The Wire - Low Power Ultra High Perfromance (LPUHP) 16W Power Amplifier

What I'm still struggling to find is a decent gap filling thermal pad that doesn't cost a fortune. I had two scrap pieces of 0.2" thick high compliance stuff from work and it's pretty much perfect as it forms around the buffers and makes good contact with the PCB which is where all the heat goes.

@454Casull see bolded section above ...

The buffer is 0.18" high and opc wants to couple heat in pcb to the heatsink
 
Last edited:
probably there is something on the pcb thats .2" higher than the buffers

edit: nope, just checked, yes opc, why is it there is a thermal gap that needs filling? from the standoffs if you use them between pcb and heatsink?

the thermal surface is facing the board instead of the heatsink. He's just trying to get heat out of the board by bonding the board to the heatsink using some kind of a "sponge" type thermal pad.

Do
 

opc

Member
Joined 2004
Paid Member
Hi Guys,

I've attached a picture that shows the basis of the problem.

The thermal resistance from the die to the outside of the case on a TO-263 pack is very high. It's of minimal use heatsinking to the top surface of the packages since the drop across the molding compound itself is so high.

The tab and copper plane it's soldered to, on the other hand, has very low thermal resistance.

Ideally, we should aim for a gap filling pad that actually has the TO-263 packages cut out of it, and then apply a small amount of thermal paste to the top of the packages and let the gap filling pad conduct heat from the top copper layer up to the heatsink.

To be honest, it's not too big of a deal unless you plan on running these full out for long periods of time. I'm running my pair without any heatsink and they're just fine (if not a little toasty at about 58C) even during moderate use.

The reason for the 0.2" height is that the buffer packages themselves are 0.180" high and the 0.2" height allows the gap filling pad to conform well around the edges and down onto the PCB itself.

Cheers,
Owen
 

Attachments

  • IMG_7022.jpg
    IMG_7022.jpg
    141.6 KB · Views: 715
Doh, of course. hey Owen, love the pimped out half brick recommendation. i use their BMR-C1 on individual to252, great stuff

i like the ribbed HC6158, but perhaps the rods have better thermal performance i'll have to check. GB anyone? i'm also thinking of contacting Richard at sumR Magnetics about a custom potted transformer to suit, as i know he has a few cans he can use to suit and will have one that will fit. we would need someone in Canada or the US to take delivery of the initial bulk order and send them out. i'm going to start a thread in the GB area in a moment for all the potential group buys we can do for this build.

the buffers and lme49990 are an obvious addition and i mean the lme49610 as i cant see a downside for using these, even if you dont need the extra swing right now. pretty sure we could buy direct given the numbers we could possibly get. with 40 boards even if only 30 of us threw in thats 480 buffers and 180 lme49990 i would buy some extras of both at the price we would get. then we have heatsinks, transformers and the main 1210 size ceramics. also LT317/337 or LT1085/1033 and the power supply caps

so who would we contact about getting a gap filler custom Die? and would it become a realistic solution with a number of us to get the cost down? you may be ok running in open air in Canadian winter, but boxed up in a 40c summers day in Australia doesnt sound like fun
 
Last edited:
Doh, of course. hey Owen, love the pimped out half brick recommendation....so who would we contact about getting a gap filler custom Die? and would it become a realistic solution with a number of us to get the cost down? you may be ok running in open air in Canadian winter, but boxed up in a 40c summers day in Australia doesnt sound like fun

How about installing the buffers vertically and soldering them to copper bars. The heat sink will go on top. You can double the buffers this way by soldering them in 4 rows.

On a second thought, the design is great as is :)
 

opc

Member
Joined 2004
Paid Member
What you have modeled would honestly be the ideal heatsink for this. Good contact with both the top of the buffer case and the PCB would allow the heatsink to provide the highest possible benefit.

The idle power is actually extremely low considering the performance (~4W for the output stage), and even at full tilt the dissipation isn't that bad. I personally would sleep better at night knowing that the buffers were as cool as possible though.


What do you think about having the heatsink threaded and screwed on from below the board?

Again, this would be ideal... threading from the bottom makes for much easier installation and a cleaner look.

I can send you the dimensions for the board, screw locations, and buffer height/location.

Would you be willing to take on making 30-50 of these? I would certainly be in for 6 of them.

Cheers,
Owen
 
How about installing the buffers vertically and soldering them to copper bars. The heat sink will go on top. You can double the buffers this way by soldering them in 4 rows.

On a second thought, the design is great as is :)

sounds fiddly wrt lead bending and you lose the effectiveness of the pcb, which already does a fairly good job. interesting idea, but construction would be a pita and perhaps expensive. glad someone mentioned under the board as a possibility, but i was going to wait till i had the boards before thinking about that. the solution Owen and Mark are talking about sounds ideal, but how do we avoid shorting the conductive parts and components? by anodising the heatsink? this is regarding the underside mounting

perhaps a slab of keratherm isolator between pcb and heatsink and just avoid the very front where the lme49990 are?

the only problem i see with the bottom is it will increase the height of the completed unit
I do LOVE the copper heatsink look
 
Last edited:
@qusp I thought it was just screws from the underside? I haven't got a clear picture of what you're describing for the underside though ... I thought HYPERTUNE was suggesting just the inverse of how the heatsink is mounted in opc's photos. ie bolt head only on the underside of the pcb

I am probably going to go for an off board transformer but definitely VERY interested in the idea of GB for lme49990 and 49610 with 3 x bal-se head amps and a pair of the LPUHP Wire all about to be built almost concurrently. I was thinking maybe 20 or 25 of each already. Also pretty keen on those heatsinks too!
 
Last edited:
A GB would be great to bring the overall cost down - especially for the SMD passives. The Susumo thin-film resistors specified are $0.40 in Qty 10, but half the cost at Qty 200.

The best prices I found for the LME parts are from Future Electronics, but they are showing no stock.

Avnet Express are a little more expensive but they DO have stock.

LME49990 = $2.72ea for Qty 25+
LME49600 = $5.65ea for Qty 25+

It would probably be best for someone from North America to handle a G.B from these suppliers, as their freight charges tend to be high to Europe or Australasia. Other than that they're just as good as Digikey in my experience.

-Len.
 
@qusp I thought it was just screws from the underside? I haven't got a clear picture of what you're describing for the underside though ... I thought HYPERTUNE was suggesting just the inverse of how the heatsink is mounted in opc's photos. ie bolt head only on the underside of the pcb

he suggested both machining a custom sink that could provide some good contact with the PCB AND the case of the devices and mounted from the top side; this doesnt present any interference problems. but the underside would be better again as far as thermal transfer because most of the heat is directed into the copper pour in the pcb.

i mentioned the problem of electrical isolation, because to get good transfer it obviously has to be in contact and there are vias on the underside that are live, so a large thermal transfer/isolation pad would need to be used to give the transfer but isolate electrically.

there are 2 ways to do this, with a high performance isolation pad like Keratherm red but just bought in a slab rather than cut for any size device, or a 3M or ideally Sekusui #5760 thermal transfer, but electrically isolated adhesive pad that covers the whole bottom of the area and you bolt down onto it OR anodize the heatsink and take your chances with arctic silver adhesive not applied over the vias.

I am probably going to go for an off board transformer but definitely VERY interested in the idea of GB for lme49990 and 49610 with 3 x bal-se head amps and a pair of the LPUHP Wire all about to be built almost concurrently. I was thinking maybe 20 or 25 of each already. Also pretty keen on those heatsinks too!

yeah i might do that too as i already have 2 x 15vac 50va potted units from sumR, but it would be neater to have it board mounted.

I really do suggest that for a GB we go for the lme49610 instead of the lme49600 but if the price is right and nobody else cares about the extra swing it has available and then again thinking about it, really its not a big deal because the input opamps would have to be changed, as there would be nothing to supply the higher swing even if the buffers were powered separately. i hadnt thought that through, however i need the 610 for my other projects so....

ok well i'll go start a thread then hey? it would indeed be better for someone else to take delivery over that side of the planet to lower shipping costs.
 
Last edited:
arrow look to be the best for the 610 at least, here only 7.68 each singly and lower in bulk if you ask, they dont have the 600 in stock they also have the lme49990 for only 2.61 in small numbers and will do a custom price for a bulk order.

otherwise digikey is actually doing free express worldwide shipping now; just like mouser if the order is over 200. they have only a limited number of the 610 but will get more if minimum order is reached, but they have enough to cover enough 600 without doing that; as well as the 9990 and every other part we would want, so it might end up being easier to go with them
 
Last edited:
ok well i'll go start a thread then hey? it would indeed be better for someone else to take delivery over that side of the planet to lower shipping costs.


Do we need a volunteer/victim on dark (at the moment) side of the planet before GB thread kick off time? Is that only most important if we go with someone other than digikey?

The lme49610 doesn't have any real advantage for my uses. I'm happy to go whichever way is most popular for those interested in the GB. Is quiescent current the main difference for this design between 610 and 600 when running less than 18V rails?

Has the digikey price on the 600 jumped up in the last few days? I swear it wasn't that expensive a few days ago.

EDIT: understand your point that there is a lot of available pcb realestate we could use for heatsinking on underside of pcb. With amount of heat we're dissipating is that going to get into territory of diminishing returns on the effort required to obviate the issues you've already pointed out?
 
Last edited:
What you have modeled would honestly be the ideal heatsink for this. Good contact with both the top of the buffer case and the PCB would allow the heatsink to provide the highest possible benefit.

The idle power is actually extremely low considering the performance (~4W for the output stage), and even at full tilt the dissipation isn't that bad. I personally would sleep better at night knowing that the buffers were as cool as possible though.




Again, this would be ideal... threading from the bottom makes for much easier installation and a cleaner look.

I can send you the dimensions for the board, screw locations, and buffer height/location.

Would you be willing to take on making 30-50 of these? I would certainly be in for 6 of them.

Cheers,
Owen
The problem with such a design is that if the height of the chips changes even slightly, then there will be a gap between the chip and the sink, or the sink and the PCB. Unless, of course, you design the sink to accommodate a compressed pad interface.
 
The problem with such a design is that if the height of the chips changes even slightly, then there will be a gap between the chip and the sink, or the sink and the PCB. Unless, of course, you design the sink to accommodate a compressed pad interface.

If this was a problem a compressed thermal interface pad could be used on both the chips / heatsink interface and the heatsink / pcb interface couldn't it? Thinking about it further this may be prudent even if the heatsink is designed to accommodate the pad above chips.
 
Last edited: