If you have your devices are mounted directly to a smooth anodized surface, that is great, and will be much better than, say, a bead-blasted surface for thermal transfer. But the surfaces will still not be perfectly mated and the junction will be LOUSY. One of our members polished his heatsink to a mirror finish and looked at it under a microscope and it was still full of scratches.
http://www.koolance.com/technical/cooling101/002.html
http://www.coolermaster.com/index.php?LT=english&Language_s=2&url_place=discovery_detail&serial=56
Even "smooth" parts need a thermal interface material. Goop is easy, silpads are easiest. Phase change material like the stuff by Aavid Thermalloy is both easy and much better than both the materials mentioned above.
If you already have an electrical insulation with your anodized heatsink... I still have lots of powerstrate available, cheap...
But if you don't want that, PLEASE! don't use 'reconstituted' heat sink grease! you'd be way better off going to ebay and spending a few bucks.
My recommendation to everyone: at some point you most certainly WILL wish you had some sort of thermal interface material to go on your project in the middle of the night when the electronics store is closed.... get some. Mica, Aluminum-whatever, silpads, powerstrate, or even a small tube of grease. Get Something! You will *never* regret having it just laying around. !
http://www.koolance.com/technical/cooling101/002.html
http://www.coolermaster.com/index.php?LT=english&Language_s=2&url_place=discovery_detail&serial=56
Even "smooth" parts need a thermal interface material. Goop is easy, silpads are easiest. Phase change material like the stuff by Aavid Thermalloy is both easy and much better than both the materials mentioned above.
If you already have an electrical insulation with your anodized heatsink... I still have lots of powerstrate available, cheap...
But if you don't want that, PLEASE! don't use 'reconstituted' heat sink grease! you'd be way better off going to ebay and spending a few bucks.
My recommendation to everyone: at some point you most certainly WILL wish you had some sort of thermal interface material to go on your project in the middle of the night when the electronics store is closed.... get some. Mica, Aluminum-whatever, silpads, powerstrate, or even a small tube of grease. Get Something! You will *never* regret having it just laying around. !
You guys take things way too seriously. This is an enthusiast forum, not mil-spec/medical applications.
I never meant to imply that "reconstituted" goop is ideal or even recommended. In my case, it got me by in a pinch while I was troubleshooting a piece of gear and didn't have all my supplies handy.
And if your design depends so critically on thermal transfer, I would follow a previously posted suggestion to spread the load over multiple devices. In certain cases, (like bipolars, not sure for FET's) it can result in better gain linearity and lower noise anyway.
I never meant to imply that "reconstituted" goop is ideal or even recommended. In my case, it got me by in a pinch while I was troubleshooting a piece of gear and didn't have all my supplies handy.
And if your design depends so critically on thermal transfer, I would follow a previously posted suggestion to spread the load over multiple devices. In certain cases, (like bipolars, not sure for FET's) it can result in better gain linearity and lower noise anyway.
People do not know when to stop.GD3 said:
On other forum some guys discussed whether TXCO with 1ppm is enough good to upgrade clock in CD. )) I think that if they were able to get clock components utilized by GPS systems or cesium oscilator they would still say about improvements in sound quality of upgraded CD player.Best regards
Jacek
> People do not know when to stop.
Please let us know whether you have done any measurement for a TO3 or TO247 transistor dissipating ca. 30W to a heatsink, how much temperature there is between the transistor case and the heat sink; i.e. how much temperature rise along is due to the mica / goop / silpad / ... or whateverelse in between.
Best regards,
Patrick
Please let us know whether you have done any measurement for a TO3 or TO247 transistor dissipating ca. 30W to a heatsink, how much temperature there is between the transistor case and the heat sink; i.e. how much temperature rise along is due to the mica / goop / silpad / ... or whateverelse in between.
Best regards,
Patrick
EUVL said:
Patrick, I have not done any measurements excluding my hands or rather fingers
I utilized Al2O3 after bad experience with mica and silicon pads but my case was "difficult". I build ( works for two years so far without any problems ) amplifier with output pair transistors in TO247 case, which both dissipate 80W power to one heatsink
on "good day" ( class A )
The base is to have proper heatsink designed to disappated power and one should not save the money for not efficient thermal pads. It is really low cost in comparison to whole project.
Best regards
Jacek
You said for yourself that you had bad experience with Silpad and Mica. It does not really matter whether you dissipate 80W and 30W per transistor. The physics is the same. People here at the Pass forum build Class A amps, so they have enough heat to get rid of.
If you look at the entire thermal chain (from transistor substrate to heatsink), then you would realise that there are 3 major contributors to thermal resistance :
1) Substrate to case (in the order of 0.8°C/W/transistor, as supplied, nothing you can do about);
2) Heatsink (assume 0.2°C/W/transistor, and 4 transistor per heatsink, then also 0.8°C/W/transistor, can be improved but costs a lot of money and space);
3) Insulation between case and heatsink.
For point (3), let say we stick to simple solutions. It does not take much effort to calculate that for a typical TO3 ot TO247 package, a single piece of mica plus 10µm of grease on each side, or a standard silpad, the temperature rise for 30W of dissipation per transistor is in the order of 12°, or 0.4°C/W/transistor.
The sum of the entire thermal train would then be something like 2°C/W. And you can easily reduce this by some 20% with minimum attention to just the choice of insulator. This is much much easier and more cost effective than changing to a larger heatsink.
Of course there are more tricks to get it even further down. But that is another story.
So to response to your comment, I, for one, know where to stop - that is when the thermal resistance across the insulator is less than 5% of the total thermal resistance.
Patrick
If you look at the entire thermal chain (from transistor substrate to heatsink), then you would realise that there are 3 major contributors to thermal resistance :
1) Substrate to case (in the order of 0.8°C/W/transistor, as supplied, nothing you can do about);
2) Heatsink (assume 0.2°C/W/transistor, and 4 transistor per heatsink, then also 0.8°C/W/transistor, can be improved but costs a lot of money and space);
3) Insulation between case and heatsink.
For point (3), let say we stick to simple solutions. It does not take much effort to calculate that for a typical TO3 ot TO247 package, a single piece of mica plus 10µm of grease on each side, or a standard silpad, the temperature rise for 30W of dissipation per transistor is in the order of 12°, or 0.4°C/W/transistor.
The sum of the entire thermal train would then be something like 2°C/W. And you can easily reduce this by some 20% with minimum attention to just the choice of insulator. This is much much easier and more cost effective than changing to a larger heatsink.
Of course there are more tricks to get it even further down. But that is another story.
So to response to your comment, I, for one, know where to stop - that is when the thermal resistance across the insulator is less than 5% of the total thermal resistance.
Patrick
Aluminium oxide insulators were already made and sold 20 years ago, but the only place i knew where to get them was Germany.
At Bürklin in Düsseldorf they were called Al2O-2 back then, AOS types.
You had silicone insulators back then as well, 0.25mm and twice the thermal resistance of the 3mm thick aluminium oxydes for a TO3.
Nowadays availability may be better but still difficult to obtain for many.
The Al2O3s i got from the US are much thinner. For big Japanese devices the nominal gain by obtaining even lower thermal resistance insulators may not be spectacular.
Insulator material with a lower thermal resistance rating AND an easier way to cut it to size for Japan devices than AL2O3 is however.
For TO247s i'd say it is definitely worth gaining 5 Cs or more by using SOTA insulator stuff.
At Bürklin in Düsseldorf they were called Al2O-2 back then, AOS types.
You had silicone insulators back then as well, 0.25mm and twice the thermal resistance of the 3mm thick aluminium oxydes for a TO3.
Nowadays availability may be better but still difficult to obtain for many.
The Al2O3s i got from the US are much thinner. For big Japanese devices the nominal gain by obtaining even lower thermal resistance insulators may not be spectacular.
Insulator material with a lower thermal resistance rating AND an easier way to cut it to size for Japan devices than AL2O3 is however.
For TO247s i'd say it is definitely worth gaining 5 Cs or more by using SOTA insulator stuff.
Hi GD3,
Fix it right, or don't fix it at all. If the repair gets the show going, then do it right, after the show. 99% of the guys won't.
Don't kid yourself, if you don't repair an amp properly, it will break down again. "Close enough for rock & roll" is the dumest phase I've ever heard. Rates right up there with "don't worry about it".
-Chris
I've had to service plenty of amplifiers where the heatsink and insulator where not cleaned and freshly greased. I've even seen chunks missing from insulators.
Fix it right, or don't fix it at all. If the repair gets the show going, then do it right, after the show. 99% of the guys won't.
Don't kid yourself, if you don't repair an amp properly, it will break down again. "Close enough for rock & roll" is the dumest phase I've ever heard. Rates right up there with "don't worry about it".
-Chris
EUVL said:
Did You experiment with one transistor dissipating about 40W when heatsink became 70-80 degree centigrade warm ? I did.
Silicon pads I had ( not only one sample ) after dismounting looked like slightly burnt. This is my experience maybe other people have better results. What concider Pass clones take into account that original solutions have pair or even more ( up to six ) transistors connected paralely. I have one transistor . These are much more lighter conditions for transistors ( and pads too ) than in my case ( locally transistors do not have to dissipate so much power in time unit than my transistors have to ) and main factor then is not only thermal resistance but thermal conductivity that was discussed earlier by me and other people.
Regards
Jacek
> Did You experiment with one transistor dissipating about 40W when heatsink became 70-80 degree centigrade warm ?
For your reference, my Class A amp has 12x TO247's each dissipating 34W with heatsink temperature at 70°C, as measured and agreed to calculated value to within 2°C. It has been running for about half a year. I have not had any burnt insulators or short circuits, as yet.
And of course I have also tried mica, silpad, Al2O3, Aluminium nitride, kapton, and various types of Kerafol before finalising my thermal concept, before building.
> main factor then is not only thermal resistance but thermal conductivity
Isn't thermal conductivity = 1 / thermal resistance, or you have a different definition ?
Patrick
For your reference, my Class A amp has 12x TO247's each dissipating 34W with heatsink temperature at 70°C, as measured and agreed to calculated value to within 2°C. It has been running for about half a year. I have not had any burnt insulators or short circuits, as yet.
And of course I have also tried mica, silpad, Al2O3, Aluminium nitride, kapton, and various types of Kerafol before finalising my thermal concept, before building.
> main factor then is not only thermal resistance but thermal conductivity
Isn't thermal conductivity = 1 / thermal resistance, or you have a different definition ?
Patrick
anatech said:
Nevermind won't, how about oops I forgot. I couldn't tell you the number of times I did a quick-fix and moved on, forgetting it wasn't 100%.
As far as going too far goes, how about this:
**No interface material: 1, maybe 2 degrees C/W
**Powerstrate: 0.008 degrees C/W
To me, that's hardly splitting straws. More like burning the hayfield. The difference between interface materials when everything is way overbuilt beyond worst-case scenarios may not mean much besides slightly shorter semiconductor life. If the design is constructed according to ideal numbers, it may be the difference between an explosion/fire and perfect operation.
> **Powerstrate: 0.008 degrees C/W
According to their website (Table 1) :
http://www.powerdevices.com/powerstrate.htm
The thermal impedance is quoted as 0.008 degC-sq.in./W, or in metric terms 3W/m/K.
Patrick
According to their website (Table 1) :
http://www.powerdevices.com/powerstrate.htm
The thermal impedance is quoted as 0.008 degC-sq.in./W, or in metric terms 3W/m/K.
Patrick
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.