yes this would be perfect for heat dissipation but electrical dangerous. the potential is on the heat sink and you have to mount this heat sink isolated from the rest!
V- is on that heat sink
i personally use ceramic insulators...
kr
chris
If you mount the LM3886 to the heat sink without an insulating washer the heat sink will be connected to VEE. That is not a good practice. Metal parts of an amplifier, especially those that can be touched by the end user, need to be grounded. So an insulating washer and shoulder washer for the mounting screw are required with the metal-back LM3886T.
Sure. Heat dissipation is better, but you create an electrical safety hazard. You also run the risk of blowing the LM3886 if the heat sink accidentally touches anything grounded while the power is on. Ask me how I know... 😉
I seem to recall that the thermal resistance from junction to case on the LM3886TF is about 2 K/W versus 1 K/W for the LM3886T. Many silicone pads have thermal resistances around 1 K/W so you basically don't gain anything by using the LM3886T + pad versus the LM3886TF. Only if you can find an alumina or Kapton pad that fits the package (the ones for TO-247 usually do) can you do better with the LM3886T + pad. The Keratherm Red (86/82) available from the diyAudio Store (and from my store) are excellent too.
For the best thermal performance with silicone pads you need to torque the mounting screws to the torque recommended by the manufacturer. For the Keratherm that's surprisingly little torque (0.45 Nm). I use a torque screwdriver for this. It's basically finger tight plus maybe 1/4 of a turn. Not much torque.
Tom
One can also use a larger block/slab of aluminum as a head-spreader, mounted directly to the chip(s), and then mount this head-spreader isolated to a main heat-sink. This has the advantage that a very thin Kapton sheet or similar can be used because the area is large. Total thermal resistance is effectively as good as direct mounting.
The thing is, though, that the insulated package option, the LM3886TF, provides the full performance so why all this geekery? I get that it's fun to geek out about things, but it's completely unnecessary.
Tom
Tom
I have worked with amplifiers for 46 years including import/distribution QC and repair 😎
Fascinating. I'm frankly at a loss for words here. That doesn't happen very often so I guess I'll go mark my calendar.
Tom
Tom
Any advantage of copper plating targeted surface area of aluminum heatsink ?
Or friction welding copper plate to aluminum surface ?
Or friction welding copper plate to aluminum surface ?
Normal hifi amplifiers are not constant voltage output devices. They have an internal resistance hence lower output into a lower external impedance unless many output devices are in parallel aka J.Curl, Haffler and many others who can deliver the voltage into lower impedances.
Clipping is not the issue as the voltage would drop across the internal resistance. However the sound spectrum changes as Mr. Pass can tell you.
Clipping is not the issue as the voltage would drop across the internal resistance. However the sound spectrum changes as Mr. Pass can tell you.
No, but they come pretty close.
Take my Modulus-86 for example. It's an LM3886 with error correction applied. At 1 kHz it has an output impedance of 12 mΩ, i.e., 0.012 Ω.
So if I dial the input amplitude such that the Modulus-86 provides 10.0 V RMS at 1 kHz with no load, it will provide 9.99 V RMS into 8 Ω and 9.97 V RMS into 4 Ω. That looks like a pretty constant voltage to me.
The amps by Rowland, Hafler, et al. will have a finite and non-zero output impedance too.
Tom
thats a good idea, as I have the genuine T versions with me, so no plans of buying new TF ver. , especialy with so many fakes floating around
Last edited:
This scenario you posted is exactly what I had in mind >
The results far exceed what I thought & expected. ChipAmps have obviously come a long way. Maybe the LM3886 has multiple output transistors (?)
Copper plating is too thin to actually help unless you are going to solder to the aluminium heatsink (which doesn't look a good idea). Friction welding a copper plate could help spread the heat again depending on thickness of the plate.
I may be wrong here but isn't that why NFB is useful to make the output resistance near zero?
The output devices will provide as much current as they can to provide the output voltage defined by NFB (that is: input signal × gain factor). Once the internal resistance is too high (or the supply voltage too low) to provide this output voltage the amp is clipping.
Wouldn't thiin copper layer transfer heat rapidly to cooler area of aluminum heatsink. I mean further away from IC ?
In a way, but heat transfer works like voltage en current. A thin layer will transfer well "through" the layer from the LM to the heatsink. But the resistance in the spreading direction is high because of it being so thin. So it won't help spreading the heat. If the layer is thick (several mm) then the thermal resistance in the spreading direction becomes low enough for it to have a contribution.
Also keep in mind that the problem with heat transfer is the contact surface imperfections and the difficulty to get the two surfaces to fit togehther perfectly over their entire surface - they nearly never do, unless you press them together with massive force up to the point where metal surfaces to deform and fit.
so usually you need a soft material to fill the gaps. It's better to fill the gaps with a mediocre thermal conductor (thermal grease, thermal foil etc) than to leave them and have thermal conductivity of air at the gaps.
Have a look at my tests here.
so usually you need a soft material to fill the gaps. It's better to fill the gaps with a mediocre thermal conductor (thermal grease, thermal foil etc) than to leave them and have thermal conductivity of air at the gaps.
Have a look at my tests here.
Yep. The output impedance is reduced by a factor of (Aß+1), where A is the open loop gain and ß is the feedback factor.
Tom