[Amps] Price per Watt Conversation

Sat Apr 22 12:37:38 EDT 2017

Hi all,

just a quick comment about a point touched by Joe:

> Several of the newer LDMOS designs are using *two* active devices
> rated at 1500W each.

I'm not aware of any LDMOSFETs rated at 1500W output in linear 
operation. I'm pretty sure that no such device is on the market yet.

Many hams are misinterpreting the datasheets of these devices. Let's 
take, for example, the BLF188XR. Its manufacturer calls it a 1400 Watt 
LDMOSFET. The application information in the datasheet gives several CW 
application examples with output powers ranging from 1200 to 1400W, but 
this is _not_ in linear service! Instead these power levels are obtained 
by building class AB circuits and driving them deeply into saturation, 
resulting in very nonlinear operation. This is fine for typical CW 
applications, where CW stands for truly "continuous wave", not for Morse 
code. It means applications such as FM broadcast transmitters, or 
industrial RF power generators for welding and other such applications.

In this saturated class AB operation, the efficiency is high, like 73 to 
85%. In terms of power dissipation, the worst case listed in the 
datasheet is 1400W output at 73% efficiency. This results in 518 watts 
dissipated as heat, which is about the reasonable limit when using 
normal heat spreaders, heatsinks and mounting techniques.

Instead if you want to run a class AB stage in a linear way, you have to 
keep the drive low enough to stay out of deep saturation. Under these 
conditions, the efficiency tends to be around 50%. In theory a class AB 
stage should be able to run at something between 65 and 70% efficiency, 
depending on how much quiescent current is used, but MOSFETs (like all 
devices) are not perfect, which reduces the efficiency somewhat. And 
what's worse, the extremely low drain impedance of these 50V high power 
transistors makes it pretty much impossible to properly couple the two 
drains together, which is a requirement for correct class AB linear 
push-pull operation. The usual, very poor implementations of the output 
transformer and feed arrangement result in a further reduction of 
efficiency and increase of IMD. That's why practical broadband linear 
amplifiers using low impedance transistors often struggle to achieve 
even 50% efficiency, at acceptable linearity.

So, you can apply the old rule that in class AB linear service you have 
to dissipate about as much power in heat, as goes into the antenna. 
Generating 1500W output will generate roughly another 1500W in heat.

And here is the core point: There is no practical way on earth to make 
any currently available LDMOSFET dissipate 1500W without overheating! 
That's why nobody can make a single device 1500W class AB linear amp yet.

Even using two of these devices, the thermal aspect is a bit marginal. A 
very good heatsinking scheme is required, and the devices operate at 
very high junction temperature, which counts against their reliability 
and life expectance.

The mentioned BLF188XR has a rated internal thermal resistance of 0.1 
K/W. For the metrically challenged among you, K/W means kelvin per watt, 
and a temperature rise in kelvin is the same as a temperature rise in 
old fashioned degrees Celsius. So, the rating means that at a given 
power dissipation, the junctions will be one tenth as many degrees 
Celsius hotter than the LDMOSFETs underside. The rest is up to the 
system designer/builder.

The maximum acceptable short term junction temperature is 225°C. So, if 
you wanted to dissipate 1500W from a single BLF188XR, you are 
responsible for keeping the external surface of the device below 75°C. 
At first sight this might seem possible, but once you do the maths 
involved in calculating the thermal resistance of the device-to-spreader 
mounting, the internal thermal resistance of the spreader, then the 
spreader-to-heatsink thermal resistance, then the thermal resistance 
inside the heatsink, then the fins-to-air thermal resistance at a given 
amount of air motion, you will realize to what degree it's impossible to 
satisfy this requirement!

And that would be just to keep the junctions at the absolute maximum 
short-term allowable temeperature! For good reliability they should be 
kept much cooler.

When spreading the 1500W dissipation over two devices, you gain in 
several ways: First, each device can be allowed to heat up to 150°C at 
its underside, before the junctions exceed the absolute limit 
temperature. Second, you have twice the surface to extract the heat 
from, and the two devices can be physically separated as much as you 
want, if you build two pallets and use a splitter/combiner arrangement. 
So you end up using two heat sinks, or two quite independent sections of 
a single heatsink, each having to extract 750W of heat while having 
permission to let the device heat up to 150°C. That's hugely easier than 
having to extract 1500W with no more than 75°C device temperature!

How much easier is it? Well, assuming that the air temperature will 
never be higher than 25°C, the single-device solution requires a total 
device-to-air thermal resistance of 0.033 K/W, and that's impossible to 
achieve. Instead the two-device solution requires each of the two 
heatsinks to have a thermal resistance of 0.167 K/W. Five times less 
stringent a requirement. It's still hard enough to achieve, but with 
good engineering it can be done.

The newest upcoming high power LDMOSFET I have heard of is the 
MRFX1K80H. It's rated at 1800W output in nonlinear CW operation. Some 
people might think that it even has headroom when operating at 1500W! 
Well, yes - but not in linear operation!

The internal thermal resistance of this device is 0.09 K/W. Only 10% 
better than the BLF188XR. And it's the same size, so whatever heatsink 
and spreader you use, it would have the same thermal resistance for 
either device. So, if your super heatsink system achieves 0.1 K/W, the 
total junction-to-air thermal resistance of the MRFX1K80H is 0.19 K/W, 
against 0.20 K/W for the BLF188XR. Just a 5% improvement, nothing more! 
And with less marvelous heatsinking schemes, the advantage brought by 
the new higher power device is even less than that.

It follows that it doesn't pay to use the highest power rated devices 
for dissipation-limited applications like class AB linear amplifiers. 
It's much better to use two or more, lower power devices. Instead the 
very high power devices come into their own when used in high efficiency 
applications such as nonlinear FM amplification.

 > They are not nearly as stressed as older
> devices run into saturation and should provide significantly better
> IMD than marginal designs like the Expert 1.3K.

As far as I know, the Expert 1.3K has a single device with a 50V supply 
and a 1:9 output transformer. This allows it to produce roughly 700 to 
800W before entering significant saturation. The fully saturated output 
power of a typical amplifier is somewhat less than twice its maximum 
unsaturated (linear) output. That explains the 1300W rating - it's the 
fully saturated, nonlinear, high efficiency output power of this amp, 
usable only in modes that don't require linearity. In SSB it should not 
be driven beyond 700 to 800W PEP, or severe splatter will result. And at 
that power level its efficiency will be roughly 50%.

At 1300W output it's likely that its LDMOSFET is dissipating _less_ 
power than at 700W output. In any case the dissipation is high enough to 
heat the junctions very close to the allowable limit.

And a 1500W linear amplifier using two devices is pushing them just as 
hard as the Expert 1.3K does when operating inside its linear range at 
750W or so. An 1500W linear amp could be rated at roughly 2600W 
saturated output power, if the power supply and output networks are up 
to it.

Oh boy, I promised to make just a quick comment! ;-)

Manfred

========================
Visit my hobby homepage!
http://ludens.cl
========================