Far from conclusion..... Take a "typical" well engineered pallet designed for 1kW output. We are lucky, assume efficiency is 80% biased in AB1. Following the load line we see efficiency would drop to ~63% at ~50% power output, we must design the cooling system for this factor, hams have a habit of either running things flat out or 1/2 power.... DC input power is 1250W at max output but dissipation is worse at lower power, assume max dissipation is 320W. Typical device has a RthetaJC of ~0.15K/W, that means the case temperature due to purely conduction should be 320*0.15=48C below the junction temperature if we do our job right. This is where things get complicated. As a design aim for "long life" which can be interpreted as operating 24/7 at full power, the junction temperature should be kept below 175C. This is an arguable point but a figure adopted by most commercial (non ham) manufacturers. Taking 175C in our example means the maximum case temperature to aim for will be typically 175-48=127C. The purpose of a spreader plate is just that. It is not immediately obvious as to the behavior of the spreader and to get a true answer requires thermodynamic tools, However I will make it easy. 0.2% tungsten doped copper (to make it easy to machine) has a conductivity close to 400W/m/K, so how big does it need to be ? Well, there is an upper bound, once past this it is diminishing returns. As it happens that value is close to the spreader size specified by many semiconductor manufacturers for this type of device, you can assume they have done their homework correctly. So for our example a spreader 125x75x15 mm has an effective RthetaSC= ~0.18K/W, if we doubled the area RthetaSC would drop to ~0.15K/W, if we made it 1mx1m it would only be ~0.13K/W. This is following the general behavior of a point source in contact with an infinite heatsink. To improve this dramatically we need a material with a much higher conductivity...e.g. diamond. So following our observations we need to keep the spreader at 127-(320*0.18)= ~70C. Taking a typical bonded-fin heatsink with a effective conductivity of 1000W/m^2/K we would need an airflow of ~0.025m^3/s in an ambient of 25C to keep the exhaust temperature below 60C. That equates roughly to a pair of 80x80x30 high efficiency fans shifting at 6.5m/s giving an effective RthetaSH of ~0.03K/W for the heatsink. We are already approaching an engineering limit for an air cooled heatsink. We can now introduce some changes and see where this leads..... Substitute the spreader with aluminium, conductivity is around 60% of copper but the effective RthetaSC for the same size spreader rises to 0.23K/W at best, given the same heatsink the exhaust temperature would have to be below 43C. Turns out our build is not quite so efficient and we get 75%, oh and the antenna VSWR is 1.15:1. Our dissipation figure has risen to 420W. Given our simplification on the spreader we rework the effective RthetaSC, it arises that the exhaust temperature needs to be BELOW ambient..... so junction temp suffers and goes thru the roof = BANG. You can see how we quickly get into trouble. One can draw your own conclusions as to why most commercial ham amps actually turn out to be quite unreliable. If the device is bolted down, rather than soldered, and ALL flatness considerations are taken into account, the variability ought to be less than 5%. Remember, heatsink compounds should only be used to fill microscopic voids, heatsink compounds like Arctic Silver 5 have quite a high thermal resistance compared to the numbers I have shown here, but of course, needless to say this remains far far better than air..... Most compounds use silver and a number of ceramic oxides, electrical conductivity is unpredictable and you need upwards of 250 hours operation at expected temperature before it does its work correctly.......not just slap it on and hope for the best. So, When buying/building your 1kW amp, ask yourself the following: How big is the spreader compared to the device manufacturers recommendations and what is it made of ? What is the specification of the heatsink and cooling fans used ? How efficient is the final pallett ? What VSWR, overcurrent protection is used, what are the trip points (deg C, Amps). Where is the temperature sensor ? (if its anywhere BUT the spreader walk away). Demonstrate this all by pulling the coax off the TX port say 1/2 dozen times at full power. 73' Keith G4FUF/XV4F