I`m wondering looking at a FEA simulation the heatsinking size of a Pure class A is accurately estimated by giving constant power dissipation and right thermal conductivity constants and we do get perfectly right with tolerance of +/-0.5 deg C when data is given properly.
Now consider a following example. I have a heatsink with which a constant power dissipation is made of 50W with mosfets on it. Then the heatsink temperature raises 15 degree above room temperature of consider 35 degree C to heatsink temp 50 deg.
Now consider the same heatsink how much it can dissipate when it comes to class AB where the idle current temp of heatsink rose by just 3deg C in AB mode. Now what I would like to know is that how much Class AB heat does the heatsink takes up for same AB.
Now consider a following example. I have a heatsink with which a constant power dissipation is made of 50W with mosfets on it. Then the heatsink temperature raises 15 degree above room temperature of consider 35 degree C to heatsink temp 50 deg.
Now consider the same heatsink how much it can dissipate when it comes to class AB where the idle current temp of heatsink rose by just 3deg C in AB mode. Now what I would like to know is that how much Class AB heat does the heatsink takes up for same AB.
Chapter 16 of Douglas Self's power amp textbook is called "Power Dissipation in Amplifiers". He covers class-A, class-B, class-AB, and class-G circuit types. He also covers the impact of reactive loads. He also covers the different power consumption (and different heat dissipation) when amplifying music versus amplifying pure sine waves. It's well worth reading! If you want to skip right to the conclusions, look at Table 16.3 on page 415.
You can find the equations and derivation I show on the Taming the LM3886 - Thermal Design page in many engineering texts. I recommend Sedra/Smith Microelectronic Circuits. As with any engineering text the current revision is stupid expensive, but you should be able to find an older version for much less. I have the 2nd, 3rd, and 5th editions. The biggest change was from the 2nd to 3rd where the treatment of JFETs was cut dramatically. Any of those would work for you.
The heat sink doesn't care if the amp is Class A, AB, C, D, G, H, whatever. All it cares about is the dissipated power, the ambient temperature, and the amount of air flow. It just so happens that a Class A power amp will dissipate much more power than any of the other amplifier classes. That's what happens when you insist that all output devices should always be conducting current.
Tom
The heat sink doesn't care if the amp is Class A, AB, C, D, G, H, whatever. All it cares about is the dissipated power, the ambient temperature, and the amount of air flow. It just so happens that a Class A power amp will dissipate much more power than any of the other amplifier classes. That's what happens when you insist that all output devices should always be conducting current.
Tom
Some points to consider - you really need to determine the heat you have to get rid of. That is dependent on the input signal characteristics as well as the output power levels.
Theoretical efficiency of a class AB is pi/4 but that only applies to a sinewave at full output. A signal might be a square wave at half peak output voltage which is then 50% efficient, and that is really not far off values achieved in practice. One estimate for music power is 10:1 peak to mean, so you could estimate the heatsinking needed for music on 10% nominal peak output power on average or 20% rms max to mean as Ed suggested earlier.That means your example heatsink (0.3C/W) which may be delivering 15W from your 50W of MOSFETs might be able to provide 60W (music) or 30W(if signals considered to be square wave worst case) or something else considering reactive loads, though the lower of these estimates would allow for some reactive loading.
It isn't that hard to evaluate currents and voltages across a transistor given the loading conditions, and signals, which is where you might delve a little more, but a reactive load would also be frequency dependent, which would mean having to know the frequency content in the music (as well as a loudspeaker load model).
Theoretical efficiency of a class AB is pi/4 but that only applies to a sinewave at full output. A signal might be a square wave at half peak output voltage which is then 50% efficient, and that is really not far off values achieved in practice. One estimate for music power is 10:1 peak to mean, so you could estimate the heatsinking needed for music on 10% nominal peak output power on average or 20% rms max to mean as Ed suggested earlier.That means your example heatsink (0.3C/W) which may be delivering 15W from your 50W of MOSFETs might be able to provide 60W (music) or 30W(if signals considered to be square wave worst case) or something else considering reactive loads, though the lower of these estimates would allow for some reactive loading.
It isn't that hard to evaluate currents and voltages across a transistor given the loading conditions, and signals, which is where you might delve a little more, but a reactive load would also be frequency dependent, which would mean having to know the frequency content in the music (as well as a loudspeaker load model).
so what I am looking for is what is the ratio of heat handing capability of continuous : music/movies.
So now take an example like John ellis said the ratio even Ed suggested is 20% of RMS heat dissipation?
The reason for this is that with FEA the accuracy of right Heatsinking estimation can be really precize we can also add more ambient temp and air flow and see how the heatsink temp is getting hot.
So now take an example like John ellis said the ratio even Ed suggested is 20% of RMS heat dissipation?
The reason for this is that with FEA the accuracy of right Heatsinking estimation can be really precize we can also add more ambient temp and air flow and see how the heatsink temp is getting hot.
You do you. Make your own assumptions and accept the consequences (or, if disaster occurs, make new assumptions!)
Mark I`m trying to accurately predict then just going with assumptions. When we know exactly what is happening with the Heatsink used with music. I think we need to calculate the average area under the music waveform and see how much that would cover in a given time will give much more clarity on how much time the transistors are on and dissipating even in the worst condition. + Idle bias current based dissipation to be added
rhythmsandy - The 20% figure that I quoted is the ratio of the heat dissipated by ideal class AB and class A amplifiers.
Class A dissipates the most heat when idle. Class AB dissipates the most heat with a sine wave at 64% of maximum amplitude.
You have to use these conditions to size the heatsinks.
Ed
Class A dissipates the most heat when idle. Class AB dissipates the most heat with a sine wave at 64% of maximum amplitude.
You have to use these conditions to size the heatsinks.
Ed
assume efficiency is not much more the 20 to 25%
so up too 80 to 75 % of power supply power, consider a loss to heat.
aint going to the speakers, so heatsink is about the only other place.
so up too 80 to 75 % of power supply power, consider a loss to heat.
aint going to the speakers, so heatsink is about the only other place.
There is no ideal, just a variety of situations for you to experiment with and see what works for you.
Nice thing about your typical amplifier is that you can adjust it for various bias values and make your own judgment.
Nice thing about your typical amplifier is that you can adjust it for various bias values and make your own judgment.
The efficiency of a radiator also depends on its geometry, material, type of surface, method (place) of installation in the case.
Use forced airflow in your experiments. Monitor the temperature of the Mosfet case.
Use forced airflow in your experiments. Monitor the temperature of the Mosfet case.
...well, only if you plan to use forced air cooling in your amp. It will give better thermal resistance figures but most of us do not use forced cooling in our amps (except if high power PA perhaps).
surely better to explore free air cooling if that is what is what you will want to end up with.
FEA might still give an accurate solution if it handles turbulence, lamina flow, surroudings (casing, holes in case for circulation) etc, but it still requires an input of the heat you want to dissipate!
Many just consider music to be a collection of sinewaves so base it on peak sinewave efficiency figures multiplied by root 2 for example for an inductive loading (in part because a speaker impedance can be lower than the stated "8 ohms" and the resistive part could be as low as e.g. 5 ohms for nominal 8).
it is up to you whether you allow for worst case loads and signals after that. That's where engineering judgement has to come to a decision.
I cover that here: https://neurochrome.com/pages/thermal-design That's the page I linked to in Post #4. I use the LM3886 as an example, but the math is the same for any other amp.
Tom
Class A with 2 output transistors on the same heatsink is very simple. Even if one transistor carries the current all of the time dissipation is the same as far as the heatsink is concerned. In practice disipation is highest in the static condition as no power is going to the loudspeaker.
😉 Suppose all know this anyway.
The link using crest factors etc. A possible argument against it. Say some one takes a piece of metal and turns the gas stove on and heats the end of the metal. Would they have to drop the metal immediately? No as heat takes time to conduct. Pass, just wondering about a tiny little semiconductor junction. RMS style testing with sine is just a metric. In practice many amps in a home environment wont ever use there full power rating even in terms of a DIN rating. Music power values may vary but DIN have a standard that uses similar ideas. How long an amp can hold some power level etc.
Perhaps the heatsink answer is to design according to power levels that are actually needed. Why DIN etc - it makes it easier to state a higher power level. Power in general - given a choice of same spec other than power would a punter choose to be attracted to a 20 or 100w amp? Many might see the 100w as bound to be better.
😉 Suppose all know this anyway.
The link using crest factors etc. A possible argument against it. Say some one takes a piece of metal and turns the gas stove on and heats the end of the metal. Would they have to drop the metal immediately? No as heat takes time to conduct. Pass, just wondering about a tiny little semiconductor junction. RMS style testing with sine is just a metric. In practice many amps in a home environment wont ever use there full power rating even in terms of a DIN rating. Music power values may vary but DIN have a standard that uses similar ideas. How long an amp can hold some power level etc.
Perhaps the heatsink answer is to design according to power levels that are actually needed. Why DIN etc - it makes it easier to state a higher power level. Power in general - given a choice of same spec other than power would a punter choose to be attracted to a 20 or 100w amp? Many might see the 100w as bound to be better.
Last edited:
It’s not that hard to construct a simple behavioral model of a class AB output stage and determine the average power dissipated - given a pink noise signal (you can get a sound card to generate this, and spit the digital output) and a given load (say 5+j3). It doesnt need to be broken down into microsecond intervals like a Spice analysis - milliseconds will do. Run it for a statistically long enough interval and do the numerical integration. THE problem is what to use for the input ”signal”. How far into clip do you run it? Thats entirely up to the user. You’ll end up with some value when the peak output voltage of your noise signal just hits the rail. It will approach some asymptote as the signal is “turned up” and scaled by a factor of X, then allowed to clip progressively more (Minimum and average values will continue to rise, maximum values will not). One could design for just the onset of clipping, absolute worst case, or something in between. The SAME methods may be used to determine the appropriate VA size for the power transformer such that it doesn’t overheat long term but isn’t bigger than it needs to be. It’s one more column of data.
If you are worried about under/over sizing and do not want to design for the max possible dissipation (2X Vcc^2/(10RL) + PQ) then it’s worth it to go through the math for at least some test cases even if you don’t do a full study. One could do a full up analysis in a circuit/system simulator - given an exact statistical test signal, and complex load over frequency. It’s done all the time for cell tower base station amplifiers, running LTE signals. That’s way more complicated than it needs to be here but the idea is the same.
Self measured the cumulative distribution function (CDF) of signal level, of actual music. His plot is attached below. At the upper left, his graph says there is >95% probability that the signal level is greater than or equal to zero. At the lower right, his graph says there is <5% probability that the signal level is its maximum.
Put this together with your particular amplifier's dissipation-vs-level curve, driving your worst-case reactive load, and you get your amplifier's heat dissipation, while playing music and while driving that load.
_
Put this together with your particular amplifier's dissipation-vs-level curve, driving your worst-case reactive load, and you get your amplifier's heat dissipation, while playing music and while driving that load.
_
Attachments
Thanking you guys there seems one calculative approach that could be done with the link provided https://neurochrome.com/pages/thermal-design by TomChr. Considering it as one of the approaches to solve the problem.
considering the amp would require 33.96W of constant heat dissipating capability need to see two more parameters on this is what would be the die temperature with the given ambient temp into consideration.
The reason to ask is that using over massive heatsink is also one problem in getting the weight and cost of the amp going up considerably. Hence thought of a calculative approach.
Any recommendations on how to calculate the perfect way for die temperature?
considering the amp would require 33.96W of constant heat dissipating capability need to see two more parameters on this is what would be the die temperature with the given ambient temp into consideration.
The reason to ask is that using over massive heatsink is also one problem in getting the weight and cost of the amp going up considerably. Hence thought of a calculative approach.
Any recommendations on how to calculate the perfect way for die temperature?
Last edited: