AndrewT
If the thermal resistance is low (correct) then the device case Tc should be within 5 to 10C degrees of the heat sink temperature.
Measured the clamp/bar, at hottest position it is 60-61°C.
Measured on the HS above fet right at sil-pad measures 55-58°C.
Measured at the junction of bar and fet/case temp is 73°C.
Within acceptable range?
--just measured again, 57°C@HS-silpad, 72°C@bar-fetcase, 60°C@center of bar clamp i.e. between fets
From my Post #2875 on 2-11-09, I see that I measured Temp at tab on mosfet was 88°C. {cannot measure at that point now, bar covers it}
Edit: also that was when I was first testing with two channels on on HS.
-----------------------------------------------------------------------------------
thanh1973
I was just having a look at the heatsink surface. It doesn't look to even/smooth from what I can tell in the picture.
Also what are you using between the mosfet and the heatsink?
Is it some sort of tape or silpad?
If the heatsink surface is too rough and the thermal pad too thin then you not going to get good heat transfer.
This Conrad MF35-151.5 heat sink mounting surface doesn't show well in the picture but it is very smooth and flat.
Yes, silpad, Bergquist BER120-NDHEATPAD TO-247 .006" K10 with "EFD (Providence, RI) Part # 70-AM Heat Sink Compound" between fet & pad. -- Is compound necessary with silpads??
I bought this HS compound about 20 years ago at the Rochester,NY Hamfest, manufacturing surplus I guess?
-----------------------------------------------------------------------------------
npapp
The bar clamp could be the problem. Although it looks nice and substantial I bet it's bending at the screw which will keep the fet from sitting flat. Even if it's not visible to the eye, the inside edge of the fet (near the screw) probably has more pressure on it which is causing the outside edge to lift.
It is best to have a screw on both sides of fet for even pressure.
Yes, no bend visible to the eye. I could be wrong but the bar is ¼ inch thick, 3½ inches long, ¾ inch wide aluminum.
-----------------------------------------------------------------------------------
thanh1973
You could achieve this by putting a whole heap of screws into the clamp. Have a look at the image to get an idea. http://www.newegg.com/Product/Produ...N82E16835108037
Or make some fins to attach to the clamp
In the future I'd like to mount the fets on an aluminum or copper bar spreader to HS.
I just might add something, a fin, to the bar clamp to aid in dissipation.
-----------------------------------------------------------------------------------
tinitus
You can mount a copper plate between transistor and heatsink
Copper plate should be as big as layout permits
Yes, future plan.
-----------------------------------------------------------------------------------
amp-guy
One more thing that can be done is to lap the heatsink surface
As mentioned above Conrad HS leaves nothing to be desired as to a smooth flat surface. I've noticed many builders using this sink seem to be happy.
-----------------------------------------------------------------------------------
Many thanks to all responders.
I greatly appreciate all your advice and considerations.
If the thermal resistance is low (correct) then the device case Tc should be within 5 to 10C degrees of the heat sink temperature.
Measured the clamp/bar, at hottest position it is 60-61°C.
Measured on the HS above fet right at sil-pad measures 55-58°C.
Measured at the junction of bar and fet/case temp is 73°C.
Within acceptable range?
--just measured again, 57°C@HS-silpad, 72°C@bar-fetcase, 60°C@center of bar clamp i.e. between fets
From my Post #2875 on 2-11-09, I see that I measured Temp at tab on mosfet was 88°C. {cannot measure at that point now, bar covers it}
Edit: also that was when I was first testing with two channels on on HS.
-----------------------------------------------------------------------------------
thanh1973
I was just having a look at the heatsink surface. It doesn't look to even/smooth from what I can tell in the picture.
Also what are you using between the mosfet and the heatsink?
Is it some sort of tape or silpad?
If the heatsink surface is too rough and the thermal pad too thin then you not going to get good heat transfer.
This Conrad MF35-151.5 heat sink mounting surface doesn't show well in the picture but it is very smooth and flat.
Yes, silpad, Bergquist BER120-NDHEATPAD TO-247 .006" K10 with "EFD (Providence, RI) Part # 70-AM Heat Sink Compound" between fet & pad. -- Is compound necessary with silpads??
I bought this HS compound about 20 years ago at the Rochester,NY Hamfest, manufacturing surplus I guess?
-----------------------------------------------------------------------------------
npapp
The bar clamp could be the problem. Although it looks nice and substantial I bet it's bending at the screw which will keep the fet from sitting flat. Even if it's not visible to the eye, the inside edge of the fet (near the screw) probably has more pressure on it which is causing the outside edge to lift.
It is best to have a screw on both sides of fet for even pressure.
Yes, no bend visible to the eye. I could be wrong but the bar is ¼ inch thick, 3½ inches long, ¾ inch wide aluminum.
-----------------------------------------------------------------------------------
thanh1973
You could achieve this by putting a whole heap of screws into the clamp. Have a look at the image to get an idea. http://www.newegg.com/Product/Produ...N82E16835108037
Or make some fins to attach to the clamp
In the future I'd like to mount the fets on an aluminum or copper bar spreader to HS.
I just might add something, a fin, to the bar clamp to aid in dissipation.
-----------------------------------------------------------------------------------
tinitus
You can mount a copper plate between transistor and heatsink
Copper plate should be as big as layout permits
Yes, future plan.
-----------------------------------------------------------------------------------
amp-guy
One more thing that can be done is to lap the heatsink surface
As mentioned above Conrad HS leaves nothing to be desired as to a smooth flat surface. I've noticed many builders using this sink seem to be happy.
-----------------------------------------------------------------------------------
Many thanks to all responders.
I greatly appreciate all your advice and considerations.
mosfets may be junk, but they are flat
I had been reading and watching the various F5 conversations for awhile and decided to finally build one . I have got just about all the bits together and started soldering the Daniels boards last night. I was able to score a 28.5x10.25x2.75 in aluminum heatsink. I am sure this will be adequate for a stereo pair and it is in good condition, but its just ugly. I would like to mount the F5 pcb and outputs to a heat spreader so I can have some flexibility in the future just screwing a “component” from one sink to another. The recent posts have been very helpful for planning that.
One thing that has me worried, I did very lightly file, then smooth with sandpaper (100 grit, then wet sand with isop. alcohol and 220 grit) all 4 mosfet outputs, not even once thinking about ESD as mentioned by amp- guy. Should I be OK? Kind of took NP "intimate contact" thing a little to far maybe. I have minimal amp building experience and would like to keep the start up paranoia low as possible.
I had been reading and watching the various F5 conversations for awhile and decided to finally build one . I have got just about all the bits together and started soldering the Daniels boards last night. I was able to score a 28.5x10.25x2.75 in aluminum heatsink. I am sure this will be adequate for a stereo pair and it is in good condition, but its just ugly. I would like to mount the F5 pcb and outputs to a heat spreader so I can have some flexibility in the future just screwing a “component” from one sink to another. The recent posts have been very helpful for planning that.
One thing that has me worried, I did very lightly file, then smooth with sandpaper (100 grit, then wet sand with isop. alcohol and 220 grit) all 4 mosfet outputs, not even once thinking about ESD as mentioned by amp- guy. Should I be OK? Kind of took NP "intimate contact" thing a little to far maybe. I have minimal amp building experience and would like to keep the start up paranoia low as possible.
ichiban said:
Caution
Forgot, but be careful to ispolate mounting screws
Mayb also those ordinary mounting angles could be used that way
One seperate for each transistor
Big isolating pads between angle and heatsink
Would also help to get amp board away from heatsink
I would not attempt this without careful grinding of alu parts, by a professional machinist
A small surface like a transistor is way more easy to make smooth contact
If these added alu/copper parts dont allign perfectly, they will do more harm than good
May NOT be a beginners job
tinitus said:
I would not attempt this without careful grinding of alu parts, by a professional machinist
A small surface like a transistor is way more easy to make smooth contact
If these added alu/copper parts dont allign perfectly, they will do more harm than good
May NOT be a beginners job
I did not want to mess the junction between the fet and the heatsink that is why I went for the copper "wings" on post 4156. They really evened out the heat and lowered the heat on the IC package. I hope someone gives something similar a try and reports back.
Nick
I have seen it used on Holfi amps
Danish, and you may not know them
An alu bar was used across the trannies
It looked a bit suspect to me
Never mind
I would use a long alu/copper bar, attached over the transistors with the normal screw
And maybe one screw on each side of transistor body as well
Carefully adjusted, not to disturb attachment
I would use LONG bars, and attach the to heatsink, as far away from transistors as possible
This way a smooth surface or alignment of the bar would be less significant, as it would be allowed to bend a bit without affecting the transistors
Further, the attachment to heatsink, far away from transistors, would be where there is most cooling available, maybe
But sure, such tricks might be worth to try
But one should be very careful not to introduce elements that could make things worse, as always
Just a thought
Danish, and you may not know them
An alu bar was used across the trannies
It looked a bit suspect to me
Never mind
I would use a long alu/copper bar, attached over the transistors with the normal screw
And maybe one screw on each side of transistor body as well
Carefully adjusted, not to disturb attachment
I would use LONG bars, and attach the to heatsink, as far away from transistors as possible
This way a smooth surface or alignment of the bar would be less significant, as it would be allowed to bend a bit without affecting the transistors
Further, the attachment to heatsink, far away from transistors, would be where there is most cooling available, maybe
But sure, such tricks might be worth to try
But one should be very careful not to introduce elements that could make things worse, as always
Just a thought
AndrewT said:
This is simply dead wrong.
The clamp is per definition going to be warmer than the heatsink.
As the surface of the clamping bar is rather limited, so is the dissipation, hence the bar will be approx. the same temperature as the device. The temperature of the device can very well be more than 10C higher than the heatsink temperature.
thanh1973 said:
No it doesn't.
As the temperature was measured by hand, even 65C is going to feel VERY hot, as in ouch!
So, in conclusion all the facts we have here is:
1) We know that the clamp has little dissipation.
2) We know that the device temperature is above approx. 55C (the ouch! limit for most people)
Magura
Magura said:
No it doesn't.
As the temperature was measured by hand, even 65C is going to feel VERY hot, as in ouch!
So, in conclusion all the facts we have here is:
1) We know that the clamp has little dissipation.
2) We know that the device temperature is above approx. 55C (the ouch! limit for most people)
Magura
I always have fans in my amplifier cases.
One that sucks air in at the front and another blowing air out of the back.
Why things should turn so complicated when micas and white qrease do the job well.
Look at this one.
http://www.6moons.com/audioreviews/firstwatt7/f5_2.html
Look at this one.
http://www.6moons.com/audioreviews/firstwatt7/f5_2.html
ichiban said:AndrewT
If the thermal resistance is low (correct) then the device case Tc should be within 5 to 10C degrees of the heat sink temperature.
Measured the clamp/bar, at hottest position it is 60-61°C.
Measured on the HS above fet right at sil-pad measures 55-58°C.
Measured at the junction of bar and fet/case temp is 73°C.
Within acceptable range?
--just measured again, 57°C@HS-silpad, 72°C@bar-fetcase, 60°C@center of bar clamp i.e. between fets
From my Post #2875 on 2-11-09, I see that I measured Temp at tab on mosfet was 88°C. {cannot measure at that point now, bar covers it}
Edit: also that was when I was first testing with two channels on on HS.
-----------------------------------------------------------------------------------
thanh1973
I was just having a look at the heatsink surface. It doesn't look to even/smooth from what I can tell in the picture.
Also what are you using between the mosfet and the heatsink?
Is it some sort of tape or silpad?
If the heatsink surface is too rough and the thermal pad too thin then you not going to get good heat transfer.
This Conrad MF35-151.5 heat sink mounting surface doesn't show well in the picture but it is very smooth and flat.
Yes, silpad, Bergquist BER120-NDHEATPAD TO-247 .006" K10 with "EFD (Providence, RI) Part # 70-AM Heat Sink Compound" between fet & pad. -- Is compound necessary with silpads??
I bought this HS compound about 20 years ago at the Rochester,NY Hamfest, manufacturing surplus I guess?
-----------------------------------------------------------------------------------
npapp
The bar clamp could be the problem. Although it looks nice and substantial I bet it's bending at the screw which will keep the fet from sitting flat. Even if it's not visible to the eye, the inside edge of the fet (near the screw) probably has more pressure on it which is causing the outside edge to lift.
It is best to have a screw on both sides of fet for even pressure.
Yes, no bend visible to the eye. I could be wrong but the bar is ¼ inch thick, 3½ inches long, ¾ inch wide aluminum.
-----------------------------------------------------------------------------------
thanh1973
You could achieve this by putting a whole heap of screws into the clamp. Have a look at the image to get an idea. http://www.newegg.com/Product/Produ...N82E16835108037
Or make some fins to attach to the clamp
In the future I'd like to mount the fets on an aluminum or copper bar spreader to HS.
I just might add something, a fin, to the bar clamp to aid in dissipation.
-----------------------------------------------------------------------------------
tinitus
You can mount a copper plate between transistor and heatsink
Copper plate should be as big as layout permits
Yes, future plan.
-----------------------------------------------------------------------------------
amp-guy
One more thing that can be done is to lap the heatsink surface
As mentioned above Conrad HS leaves nothing to be desired as to a smooth flat surface. I've noticed many builders using this sink seem to be happy.
-----------------------------------------------------------------------------------
Many thanks to all responders.
Hello Ichiban,
Good effort for measuring the temps.
Silicon grease/ thermal compound is unecessary with silpads. and will add a finite amount of thermal resistance.
in between your clamp and the mosfet however it may prove beneficial.
were the temperatures measured at thermal equilibrium?
-Dan
Heatsinks:
If the clamp is massive and has little dissipation then it works as a heat accumulator and Tº will build-up. Not good.
If you think (the owner of the amp) you need more heatsinking, why not try a little a frontal heatsink for heach Mosfet. Something like:
http://www.futurlec.com/Heatsinks/TO3B.shtml
Just an idea.
Good luck.
M.
If the clamp is massive and has little dissipation then it works as a heat accumulator and Tº will build-up. Not good.
If you think (the owner of the amp) you need more heatsinking, why not try a little a frontal heatsink for heach Mosfet. Something like:
http://www.futurlec.com/Heatsinks/TO3B.shtml
Just an idea.
Good luck.
M.
I'll try to make a neutral remark.
Thermal flow (Power in W) is analogous to electrical current (A). So consider the heat dissipation in the device as a current source, and the atmosphere is Ground. Temperature difference is like voltage, thermal resistance (K/W) is like resistance (ohm)
The heatsink itself has a certain thermal resistance (in K/W). So for a certain dissipation, it results in a certain temperature rise. This seems obvious, but many people do not realise that heatsink is a VERY non-linear resistor. The figures given by many heatsink manufacturers are based on certain assumptions (e.g. natural convection, vertical fins, black anodised, .... ), but the most important of all -- temperature rise. Many manufacturers quote figures based on 40°C or even 60°C temperature rise, and you will be surprised how much more resistance you have when reducing to 25°C temperature rise.
In a normal layout, you then have the insulation (e.g. Mica & thermal grease), the metal backing of the device (back side), a 0.1mm thick silver epoxy layer the size of the die, and the die itself (0.8mm thick silicon). Now the heat is generated in the die, and 0.8mm silicon has next to no thermal resistance. So do your thermal resistance calculations, and you will find out that the highest resistance in the series are the silver epoxy (because of its small area) inside the transistor, and the insulation (because of its low thermal conductivity). The former you have to live with (unless you use transistors in parallel and split the bias current between them), and the latter you can do something about, but also somewhat limited for most people who wants to use commercial stuff and not go to the extreme.
So what happens if you put a copper clamp, like shown in the last page ?
You can ignore the thermal storage effect if you just wish to consider final steady state temperature, because thermal capacity only affects time constant and not resistance.
The clamp does offer a parallel path for thermal flow. And this parallel chain, in parallel to the one through the metal backing of the transistor, now consists of the silicon die, 3mm of encapsulation compound on the top part of the transistor (the black plastic stuff with the markings), a thin layer of grease between encapsulation and the copper clamp, the clamp itself, the contact resistance between the clamp and the heatsink, and then the heatsink.
Guess which is the highest resistance -- 3mm of encapsulation. The rest you can ignore in comparison.
Current always prefers to flow through the path with the least resistance. So I bet they want to go through 0.1mm of mica plus 0.1mm of silver epoxy, than 3mm of encapsulation compound.
Not that it hurts, but it is not going to solve your temperature problem. But as ichiban already said, it is a very good way for fixation. And I always use clamps (see photos of my balanced F5). Note that I use 2 M3 screws to clamp.
Using mica & grease, you probably get something like 8°C difference between transistor (metal tag) and heatsink for 25W on a TO247 package. If you are ready to take risk, follow Nelson and get rid of the mica. Just use the anodisied layer of the heatsink as electrical insulation. But you better be sure there are no scratches. You might get down to something like 4-5°C then, using thermal grease btw transistor and heatsink. Want some more risk, use arctic silver in place of grease. That will get you another 1°C or so.
The best I have acheived so far is 2°C difference between transistor (metal tag) and heatsink at 32W. But it is way too complicated.
Kerafol will get you down to 5°C or less, without all those risks.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=144076
http://www.diyaudio.com/forums/showthread.php?postid=1032023#post1032023
http://www.diyaudio.com/forums/showthread.php?postid=1034956#post1034956
Patrick
Thermal flow (Power in W) is analogous to electrical current (A). So consider the heat dissipation in the device as a current source, and the atmosphere is Ground. Temperature difference is like voltage, thermal resistance (K/W) is like resistance (ohm)
The heatsink itself has a certain thermal resistance (in K/W). So for a certain dissipation, it results in a certain temperature rise. This seems obvious, but many people do not realise that heatsink is a VERY non-linear resistor. The figures given by many heatsink manufacturers are based on certain assumptions (e.g. natural convection, vertical fins, black anodised, .... ), but the most important of all -- temperature rise. Many manufacturers quote figures based on 40°C or even 60°C temperature rise, and you will be surprised how much more resistance you have when reducing to 25°C temperature rise.
In a normal layout, you then have the insulation (e.g. Mica & thermal grease), the metal backing of the device (back side), a 0.1mm thick silver epoxy layer the size of the die, and the die itself (0.8mm thick silicon). Now the heat is generated in the die, and 0.8mm silicon has next to no thermal resistance. So do your thermal resistance calculations, and you will find out that the highest resistance in the series are the silver epoxy (because of its small area) inside the transistor, and the insulation (because of its low thermal conductivity). The former you have to live with (unless you use transistors in parallel and split the bias current between them), and the latter you can do something about, but also somewhat limited for most people who wants to use commercial stuff and not go to the extreme.
So what happens if you put a copper clamp, like shown in the last page ?
You can ignore the thermal storage effect if you just wish to consider final steady state temperature, because thermal capacity only affects time constant and not resistance.
The clamp does offer a parallel path for thermal flow. And this parallel chain, in parallel to the one through the metal backing of the transistor, now consists of the silicon die, 3mm of encapsulation compound on the top part of the transistor (the black plastic stuff with the markings), a thin layer of grease between encapsulation and the copper clamp, the clamp itself, the contact resistance between the clamp and the heatsink, and then the heatsink.
Guess which is the highest resistance -- 3mm of encapsulation. The rest you can ignore in comparison.
Current always prefers to flow through the path with the least resistance. So I bet they want to go through 0.1mm of mica plus 0.1mm of silver epoxy, than 3mm of encapsulation compound.
Not that it hurts, but it is not going to solve your temperature problem. But as ichiban already said, it is a very good way for fixation. And I always use clamps (see photos of my balanced F5). Note that I use 2 M3 screws to clamp.
Using mica & grease, you probably get something like 8°C difference between transistor (metal tag) and heatsink for 25W on a TO247 package. If you are ready to take risk, follow Nelson and get rid of the mica. Just use the anodisied layer of the heatsink as electrical insulation. But you better be sure there are no scratches. You might get down to something like 4-5°C then, using thermal grease btw transistor and heatsink. Want some more risk, use arctic silver in place of grease. That will get you another 1°C or so.
The best I have acheived so far is 2°C difference between transistor (metal tag) and heatsink at 32W. But it is way too complicated.
Kerafol will get you down to 5°C or less, without all those risks.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=144076
http://www.diyaudio.com/forums/showthread.php?postid=1032023#post1032023
http://www.diyaudio.com/forums/showthread.php?postid=1034956#post1034956
Patrick
Interesting, thanks.
Sorry Magura
Andrew was not dead wrong. He was slightly wrong.
The only thing that he said that was wrong, was when he said that "the clamping bar should be much cooler than the device case"
And even this comment although wrong was based on the real possibility that there was poor thermal contact between the rough plastic case and the copper bar as he explained.
Remember his comments were made before any real measurements were made.
His comments were made based on finger measurement ie "hot or bloody hot".
Your comments were made after proper temperature measurements were performed.
So I think you are being a little bit hasty to criticise here.
Andrew was not dead wrong. He was slightly wrong.
The only thing that he said that was wrong, was when he said that "the clamping bar should be much cooler than the device case"
And even this comment although wrong was based on the real possibility that there was poor thermal contact between the rough plastic case and the copper bar as he explained.
Remember his comments were made before any real measurements were made.
His comments were made based on finger measurement ie "hot or bloody hot".
Your comments were made after proper temperature measurements were performed.
So I think you are being a little bit hasty to criticise here.
like I said, you have a problem. A very serious problem.ichiban said:
Tc ~88degC, the temperature of the tab.
Ts ~57degC the temp of the sink underneath the device.
You have 31Cdegrees of temperature difference bewteen sink and device.
This indicates that the thermal resistance, Rth s-a, is enormous.
Magura,
if you believe what you posted in reply to my post, then you are never going to understand thermodynamics in it's simplest form.