Simulated Die-IHS-WB

[Groth]

Oh great, now I'm gonna be doin' multiple runs with assort bp's and h's to see how well they match.

[BillA]

just establishing confidence limits
- seeing how far we can entice you

[Groth]

I'd have more confidence if I could get at least one decent run through. This last re-run resulted in temperature data for the die and wb, but 3/4 of the ihs data is missing. I don't yet know what quarter of it I have; my fingers are crossed that it's a full quadrant and not a some useless strip...

Time to work out some frustrations on the bicycle.

[Les]

For future work.
Think may be good idea to to discard TIMa.(35 w/cm^2*K)) and use TIMb(20w/cm^2*K) for all joints:-
1) Think is more representative.
2) Less confusing in modeling(at least for me) and interpretation of IHS and IHSless results.

Time for a quick hobble,on my crutches, to the pub.

[Groth]

Posted graphs have been updated with the newer data (correct behavior of silicon).

Les, do you have an exact number off your waterloo graphs? For my best-so-far run, I've got delta-T(die underside - water) of min=19.00668 max=24.21851 mean=21.96613. Appears to be a good 3° lower than yours.

And delta-T(die topside - water): min=12.70811 max=18.99964 mean=16.56051

[Les]

Good morning.
26.5828
Suspect this is an underestimate. Dodgy boundary conditions at wb/TIMb interface.
Much more confidence in the non-IHS temperatures.
DeltaTunderside-water = 22.5300 (24k, 6mm) , 29.4428 (24k,1mm)

Still cannot understand DeltaT Si.
I use the non-spreading value of 5.4054
Being calculated with zero spreading-resistance this if anything an underestimate.

Remember I am a thermo dilettante and no authority.Have not done any serious science for 25 years.

More coffee.

[Groth]

It's very easy to explain that DeltaT in my Si: wrong matrix. Had used a set of nodes a quarter millimeter from the top. Fixed.

If the trend continues, I won't even attain the status of thermo dilettante.

Less coffee, more ice cream.

[Les]

Originally Posted by Groth

It's very easy to explain that DeltaT in my Si: wrong matrix. Had used a set of nodes a quarter millimeter from the top. Fixed. ...........
Less coffee, more ice cream.

Good , at least our basics are similar.
Does this fix not add ~2c to DeltaTmax(underside -water).
Think my fumble(excellent adjective Cathar) sums refer to max.

My first preference is "choc Cake and cream" with ice cream being a good cream substitute

[Les]

Ah, your 7:44 am edit has done this?
Now DeltaTmax(underside - water) =24.21851 was less?

[Groth]

Gets confusing when subbing pictures and editing old posts...hmmm, think I'll do it some more.

The deltaTmax(underside - water) remains 24.21851. The error in the previous deltaT(die) was with topside of the die, specifically that I didn't use the actual top of the die.

A pint of Ben and Jerry's "Karmal Sutra" does wonders. The heatspreader is being more cooperative.

[Les]

Know the confusion well.
Just been correcting a link in a 2 day old post .
I try to be as specific as possible but this leaves one very vulnerable to errors.

Would be interested in the no IHS case if could be done without too much effort.
I feel this would give a reference point for the divergence of the our methods.
Have fair confidence in my interpretation of Waterloo for no IHS.

I don't know "Ben and Jerry's "Karmal Sutra"".
However I will be in the John Bull this afternoon
Will be shaking, quaking , and brain dead Saturday.

Edit : Better pic of the John Bull

[Groth]

The non-IHS has been running for a while now. I had thought it would be quicker, but it looks to want another 20 hours (~62 total). Oh, what I wouldn't do for 10 GB of very fast RAM. I'll run with a couple different h, too, and make a curve.

I had the spelling wrong: Karamel Sutra.

[Les]

May be done by the time I am compos mentis(ish) again

Did think you were referring to beer.
Dunno whether is available in UK ,but will look.
Have updated link to the John Bull

[Groth]

The sans-IHS runs were crap. I had tried a new method of putting the pieces together, in hopes of being more able to reuse chunks old models in the newer. Equilibrium was never attained. Back to the drawing board.

In other news, the with-HS runs are not showing the same response to increased h(eff) as in Les's. Will have enough points to graph shortly.

[Les]

Have been looking for a number check.
In particular at the predictions at infinite h.
Initially thought should give the 1-D results however now have doubts.
Needs more thought.

My results, for possible inclusion in any graph, for IHS inclusive case.

[Groth]

I figured out why the sans-IHS wouldn't converge, a rounding error. That I've found five bugs in Linear's SPICE program bothers me, but I know that I'm not using it as it was intended be used, so...ehh :shrug:

While waiting for a fix or work around, now I have to go back and search for any subtle effects it might have had on earlier bits.