Upupa Epops said:
Actually, the holes are not there for looks and I'm aware that some of them are too small for ventilation, but that's what experimentation is all about.
The bronze heatsink was initially much bigger and had only one row of holes in a center. Initial using indicated that this is not the best approach. Because of low thermal conductivity and high mass, the bronze piece didn't work as heatsing but as a heat tank, accumulating thermal energy but not relasing it (the two adiitional copper panels are supposed to release the heat from bronze).
Later, I decided to reduce the "heatsink" mass and storage capacity, that's why those additional smaller holes. It works better now, but still not efficient as I wished.
I might consider adding additional metal plates between two copper panels. That will both improve shielding and increase the dissipating surface.
After changing the shape of bronze piece, I noticed slight difference in sound as well. So it's not only the material itself, but also mass and shape.
You are spending an awful lot of time on that stuff, now ain't ya??? I must admit, I did pull most of it from where the sun don't shine, but it is stuff I use everday.metalman said:Dang nabbit! I realized I didn't say it right. The thermal flux does decrease, but the thermal gradient does increase in response to a thermal pulse. Not that anyone actually cares, but I hate leaving my mistakes uncorrected. Less embarrasing than having someone else correct me later!
Thermal diffusivity is a function of thermal conductivity and heat capacity. The more conductive, the higher the diffusivity. The lower the heat capacity, also the higher the diffusivity.
Copper at 4.5 Kelvin has incredibly high diffusivity, for example. This is because it's thermal conductivity remains the same (relatively) as it is at room temp, but it's heat capacity drops 2 to 3 orders of magnitude..heat flies through it at liquid helium temperatures.
Cheers, John
PS...ok, I admit...I woulda had fun wit youzes errahs..
In the old days, when dinosaurs ruled the world, most of the packaged silicon was put into non isolated packages. Kovar (magnetic), aluminum (god forbid), and now, for the most part, copper.
The isolated packages have a plastic layer on the bottom of the copper, and this has low thermal conductivity, essentially forcing the die to act as an isothermal island. This can be good for high power devices, as it lessens the thermal gradients which can form across the die. It does, of course, impact derating.
There are significant differences between the isolated and non isolated packages..It is unknown if those differences show as sonic, but one cannot easily brush the possibility aside.
Capacitance is changed. Thermal pulse response is changed. Indeed, proximity of the heatsink to the silicon die changes the system inductance as well as the coupling between the output currents and the input gain stages.
I can figure out at least three ways changing the heatsink material could conceivably alter the transfer function of an "all in one silicon die"..
Vibrational aspects are not one of them. Skin depth, thermal diffusivity, and thermal conductivity certainly, but not what I would consider silly.
The fact that someone reports a diff, well..are we all so sure? I'm not.
Cheers, John
The isolated packages have a plastic layer on the bottom of the copper, and this has low thermal conductivity, essentially forcing the die to act as an isothermal island. This can be good for high power devices, as it lessens the thermal gradients which can form across the die. It does, of course, impact derating.
There are significant differences between the isolated and non isolated packages..It is unknown if those differences show as sonic, but one cannot easily brush the possibility aside.
Capacitance is changed. Thermal pulse response is changed. Indeed, proximity of the heatsink to the silicon die changes the system inductance as well as the coupling between the output currents and the input gain stages.
I can figure out at least three ways changing the heatsink material could conceivably alter the transfer function of an "all in one silicon die"..
Vibrational aspects are not one of them. Skin depth, thermal diffusivity, and thermal conductivity certainly, but not what I would consider silly.
The fact that someone reports a diff, well..are we all so sure? I'm not.
Cheers, John
Re: Basic Thermal Management of Power Semiconductors
Thanks for the link.
It is a shame that on the first page is a big glaring error.
Look at the picture of the packaged die..lower left hand side of the first page..
They drew the die with the die bottom surface on the same plane as the package case bottom????
That doesn't happen. They shoulda drawn the package base thickness there, as the die is soldered to that base. Typically, this thickness is about 40 mils, and increases the effective heat source dimension 80 mils by 80 mils. For die in the sub 250 mil range, this is a significant error component in calculation of thermal resistance from the junctions on the top of the die to the heatsink.
Without this heat spreader, as it were, the calculated heat flux through the insulator will be over estimated, the thermal drop overestimated...life as we know it would end!!!!!!!!!!
Sloppy, shame on them..
Cheers, John (aka, the puppeteer)
ps...of course, my little nitpick tirade has nothing to do with the article's intent...it is a very good article..
pps...I built me one of dem thermal testers, I like the fact that they specify measurement of the parameter of interest, as nuttin generally runs 2.2 mV/degree C. Power diodes generally run from .6 to 1.2 mV/C. I ran up to 100 amps DC, and had the choice of method...eitehr forcing current/measuring temp, or forcing temp with a feedback loop.. It was really neat, a bridge running 45 amperes and then unbolting the bridge from the sink, and watching the machine react by dropping the current..
Elso Kwak said:Hi, it would do no harm to read this:
http://www.onsemi.com/pub/Collateral/AN1083-D.PDF
Basic Thermal Management of Power Semiconductors
Most is said already in the 1st paragraph!
Thanks for the link.
It is a shame that on the first page is a big glaring error.
Look at the picture of the packaged die..lower left hand side of the first page..
They drew the die with the die bottom surface on the same plane as the package case bottom????
That doesn't happen. They shoulda drawn the package base thickness there, as the die is soldered to that base. Typically, this thickness is about 40 mils, and increases the effective heat source dimension 80 mils by 80 mils. For die in the sub 250 mil range, this is a significant error component in calculation of thermal resistance from the junctions on the top of the die to the heatsink.
Without this heat spreader, as it were, the calculated heat flux through the insulator will be over estimated, the thermal drop overestimated...life as we know it would end!!!!!!!!!!
Sloppy, shame on them..
Cheers, John (aka, the puppeteer)
ps...of course, my little nitpick tirade has nothing to do with the article's intent...it is a very good article..
pps...I built me one of dem thermal testers, I like the fact that they specify measurement of the parameter of interest, as nuttin generally runs 2.2 mV/degree C. Power diodes generally run from .6 to 1.2 mV/C. I ran up to 100 amps DC, and had the choice of method...eitehr forcing current/measuring temp, or forcing temp with a feedback loop.. It was really neat, a bridge running 45 amperes and then unbolting the bridge from the sink, and watching the machine react by dropping the current..
Re: Re: Basic Thermal Management of Power Semiconductors
Yes the package metal is uncredible thin under the die in the picture. Must be a drawing error, as I think it is as thick as the tab there.
I my carreer as a repair tech I replaced so many transistors as they became too hot. Too small heatsink! Also in my Gas Ampzilla pre-drivers became way too hot; crazy!
jneutron said:Thanks for the link.
It is a shame that on the first page is a big glaring error.
Look at the picture of the packaged die..lower left hand side of the first page..
They drew the die with the die bottom surface on the same plane as the package case bottom????
That doesn't happen. They shoulda drawn the package base thickness there, as the die is soldered to that base. Typically, this thickness is about 40 mils, and increases the effective heat source dimension 80 mils by 80 mils. For die in the sub 250 mil range, this is a significant error component in calculation of thermal resistance from the junctions on the top of the die to the heatsink.
Without this heat spreader, as it were, the calculated heat flux through the insulator will be over estimated, the thermal drop overestimated...life as we know it would end!!!!!!!!!!
Sloppy, shame on them..
Cheers, John (aka, the puppeteer)
ps...of course, my little nitpick tirade has nothing to do with the article's intent...it is a very good article..
pps...I built me one of dem thermal testers, I like the fact that they specify measurement of the parameter of interest, as nuttin generally runs 2.2 mV/degree C. Power diodes generally run from .6 to 1.2 mV/C. I ran up to 100 amps DC, and had the choice of method...eitehr forcing current/measuring temp, or forcing temp with a feedback loop.. It was really neat, a bridge running 45 amperes and then unbolting the bridge from the sink, and watching the machine react by dropping the current..
Yes the package metal is uncredible thin under the die in the picture. Must be a drawing error, as I think it is as thick as the tab there.
I my carreer as a repair tech I replaced so many transistors as they became too hot. Too small heatsink! Also in my Gas Ampzilla pre-drivers became way too hot; crazy!
My personal favorite was your von Mises reference. Wonderfully obscure! My clue to the sarcasm was that anyone who could reference the von Mises stress criteria pretty much by definition would have to understand it, and for Peters chassis and the temperature differences we're considering, the differential thermal expansions wouldn't come anywhere close to the von Mises stress.
If it weren't for the character limitations here, I'd post the differential equation for solving transient conduction in a semi-finite solid. Then we could build a finite element model and simulate the different materials. Ooh Ooh, who's with me!!!
Everything sounds according to Peter. You ought to know that by now, PavelUpupa Epops said:Sound of heatsinks ? Do you really mean it seriously ?
hard to argue against that..
Diamond! This is a very nice material and you don't have to use any insulation material. Diamond heatsinks sound really incredible, don't you agree Peter?peranders said:I have a good material for heat transfer. It's an insulator and conducts heat better than silver. Which material?I'll let you think about this for a while.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.