LM3875 Heat Sink Issues

[cjv998]

Hey everyone, I recently bought some heatsinks (Wakefield model 677-10ABP, bottom of the page numbered 47 on this catalog: http://www.wakefield.com/pdf/Board_Level_Heat_Sink.pdf) for my LM3875's, only to realize the fins on the side wont clear the capacitors on the audiosector.com PCB. I've already cut the leads on the capacitors, so I can't move them. I tried bending the LM back so it's parallel with the board (I'm afraid if I bend it much more, the pins on the chip will break), but then the pins on the bottom of the heatsink don't clear the PCB. I could cut these pins off I suppose, but I feel I'd need some way to support the weight of the heatsink (regardless of whether the PCB is mounted horizontally, which I'd prefer, or vertically...maybe if I leave the PCB horizontal, and brace the top flat face of the heatsink up against the side of the case?). So basically, my options are:

1. Cut the pins off the bottom of the heatsink and come up with a way to support their weight.

2. Come up with an alternate heatsink method (my case is 2mm thick aluminum, I don't think that would be sufficient, but maybe I'm wrong?)

Thanks for the help!

[sek]

Hi,

how hard will you be driving the 3875? E.g. what voltage and what speaker impedance?

Depending on your enclosure arrangement and distance between the two channels, the enclosure might already be enough.

One possibility - in case you're not satisfied with 2mm alu - is using aluminium plates as heat spreaders between chips and enclosure. Just make sure your enclosure gets enough air convection on it's outside.

Another possibility would of course be to machine fit the fins...
Your weight support problem could be solved with threaded holes in the heatsinks.

Cheers,
Sebastian.

[cjv998]

Sorry, I guess more specs would help. I'll be powering the amps with around 35V, and the speakers are 8 ohm loads, and I can't imagine I'll ever need to crank up the volume more than halfway. What if I do as I mentioned before, with leaving the PCB horizontal, bending the LM so it's parallel with the board, cutting the pins off the bottom of the heatsink, and then using a standoff to support the weight of the heatsink? I think that might just work!

[jackinnj]

the heat sink is used to dissipate the energy which isn't used to produce sound -- the thermal impedance required for the sink (for the overture chips) is a function of the square of the absolute voltage on the rails and impedance of the load. To make life easy National has an Excel spreadsheet calculator on their website which will allow you to choose an adequate heat sink. The one you have chose has a thermal impedance of about 8.5 degrees C per watt. This is inadequate for an LM3875 with +/- 35V rails.

With heat sinks the area exposed to the movement of air is the most critical factor in determining thermal impedance -- not the thickness of the material. But the math gets beyond exposition in a one paragraph response.

[cjv998]

Originally posted by jackinnj
the heat sink is used to dissipate the energy which isn't used to produce sound -- the thermal impedance required for the sink (for the overture chips) is a function of the square of the absolute voltage on the rails and impedance of the load. To make life easy National has an Excel spreadsheet calculator on their website which will allow you to choose an adequate heat sink. The one you have chose has a thermal impedance of about 8.5 degrees C per watt. This is inadequate for an LM3875 with +/- 35V rails.

With heat sinks the area exposed to the movement of air is the most critical factor in determining thermal impedance -- not the thickness of the material. But the math gets beyond exposition in a one paragraph response.

I'm curious how you found the natural thermal resistance of the heatsink. I saw the forced resistance was 3.1 C/W at 200LFM, but couldn't figure out the natural resistance. On a side note, I bet I'll never turn the volume up more than halfway, but from the 3875's datasheet, I see that doesn't change the power dissipated very much. This doesn't make sense to me, as I would think that power dissipated as heat would scale relatively linearly with output power, but it seems to plateau at ~20W output power for an 8 ohm load with 35V powering the chip. At any rate, I'll take National's word that it doesn't work that way.

Okay, so assuming the heat sinks I have won't cut it, I'm willing to bet that the 2mm thick aluminum chassis won't work out either, at least not without the chassis getting very warm after a while. Does anyone have any more thoughts or suggestions? I know if I just got a solid piece of aluminum, that would work...except that's not something I have readily available.

[gootee]

You might want to try looking at a local hardware store. Most of them sell aluminum bar stock. Some of them have some fairly thick, wide pieces. I think I have seen it up to 1/4-inch thick, three or four inches wide, and several feet long, at someplace like True Value Hardware, or a similar store. In your area, a farm-supply store might be worth checking, too.

- Tom Gootee

http://www.fullnet.com/~tomg/index.html

[sek]

the area exposed to the movement of air is the most critical factor in determining thermal impedance -- not the thickness of the material.

I wasn't talking about a mere distance plate.

I was rather thinking of what Tom also suggested, sheet or bar material.

In my area, even a home depot or Lowes would stock that - had we home depots or Lowes over here.

All this doesn't help cjv988, I'm afraid, as he's still bugged by the question wether he should use a custom tailored sink or his enclosure walls.

This doesn't make sense to me, as I would think that power dissipated as heat would scale relatively linearly with output power

In a class-AB amplifier (such as a chip amp) the power dissipation has it's maximum at or around 50% output power, because efficiency doesn't scale linear with output.

Rod Elliot wrote an article about it, particularly interresting for us is the class-b case.

Cheers,
Sebastian.

[jackinnj]

The 677-10ABP has a thermal impedance of 52/6 = 8.7 in still air. (Still air actually moves at approximately 32 LFM).

For an amplifier of this type the amount of energy dissipated by the chip (and that which must be conveyed away) is:

PdMax = [(V+ + |v-|)^2]/2 pi RL

Here's Nat Semi's inter-active design tool for the Overture series chips:
http://www.national.com/appinfo/audi...gn_Guide15.xls

If you want to use a sheet of aluminum or copper:

[cjv998]

That table is excellent jackinnj. So if I go by Home Depot or Lowe's or something, they may have aluminum I can buy? I'll check this afternoon after work. If I get a 3/16" or 1/4" thick piece, I can get a piece that will separate my case into two halves (the piece would be right under 3" by 8" in size), and cut down on EMI from the transformer as well. One stupid question though: how am I going to secure this large piece of metal in my case? If I got screws with a large enough thread, I suppose they could hold it in there.

[cjv998]

Update: I called some hardware stores in the area looking for aluminum. Home Depot and Lowes don't have anything, Do-It Best has up to 1/8" thick, and up to 2" wide, and True Value has 6" by 24" aluminum sheets, but the associate said they didn't look like they were 1/4" thick (I bet they're 1/8" thick as well). So I'm still having trouble.