On 8/21/2011 9:00 PM, Manfred Mornhinweg wrote:
> Roger,
>
>> You do need a certain amount of metal in there above the minimum
>> required for heat transfer, or thermal inertia to take care of cooling
>> failures.
> If you want that thermal mass as a safety measure against failure of the
> cooling circuit, why not use water for the thermal mass too? After all
> water has a specific heat per weight that is more than ten times better
> than that of copper, and even when comparing specific heat per volume,
> water is still better than copper! And water is quite a bit less
> expensive than copper...
True, but you are far less likely to lose that copper than you are water
flow. Loss of flow in water cooled systems is probably the most common
failure, hence you need something to back it up and thermal mass is the
cheapest and most reliable. Lose water to the device and it gets hot, fast.
>
> You can have a certain amount of water enclosed in the cooling block,
> and this water would be separate from the cooling circuit, just for
> thermal mass. You can leave a vent, so if the cooling circuit fails, the
> reserve water will alert you by boiling and hissing - that is, as long
> as the transistors can survive operating with their mounting surface a
> bit above 100 degrees Celsius,
The transistors are designed to run as high as 150C, but with reduced
life. The problem with using water as a backup is the stuff has a way of
getting into places you don't want and getting out of places where you
want it to stay. I say this after having spent nearly 26 years working
with liquid cooling on RF generators running the range of 5KW to over
200KW. Smaller units were air cooled.
> which depends simply on the power they
> are dissipating, compared to their ratings. Or better, have a
> temperature sensor shut down the beast if it overheats, and the enclosed
> water will give it lots of time to do that.
You can make the device the low spot in the system and shut down using
pressure differential, but the most reliable by far is a passive heat
sink with shut down based on both water flow and device temp. That at
least gives some redundancy. I can guarantee that cooling failures will
occur in the most inopportune locations. Also cooling shutdown needs to
be based on flow and not pressure. Ideally it would also shut down with
loss of power to the pump.
>> BTW when you figure 20% duty cycle for SSB AND the typical 10 minutes
>> out of 30 transmit time the required water becomes miniscule.
Except with a very small thermal mass you can use the duty cycle, but
not the transmit time when the device will overheat from loss of cooling
in just a few seconds. With air cooling and large tubes you have a lot
of thermal inertia. Even water cooled tubes of sufficient size give you
a reasonable time to shut down, but with a tiny transistor with little
mass it depends almost entirely on the thermal mass of its heat sink.
The problem with water is you can never depend on it being there and
even if it's there you are likely to only have a few grams. Several
ounces of copper will outdo a few grams of water as far as a heat sink.
Even though the water has a much higher specific heat the copper heat
sink has far more mass which gives it more heat capacity than a few
grams of water. The main concern would be the delta T across the copper
to the water. You need enough copper for a safe heat sink, but not so
thick as to impede heat flow to the water. More than likely the
junction between the copper and device is going to be the critical
point. I believe most are now using "Artic Silver" heat sink compound
at this point. It's a bit pricey for heat sink compound but very effective.
> That's right. I calculated that a 10 liter bucket under the desk would
1.5KW continuous = ~1,289,000 calories/hr (depending on class of
operation) and 10L would have a 128 deg rise except water boils at
100C. Fortunately running a 20% duty cycle (SSB) would give a tad over
25C rise in one hour neglecting receive time. Fine for casual SSB or
even CW but a bit anemic for contesting or digital.
> be enough for even the most intense operating I ever do. I wouldn't need
> any active cooling. Just starting the water from room temperature (so
> there is no risk of condensation), and having it warm up a few degrees
> during an operating session, then cool down naturally over the next
> several hours.
Open water cooling systems come with their own problems with algae being
only one. I would use a completely closed system of about 10 gallons. I
would never use an open system as they collect dust and start growing
things quite soon. I say that after working with these systems for well
over 20 years. Dumping one KW, which is not unreasonable at the legal
limit, into 10 liters, or 10,000 grams of water can heat it up
considerably in several hours of contest operation
The same is true for emergency operations, you need lots of reserve.
OTOH it is fine for casual operation although the water should be
changed out regularly.
It's probably due to having spent over 25 years of my life working with
RF and particularly water cooled high powered systems that I tend to try
to err on the conservative side. Algae, bacteria, dust and other foreign
materials, DI water eating ions out of nearly any metal in the system,
non compatible chemicals in hard water... All can be problems for water
cooling.
My next planned tube amp will be running the legal limit, class A
with enough overhead for long digital transmissions. That means about
6000 watts of dissipation for 1.5 KW out. Although it'll start out with
air cooling, that much air is going to be noisy so it *may* eventually
end up with water cooling. With nearly 300 watts of filament power added
to the plate dissipation it will likely dump around 6 KW out in heat.
(Class A is clean, but not very efficient power wise) As an added
note, water ions really like metal ions.
73
Roger (K8RI)
> 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
|