Did you look at the datasheet and determine from the plot, what the chipamp dissipation is for a 3886 on +-35Vdc, when outputting 10W into 8r0 dummy load resistor?
Then try to estimate the chip dissipation for 5W of output and again for 2.5W of output.
In all of these cases the chipamp dissipation must be greater than the Pq value for zero output power. 60mA from +-35Vdc is ~4.2W
The lowest point in the plot must be around 4.2W when output power into a resistive load is zero, the left hand edge of the plot.
Then try to estimate the chip dissipation for 5W of output and again for 2.5W of output.
In all of these cases the chipamp dissipation must be greater than the Pq value for zero output power. 60mA from +-35Vdc is ~4.2W
The lowest point in the plot must be around 4.2W when output power into a resistive load is zero, the left hand edge of the plot.
After you have done that, try for 2W into 8r0 from a +-37Vdc powered chipamp with 60mA of quiescent current.
Then add on for 1W during the same conditions.
I estimate the chipamps had to be dissipating more than 10W each, to raise the sink by more than 30C degrees. And that for outputting a total of just 3W into resistor loads.
Music into reactive loads will be worse.
The only advantage to music reproduction are the gaps between the signal peaks where the dissipation drops back to very close to, but slightly above the quiescent dissipation (Pq) of the two chipamps.
Then add on for 1W during the same conditions.
I estimate the chipamps had to be dissipating more than 10W each, to raise the sink by more than 30C degrees. And that for outputting a total of just 3W into resistor loads.
Music into reactive loads will be worse.
The only advantage to music reproduction are the gaps between the signal peaks where the dissipation drops back to very close to, but slightly above the quiescent dissipation (Pq) of the two chipamps.
Yep, the previous poster is absolutely correct.
A class AB amplifier dissipates a lot more into the heatsink than it outputs at low power levels.
This is taken from the LM3886 datasheet. At about 2 watts into 4 Ohms on +/-35 volt rails, the IC will be dissipating something in the region of 20W.
This will make most heatsinks warm, and a small one very hot indeed!
Don't confuse "output power" with device dissipation...
A class AB amplifier dissipates a lot more into the heatsink than it outputs at low power levels.
This is taken from the LM3886 datasheet. At about 2 watts into 4 Ohms on +/-35 volt rails, the IC will be dissipating something in the region of 20W.
This will make most heatsinks warm, and a small one very hot indeed!
Don't confuse "output power" with device dissipation...
yes i did look.. if im looking at the right graph its about 28w dissipation for 10w, about 22w for 5w and not much less for 2.5w.. gets a bit bunched up in the graph there. i understand the concept and the minimum consumption etc.
i also notice the dissipation peaks at around 35w when the output is 40w, and then decreases again as the output increases beyond that.. this i admit makes no sense to me.
believe me i accept this need for a large heatsink.. im not going to argue with the datasheet or the brains here... it just seems a lot more heatsink is necessary than people generally use, or the kit designers advise. if id realised (planned properly) this was the case id have probably chosen a different route.
my planning went as far as:
1; this case was fine (cool to the touch) with a 90wpc t-amp at 95% efficiency
2; chipamps are not as efficient but still quite good..circa 80 something percent (read on a thread here somewhere)
3, this chipamp is a bit more than half the output of my t-amp
4 my case should be fine.
obviously this was slightly flawed reasoning.
i also notice the dissipation peaks at around 35w when the output is 40w, and then decreases again as the output increases beyond that.. this i admit makes no sense to me.
believe me i accept this need for a large heatsink.. im not going to argue with the datasheet or the brains here... it just seems a lot more heatsink is necessary than people generally use, or the kit designers advise. if id realised (planned properly) this was the case id have probably chosen a different route.
my planning went as far as:
1; this case was fine (cool to the touch) with a 90wpc t-amp at 95% efficiency
2; chipamps are not as efficient but still quite good..circa 80 something percent (read on a thread here somewhere)
3, this chipamp is a bit more than half the output of my t-amp
4 my case should be fine.
obviously this was slightly flawed reasoning.
switching power amplifiers may get up to around 95% (not many can get that high) efficiency when delivering full power.
They, just like non switching power amplifiers, have VERY much LOWER efficiencies when output power is below maximum.
Amplifiers reproducing music and audio almost never operate at full power.
Amplifiers playing music and audio through speakers almost never see resistance as the load.
All (non ClassA) amplifiers have pretty much similar efficiencies when outputting <5% of maximum power into reactive loads.
Expect 1% to 5% efficiency from the power amp when reproducing music and maybe similar efficiencies for the speakers converting the electrical energy into acoustic energy.
Who mentions the overall efficiency of typical audio systems (including radios, TVs and computers) in converting electrical energy into acoustic energy? How many times have you seen data being published? Does any Member have a link or two?
Listening to music can never be efficient. Very different from heating your house with old fashioned incandescent light bulbs where efficiency is 100%
All the energy delivered to the bulb eventually comes out as heat. Is that why the authorities banned them?
They made everything else look bad?
They, just like non switching power amplifiers, have VERY much LOWER efficiencies when output power is below maximum.
Amplifiers reproducing music and audio almost never operate at full power.
Amplifiers playing music and audio through speakers almost never see resistance as the load.
All (non ClassA) amplifiers have pretty much similar efficiencies when outputting <5% of maximum power into reactive loads.
Expect 1% to 5% efficiency from the power amp when reproducing music and maybe similar efficiencies for the speakers converting the electrical energy into acoustic energy.
Who mentions the overall efficiency of typical audio systems (including radios, TVs and computers) in converting electrical energy into acoustic energy? How many times have you seen data being published? Does any Member have a link or two?
Listening to music can never be efficient. Very different from heating your house with old fashioned incandescent light bulbs where efficiency is 100%
All the energy delivered to the bulb eventually comes out as heat. Is that why the authorities banned them?
They made everything else look bad?
Last edited:
hehe now im even more confused.
if efficiencies are broadly similar between chipamps and t-amps,how is it possible that my ta2022 90wpc t-amp was completely happy in the case under discussion, without added heatsinking, barely getting warm even after running at relatively high loads for extended periods (in a cupboard) , but the chipamps of max 68 wpc need vastly more cooling..?
is there something other than electrical efficiency to consider?
i will add the t-amp did burn out spectacularly, but that was after being left on for 4 days with no source attached, so i assume it wasnt a heatsinking issue.
if efficiencies are broadly similar between chipamps and t-amps,how is it possible that my ta2022 90wpc t-amp was completely happy in the case under discussion, without added heatsinking, barely getting warm even after running at relatively high loads for extended periods (in a cupboard) , but the chipamps of max 68 wpc need vastly more cooling..?
is there something other than electrical efficiency to consider?
i will add the t-amp did burn out spectacularly, but that was after being left on for 4 days with no source attached, so i assume it wasnt a heatsinking issue.
http://www.datasheet-pdf.com/datasheet-html/L/M/3/LM3886_NationalSemiconductor.pdf.html
On Page 11 you will see that the worst case scenario is the chip dissipating nearly 50W of heat.
A 1 degrees C/Watt heatsink will rise 50 degrees above the ambient temperature of the room. If your heatsink is sitting at 25 degrees C the chip will heat it a further 50 degrees above that = 75 degrees C.
A 0.5 degrees C/Watt heatsink will rise only 25 degrees above the ambient temperature.
A heatsink can never be too big but it can easilly be too small.
For reliabilty we generally try to keep the case of the IC below 65 degrees.
On Page 11 you will see that the worst case scenario is the chip dissipating nearly 50W of heat.
A 1 degrees C/Watt heatsink will rise 50 degrees above the ambient temperature of the room. If your heatsink is sitting at 25 degrees C the chip will heat it a further 50 degrees above that = 75 degrees C.
A 0.5 degrees C/Watt heatsink will rise only 25 degrees above the ambient temperature.
A heatsink can never be too big but it can easilly be too small.
For reliabilty we generally try to keep the case of the IC below 65 degrees.
Last edited:
I understand ( i should by now 😉 that i need a big heatsink for these chipamps.. reading the datasheet again wont help. I decided to read the datasheet for the ta2022 instead to see why that ran fine in my case. Must be significantly more efficient than the 3886 as it only dissipates 32w of heat running a total of 160w output power.
32W dissipated heat for 160W of output (83%) compares to the 3886 @ 35W of dissipated heat for 85W of output (71%).
salvage a computer heatsink.
the 95 watt tdp rated cpu heatsinks should keep an lm3886 cool.
most probably one can get them for quite cheap 🙂
the trick to enhance them is to use acid to beef up the heatsink.
do take care if you do it, at most petrol stations here sulphuric acid -to refill lead-acid battery- can be bought. the same acid works to get eloxated aluminium.
needs a plastic container, make sure only the fins get in contact with the acid.
and make sure the face of the heatsink -where it will be in contact with the chip- is not.
do make sure to properly polish the face of the heatsink and the chip heatsink tab.
then use computer heat "goo"/paste to even the contact surface even more.
a simple bimetallic switch can switch the fan on/off if needed, the ones rated for 65C will do just fine.
better option to get some 125w tdp rated cpu heatsinks. those are bigger.
the 95 watt tdp rated cpu heatsinks should keep an lm3886 cool.
most probably one can get them for quite cheap 🙂
the trick to enhance them is to use acid to beef up the heatsink.
do take care if you do it, at most petrol stations here sulphuric acid -to refill lead-acid battery- can be bought. the same acid works to get eloxated aluminium.
needs a plastic container, make sure only the fins get in contact with the acid.
and make sure the face of the heatsink -where it will be in contact with the chip- is not.
do make sure to properly polish the face of the heatsink and the chip heatsink tab.
then use computer heat "goo"/paste to even the contact surface even more.
a simple bimetallic switch can switch the fan on/off if needed, the ones rated for 65C will do just fine.
better option to get some 125w tdp rated cpu heatsinks. those are bigger.
Not exactly, don't mix power into the load with chip dissipated power.
But anyway, let's go to the datasheet itself.
I chose some realistic supply and load values, and it states that for +/-30V rails, 8 ohms load, 40 degC ambient temperature (25C is unrealistic) , *maximum* dissipation requires a heat sink with an (interpolated value) halfway between 3.2 and 4.3 degC/W.
Let's say 3 degC/W per chip should be ample.
The datasheet does not specify "Music Program", "IHF" or any such "lighter" spec; coming from no nonsense TI Engineers I consider that continuous sinewave power into the load, unless specifically told otherwise.
So I can guess that with your case and the muysic proigram I think you'll listen to, you'll be fine.
Add a 75C thermal switch bolted to the chassis, between both chips, for peace of mind.
Maybe that explains the ton of properly working amplifiers with simple flat aluminum heatsinks, usually the chassis itself.
I can not understand how a 1C/W heatsink becomes "untouchable" after 1 minute 2 1W into the load, unless said amp was fiercely oscillating, or some other gross explanation.
As of DIYAudio itself, it's customary to find way overspec'd parts in use.
Nothing against, of course, unless it instills (undeserved) fear in users who shy away from building stuff which is really within their reach.
There's a chipamp gallery in the Forum, look at ait to have some ideas.
Also look at *commercial* equipment using LM3886 🙂
But anyway, let's go to the datasheet itself.
I chose some realistic supply and load values, and it states that for +/-30V rails, 8 ohms load, 40 degC ambient temperature (25C is unrealistic) , *maximum* dissipation requires a heat sink with an (interpolated value) halfway between 3.2 and 4.3 degC/W.
Let's say 3 degC/W per chip should be ample.
The datasheet does not specify "Music Program", "IHF" or any such "lighter" spec; coming from no nonsense TI Engineers I consider that continuous sinewave power into the load, unless specifically told otherwise.
So I can guess that with your case and the muysic proigram I think you'll listen to, you'll be fine.
Add a 75C thermal switch bolted to the chassis, between both chips, for peace of mind.
Maybe that explains the ton of properly working amplifiers with simple flat aluminum heatsinks, usually the chassis itself.
I can not understand how a 1C/W heatsink becomes "untouchable" after 1 minute 2 1W into the load, unless said amp was fiercely oscillating, or some other gross explanation.
As of DIYAudio itself, it's customary to find way overspec'd parts in use.
Nothing against, of course, unless it instills (undeserved) fear in users who shy away from building stuff which is really within their reach.
There's a chipamp gallery in the Forum, look at ait to have some ideas.
Also look at *commercial* equipment using LM3886 🙂
Attachments
Perhaps I need to add a section on music vs sine wave signals on my site...
The difference between the peak amplitude of a signal and its RMS (root mean square) value is called the crest factor. A sine wave has a crest factor of sqrt(2) = 1.4 (2.9 dB), i.e. the peak amplitude is 1.4x the RMS value. In The Good Old Days™, the crest factor for music was commonly assumed to be about 10x (20 dB). For modern, compressed music, you'd be hard pressed to find anything below 6~10 dB crest factor. This implies that the RMS power delivered to the speaker when the amp is operating with a music signal just below clipping on the peaks is somewhere between 0.25x (-6 dB) and 0.10x (-10 dB) below the peak power.
On ±35 V rails, the LM3886 should be able to deliver about 120 W peak into 8 Ω (assuming a few volt drop across the LM3886 as it approaches the rails). This means that if the amp is running flat out, the RMS power should be around 12~30 W depending on the crest factor of the music. According to Figure 36 in the data sheet, this results in a power dissipation of 37~45 W. You can do the math from there to find the required heat sink.
So why do so many build LM3886-based amps with itty-bitty heat sinks? Well... First off, they obviously don't do the math. Secondly, many of them wonder openly why their amps sound like crap when they turn up the volume (the amp is hitting the thermal protection). And, finally, if designed properly, there is no reason a (well-designed) LM3886-based amp shouldn't be able to operate on a relatively small heat sink.
I suggest approaching the design from a system-level perspective: How much power is needed? In a normal size living room using medium efficiency speakers (say, 87 dB/W @ 1 m) you need a few watt peak. Let's design for, say 5 W peak. Allow 3 dB for headroom --> 10 W peak. This means the amp will need to be able to deliver 8.9 V peak into 8 ohm. Allowing for a few volt across the LM3886 (check the data sheet for the actual numbers), this means the amp can operate from ±12 V rails. You may have to consult a textbook to work out the power dissipation for such low rail voltages as the lowest provided in the LM3886 data sheet is for ±20 V. The power dissipation will be much lower than the 37~45 W mentioned for ±35 V rails. Now, suddenly, the thermals become more manageable.
Following this design philosophy you'll end up with a design that is solid and will perform well.
In the end the laws of physics - including those of thermodynamics - apply to all of us. Even so-and-so on the Internet who built a "200 W" LM3886 amp with a heat sink the size of a deck of cards and claimed it "worked fine".
~Tom
I can guarantee that neither chipamp was oscillating.
The amps were still on the workbench with the scope still monitoring outputs.
The load resistors were still cool (8r02, 450W each).
I gave some details for dissipated heat in a post yesterday
Look at the plots in the datasheet.
Estimate the dissipation @ 2W output into 8r0. It's probably around 10W.
The dissipation cannot fall below 5W in my rail voltage and Iq case.
The total dissipation for a total of 3W from the two chipamps is probably around 20W.
A 1C/W sink rated @ deltaT of 75Cdegrees will rise to Ts = Ta + 20*1C/W *DF (~1.5)
i.e. Ta+30C giving Ts ~ 50°C
Tc = Ts + 10W * 1 C/W =~60°C
These datapoints can all be found in the Nat info, or measured in my workshop (spare upstairs bedroom)
Don't you believe that I can't hold my finger on 60°C chip tab?
BTW,
carrying out this (3W total) test in Scotland's "high summer" will raise measured temperatures by about 7Cdegrees for a Ta=28°C
Last edited:
hmm im not sure which direction that link was supposed to be inspiring me.. its an internal heatsink, so not particularly well ventilated, and way smaller than the discussion has suggested...
in any case ive decided to go with the 0.9 c/watt 20 cm long heatsink i linked to earlier, with another section of 7.5 cm, rated 1.7 c/watt to fill the rest of the case lid.. along with the case itsself, which dealt fine with the 30 watt dissipation of the ta2022.....
not gonna be reaching the ideal heatsink sizing, but it wont be far off, i wont be running anywhere near max output, and its a damn sight more heatsinking than ive seen on many other chipamps.
i could get a 0.5 c watt one on there, but it costs 50 euros (almost the same as the amp kit) and would be almost the same volume as the whole case.
hopefully ill be fine. thanks a lot for the input from everyone.. most appreciated and im learning a lot. im sure ill be back on here when the kit arrives.
one advantage is that if i dont like the chipamp for whatever reason, i could cram a whole lot of t-amp in the case with that cooling.
in any case ive decided to go with the 0.9 c/watt 20 cm long heatsink i linked to earlier, with another section of 7.5 cm, rated 1.7 c/watt to fill the rest of the case lid.. along with the case itsself, which dealt fine with the 30 watt dissipation of the ta2022.....
not gonna be reaching the ideal heatsink sizing, but it wont be far off, i wont be running anywhere near max output, and its a damn sight more heatsinking than ive seen on many other chipamps.
i could get a 0.5 c watt one on there, but it costs 50 euros (almost the same as the amp kit) and would be almost the same volume as the whole case.
hopefully ill be fine. thanks a lot for the input from everyone.. most appreciated and im learning a lot. im sure ill be back on here when the kit arrives.
one advantage is that if i dont like the chipamp for whatever reason, i could cram a whole lot of t-amp in the case with that cooling.
0.5C/W is very large.
That would run two 3886 from +-40Vdc using Ta ~ 40°C
Nat's datasheet shows 1.6C/W for the same conditions.
Two would require 0.8C/W and my recommendation to double takes it to 0.4C/W (pretty close enough).
0.9C/W for two internal is workable if you ensure sufficient ventilation
Nat shows 2.5C/W for +-35Vdc @ Ta = 40° Two would be 1.25C/W
Go along to Ta=60°C and you get 1.8C/W, which matches your size.
That tells you the maximum ambient temperature for Tjmax = 150°C.
If you have an ambient of 40°C, then your Tjmax is 20 Cdegrees lower, i.e. Tj=130°C
The chips will not blow up. But spike will come in far earlier than if you could run at Tj<100°C
If the rail sag below +-35Vdc when you work the amps hard that will reduce the dissipation.
If you listen to music a bit quieter that will reduce the dissipation.
If you run the sink cool, the Spike protection gets closer to the cold values quoted through out the datasheet.
That would run two 3886 from +-40Vdc using Ta ~ 40°C
Nat's datasheet shows 1.6C/W for the same conditions.
Two would require 0.8C/W and my recommendation to double takes it to 0.4C/W (pretty close enough).
0.9C/W for two internal is workable if you ensure sufficient ventilation
Nat shows 2.5C/W for +-35Vdc @ Ta = 40° Two would be 1.25C/W
Go along to Ta=60°C and you get 1.8C/W, which matches your size.
That tells you the maximum ambient temperature for Tjmax = 150°C.
If you have an ambient of 40°C, then your Tjmax is 20 Cdegrees lower, i.e. Tj=130°C
The chips will not blow up. But spike will come in far earlier than if you could run at Tj<100°C
If the rail sag below +-35Vdc when you work the amps hard that will reduce the dissipation.
If you listen to music a bit quieter that will reduce the dissipation.
If you run the sink cool, the Spike protection gets closer to the cold values quoted through out the datasheet.
Last edited:
im not sure if i misunderstood your answer, but the heatsink is external, on the top of the box. im guessing once you include the second heatsink and case itself into the equation, things look a bit better to.
i shall let everyone know how i get on.. like i said before, i have 6 enormous copper server heatsinks with a very fine fin pitch.. weigh about 1.5 kg each. so if i wanted to rehouse the amp i just need to build a case around a few of those. hopefully this solution will be fine though.. and a lot neater.
i shall let everyone know how i get on.. like i said before, i have 6 enormous copper server heatsinks with a very fine fin pitch.. weigh about 1.5 kg each. so if i wanted to rehouse the amp i just need to build a case around a few of those. hopefully this solution will be fine though.. and a lot neater.
repeating what I said just earlier.
If you adopt the sink size that National recommend, you know you will hit the operational conditions set by National.
One of those operational conditions is that Tj hits Tjmax = 150°C
If you have to raise the Ta to 60°C to exactly match the Nat guidance. You now know that during worst case conditions equal to Nat's you have Ta=60 and Tj=150.
If your heatsink is external and you know it's Ta temperature is 20°C, then you are running ALL the temperatures during that specified set of worst case conditions at 40 C degrees below what the table predicted.
i.e. the chip is running with Tj = 110°C instead of 150°C
You are on a winner already.
If your "normal" operating conditions are less dissipative than Nat's worst case, then you know YOUR chip will be running at < 110°C and if you are listening to music in domestic conditions, with normal domestic speakers, with normal hearing, then your Chip will be running at <<<110°C
You have won !!!!
Last edited:
At least you're a winner until someone cranks the volume beyond "normal" and the amp hits thermal shutdown. If you're comfortable with that kind of design, great... If not, dig out the calculator.
~Tom
~Tom
- Status
- Not open for further replies.