Amps
[Top] [All Lists]

Re: [Amps] Future of RF AMPS

To: amps@contesting.com
Subject: Re: [Amps] Future of RF AMPS
From: Manfred Mornhinweg <manfred@ludens.cl>
Date: Fri, 13 Sep 2013 19:14:14 +0000
List-post: <amps@contesting.com">mailto:amps@contesting.com>
Jim,

## What about these freescale devices ? That company in OH land manages to get 800w to 1 kw out of just one device. I saw on U tube, 1 of em being used in a 1250w CCS FM broadcast TX, 88-108 mhz
 variety.

That's in saturated service. Dissipation should be no more than perhaps
400 watts. That can be handled.

Also on U tube, was a 2.5 kw CCS eme  144 mhz amp... using 2 of em.

I would think that this one also operated in saturation, at high
efficiency. So it was not linear. Surely it was used in modes that don't
require linearity, like CW or some digital modes that use slowly single
tones. Was it that way?

The cooling used was unique on all 3 of them.

The biggest problem when one designs cooling systems for such devices is
getting the heat away from the small device. There is just simply an
awful lot of heat coming out of their small mounting surface. A copper
spreader is the usual answer, but to get the heat into the bulk of a
heatsink large enough, it might have to take a 10cm long path through
copper, and the device end of this path is narrow and tight. The thermal
resistance in this area is already so much that it severely limits the
maximum safe total dissipation.

I did a design where I got water flow as close as possible to the
device. The trick was getting enough water/copper interface surface,
with short enough in-copper path length.

As soon as the amount of heat is less, or the device mounting surface is
larger, or it's spread out among several devices, or the device can run
hotter, things get easier.

Then the 3 x 150 ohms are all in parallel to make 50 ohms. They are rated at 800w CCS each. The air cooling was not up to snuff..and one exploded. Plan B was to lower the fins face down into a tub of ice water..problem solved.

But when it comes to power transistors instead of resistors, even that
trick is often not good enough. Instead you need copper, not aluminium,
and you need to move the water fast, to have enough turbulence to get a
low thermal resistance between the metal and the water.

And just a little further up the scale you near the point where the
copper between the device and the water is the main limiting factor
(other than the internal resistance of the device, of course). At that
point, it's time to say "enough"!

+ 69dbm was being used at the time.

And may I ask what's the connection between +69dBm and ham radio?


Dick,

M2 markets both a 6 and 2 Meter Monoband Amp, 1250 watts out, uses a
single device. Granted there's a time limit on AM and JT modes, but
there are no limits on SSB.

I just had a look at 2m version's manual. I see that it switches between class AB and class C, depending on the drive envelope. In CW and such it runs 950-1000W in class C, while in SSB it runs 1250W in class AB. But drain matching doesn't change. I didn't check if perhaps they lower the supply voltage when in class C. That's likely - otherwise I don't see how they could get the 65-80% efficiency of teh rating!

Anyway, with 1000W output and 60% efficiency in class C, the FET might be running barely in the green area, as long as the heat sink remains cool enough.

But with 1250W in class AB, never! It has 1:4 matching, so it has a swing of +/-88V at 1250W output, which at 50V supply is OK for a push pull stage - but results in only slightly over 50% efficiency. That means about 1200W dissipation at the envelope peaks, and an average dissipation not very much lower, given that efficiency falls at lower amplitudes. During sustained speech in SSB, this might result in about 900-1000W of dissipation - if the SWR is perfect. The device is rated at 1333W dissipation, if it is kept under 25 degrees Celsius. M2 activates thermal protection at 90 degrees - I don't know if they measure this at the heat sink, the spreader, how close to the FET, or the case of teh FET proper - but certainly the mounting surface of the FET will be hotter! Even if the spreader had zero thermal resistance, so that the maximum possible temperature at the FETs mounting surface would be 90 degrees, that would still limit maximum dissipation to 900W - and that is for 225 degrees junction temperature, which is higher than most manufacturers would rate their devices to survive!

All this means that unless I'm missing something, this amplifier stressed the MOSFET to its absolute limit when operating with a very light hand (no speech processor, short transmissions, lots of RX time, low ambient temperature). In any other case, the MOSFET gets stressed beyond its absolute maximum thermal ratings.

I hope I am wrong, but I would expect those amplifiers to fail easily, due to MOSFET meltdown.

> Look at the W6PQL site for info on cooling one of these LDMOS devices.

He solders the MOSFETs to a thick but rather smallish copper heat spreader, and attaches that to a rather small heat sink.

Soldering is good, far better than using screws and thermal grease. The spreader is thick, but too small. The edges of the heat sink will be a several degrees cooler than the middle, wasting capacity.

Want to do some math? Let's do it, very simply and not very accurately, but enough to get an idea:

The thermal resistance inside the device is 0.15 degrees per watt. This part is easy, it's in the data sheet.

The thermal resistance of the soldered connection is probably low enough to ignore.

The copper spreader seems to be about 1cm thick. The hard part is estimating the thickness of the heath path through the copper. It starts as small as the MOSFET is, about 3cm^2, and expands to about 50cm^2 at the spreader's edges. Heat leaves it all along that path. The average bit of heat might travel about 5cm through the copper - this is a coarse approximation! Actually we should be using integral calculus here, and that's beyond my range of most beloved hobbies. So let's take a 5cm long copper path, that starts with 3cm^2 cross sectional area, and ends with about 20cm^2, which is what it would have at the 5cm path length periphery.

A cube of copper, 1cm on each side, has a thermal resistance across it of about 0.27 degrees per watt. The spreader starts with about 3 such cubes side by side under the MOSFET, so that gives about 0.09 degrees per watt. As we get farther away from the MOSFET, the path widens, and resistance per length drops, all the way to about .013 degrees per watt per cm at the periphery of our mean path length. The average will be around .03 degrees per watt per cm along the path, it's 5cm long, so we get about 0.15 degrees per watt total thermal resistance for the spreader. Remember that this is VERY CRUDE, and it would be good to make the exact math. If anyone knows how to do that, I would like to learn!

So, we have 0.15 degrees per watt inside the device, and another 0.15 in the spreader! Now comes the interface to the heatsink, which will add very little, because it's a large surface. Then comes the thermal resistance of the heatsink, considering that we have to take a value slightly worse than its rated one, because the heat isn't spread out evenly over the full mounting surface!

That heat sink used by W6PWL looks like about 0.2 degrees per watt, if it gets a good air flow from strong fans. Let's de-rate it to 0.25, because of the non-even heat distribution. Again this is pure guesswork, but based on some practical experience.

So we would have a total thermal resistance of 0.15+0.15+0.25=0.55 degrees per watt.

How warm will the air be? Let's suppose, at most 35 degrees Celsius. And the MOSFET is rated to be able to work at up to 225 degrees. So we have 190 degrees difference. With 0.55 degrees thermal resistance, this allows continuously dissipating 345 watts. And nothing more!!!

In ICAS one can go higher, because we can store some heat in the spreader and heatsink and release it later. But we cannot go very much higher, because some important part of the total thermal resistance is very close to the MOSFET, even inside it, where the thermal mass is very small! The heat storage within a few millimeters of the chip is great to inflate the pulse power capability, but not the dissipation over a one minute transmission! The spreader and heatsink can store quite some heat for one minute, though.

So, this arrangement can dissipate about 345W continuously, perhaps 500W for one minute, and well over a kilowatt for a few milliseconds. This is enough for 1kW continuous output in high efficiency, well saturated class C, well over 1kW in class E or class F, but only about 800W or so PEP in brief, processor-less SSB when running class AB, and even less when being as long winded on the air as I am in this post! ;-)

Don't forget that I did some gross estimations. Maybe I'm wrong. But then I would like someone to PROVE me wrong, either by doing the precise math, or by actually measuring the temperature of the MOSFET's mounting surface (not the top of its case!!!) during full power operation!

The problem is that this kind of design leads to amplifiers that seem to work just fine, but then tend to fail very easily, or to fail consistently after just a few hundred hours of use, for the simple reason that the MOSFETs are running far too hot.

Phew!

Manfred

========================
Visit my hobby homepage!
http://ludens.cl
========================
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps

<Prev in Thread] Current Thread [Next in Thread>