On 7/23/2011 11:56 AM, Manfred Mornhinweg wrote:
> Paul,
>
>> Over the years I've heard many people say that adequately cooling a
>> transistor is difficult because it is so small.
> That's only half of the picture. The other half is that it needs to be
> kept very much cooler than a tube!
>
>> It seems to me the dissipated
>> power per area is roughly the same when looking at the devices minus
>> their coolers.
> In some cases, comparing the more compact tubes with the larger
> transistors, that could actually be true. But there is one VERY BIG
> difference: The maximum allowable dissipated power is rated at a core
> temperature of typically 200 to 250 degrees Celsius for the tube, and at
> only 25 degrees Celsius for the transistor! This is what makes cooling
> transistors harder.
Unless something has changed we used to run the core temps of Silicon
power transistors and CPUs as high as 90C, but how well that works
depends on the doping and construction techniques of the Silicon die.
The devices can run quite a bit hotter than that and survive. In the
higher temp ranges, but less than what would cause catastrophic failure,
the doping (N and P type) tends to migrate across junctions causing a
slow degradation in the transistor. Fortunately with CPUs, most actually
have the ability to output the core temp. Unfortunately with power
transistors, I don't know of any that do.
I think we may be into semantics, but I'd not call it a core temp
problem in tubes and they can run considerable hotter than 250 C
internally. They have two limitations: The external *seal* temp which
is normally 200 to 250C with the lower temp *usually* associated with
older designs and construction. *Internally* the grid and screen (if it
has one) are limited only by temps that will allow/cause secondary
emission and the power (if any) they can dissipate. The leads for both
grid and screen are usually small and poor at heat conductivity which is
good for seal temp.
IF you can keep the seals cool the only real limit as to how much power
a tube can run is cathode emission, voltage breakdown, and the emission
limits of the grid and screen. This is why the larger power tubes are
water cooled. It's also why we are seeing an increase in water cooled
heat sinks in high power, solid state amps. OTOH as I and others have
already said, there is a limit as to how low a temp the device will
still operate efficiently and the dangers of condensation.
The main problem with high power transistors is the thermal mass and
thermal inertia. In a transistor this consists almost entirely of the
small Silicon die inside (a few grams), which itself has a high thermal
resistance (does not transmit heat well). In a tube it consists of
almost the entire tube mass as the anode represents on the order of 90%
(or more) of the tube's mass. Hence with a tube (or transistor with
limitations) that has a 1000W dissipation running 70% efficient we can
get roughly 1500 watts output. However if the tube has the emission
capability we could easily run over 2 KW PEP out. The same does not
hold true for the transistor. With the tiny thermal mass you are pretty
well stuck with the maximum dissipation as a limit, unlike the tube. We
could even add water cooling to the tube and increase it's power
capability considerably (If it has the emission). With the transistor we
need all the cooling we can get (water cooled heat sink) to even safely
approach the maximum power rating. As has already been said, we
typically have to derate the power transistor just for normal operation.
You can even exceed the tube's seal temp rating to a point. Too much and
the seal where the ceramic is soldered to the metal will come unsoldered
and the tube will lose vacuum.
73
Roger (K8RI)
> You can rather easily keep a ceramic tube's core below the maximum rated
> temperature, while the tube is producing its full rated heat, by blowing
> a modest amount of air through a modest heat sink (which usually comes
> as an integral part of the tube). At a top room temperature of maybe 30
> degrees on a hot summer day, you still have roughly 200 degrees of
> difference. At that temperature difference, even a small amount of air
> will remove a lot of heat.
>
> But you absolutely cannot keep a transistor at 25 degrees while it is
> dissipating its full rated power, unless you are working outdoors in the
> arctic in mid winter, or you use cryogenic techniques. So the
> dissipation rating of transistors ALWAYS has to be drastically de-rated,
> to account for the actual temperature at which you can hold it. In
> practice, this might look like a nominal "1200 Watt" transistor (at 25
> degrees Celsius) being derated to 600 watts, which allows it to operate
> safely at roughly 90 degrees at its mounting surface, and then you need
> to remove that amount of heat while having only 60 degrees of
> temperature difference between the transistor and the air. That will
> require a much larger heatsink than for a tube dissipating 500 watts,
> and a much larger airflow too.
>
> This is not a "terrible" problem, but one must understand it and design
> the thermal side of an amplifier as carefully as the rest, or the
> transistor will likely overheat and fail. I have seen many ham
> homebrewers just guessing the proper size of a heatsink, and burning out
> their transistors because it was much too small, or because they mounted
> the transistors on insulators that could never transfer heat well
> enough. I have seen this happening even with many homebrew power
> supplies using 2N3055 transistors, which have a really low power density
> compared to many RF transistors!
> And that's why correctly designed solid state amplifiers often look like
> one VERY BIG HEATSINK with a tiny little bit of circuitry attached to it!
>
> And the compound the problem, the highest power transistors actually
> have a higher power per contact area than a tube's core. That's where we
> run into the limitations of thermal conductivity of metals, and only the
> best materials will do. This is why high power transistors usually need
> a heat spreader made of copper, before the heat can flow through a
> larger cross section of a cheaper material, such as aluminum. Even then
> you might still end up with 30 degrees of thermal gradient between the
> transistor's mounting surface and the main part of the heatsink!
>
> > What would be interesting is if a transistor
>> manufacturer took a page from the tube world and integrated similar
>> cooling.
> It's not convenient. The larger size of heatsink required for a
> transistor makes it a larger portion of the total device cost, than in
> the case of a tube. If I had to throw away a big, expensive heat sink,
> because the tiny chip of silicon attached to it burned out, I wouldn't
> like that! I prefer to just replace the tiny little piece of silicon,
> with as little packaging around it as possible. And since most people
> agree with me in this, the market for power semiconductors with built-in
> heatsinks is small.
>
> Paul, to understand the problem of having lots of heat in a small area,
> and having to keep this area pretty cool, I suggest that you look up the
> thermal resistivity of copper (which is the best material available at
> acceptable cost), and that of aluminium, of which most heatsinks are
> made, and the thermal capacity of air, and so on, and do some real-world
> calculations about how to keep a 1x3cm mounting surface of a transistor
> below 90 degrees Celsius, while it is producing 600 watts of heat. Be
> sure to include ALL the thermal resitances: The interface between the
> transistor and the spreader, that of the spreader, that between the
> spreader and the heatsink, the internal one of the heatsink from the
> spreader to the fins, and that from the fins to the air, at a certain
> airflow. I don't know if you enjoy such number play, but if you do, it
> will be enlightening!
>
> I have done it many times, and usually arrive at the conclusion that
> using a high power transistor even at half the rated power dissipation
> can be quite a challenge! As you relax the power level, or use devices
> with lower power density, it gets much easier.
>
> Manfred.
>
> ========================
> Visit my hobby homepage!
> http://ludens.cl
> ========================
> _______________________________________________
> Amps mailing list
> Amps@contesting.com
> http://lists.contesting.com/mailman/listinfo/amps
>
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|