Now just be careful that you don't inadvertently build yourself a thermal barrier. Every surface must transfer heat to the next layer of the sandwich, ie IC body to chassis to heatsink to air.
If any of the layers of the sandwich are a poor conductor of heat the heatsinking will be compromised.
If the case is steel for example, don't sandwich the case between the IC and the heatsink, cut a hole and mount the IC directly on the heatsink.
Tom
I think you have misunderstood my explanation.
National's guidance uses Tj = Tj max = 150°C.
They adopt worst case loading for a resistive load.
The chip cannot destroy itself in the conditions adopted by National.
If the temperature of ambient is lower than that adopted in the National worst case resistive loading, then the WHOLE amplifier and it's sink runs at a LOWER temperature.
Ta and Ts and Tc and Tj, all get reduced by the same difference.
There are two situations that Nat do not address,
a.) reactive loading.
b.) deliberate clipping of the signal.
Sensible operation avoids b.
The effects of a. are reduced by operating at lower temperatures and by taking stability enhancement procedures to avoid oscillation/ringing, i.e. adopt most of the optional stability components.
Last edited:
Andrew - I think a little explanation of your giant leap might help him.
I'll leap one stage further and then retrace a few lessons.
Tj - This is the all important parameter is the temperature of the chip die, it is not possible to measure as it is inside the IC package. This is generally assumed to be 200 degrees C max but is reduced for reliability reasons.
Ta - This is the ambient temperature in which the heatsink may operate. If you assume 40 degrees C you will allow for most conditions.
Ѳjc - This is the thermal resistance from the IC die (junction) to the case of the IC.
Ѳcs - This is the thermal resistance from the IC case to the heatsink.
Ѳsa - This is the thermal resistance from the heatsink to the air. This is the figure quoted for the heatsink in its datasheet
Ѳcc - This is thermal resistance of any additional layers of the thermal sandwich.
FORMULA
Tj = Ta + (Ѳjc + Ѳcs + Ѳcc + Ѳsa)P
Where (P) is the power being dissipated by the IC.
Ѳjc can be obtained from the datasheet but is generally in the order of 1.5 deg C/W.
Ѳcs is down to the insulating material between the IC and the heatsink and is generally in the order of 1 deg C/W.
Lets assume that you have also sandwiched a plain alluminium chassis into the equation so Ѳcc can be about 2 deg C/W.
Now let's rearrange the formula allowing the LM3886 to dissipate 50W of heat and keeping Tj to a more reliable 150 degrees C.
(Tj - Ta) / P = Ѳjc + Ѳcs + Ѳcc + Ѳsa
or
(150 - 40) / 50 = The thermal sandwich = 2.2 deg C/W
Bear in mind this is for each IC.
Now 2.2 deg C/W isn't a tiny heatsink. BUT, you've still got to allow for the sandwich.
Ѳsa = 2.2 - Ѳcs (1 deg C/W) - Ѳcc (2 deg C/W) - Ѳjc (1.5 deg C/W).
You can see that it doesn't compute - the IC cannot run at Tj = 150 degrees.
Let's get rid of Ѳcc and mount the chip directly onto the heatsink with an insulating washer and allow Tj = 200 degrees C.
(200-40)/50 = 3.2 deg C/W
Optimally Ѳsa still needs to be 0.7 deg C/W.
If it's a cool day and the heatsink is only at 20 degrees C then things become more realistic. (200-20)/50 = 3.6 deg C/W. Minus the junctions gives you a realistic 1.1 deg C/W heatsink.
OK, these figures are unrealistic because the IC will never be delivering a pure sine wave into a reactive load at full power continuously.
What they do show is how important the heasinking is to the poor silicon junction inside that plastic case.
I'll leap one stage further and then retrace a few lessons.
Tj - This is the all important parameter is the temperature of the chip die, it is not possible to measure as it is inside the IC package. This is generally assumed to be 200 degrees C max but is reduced for reliability reasons.
Ta - This is the ambient temperature in which the heatsink may operate. If you assume 40 degrees C you will allow for most conditions.
Ѳjc - This is the thermal resistance from the IC die (junction) to the case of the IC.
Ѳcs - This is the thermal resistance from the IC case to the heatsink.
Ѳsa - This is the thermal resistance from the heatsink to the air. This is the figure quoted for the heatsink in its datasheet
Ѳcc - This is thermal resistance of any additional layers of the thermal sandwich.
FORMULA
Tj = Ta + (Ѳjc + Ѳcs + Ѳcc + Ѳsa)P
Where (P) is the power being dissipated by the IC.
Ѳjc can be obtained from the datasheet but is generally in the order of 1.5 deg C/W.
Ѳcs is down to the insulating material between the IC and the heatsink and is generally in the order of 1 deg C/W.
Lets assume that you have also sandwiched a plain alluminium chassis into the equation so Ѳcc can be about 2 deg C/W.
Now let's rearrange the formula allowing the LM3886 to dissipate 50W of heat and keeping Tj to a more reliable 150 degrees C.
(Tj - Ta) / P = Ѳjc + Ѳcs + Ѳcc + Ѳsa
or
(150 - 40) / 50 = The thermal sandwich = 2.2 deg C/W
Bear in mind this is for each IC.
Now 2.2 deg C/W isn't a tiny heatsink. BUT, you've still got to allow for the sandwich.
Ѳsa = 2.2 - Ѳcs (1 deg C/W) - Ѳcc (2 deg C/W) - Ѳjc (1.5 deg C/W).
You can see that it doesn't compute - the IC cannot run at Tj = 150 degrees.
Let's get rid of Ѳcc and mount the chip directly onto the heatsink with an insulating washer and allow Tj = 200 degrees C.
(200-40)/50 = 3.2 deg C/W
Optimally Ѳsa still needs to be 0.7 deg C/W.
If it's a cool day and the heatsink is only at 20 degrees C then things become more realistic. (200-20)/50 = 3.6 deg C/W. Minus the junctions gives you a realistic 1.1 deg C/W heatsink.
OK, these figures are unrealistic because the IC will never be delivering a pure sine wave into a reactive load at full power continuously.
What they do show is how important the heasinking is to the poor silicon junction inside that plastic case.
Last edited:
If using the National datasheet as your guide to heatsink size, then National have done all the arithmetic (and some mathematics) for you.
After you have done that, then just double the size of the heatsink.
Simple arithmetic.
After you have done that, then just double the size of the heatsink.
Simple arithmetic.
When TI took over Nat Semi, they didn't take "The Overture Design Guide" with them (at least I can't find it on TI's site) -- it's just an Excel spreadsheet which yields the maximum thermal impedance for specified load and rail voltage. Anyone needing a copy PM me with your email address.
ok so a quick follow up, having build the amp.
i used the aforementioned 0.9c/watt heatsink (plus another 1.7c/watt heatsink of the same design but shorter, to fill the rest of the case top nicely)
i cut through the case and mounted directly to the heatsink, as i was unable to get a prefect contact between the case and sink, as the case top had a slight bow in it.
i did fill any gaps as well as possible with keratherm heat paste, but basically the transfer from the heatsink to the case and second heatsink is likely limited by this.
anyway, with no load, the sink (and case) get a bit warm to the touch, maybe body temperature, after a few hours.
with music playing louder than i normally listen (max volume on ipod, 75% on vol pot of amp) , the sink got quite warm ( but not hot) after an hour of listening.
no sound degradation noticed.
im sure ill test further, but those are my preliminary results. seems for my needs at least, 0.9 c/watt is fine.
if i do ever have a party, a pair of 120mm pc case fans will sit nicely on top of the heatsink, and i have a spare 8v feed from the transformer i can use.
cheers for all the advice, Robin.
i used the aforementioned 0.9c/watt heatsink (plus another 1.7c/watt heatsink of the same design but shorter, to fill the rest of the case top nicely)
i cut through the case and mounted directly to the heatsink, as i was unable to get a prefect contact between the case and sink, as the case top had a slight bow in it.
i did fill any gaps as well as possible with keratherm heat paste, but basically the transfer from the heatsink to the case and second heatsink is likely limited by this.
anyway, with no load, the sink (and case) get a bit warm to the touch, maybe body temperature, after a few hours.
with music playing louder than i normally listen (max volume on ipod, 75% on vol pot of amp) , the sink got quite warm ( but not hot) after an hour of listening.
no sound degradation noticed.
im sure ill test further, but those are my preliminary results. seems for my needs at least, 0.9 c/watt is fine.
if i do ever have a party, a pair of 120mm pc case fans will sit nicely on top of the heatsink, and i have a spare 8v feed from the transformer i can use.
cheers for all the advice, Robin.
The heatsink is only too hot if you cant hold your hand on it for more than a few seconds. If it feels fookin hot but you can hold your hand on it - that's OK.
For my hands and pain threshold, that's about 60 ºC. I adopt that as safe for an external heat sink, but I wouldn't want it hotter than that.
~Tom
For the past year+ I've been running a 3886 Amp on 36v rails.
Heat sinks, one per 3886, are 6"x 3" x 1/8" thick vertically mounted black flat ali plates.
Internal to the case... which is perforated mesh, in Tube amp style
These things have.. Not Yet!... gotten past slightly warmer than ambient.
Albeit using 94 db speakers, often at half pot/volume rotation 🙂 my 87db ones, although being smaller drivers do similar.
In MINE... heat has proved a non issue... thankfully 🙂
An initial issue of concern.. as it was suggested the heat sinking was marginal...
it seems as not the case .. in mine.
Just a differing pov. 🙂
Heat sinks, one per 3886, are 6"x 3" x 1/8" thick vertically mounted black flat ali plates.
Internal to the case... which is perforated mesh, in Tube amp style
These things have.. Not Yet!... gotten past slightly warmer than ambient.
Albeit using 94 db speakers, often at half pot/volume rotation 🙂 my 87db ones, although being smaller drivers do similar.
In MINE... heat has proved a non issue... thankfully 🙂
An initial issue of concern.. as it was suggested the heat sinking was marginal...
it seems as not the case .. in mine.
Just a differing pov. 🙂
Last edited:
yes given all the stressing over undersized heatsinks etc.. mine has been cranked up to max for several hours, with 35v rails (24v toroid) 6 ohm 87 db mission speakers.. heatsink is very warm but not hot. no sound issues to report. and this is in 30 degree ambient, heatsink mounted horizontally (not ideal) with the amp in a small wooden cupboard (door open)
seems from my experience that the theoretical requirements are way above my real world needs. wonder how many watts im actually pushing.
i need batteries for my digital thermometer but ill be interested to take some measurements when i get a chance.
seems from my experience that the theoretical requirements are way above my real world needs. wonder how many watts im actually pushing.
i need batteries for my digital thermometer but ill be interested to take some measurements when i get a chance.
its as i mentioned.. both chips bolted thru a hole in the case to the 0.9 c watt heatsink.
rest of heatsink is bolted to the case with heat transfer paste in the tiny gap. case is 3mm aluminium. for mainly cosmetic reasons i have another section of heatsink, with a rating of 1.7 deg watt bolted to the case top too (also done with heat transfer paste). this is so the entire top is covered in a matching heatsink, instead of having a big gap at the front.
obviously the 0.9 c watt heatsink will be doing most of the work, especially since the connection between the heatsinks and case surface isnt perfect (bow of 0.2 mm or so in case surface that i couldnt completely remove with bolt tightening) .
having said that, the case and front heatsink, while not as warm as the main heatsink, still warm up significantly, so they are contributing somewhat to dissipation.
rest of heatsink is bolted to the case with heat transfer paste in the tiny gap. case is 3mm aluminium. for mainly cosmetic reasons i have another section of heatsink, with a rating of 1.7 deg watt bolted to the case top too (also done with heat transfer paste). this is so the entire top is covered in a matching heatsink, instead of having a big gap at the front.
obviously the 0.9 c watt heatsink will be doing most of the work, especially since the connection between the heatsinks and case surface isnt perfect (bow of 0.2 mm or so in case surface that i couldnt completely remove with bolt tightening) .
having said that, the case and front heatsink, while not as warm as the main heatsink, still warm up significantly, so they are contributing somewhat to dissipation.
Can you imagine what temperatures are inside the chips with your current cooling?
Now extrapolate that to what would be occurring with only the 1.7C/W sink serving the two chipamps.
The Spike protection would be entering current limiting regularly due to the SOAR detection built into these chips. And that would be without getting near Tj overtemp shutdown.
Now extrapolate that to what would be occurring with only the 1.7C/W sink serving the two chipamps.
The Spike protection would be entering current limiting regularly due to the SOAR detection built into these chips. And that would be without getting near Tj overtemp shutdown.
the chips are fine, not too hot to touch, even after running the amp loud for some hours. why should there be any fundamental problem?
ive got the chips bolted directly to a 0.9 c watt heatsink with expensive thermal paste.. it was already concluded that this would likely be sufficient for normal loads, which mine is. i also have (although with questionable thermal contact) a load more heatsinking in the form of another 50% of heatsink, and a thick aluminium extruded case. heat is obviously being transferred to those with some effectiveness as they warm up along with the heatsink, just a bit less.
im not arguing with the maths, my heatsink is less than the recommended amount for full rated output.. but its probably not far off..
just saying maybe im not pushing the amp as hard as feared.. maybe my idea of loud music differs from others... maybe my music is putting significantly less load on the chips than a pure sinewave.. this is the key i think. and the reason that i have half the people telling me their amp is fine with less heatsinking than ive got, including the kit designer (brianGT, who seems well respected here.) and the other half telling me my amp is gonna melt, and to do the worst case calculations.
my amp works fine. even cranked up. it doesnt get too hot, even with the conditions mentioned before. i cant hear any degradation in quality that i might attribute to the spike protection.
thats the only thing that matters imho.
ive got the chips bolted directly to a 0.9 c watt heatsink with expensive thermal paste.. it was already concluded that this would likely be sufficient for normal loads, which mine is. i also have (although with questionable thermal contact) a load more heatsinking in the form of another 50% of heatsink, and a thick aluminium extruded case. heat is obviously being transferred to those with some effectiveness as they warm up along with the heatsink, just a bit less.
im not arguing with the maths, my heatsink is less than the recommended amount for full rated output.. but its probably not far off..
just saying maybe im not pushing the amp as hard as feared.. maybe my idea of loud music differs from others... maybe my music is putting significantly less load on the chips than a pure sinewave.. this is the key i think. and the reason that i have half the people telling me their amp is fine with less heatsinking than ive got, including the kit designer (brianGT, who seems well respected here.) and the other half telling me my amp is gonna melt, and to do the worst case calculations.
my amp works fine. even cranked up. it doesnt get too hot, even with the conditions mentioned before. i cant hear any degradation in quality that i might attribute to the spike protection.
thats the only thing that matters imho.
With the 0.9C/W and partial use of the extra 1.7C/W your temperatures are fine.
A thinking exercise to what could be happening if you simply adopted the National guidelines.
Not an invitation to try, just think it through.
National guideline says for +-35V 8r load and Ta=25°C use 3C/W for their chip.
You have two so that equates to 1.5C/W
For Ta=30°C probably need to increase this to 1.4C/W
That is considerably less then you have fitted.
not
A thinking exercise to what could be happening if you simply adopted the National guidelines.
Not an invitation to try, just think it through.
National guideline says for +-35V 8r load and Ta=25°C use 3C/W for their chip.
You have two so that equates to 1.5C/W
For Ta=30°C probably need to increase this to 1.4C/W
That is considerably less then you have fitted.
not
Last edited:
sorry crossed wires, i though you were suggesting i was still short of heatsinking.
however when i say "my heatsinking is less than the recommended amount for full rated power output, i was going more by
Neurochrome.com : : Audio : Taming the LM3886 Chip Amplifier
which makes the datasheet recommendations look tame.
however when i say "my heatsinking is less than the recommended amount for full rated power output, i was going more by
Neurochrome.com : : Audio : Taming the LM3886 Chip Amplifier
which makes the datasheet recommendations look tame.
Maybe so.. i read the heatsinking section several times, and the only final calculation i can see for two isolated chips is the requirement for a heatsink of 0.44c/watt when using 25v rails.. mine are 35, but the load is 6 ohm not 4. What did i miss?
- Status
- Not open for further replies.