I'd expect you still need fairly close mechanical coupling between pipe and bock. Thermal paste is not that good if it is a thick layer. (Where thick can still be quite small)
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Heat pipe cooling in power amps
- Thread starter MCH
- Start date
Yes, the fit was very tight. It used screws to clamp the pipe and the machined block together.
Prana Ferox
Addicted to Fun and Learning
I would think this would be easier to do Hypex-style where you thermally couple the bottom of the board underneath the chip amps to the case (or a heat sink functioning as such.)
Yes, it is a very effective and cheap on the fly method. I specialized in high end car audio service for some years and saw many amps using the clamshell clamping design and it was the worst for thermal tracking and thermal runaway. Hypex made serious improvements to their clamshell design and I hope it works out well for them.
There are DIY heatpipe systems for GPUs and even for using computer casing as the heatsink. No problem with those.
Heatpipes only transfer heat from one solid to another, they don't transfer to the air. Fins transfer to the air. Heatpipes are used to send heat to faraway fins. And the cooling of amps doesn't even require that many fins.
Heatpipes only transfer heat from one solid to another, they don't transfer to the air. Fins transfer to the air. Heatpipes are used to send heat to faraway fins. And the cooling of amps doesn't even require that many fins.
I had a little Denon receiver - DRA-350 - that used heat pipes too. Made in 1985. It was a nice feature but probably overkill for its modest 36W/ch rating.
@restorer-john probably has the most insight. In the late 70s and early 80s, it seems like heat pipes were common with Sony, Marantz, Luxman, Kenwood and Sansui. I feel like back then, there was more emphasis on small and slim components and products were actually kept in bookshelves.
I am not sure if cost savings or the desire for “bigger” amps changed the game. People talk about how light class D amps are, but stuff like the Adcom GFA-555 were surprisingly light for the amount of power they could put out. Crown commercial PA amps put out a ton of power for their size too.
- Thread Starter
- #28
The initial idea (post 1) was to use CPU like clamps to ensure a good contact. Based on the answers I received, I will go for a mechanical attachment but not so sophisticated - some sort of clamp, but simpleri need to figure it out. I didn't draw the holes for the mechanical attachment in the PCB for simplicity. It is just a draft.
I don't think it is possible to DIY solder heat pipes, they might explode, I plan to use heat transfer tape at the amp side. At the radiator side, they sell radiators with holes for the pipes and a small side dent to refill with thermal paste:
The hole point of the project is not to use the enclosure as heat sink and make a multichannel amp as small as possible. If i see it is feasible, I will even 3D print the enclosure with the minimum parts made not of thermoplastic, what would give me a lot of design freedom. This is the main objective, see if it can be done.
Btw last night I was rearranging parts and am in 10x6.5 cm now, the big caps being the limitation at the moment.
I am still trying to figure out if I will need to operate the 4 amps in master/slave mode and if it is worth having the microcontroller for fail signal reporting and most specially for reset. Advice is welcome (there will be a microcontroller anyways to control the pem fan).
I don't think it is possible to DIY solder heat pipes, they might explode, I plan to use heat transfer tape at the amp side. At the radiator side, they sell radiators with holes for the pipes and a small side dent to refill with thermal paste:
The hole point of the project is not to use the enclosure as heat sink and make a multichannel amp as small as possible. If i see it is feasible, I will even 3D print the enclosure with the minimum parts made not of thermoplastic, what would give me a lot of design freedom. This is the main objective, see if it can be done.
Btw last night I was rearranging parts and am in 10x6.5 cm now, the big caps being the limitation at the moment.
I am still trying to figure out if I will need to operate the 4 amps in master/slave mode and if it is worth having the microcontroller for fail signal reporting and most specially for reset. Advice is welcome (there will be a microcontroller anyways to control the pem fan).
Very yucky stuff the tape is. The thermal tracking and thermal stability will be very poor with thermal tape. You are creating a thermal runaway situation by using thermal tape. Use proper thermal paste.
Use thermal epoxy. It is very effective if you keep it very thin and don't use gobs of it.
- Thread Starter
- #30
Will leave a big pad without solder mask, but will only be possible if the legs of the through hole inductors allow. If not, I will try to use Würth inductors that are surface mount, are cheaper, and TI claims perform better. But we'll, they also claim they have the smallest footprint... I guess they where a bit drunk when they wrote the technical note....
Well noted, thanks.
Right - precise machining requred. This may be more difficult than soldering for a one off homebrew attempt.
But again - unless you use some sort of machined block, you can't get thermal coupling from a flat surface to a heat pipe. I'm going to suggest you do some experiments. Get a power resistor with a flat surface. Get a heatpipe. Get a heatsink.
See how much heat you can get from the resistor to the heatsink using the heat pipe, with your proposed methods of coupling. If it is working well enough, you should expect the heatsink surface next to the heat pipe to be close in temperature to the resistor surface next to the heat pipe. Further, if the resistor is dissipating the sort of heat your amp is, then you should be able to see how effectively you can cool it via the heatsink. (How hot without the heatpipe/heatsink connected vs how hot with)
Otherwise you risk spending a lot of time building something that can't work thermally.
Last edited:
Sokel
Master Contributor
- Joined
- Sep 8, 2021
- Messages
- 6,157
- Likes
- 6,251
That is also a think for some fanless PC:
silentpc.com
H1 Fanless Mini PC
The HD Plex H1 Fanless Mini PC is a true accomplishment in small, fanless (no-noise) computing, utilizing up to a 8-core CPU.
- Thread Starter
- #35
This is something I need yet to decide, if I go for flat pipes (disadvantage is more difficult attachment to the radiator) or if I try flatten them myself at the section that touches the amp, but in any case the pipes will be flat at the amp side, what together with the paste and the mechanical clamping makes me hope it will be enough contact.
In the very last case, the same guys that sell the radiators also sell aluminium blocks with semicircular indents to fit the pipes, but I was hoping to be able to avoid this.
It won't be.
In a former life I project managed design of power electonics. Heatsink mating surfaces need flat to quite a tight tolerance. Flattening a cylindrical heat pipe won't hack it. Again I suggest you try for yourself experimentally before you go down any of these Heath Robinson routes without a machined block.
As things stand you are going to get much better results using the housing for your heatsink.
- Thread Starter
- #39
Thanks for your insight. Actually the whole point of the thread was to get this kind of feedback. The previous answers were mostly on the "you won't need much heat transfer" side, that's why I downgraded from CPU clamp to DIY clamp approach.
Will do some experiments as you suggest. Might do them with a prototype of the amp directly. Will also help me to troubleshoot the amp build itself (will be my first one).
Last edited:
- Thread Starter
- #40
The radiator will be fan cooled, laptop style
Similar threads
