10 lpm In general possibly looking at an h(eff) range of 10,000 to 300,000(call it 500,000).and areas 10x10 to 60x60 with bps 0.1 to 10 Change in Ls changes nothing much = Can distinquish a different curve Last edited by Les; 10-19-2005 at 10:26 AM. |
|
10-13-2005, 06:28 AM | #75 | |
Cooling Savant
Join Date: Sep 2003
Location: Vallentuna, Sweden
Posts: 410
|
Quote:
|
|
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
|
|