Bootstrapped capacitances can cause problems at RF. For instance a transistor buffer with its collector bootstrapped to its output may become unstable with certain source impedances. The problem with bootstrapping is that you don't damp the reactance, you just bend it over backwards which can make the problem worse. A buffer can be damped at its input, but for a power amp the goal is to not need damping inside the circuit, because that is often costly in terms of performance.
Can you explain your "correct"?
Most people would take that to mean closer to 'real life'.
If you subscribe to this view, can you provide evidence for this case?If you don't, please ignore everything I say.
I've done this for prototype amps. Our Service Dept. argued strongly against it. They said most Service Engineers don't RTFM and would assume it was isolated.
Perhaps I misunderstand but I don't see what the Service Department was worried about.
What is the problem if the service tech. assumes the heat-sink is isolated?
Obviously a risk if the situation is reversed and the heat sink is live and it looks like it is earthed, or looks like it should have been earthed and so invites a catastrophic "fix" of the "faulty" earth connection.
Best wishes
David
I can understand that it may be unsuitable for mass market production.
Mr Lee's Service Dept. was probably correct.
Inevitably, some well intentioned service techs would try to make a nice, clean connection to the heat-sink.
This is just for my own use, so different trade-off.
And I just meet a friendly bloke who owns an anodization factory.
Obviously destiny wants me to do this
Maybe if I can create a simple fail-safe connection into the protection system?
Best wishes
David
Last edited:
I meant isolated from the Output devices.
I have to admit I blew up one of my own prototypes poking around
That's what convinced me that the Service Dept was right.Still not sure we understand each other.
I meant to have a normal, earthed heat-sink, isolated from the output devices.
The only difference is that instead of, for instance, Kapton washers I consider to heavily anodize the base of the heat-sink and use the anodization as the insulator. Aluminium oxide is an excellent electrical insulator and a fine thermal conductor and is seamless with the heat-sink, so no interface thermal resistance. But how reliable in practice, I wonder?
Best wishes
David
Dear Dave,
don't do this as the anodization can't guarantee a constant oxide thickness. Due to small errors during cleaning for anodization there can be small channels of pure aluminium reaching nearly the surface.
BR, Toni
I don't think this is the core issue, but the cost of thick anodizing. Much more expensive than a few mica (or whatever other thermal conductive material) washers, while the thermal performance is not that bad.
Ultimately, throwing in another pair of output devices (to increase the output stage SOA) is much cheaper than any sophisticated thermal management solution, anyway.
You are right about the additional output devices, but I don't agree with your comment about direct mounting onto anodizing.
If cost was not the issue but performance (incl reliability) you'd see it in military, auto, telecom and other demanding applications. I don't know of any that do this.
I've seen thermally conductive epoxy used to bond power mosfets to a heatsinks for Auto ABS (now they mainly use bare die on an alumina substrate) but never straight onto anodizing.
Just don't do it.
I would not compare the reliability levels required for military, auto or telecom with those for consumer products.
I've seen demanding military applications with diamond or sapphire substrates for thermal transfer management, from the chip down to the heat sink.
Coming from a materials side I think you would have to look at the thermal resistance of a mica or other insulator and the filled epoxy system. Don't forget that even in that epoxy the filler will be usually an aluminum oxide filler and not pure aluminum. So I doubt there is any advantage to the filled epoxy except that perhaps that is used as the mechanical attachment method and no other mechanical fastener is used to attach the device. If anyone used pure aluminum filler in an epoxy as a filler you would have the same situation you are talking about with an anodized aluminum finish with the possibility of leakage paths if any of the aluminum made a direct connection across the isolation, if there becomes a connection, not a good idea. You also have to look at the thermal conductivity of an aluminum filled epoxy, it is not very good really. It is dominated by the epoxy resin, I have had to have this discussion many times with plastics molding, it is never a good solution.
Last edited:
Isn't an advantage of epoxy systems or elastic type washers that they increase the effective contact area.
For metal on metal (with or without annodize) the effective contact area is a function of the fastener pressure as well as material finish. Hard annodized finish can sometimes have worse thermal conductivity across the interface, the theory being that with plain aluminum its "ridges" get squashed down when torqued increasing microscopic contact area.
For epoxies Diamond fillers are common but expensive and when used in very thin applications are't all that worth it compared to alumina or boron nitride.
Thanks
Antonio
For metal on metal (with or without annodize) the effective contact area is a function of the fastener pressure as well as material finish. Hard annodized finish can sometimes have worse thermal conductivity across the interface, the theory being that with plain aluminum its "ridges" get squashed down when torqued increasing microscopic contact area.
For epoxies Diamond fillers are common but expensive and when used in very thin applications are't all that worth it compared to alumina or boron nitride.
Thanks
Antonio
Even mounting straight to the anodization you'd want to use thermal paste because your surface probably isn't flat enough to contact the transistor very well.
I've mounted a 2SK955 to a heatsink with a <0.75mil gap, filled with NTE424 thermal compound. Since my multimeter thermal sensor was not good enough, I used the diode function and assumed a 2.2mV/C tempco for a 1N4148. It was at the measurement limit, but I would say the thermal resistance was =<0.06C/W.
I measured the capacitance to be 190pF, which is much smaller than I thought it would be. I would think the capacitance would be 10 times higher for a gap more than 10 times thinner than a typical sil-pad. Perhaps the sil-pads are much more capacitive.
I've mounted a 2SK955 to a heatsink with a <0.75mil gap, filled with NTE424 thermal compound. Since my multimeter thermal sensor was not good enough, I used the diode function and assumed a 2.2mV/C tempco for a 1N4148. It was at the measurement limit, but I would say the thermal resistance was =<0.06C/W.
I measured the capacitance to be 190pF, which is much smaller than I thought it would be. I would think the capacitance would be 10 times higher for a gap more than 10 times thinner than a typical sil-pad. Perhaps the sil-pads are much more capacitive.
We could say that ideally you'd want no more than 10% extra thermal resistance for the transistor package, in which case if your transistor has 1C/W Rjc, then you'd want less than 0.1C/W insulator resistance. If you use a 1C/W insulator with such a device your max continuous dissipation is halved.
The problem with solid insulators is that they are usually way too thick. For audio you don't need more than 300V/mil puncture voltage. Using thermal paste instead allows us to choose the gap width and use very thin gaps inexpensively. For this reason I've always been very interested in this idea:
http://www.diyaudio.com/forums/pass-labs/149707-another-way-insulate-transistors-heatsinks.html
But now I've found the same can be done without using a heat spreader bar.
The problem with solid insulators is that they are usually way too thick. For audio you don't need more than 300V/mil puncture voltage. Using thermal paste instead allows us to choose the gap width and use very thin gaps inexpensively. For this reason I've always been very interested in this idea:
http://www.diyaudio.com/forums/pass-labs/149707-another-way-insulate-transistors-heatsinks.html
But now I've found the same can be done without using a heat spreader bar.
Anodize is not a cost problem in my case.
The elimination of lots of fiddly little washers would be nice.
But not if anodize is as problematic as Andrew and Toni think.
Another pair of output devices has its own costs.
For my main amps with +-75 V rails and 200 mA quiescent that adds 30 W for each channel.
For a 5.1 system with Bi-amps that's almost 300 W, not trivial.
Also the PCB becomes more expensive as it widens to fit the extra transistors because the output array sets the critical dimension.
Plus the extra capacitance for the driver transistor load.
More coupled transistors are potentially less stable etc. etc.
Trade-offs as usual, but I would prefer to keep the transistor count unaltered.
That's not what I discussed, or what was in Bob's post that I quoted. Hence the confusion.
It is however an idea I have considered, except with copper bus bars.
Yes, once the insulator is a small fraction of the total thermal resistance then it doesn't matter much if it's a little more or less.
The best insulator I have found so far is the Laird k52-1
Link here.
It's this or the anodize.
Any one know a lower thermal resistance product?
Best wishes
David
Last edited: