I am going to build an LM3886 chipamp on chipamp.com boards. About 30-34VDC on the rails (22vac tranny) with an 8 ohm speaker load.
I have 2 each of these heatsinks but no way to calculate if they are adequate. I will make the enclosure and can make it to fit.
The small heatsinks are 45mm x 50mm x 20mm. One LM3886 on each heatsink.
The large heatsinks are 270mm x 100mm x 50mm. Maybe put both chips on one heatsink?
Do you think the smaller heatsinks will work? If not would both chips on one large heatsink be sufficient?
I know this is somewhat of a guess without know the characteristics of the heatsinks but it is what I have.
Thanks
Small heatsinks:
Large
I have 2 each of these heatsinks but no way to calculate if they are adequate. I will make the enclosure and can make it to fit.
The small heatsinks are 45mm x 50mm x 20mm. One LM3886 on each heatsink.
The large heatsinks are 270mm x 100mm x 50mm. Maybe put both chips on one heatsink?
Do you think the smaller heatsinks will work? If not would both chips on one large heatsink be sufficient?
I know this is somewhat of a guess without know the characteristics of the heatsinks but it is what I have.
Thanks
Small heatsinks:
Large
Last edited:
For 2 channel P19 and 3886DR I used Conrad heatsink MF20-75 rated 0.55 deg C/W.
https://www.altronics.com.au/p/h053...L21wbJoVW8TGEmBWOMCVLs_8z8FUaAy8aAqm3EALw_wcB
https://www.altronics.com.au/p/h053...L21wbJoVW8TGEmBWOMCVLs_8z8FUaAy8aAqm3EALw_wcB
I presume that's two amps on one, running similar power rabbitz?
I bet it stayed pretty cool. I'm looking at the OPs choices, and his big heatsink looks very ample.
On my bench I have a hifi amp, 225rms onto 4 ohm, with a similar brick like heatsink. 160x60x75
I don't imagine talk of amp classes can make a 100w? of solid state, need 270x50x70
Edit: I did an image search, to look at peoples builds, and what kits from China might try and get away with. I wouldn't bother with the little heatsinks, but the big one, is big
I bet it stayed pretty cool. I'm looking at the OPs choices, and his big heatsink looks very ample.
On my bench I have a hifi amp, 225rms onto 4 ohm, with a similar brick like heatsink. 160x60x75
I don't imagine talk of amp classes can make a 100w? of solid state, need 270x50x70
Edit: I did an image search, to look at peoples builds, and what kits from China might try and get away with. I wouldn't bother with the little heatsinks, but the big one, is big
Last edited:
Running 2 amps per heatsink runs well even with 4 ohm loads but I only use 25VDC rails to allow for lower impedance capabilities.
For what's needed look for Overture_Design_Guide15.xls as will indicate heatsink requirements for rails, load etc.
For what's needed look for Overture_Design_Guide15.xls as will indicate heatsink requirements for rails, load etc.
I love heatsinks, why are they so cool looking?
Anyways the big heatsink shown would be good.
Anyways the big heatsink shown would be good.
I checked the ESP P19 BOM and Rod Elliott recommends 1.5 deg C/W minimum for 2 channels so it can be much smaller than the one I've shown. I tend to go for a bit of overkill and bugger factor.
Looking at the pics in the first post, the larger heatsink should do the job for both channels.
Looking at the pics in the first post, the larger heatsink should do the job for both channels.
First have a look here: https://neurochrome.com/pages/thermal-design
Then decide whether you want the amp to survive music reproduction on continuous sine wave reproduction. Music reproduction is much less demanding than sine wave reproduction.
According to my math, 0.7 K/W is appropriate for a stereo amp intended for music reproduction on ±30 V supply rails with 4 Ω load.
You can use this online calculator to estimate the thermal resistance of your heat sinks: https://heatscapecal.com/natural
I suspect you'll find that the small one is too small and the larger one is about right. Even if it is a touch small, you can always either accept that it'll be a bit hotter than the 60-65 ºC I recommend and design for, or you can fit it with a thermal switch that cuts the power once the heat sink gets too hot.
Tom
Then decide whether you want the amp to survive music reproduction on continuous sine wave reproduction. Music reproduction is much less demanding than sine wave reproduction.
According to my math, 0.7 K/W is appropriate for a stereo amp intended for music reproduction on ±30 V supply rails with 4 Ω load.
You can use this online calculator to estimate the thermal resistance of your heat sinks: https://heatscapecal.com/natural
I suspect you'll find that the small one is too small and the larger one is about right. Even if it is a touch small, you can always either accept that it'll be a bit hotter than the 60-65 ºC I recommend and design for, or you can fit it with a thermal switch that cuts the power once the heat sink gets too hot.
Tom
There is minimal requirements of course. The fun of Diy is there is no such thing as a heatsink that is too big.
You will see rather excessive large heatsinks. No harm in overkill.
I often buy heatsinks with no design goals lol. Ebay is full of large heatsinks. they are beautiful.
And big ugly old transistors are sexy to, including the heatsinks they go on.
Feel free to indulge.
Other than that take a peak into typical old school generic, home stereo amps. with similar power.
Many will have somewhat larger sinks than others. But youll will understand the minimal from known established designs.
Aside from common sense. Inside a case needs large venting for airflow, outside the case or not enclosed has benefits from less restriction.
You will see rather excessive large heatsinks. No harm in overkill.
I often buy heatsinks with no design goals lol. Ebay is full of large heatsinks. they are beautiful.
And big ugly old transistors are sexy to, including the heatsinks they go on.
Feel free to indulge.
Other than that take a peak into typical old school generic, home stereo amps. with similar power.
Many will have somewhat larger sinks than others. But youll will understand the minimal from known established designs.
Aside from common sense. Inside a case needs large venting for airflow, outside the case or not enclosed has benefits from less restriction.
What I normally do is to calculate the resistencies and set a maximum temperature for the heatsink, 80C for me and check the junction temperature.
From datasheet:
Rjc=1C/W (junction to case)
Rch=0.2C/W (case to heatsink)
Rjh=1.2C/W (junction to heatsink)
Tjmax = 150C
V=+/-35V and RL=8ohms => MaxPowerDissipated by 1 channel = 35W (continuous sine wave)
If heatsink max temp is 80C (my decision) and ambient temp is 25C, we have:
80-25 = 55C => Rha=55/35=1.57C/W (heatsink to ambient)
Rja=1.2+1.57=2.77C/W (junction to ambient)
Checking Tjmax => 35(W) *2.77 + 25(Tamb) = 122C (<150C ok!)
So you need, in the worst case scenario, a 1.57C/W heatsink per channel (continuous sine wave).
For both channels, you need single heatsink with 0.78C/W.
You can check on these online calculators, but I think the your bigger heatsink (270mm x 100mm x 50mm) has resistance below 1C/W.
I'm considering external heatsink installation.
Depending on what kind of music you listen to, you may derate max disspated power by some factor.
I listen to pop music with high power kick drum, so in my case the proportion is around 1:4 or even higher (divide max power by 4, so 35/4=9W per channel).
In my case, this heatsink would reach around 0.78*(70/4)+25=38.6C
But if you listen to highly compressed and very loud music (e.g. Highway to Hell AC/DC) you may need a smaller factor or consider the worstcase.
And I agree with @WhiteDragon, heatsinks are beautifull!!! And I also buy some heatsinks just cause I want them 🙂
From datasheet:
Rjc=1C/W (junction to case)
Rch=0.2C/W (case to heatsink)
Rjh=1.2C/W (junction to heatsink)
Tjmax = 150C
V=+/-35V and RL=8ohms => MaxPowerDissipated by 1 channel = 35W (continuous sine wave)
If heatsink max temp is 80C (my decision) and ambient temp is 25C, we have:
80-25 = 55C => Rha=55/35=1.57C/W (heatsink to ambient)
Rja=1.2+1.57=2.77C/W (junction to ambient)
Checking Tjmax => 35(W) *2.77 + 25(Tamb) = 122C (<150C ok!)
So you need, in the worst case scenario, a 1.57C/W heatsink per channel (continuous sine wave).
For both channels, you need single heatsink with 0.78C/W.
You can check on these online calculators, but I think the your bigger heatsink (270mm x 100mm x 50mm) has resistance below 1C/W.
I'm considering external heatsink installation.
Depending on what kind of music you listen to, you may derate max disspated power by some factor.
I listen to pop music with high power kick drum, so in my case the proportion is around 1:4 or even higher (divide max power by 4, so 35/4=9W per channel).
In my case, this heatsink would reach around 0.78*(70/4)+25=38.6C
But if you listen to highly compressed and very loud music (e.g. Highway to Hell AC/DC) you may need a smaller factor or consider the worstcase.
And I agree with @WhiteDragon, heatsinks are beautifull!!! And I also buy some heatsinks just cause I want them 🙂
I actually don't use a lot of power. My speakers are Klipsch La Scalas, sensitivity 104db, and I typically don't listen very loud (80-85db). Right now I am using a 10wpc tube amp I built and it drives the La Scalas to ear splitting level if needed.
You're using the wrong number for the power dissipated, though.
You are correct that with ±35 V rails and 8 Ω load you're looking at about 35 W dissipated in the LM3886 during sine wave operation at the worst case operating point. That supply voltage should give you around 60 W of output power.
To account for the crest factor of music you need to calculate the dissipated power when the amplifier delivers 60/4 = 15 W of output power. Figure 36 in the LM3886 data sheet is helpful:
You'll notice that you still have about 30 W dissipated in the IC. That's not a dramatic reduction from the 35 W.
Sound-on-Sound analyzed 2500+ tracks a while back and found an average crest factor of 14 dB, so if your taste in music is similar to the tracks they analyzed, you're probably looking at something like 18 W dissipated. Now we're getting somewhere. Now you can live with twice the heat sink thermal resistance.
The math so far has assumed that the amp is operated with music signal with the peaks just below clipping. Few of us probably run at those volume levels, which is why the DIY builds with tiny heat sinks survive.
I'we tested my math by running the amps I build with a 32-tone signal that has a crest factor of 10 dB. I design for a maximum heat sink temperature of 60 ºC (assuming 14 dB CF) and in the multi-tone test the amp reaches 62-65 ºC after an hour of operation. I think that's a pretty good correlation between math and reality.
Tom
Sure, I agree.
The number 35W was a rough estimation and I took the maximum from the same curve on figure 36 you've printed to be conservative on purpose.
If we will further divide the base number by 3, 4, 5 or up to 10, it doesn't change much the result if we go with 30 or 35W - it will be a very wide range estimation. The idea was to point out the big difference between continuous sine wave and normal music listening.
I've took several measurements myself by sampling songs I normally listen too and getting the average power of them comparing to the peak power.
I've mentioned 1:4 just to give an idea. The numbers I've got, varies from 3 to almost 10 depending on the song and the equalization.
Music with a lot of bass and very dynamic without much average energy (few instruments being played at the same time) can lead to really low average power dissipated.
The number 35W was a rough estimation and I took the maximum from the same curve on figure 36 you've printed to be conservative on purpose.
If we will further divide the base number by 3, 4, 5 or up to 10, it doesn't change much the result if we go with 30 or 35W - it will be a very wide range estimation. The idea was to point out the big difference between continuous sine wave and normal music listening.
I've took several measurements myself by sampling songs I normally listen too and getting the average power of them comparing to the peak power.
I've mentioned 1:4 just to give an idea. The numbers I've got, varies from 3 to almost 10 depending on the song and the equalization.
Music with a lot of bass and very dynamic without much average energy (few instruments being played at the same time) can lead to really low average power dissipated.
Ok! So you'll probably be safe installing the 2 channels in that bigger heatsink.
Note that in my calculation, I still was conservative by considering the junction temperature at 122C. It can go up to 150C.
In this situation, heatsink could reach a bit more than 100C and still keep the junction at 150C provide the ambient temperature is 25C.
For residential and for conservative listeners, as I think it's your case, heatsinks can be more relaxed indeed.
It's just good to keep in mind that when you design heatsinks to music, the amp cannot be used for a PA or musical instrument environment, for example.
The audience here is broad, so we need to emphasize these details since we never know who is reading.
That is what I was hoping to do. I build my own enclosures and am thinking of top mounting the heatsink.
Right. But what you want to do is to divide the output power by the crest factor and then calculate the dissipated power at that output power. What you did – at least if I understand you correctly – was to divide the dissipated power by the crest factor, which is incorrect.
You're correct that there's a fair amount of wiggle room in the math. Ultimately, it wouldn't surprise me if the thermal mass of the heat sink was more important than the thermal resistance, as long as the thermal resistance is low enough to allow the heat sink to get rid of the average power dissipated.
Tom
As long as you mount it in the orientation shown in your pic and allow for good airflow with vents top and bottom. Install the chips on the lower section of the heatsink.
Hello @tomchr !
You're correct regarding what I've posted. I took the maximum dissipated power for the power supply rails and divided it by "my" factor (1:4).
The idea was to provide a rough estimation about real world when we are listening to music and not torturing the amp with sine waves.
What you say is that we should apply the factor to the output power and then with this new output power number look up in the figure 36 in order to get the dissipated power. We may have to think about it.
That curve on figure 36 is probably based on continuous sine wave applied to a 8ohm resistor (no reactive load varying with frequency).
And this is not the scenario when we consider a musical wave form. I'm not sure, without going through deep thinking, calculations, research or even simulations, if we can use that curve the way you propose for generic and dynamic wave form. I'm not able to say yes or no at this moment.
If you have arguments, please comment - this is an interesting subject.
On top of that, things can get even more complex in real world, if we consider the real impedance curve that might vary a lot from speaker to speaker.
Imagine a music with the energy distribution in certain frequency range. Then combine that with the speaker impedance curve.
You might have a lot of energy in a frequency range where a certain speaker impedance is very high while in other speaker is very low.
The heat on heatsinks will be different - same music and same nominal 8ohms speaker.
It's not difficult to happen, specially in bass where frequencies can be around the loudspeaker A resonance or around the lowest impedance range of loudspeaker B.
And just to mention, "my" factor was measured using the following method, again rough estimation to get an idea of continuous power and dynamic power that speakers are subjected too. So the equivalent "heat" generated by each situation.
I sampled 10sec of different music styles and took 2 measurements using a scope:
-RMS voltage considering the whole period of 10sec.
-Maximum RMS voltage during that period (peak RMS voltage)
Then I've calculated the power values of each voltage over the impedance [P=(Vrms^2)/Z].
The proportions I've got varied from 3 to 10 depending on music and equalization.
My main system, which is very good at low frequencies (<30Hz) is driven flat - no equalization. Factor around 3 to 5.
My small PC speakers requires a lot of low frequency equalization to be minimum flat such as +18dB at 25Hz, +15dB at 42Hz and so on. Here, the factor moves to something around 6 to 10.
You're correct regarding what I've posted. I took the maximum dissipated power for the power supply rails and divided it by "my" factor (1:4).
The idea was to provide a rough estimation about real world when we are listening to music and not torturing the amp with sine waves.
What you say is that we should apply the factor to the output power and then with this new output power number look up in the figure 36 in order to get the dissipated power. We may have to think about it.
That curve on figure 36 is probably based on continuous sine wave applied to a 8ohm resistor (no reactive load varying with frequency).
And this is not the scenario when we consider a musical wave form. I'm not sure, without going through deep thinking, calculations, research or even simulations, if we can use that curve the way you propose for generic and dynamic wave form. I'm not able to say yes or no at this moment.
If you have arguments, please comment - this is an interesting subject.
On top of that, things can get even more complex in real world, if we consider the real impedance curve that might vary a lot from speaker to speaker.
Imagine a music with the energy distribution in certain frequency range. Then combine that with the speaker impedance curve.
You might have a lot of energy in a frequency range where a certain speaker impedance is very high while in other speaker is very low.
The heat on heatsinks will be different - same music and same nominal 8ohms speaker.
It's not difficult to happen, specially in bass where frequencies can be around the loudspeaker A resonance or around the lowest impedance range of loudspeaker B.
And just to mention, "my" factor was measured using the following method, again rough estimation to get an idea of continuous power and dynamic power that speakers are subjected too. So the equivalent "heat" generated by each situation.
I sampled 10sec of different music styles and took 2 measurements using a scope:
-RMS voltage considering the whole period of 10sec.
-Maximum RMS voltage during that period (peak RMS voltage)
Then I've calculated the power values of each voltage over the impedance [P=(Vrms^2)/Z].
The proportions I've got varied from 3 to 10 depending on music and equalization.
My main system, which is very good at low frequencies (<30Hz) is driven flat - no equalization. Factor around 3 to 5.
My small PC speakers requires a lot of low frequency equalization to be minimum flat such as +18dB at 25Hz, +15dB at 42Hz and so on. Here, the factor moves to something around 6 to 10.
Well... If you wanted to geek out completely, you would simulate the power dissipated in your amplifier with the .wav file loaded into the simulator and used for the stimulus. You would then know the amount of power dissipated for that track with that amplifier. Repeat for all the tracks in your music collection. This could take a while...
And, as you point out, you then need to account for the actual impedance of various speakers, etc.
So basically you have a system of equations with an infinite number of solutions. Congratulations!
You now know why most of us do the math with sine wave excitation and put a thermal sensor on the heat sink so the end user doesn't burn their fingers. 🙂
Tom
And, as you point out, you then need to account for the actual impedance of various speakers, etc.
So basically you have a system of equations with an infinite number of solutions. Congratulations!
You now know why most of us do the math with sine wave excitation and put a thermal sensor on the heat sink so the end user doesn't burn their fingers. 🙂
Tom
Sure, Tom - That's why I go for rough estimations and use experience in order to build things that work, have reasonable cost/size and don't get burned.
Sometimes more relaxed (my own use) sometimes more conservative for things that I build for other people.
This is a very interesting discussion! 🙂
When I have time, I'll try to load a wave file as stimulus just to take a look on Spice and check the power dissipation. I never tried to do that.
Audacity can export wave files as text - huge files, since on each second we have 44100 samples.
I'm at the moment focussed on build speakers for studio and residential use, so I'm around with the difficult task of estimating mid-range, tweeter and passive crossover components average power and current. I'm gathering my own data and defining factors, double checking with a thermometer once ready to use.
100W system power - easy to choose the woofer which is simply 100W.
But mid-range and tweeter not easy. I know I use mid that supports 30W and tweeter that support only 5W, but that's ok for 100W system power.
We need to apply dynamics and also consider frequency range distribution.
Crossover resistors maximum power. Simulator indicates unnecessary gigantic resistors that are calculated for the continuous sine wave at specific frequency - not even the mid-range or tweeter supports that power by far. With music, it will never dissipate that power.
Crossover capacitors average current - same thing. I have to define an average RMS current in order to keep the capacitor within a reasonable life span.
Sometimes more relaxed (my own use) sometimes more conservative for things that I build for other people.
This is a very interesting discussion! 🙂
When I have time, I'll try to load a wave file as stimulus just to take a look on Spice and check the power dissipation. I never tried to do that.
Audacity can export wave files as text - huge files, since on each second we have 44100 samples.
I'm at the moment focussed on build speakers for studio and residential use, so I'm around with the difficult task of estimating mid-range, tweeter and passive crossover components average power and current. I'm gathering my own data and defining factors, double checking with a thermometer once ready to use.
100W system power - easy to choose the woofer which is simply 100W.
But mid-range and tweeter not easy. I know I use mid that supports 30W and tweeter that support only 5W, but that's ok for 100W system power.
We need to apply dynamics and also consider frequency range distribution.
Crossover resistors maximum power. Simulator indicates unnecessary gigantic resistors that are calculated for the continuous sine wave at specific frequency - not even the mid-range or tweeter supports that power by far. With music, it will never dissipate that power.
Crossover capacitors average current - same thing. I have to define an average RMS current in order to keep the capacitor within a reasonable life span.