[Elecraft] LDMOS for QRO [OT]

[David Cutter]

Hi Dick

Start from the LDMOS data sheet and work out the temperature you wish to 
keep within at the junction. Because water cooling is so good, you can 
choose either to run the device cooler than you would with air cooling to 
improve reliability, or stick to the same temperature and get more power / 
use a smaller cooling plate / use less water. There's a direct relationship 
between reliability and temperature, but it's not linear. Then use the 
temperature rise per W rating of the device to get to the surface 
temperature.

In work I did >10 years ago all devices were directly bolted to the 
liquid-cooled plate without an intermediate spreader, however, these LDMOS 
devices are so small (ie very high heat density) there is a good reason to 
spread the heat out first before cooling proper takes place, I'm somewhat 
hazy what thickness, but you must achieve good flatness of contact against 
the plate. Some very large devices eg IGBTs and rectifiers the size of your 
open hand are supplied curved and the bolting-down process achieves the 
flatness with the correct torque setting on the bolts.
Then you do the same sums as you do with air cooling ie temperature rise v 
watts dissipated from the heat sink data.

Say you want to dissipate 1kW of heat continuously (eg in a data contest) 
then a small 2 pass model would give you around 90 to 100K rise at the 
surface of the plate with 1 US gallon per minute, whereas a 4 pass model 
would give you around 20K rise on a 152mm length plate. This of course 
assumes that the heat is being delivered into the plate evenly over the 
whole surface, ie using a spreader. Heat sink paste adds a little to the 
thermal gradient and is needed in very small amounts, evenly spread.

Do the sums several times with different criteria until you get to the one 
you feel comfortable with. If you live in a cold climate you can dump the 
heat into a small central heating radiator to keep the shack warm and no 
fans required just an aquarian pump to run it; if you live in a hot place, 
then put the radiator on the shade side of the house or even bury it.  If 
water is abundant, eg river water or a pond, you can re-cycle it back to the 
source.

You can make your own water cooling plate, see here a small example cooling 
a dozen TO-220 devices:
https://www.youtube.com/watch?v=WGpau-raMho
Somebody here will check my sums I'm sure.
73
David
G3UNA

----- Original Message ----- 
From: "Richard Solomon" <w1ksz at earthlink.net>
To: <elecraft at mailman.qth.net>
Sent: Tuesday, December 16, 2014 4:44 PM
Subject: Re: [Elecraft] LDMOS for QRO [OT]


> If one wanted to use one of these to cool an LDMOS VHF KW, where would
> one find the design info to calculate which cold plate would provide 
> sufficient
> cooling ?
>
> I envision a stack something like this:
>
> LDMOS PC Board
> Copper Heatsink (thickness need be determined)
> Cold Plate
> Aluminum Heat Sink (is this really necessary ?).
>
> 73 es HH, Dick, W1KSZ
>
>
> On 12/14/2014 4:32 PM, David Cutter wrote:
>> I'm a little surprised that folks in this group haven't suggested liquid 
>> cooling for this modest application. Semiconductor cold plates have been 
>> around for a long time, are economical to use and in my view a much 
>> better solution than forced air cooling. They are compact, quiet, require 
>> far less cabinet space, keep junctions cooler and more stable than air 
>> could ever and enable higher reliability.
>>
>> Look at Aavid for instance, whose devices I used on many occasions:
>> http://www.aavid.com/sites/default/files/products/liquid/pdf/liquid-cold-plate-datasheet-hicontact.pdf
>>
>> If you play your cards right, you can cool the amplifier and the power 
>> supply on a short 4-pass plate. Put the heat somewhere convenient, not in 
>> your shack.
>>
>> 73
>>
>> David
>> G3UNA
>>