PRELUDE (optional): If you are the person, who estimates the most optimal thermal resistance needed for dissipating the heat of your chip, before buying fancy-CPU-type-of heatsink, then you won't find anything useful in here. BUT please do read it, as you might add something usefull, or simply find a mistake.
INTRODUCTION (optional): I am newbie for all diy audio stuff. In fact I am so "rookie", that I have decided to start with simple chipamp kit. So I have just recently ordered a well known LM3886 kit. I also ordered the transformer, which was recommended by the manufacturer. Now I am at the stage, where you start thinking of the enclosure and heatsinks. I started closely look at posts that include something about heatsinks in diy projects, and my impression is following: the vast majority of comments describe a heatshink with single quantity - dimension (e.g. comment: your heatsink is too small!). I also came across one genuinely amiable comment, quote: "Anything that looks like it could be used for a CPU should work. Since your average processor will dissipate magnitudes more power than the LM3886, most of those heatsinks WITHOUT the fan should still work for you". At least, I found it interesting.
At its best, heatsink should dissipate enough power to keep chip in operating temperature. And that's about it. It should not be fancy, should not be large or be manufactured to CPU market🙂 And that is the scope of this tutorial-like post, on how to choose the heatsink strictly based on dead-easy well-known quantities of your chip - Max Dissipated Power of your chip, Operating Temperature and few Termal Resistances (more detail on this below). Nothing beats the precision - isn't it cool, when you have the cheapest yet the most effective solution?
THEORY (optional but recommended): Long time ago it was found that heat is similar to electrical current, and for solving some "termal problems" the Ohm's law can be used, with some basic substitution of concepts:
1. Voltage = Temperature
2. Current = Power
3. Electrical resistance = Thermal resistance (this one trivial🙂 )
Now, the law is law, and therefore we cannot be more right than saying that for a particular component in the electrical circuit,
Temperature = Dissipated Power * Thermal Resistance
Note 1: "ground" in this "thermal" law is equal to the ambient temperature, Ta. Note 2: Rja, this is thermal resistance of "junction-ambient air" layer, or simply the housing of the chip.
PRACTICE: At this stage we know all we need for diving in calculations. Let's assume our amplifier has LM3886 chip. Let's assume we the chip is supplied by +/-20V. And let's assume we have neat speakers with 4 Ohms resistance. How do we select the most optimal heatsink?
First of all, let's figure out why do we need one at all. I know, this may sound ridiculous, but I encourage you for a moment to not trust anyone, and check it yourself. Using the above "Thermal" Ohm's law and schematic we have equation:
Tj = P*Rja + Ta
Rja is physical constant of device, and it can be obtained from LM3886 datasheet (page 3, Absolute Maximum Ratings http://www.ti.com/lit/ds/symlink/lm3886.pdf) - according to the document it is equal to 43 C/W.
Now, with dissipated power it is trickier - obviously the dissipated power on the chip will at least depend on the Voltage supply. In reality there are many factors which affects dissipated power of the chip, but thankfully, manufacturers try their best to make our life easier - look at page 14, Figure 35. We can see that this graph is given for the load of 4 Ohms (we assumed before that this is our load) - it is Power Dissipation vs. Output Power curves for different supplied voltages. In our case the curve have its peak on 20W, that is Maximum Power Dissipation of the chip (by the way, look at the graph - our imagery amp is running on only 50% efficiency!). Okay, so now we have all numbers for equations, therefore the junction temperature of the LM3886 without any heatsink (let's make ambient temperature be 30C), supplied the +/-20V and giving the 20W to the 4 Ohms speakers, will be:
Tj = 20W * 43 C/W + 30C = 890C !!! Can you imagine that? 🙂 Of course you will never reach anything near to this - the worse case chip will fry out in between 200C-250C and will die, but ideally the thermal protection should kick in at some point before that, turn the functionality off until temperature won't drop in the operating temperature range.
So there is another important measure - operating temperature. In datasheet, page 3 it said that maximum operating temperature for the junction is 150C. So what we want is to reduce Tj to the point, that thermal protection won't disturb us by turning off the amp while we enjoying the music. This might be about 120C. I think if we would go for less, that would be a big overcompensation, which is not efficient. Remember - you want cheapest heatsink which does it job. If 20g of aluminium work cools my amp well, why would I bother to pay for 30g? 🙂
Okay, so we have decided to add heatsink. This changes our "equivalent thermal circuit" as follows:
Rjc - is thermal resistance of junction-case layer, that is also the constant which is given in datasheet - for LM3886 it is given as 1C/W.
Rcs is thermal resistance of "something" which is between the chip case and sink - this "something" can be thermal compound, or thermal "gum" or even gap of air if you like. Most common is compound, its resistance varies with many parameters like gap width, etc. Datasheet refers to some compound brand which has 0.2C/W, I would say the temperature drop on this layer is small and hence this resistance can be neglected.
Rsa is thermal resistance between heatsink and air. As with others, there are plenty factors affecting this resistance, like geometry, air flow and etc. This value is usually given by manufacturers (although it is not so hard to measure it by yourself with high power resistor and thermocouple😉 )
Finally, question "which heatsink should I buy" in engineering terms would be "what is the heatsink-air thermal resistance do I need for ...(here we need to strictly specify Power, Voltage, Desired Juntion Maximum Temperature,Load and Ambient Temperature)?".
Solution:
120C = 20W * (1C/W + 0.2C/W + Rsa) + 30C
Rsa = [(120C - 30C) / 20W] - 1C/W - 0.2C/W = 3.3C/W
This is our magic number. As suggested by DUG in the comments below, it is right to say that we are looking for 3.3C/W or lower, i.e. 3.0C/W would be more preferable than say 3.5C/W. We don't care is it large, is it meant to be for CPU or whatever - we have the exact specification for our heatsink, which will allow LM3886 to play under conditions described above continuously, without thermal breakouts. And filtering out the product range offered by electronics suppliers, I can freely choose the form factor of the heatsink for my amp's enclosure😉
Good luck in your designs!
UPD: Sorry my English, I will try to polish my text to make it more readable.
INTRODUCTION (optional): I am newbie for all diy audio stuff. In fact I am so "rookie", that I have decided to start with simple chipamp kit. So I have just recently ordered a well known LM3886 kit. I also ordered the transformer, which was recommended by the manufacturer. Now I am at the stage, where you start thinking of the enclosure and heatsinks. I started closely look at posts that include something about heatsinks in diy projects, and my impression is following: the vast majority of comments describe a heatshink with single quantity - dimension (e.g. comment: your heatsink is too small!). I also came across one genuinely amiable comment, quote: "Anything that looks like it could be used for a CPU should work. Since your average processor will dissipate magnitudes more power than the LM3886, most of those heatsinks WITHOUT the fan should still work for you". At least, I found it interesting.
At its best, heatsink should dissipate enough power to keep chip in operating temperature. And that's about it. It should not be fancy, should not be large or be manufactured to CPU market🙂 And that is the scope of this tutorial-like post, on how to choose the heatsink strictly based on dead-easy well-known quantities of your chip - Max Dissipated Power of your chip, Operating Temperature and few Termal Resistances (more detail on this below). Nothing beats the precision - isn't it cool, when you have the cheapest yet the most effective solution?
THEORY (optional but recommended): Long time ago it was found that heat is similar to electrical current, and for solving some "termal problems" the Ohm's law can be used, with some basic substitution of concepts:
1. Voltage = Temperature
2. Current = Power
3. Electrical resistance = Thermal resistance (this one trivial🙂 )
Now, the law is law, and therefore we cannot be more right than saying that for a particular component in the electrical circuit,
Temperature = Dissipated Power * Thermal Resistance
An externally hosted image should be here but it was not working when we last tested it.
Note 1: "ground" in this "thermal" law is equal to the ambient temperature, Ta. Note 2: Rja, this is thermal resistance of "junction-ambient air" layer, or simply the housing of the chip.
PRACTICE: At this stage we know all we need for diving in calculations. Let's assume our amplifier has LM3886 chip. Let's assume we the chip is supplied by +/-20V. And let's assume we have neat speakers with 4 Ohms resistance. How do we select the most optimal heatsink?
First of all, let's figure out why do we need one at all. I know, this may sound ridiculous, but I encourage you for a moment to not trust anyone, and check it yourself. Using the above "Thermal" Ohm's law and schematic we have equation:
Tj = P*Rja + Ta
Rja is physical constant of device, and it can be obtained from LM3886 datasheet (page 3, Absolute Maximum Ratings http://www.ti.com/lit/ds/symlink/lm3886.pdf) - according to the document it is equal to 43 C/W.
Now, with dissipated power it is trickier - obviously the dissipated power on the chip will at least depend on the Voltage supply. In reality there are many factors which affects dissipated power of the chip, but thankfully, manufacturers try their best to make our life easier - look at page 14, Figure 35. We can see that this graph is given for the load of 4 Ohms (we assumed before that this is our load) - it is Power Dissipation vs. Output Power curves for different supplied voltages. In our case the curve have its peak on 20W, that is Maximum Power Dissipation of the chip (by the way, look at the graph - our imagery amp is running on only 50% efficiency!). Okay, so now we have all numbers for equations, therefore the junction temperature of the LM3886 without any heatsink (let's make ambient temperature be 30C), supplied the +/-20V and giving the 20W to the 4 Ohms speakers, will be:
Tj = 20W * 43 C/W + 30C = 890C !!! Can you imagine that? 🙂 Of course you will never reach anything near to this - the worse case chip will fry out in between 200C-250C and will die, but ideally the thermal protection should kick in at some point before that, turn the functionality off until temperature won't drop in the operating temperature range.
So there is another important measure - operating temperature. In datasheet, page 3 it said that maximum operating temperature for the junction is 150C. So what we want is to reduce Tj to the point, that thermal protection won't disturb us by turning off the amp while we enjoying the music. This might be about 120C. I think if we would go for less, that would be a big overcompensation, which is not efficient. Remember - you want cheapest heatsink which does it job. If 20g of aluminium work cools my amp well, why would I bother to pay for 30g? 🙂
Okay, so we have decided to add heatsink. This changes our "equivalent thermal circuit" as follows:
An externally hosted image should be here but it was not working when we last tested it.
Rjc - is thermal resistance of junction-case layer, that is also the constant which is given in datasheet - for LM3886 it is given as 1C/W.
Rcs is thermal resistance of "something" which is between the chip case and sink - this "something" can be thermal compound, or thermal "gum" or even gap of air if you like. Most common is compound, its resistance varies with many parameters like gap width, etc. Datasheet refers to some compound brand which has 0.2C/W, I would say the temperature drop on this layer is small and hence this resistance can be neglected.
Rsa is thermal resistance between heatsink and air. As with others, there are plenty factors affecting this resistance, like geometry, air flow and etc. This value is usually given by manufacturers (although it is not so hard to measure it by yourself with high power resistor and thermocouple😉 )
Finally, question "which heatsink should I buy" in engineering terms would be "what is the heatsink-air thermal resistance do I need for ...(here we need to strictly specify Power, Voltage, Desired Juntion Maximum Temperature,Load and Ambient Temperature)?".
Solution:
120C = 20W * (1C/W + 0.2C/W + Rsa) + 30C
Rsa = [(120C - 30C) / 20W] - 1C/W - 0.2C/W = 3.3C/W
This is our magic number. As suggested by DUG in the comments below, it is right to say that we are looking for 3.3C/W or lower, i.e. 3.0C/W would be more preferable than say 3.5C/W. We don't care is it large, is it meant to be for CPU or whatever - we have the exact specification for our heatsink, which will allow LM3886 to play under conditions described above continuously, without thermal breakouts. And filtering out the product range offered by electronics suppliers, I can freely choose the form factor of the heatsink for my amp's enclosure😉
Good luck in your designs!
UPD: Sorry my English, I will try to polish my text to make it more readable.
Last edited:
Thanks so much for this post.
While its a little over my head.. its definitely something for me to chew on.
While its a little over my head.. its definitely something for me to chew on.
honeycomb0
A comment can be added to the final result of 3.3C/W that 3.0C/W is better (Junction temperature lower) and 3.5C/W is not better (Junction temperature higher).
Just in case someone does not understand which way to go on the heat sink requirement.
🙂
A comment can be added to the final result of 3.3C/W that 3.0C/W is better (Junction temperature lower) and 3.5C/W is not better (Junction temperature higher).
Just in case someone does not understand which way to go on the heat sink requirement.
🙂
Thanks you very much!
Just as I am about to post asking people about this topic, your post proved timely.
Just as I am about to post asking people about this topic, your post proved timely.
- Status
- Not open for further replies.