Connect with us

Heat Sink for a LM4780?

Discussion in 'General Electronics Discussion' started by solo2racr, Dec 13, 2013.

Scroll to continue with content
  1. solo2racr

    solo2racr

    142
    0
    Aug 21, 2013
    I tried asking this at DIYAudio.com forum and got the typical answer........None. I guess unless your name is Nelson Pass, getting an answer is like pulling teeth.

    Anyway.......I'm within about a week or two from ordering a LM4780 kit from AudioSector and the rest of the parts needed to build a 4ch. amp. I have been working on case design and am wondering about the heatsink. I have some 1.5" x 3" aluminum channel I was thinking about using for the sides and mounting the chips to. At about 7" long, what is the opinion about using them as heatsinks? One on either side for each chip. It's just simple channel without fins. But, it is a pretty good size (1.5" X 3" x 7"). If this is enough, I will go with it as I already have the channel.

    Alternatively, I could go for a 3" length of the 7.28" wide heatsink from Heatsinks USA. Store Surely this would be enough. If this is the route everyone thinks I should go, what would be a minimum length? I am thinking that the 7.28" would be plenty.

    I have tried to calculate the channel to see if it is enough but, The numbers are beyond what I can understand.

    Here is a link to the datasheet for the LM4780 (It's to big to attach.)

    http://www.ti.com/lit/ds/symlink/lm4780.pdf
     
  2. (*steve*)

    (*steve*) ¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd Moderator

    25,505
    2,852
    Jan 21, 2010
    Do you know the rating of the heatsink you require? You might be able to run some tests to determine if your aluminium channel meets those specifications.

    Depending on various factors you may need to insulate the device from the heatsink (see note 1 on page 3 of the datasheet), and you will want to use thermal compound.
     
  3. solo2racr

    solo2racr

    142
    0
    Aug 21, 2013
    All I know is that there are specs on page 4 of the datasheet and formulas to calculate heatsinks on page 17. I just don't understand how to read the specs, let alone understanding what answer I may get from the formula.

    I had planned on insulating the chip from the heatsink. I am also going to mount it with a strap, as opposed to screws directly on either side of the chip die.
     
  4. (*steve*)

    (*steve*) ¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd Moderator

    25,505
    2,852
    Jan 21, 2010
    OK, the maximum package dissipation shown on any of the graphs is 70W, so I'll go ahead on the assumption that you're going to drive it no harder than that.

    And I'll assume that it's running continuously at this power (even though it won't) and is allowed to get to a die temperature of 130C. I'll also assume the ambient temperature is 25C (which means highly adequate airflow)

    ok, so we have a 1.5 DegC/W budget. The device's junction to case thermal resistance is 0.8 degC/W, insulation will add about 0.5 degC/W, so we have used up 1.3degC/W. 0.2DegC/W is simply not going to be achievable with a normal heatsink, so you'll need forced air cooling or you'll need to tweak my assumptions.
     
  5. solo2racr

    solo2racr

    142
    0
    Aug 21, 2013
    Thanks Steve. I got more out of reading your last post a couple times and thinking about it for a few minutes than an hour previously spent on tech that was over my head.

    Looking over the data sheet again, in this case 65W would be a closer call on maximum package dissipation (+/-25v @4ohm load). Of course, they will be no where near this load continuously. Each chip will each see the same load. Left chip high and low freq and right chip the same. Of course, the "high" freq side will not require the same power as the "low" freq side of a chip. With that, referencing that last amp I used which had VU meters on the front, typical levels observed where peaking in the range of 5W to 8W on the low freq side. I don't know if they were measuring peak or RMS values but, with the "peak hold" button on the meters were in that range. If they measured RMS, then peak would be in the range of 12W. Of course, some margin above this would be good from a heatsink POV but, as you point out, running continuously at the chips peak output will never happen. A sine wave just before clipping isn't pleasant to listen to. :)

    The part I didn't understand was "1.5 DegC/W budget". If this is a reference to what a heatsink will dissipate, I don't understand where/how you arrived at that figure.

    Another part I don't understand is the specs on the heatsinks from HeatSink USA. They seem, to me anyway, to be backwards. The 7.28" section, @ a 3" length says....

    "C/W/3": approximately 1.30"

    Where as the 3.5" section @ 3" length says...

    "C/W/3": approximately 2.50"

    I am sure I am interpreting this wrong but, it seems like the smaller section dissipates more heat with the high C/W number. Here are links to the above mentioned heatsink sections.

    7.28 section

    3.5 section

    From what I have read, "C/W" is Celsius/Watt. I am thinking (apparently wrongly) that a higher number cools more.

    I have been playing with the HeatSink Calc but I'm not sure what it all means. Attached is a screenshot of what I entered. The bottom of the left hand column asks for an enter named "Watts". I used "30" from looking at the datasheet on page 4 where it says "Thermal Resistance" and says "30 C/W". On this, I have NO idea if it is right or not.

    The dimensions I used in the HeatSink Calc are those of the 3" aluminum channel I mentioned before. Looking at the numbers the calc generated, I am clueless as to the channels ability to pull sufficient heat away from the LM4780 chip. Plus, I don't know how to allow for the insulating pad under the chip.

    Thanks again for all the help, Steve. Any more info you can pass on, I'm sure, will be enlightening.
     

    Attached Files:

  6. BobK

    BobK

    7,682
    1,689
    Jan 5, 2010
    The C/W rating of a heat-sink tells you how may degrees C the temp will rise per Watt of dissapated power. So the lower the number, the more heat it can dissapate.

    For example. if you are dissapating 10W the 1.5 C/W will rise in temp by 15C, the 2.5C/W will rise by 25C.

    Bob
     
  7. (*steve*)

    (*steve*) ¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd Moderator

    25,505
    2,852
    Jan 21, 2010
    Let's say we have an ambient temperature of 20C (that means the chip is at 20C when it's dissipating no power.

    Let's say we want to keep the chip under 120C (and that's the silicon die within the package)

    This means we can allow the temperature to rise by 120 - 20 - 100 degrees C

    Now let's assume the chip is dissipating 25W.

    If we divide the allowable rise in temperature by the maximum power dissipation, we get the maximum thermal resistance between the chip and ambient temperature air.

    In this case it's 100/25 = 4, and the units are degrees C per watt.

    So that's our budget, we need to keep the entire system within this thermal budget or the chip will get too hot.

    Thermal resistance from the chip to the air is made up of a few things. First is the thermal resistance from the junction (the chip inside) to the case. Then there's the resistance between the case and the heatsink. Finally there is the thermal resistance between the heatsink and the air.

    We have a budget of 4 degC/W.

    Let's assume the thermal resistance between the junction and the case is 1 degree C/W, and the thermal resistance between the case and the heatsink is 0.5 deg C/W. Now, out of our 4 degC/W budget, we've used 1.5 degC/W. SO the heatsink needs to be 4 - 1.5 - 2.5 deg C/W or better (better is a lower number)
     
  8. solo2racr

    solo2racr

    142
    0
    Aug 21, 2013
    OK...I think it's starting to come together in my head. Going by the Heatsink calc, the aluminum channel is good for 3.779C/W. And as per the datasheet, the die case is at .8C/W and the thermal compound is at .2C/W. So this gets me to 4.779C/W. Chip protection comes in at 150C. 150 / 4.779 = 31.38watts. In theory, the protection would kick in at a sustained power output of 31.38watts or above.

    Now, same numbers but with the 7.280" heatsink sited above, 1.3C/W + .8C/W + .2C/W gives 2.3C/W. 150 / 2.3 = 65.21watts.

    Given that the chip really starts to distort at 50watts (+/-25v on a 4ohm load), the 7.28" heatsink from Heatsinks USA would cover the chip and then some, under any conditions.

    Real world, the channel would probably work as well. At 31.38watts, that's more than I would push it in normal use. In fact, it would probably about peak wattage when I push it.

    This amp will be used to power a pair of near field monitors. They are less than 1meter from my head. Efficiency is listed as 91.5dB/m for the woofer and 90dB/m for the tweeter. Between actual dB measurements, the dB specs for the drivers and the VU meters on the previous amp, I have to believe what the meters have told me. That being, when I push them, the most I will use is about 30-35watts, and 5-8watts under normal use.

    So, do I have this sorta figured out or am I still lost? If I am about right on the calcs, what would you do, regarding the aluminum channel vs the proper finned heatsink?
     
  9. (*steve*)

    (*steve*) ¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd Moderator

    25,505
    2,852
    Jan 21, 2010
    I think you've got the idea of the calculations.

    Now to let you in on a few other secrets:

    1) heatsinks are (almost) never as good as they're advertised. Something which claims to be 1.3 degC/W might be 1.3 degC/W if mounted in a specified position with a large amount of free air around it and without any dust on it. In practice, it may be better to regard it as half this efficiency (i.e. 2.6 degC/W) if you have less than perfect placement and where dust may build up over time.

    2) Average power is typically far less than maximum power. Whilst this varies hugely with the material you're playing, about the only thing which imposes a constant, predictable load is a pure sine wave (or really excessive clipping). Most other material will have a range of volume levels and the average power might be 1/10 of the peak power.

    3) Thermal resistance is more complex than I've said it is. Things have thermal mass too. This means that the temperature does not rise to a given level immediately the power dissipation changes. The change in temperature is a smooth process. the speed being determined by the thermal mass of the system (as well as the amount of the power change). This means that a smaller heatsink is OK as long as the heatsink can continue to absorb the extra heat during the peaks and release it during troughs.

    These three things combine to mean that with audio amplifiers (and other devices having dynamically variable loads), whilst you need to understand the characteristics of what is being amplified, often you can get by with a heatsink quite a bit smaller than the calculations initially suggest.

    The best thing to do is to operate the amplifier whilst probing the chip for temperature. This will tell you how hot it's getting. Remember that the junction to case thermal resistance means that the temperature of the junction will be hotter than the case though!

    After you've measured it, you can tweak the heatsinking.

    If you're only building a single unit, it's probably easier to design it as if it were going to have a constant load, knowing that it will run a lot cooler than you calculate.
     
  10. solo2racr

    solo2racr

    142
    0
    Aug 21, 2013
    Thanks for the insights, Steve. I wasn't aware of number 1 but do know about 2 & 3.

    Regarding your last sentence...

    "If you're only building a single unit, it's probably easier to design it as if it were going to have a constant load, knowing that it will run a lot cooler than you calculate."

    That is the case here. Just a single unit. Since my last post, I have given it some more thought and decided that the 7.280" X 3.5" (2RU high) heatsink from HeatSink USA would be the best route. Not that the channel wouldn't work day in and day out. But why push it? There is no such thing as to much heatsink, where the chip is concerned. I can fit them comfortably on either side of the enclosure and they, I think, will be aesthetically pleasing.

    One last question. Thermal paste or insulating pad? I know the sinks will be electrically "hot" with the paste and not with the insulating pad. Would it be a problem if the two chips (and their own heatsink) were tied together electrically? The back of the die is tied to the V- pin out of the chip. (ref.page 3 from the datasheet).

    Here is a link to the schematic from the AudioSector kit if it is helpful.
     
  11. (*steve*)

    (*steve*) ¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd Moderator

    25,505
    2,852
    Jan 21, 2010
    There may not be an either/or with the thermal paste and insulating pad. Some insulators will require a smear of heatsink goo on either side.

    If you are putting on heatsink compound, remember that you only need enough to full the air spaces in the rough surfaces of the device and the heatsink. Obviously both surfaces are pretty smooth, so the amount you need is *tiny*. If you put too much on, things get worse!

    I would be tempted to place the channels in such a way that it makes a tunnel. If you can arrange this vertically with openings at the top and bottom of your case, you will get a lot of assistance from convective airflow.

    In addition, if it ever needs more cooling, it provides an obvious place to connect a fan!

    You can probably have the heatsinks connected together if both devices have the same V+. The important part is that they must either be *really* connected together, or insulated. If they are kinda connected, then voltage drops can cause currents to flow in such a way as to introduce noise. My inclination would be to insulate them from each other (a 3mm gap between them, or some thin spacer between them would be sufficient as they are effectively at the same potential.
     
  12. solo2racr

    solo2racr

    142
    0
    Aug 21, 2013
    Attached is a quick drawing of what I had in mind, as viewed from the top. The heatsinks make up the right and left hand sides of the enclosure. The question about them being tied together comes from the use of what material to use to tie the enclosure all together, aluminum or wood/plastic. The front will be a nice piece of wood. It's the back, top & bottom where I have the choice. For me, aluminum is the easiest to work with and to get.
     

    Attached Files:

  13. (*steve*)

    (*steve*) ¡sǝpodᴉʇuɐ ǝɥʇ ɹɐǝɥd Moderator

    25,505
    2,852
    Jan 21, 2010
    OK, I see, that's a fairly conventional layout for the heatsinks.

    Just make sure there is enough roome aboce and below to allow adequate airflow, and as they're exposed I recommend you insulate the devices from the heatsinks.
     
  14. solo2racr

    solo2racr

    142
    0
    Aug 21, 2013
    I was just going over the supplied parts list from AudioSector for the LM4780 kit and they do include insulating pads. My plan is to just sit the amp alone on a shelf between the two speakers and behind my computer monitor. Though, I am considering adding removable rack mount ears for the possibility that I may put it in a rack sometime in the future.
     
  15. solo2racr

    solo2racr

    142
    0
    Aug 21, 2013
    Thanks for all the help, everyone. I was going to post the build of this amp when I get all the parts but, I guess that would be better for a new thread.

    Thanks again :D
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-