[Amps] Tubes vs. Solid State

Al Kozakiewicz akozak at hourglass.com
Tue May 1 12:48:28 PDT 2012


If you don't mind, could you briefly describe your cooling system?  Is it tap water through the heatsink straight to the sewer or a closed loop?  If the latter, could you please describe the components used?


-----Original Message-----
From: amps-bounces at contesting.com [mailto:amps-bounces at contesting.com] On Behalf Of Manfred Mornhinweg
Sent: Tuesday, May 01, 2012 3:36 PM
To: amps at contesting.com
Subject: Re: [Amps] Tubes vs. Solid State

Dear all,

I have been following this thread, and at some moments shuddered at what some of you opine, and assented at what others write, and several times I would almost have replied to something, to set it straight, but always somebody did it before I got a chance.

Interesting thread, but I can't help the impression that most of the people who give their opinions have never in their life actually designed some piece of equipment, let alone a high power solid state circuit of any sort. And many are also a bit weak in the field of vacuum state power electronics.

I would like to tell you that I have been working over the last months, as a background hobby work, on a project of a legal limit HF solid state power amplifier. At this time I'm getting a pretty clean 1200 watts output, and saturated 1700 watts, from 160 to 15 meters, with slightly less power on 10 meters, from a set of MOSFETs costing a total of 75 dollars. No, the number is NOT missing a zero.

I have tried several slightly different but generally similiar MOSFETs from different manufacturers, so that if one device is pulled from the market, I have a choice of others to fall back upon. And while I have been working on this circuit, newer, better and even cheaper devices have been showing up.

You might also be interested in learning that ALL of the parts for this amplifier, including the power supply, heat management, protection, low pass filtering, and automatic bandswitching, will likely end up costing below 1000 dollars total. That's using all new parts. An antenna tuner is NOT included in this amplifier, so if you want to use antennas having an SWR higher than 1.5 or so, you would need an external tuner.

Working on my little pet project, and reading how people write that solid state amps are not for the ham, or not for the home builder, I tend to smile, and sometimes I tend to become slightly angry too. People who say it can't be done shouldn't be discouraging people who are doing it.

Some of the last messages regarding water and vapor cooling have shown that some people here don't understand the thermal specifications of semiconductors. So let me explain: The maximum power dissipation, given in the datasheets, in most cases refers to the maximum power a device can safely dissipate AS LONG AS _YOU_, the designer/builder, make sure that the case of the device will never exceed 25 degrees Celsius. In practice that's almost always imnpossible, and for that reason the actual power that a device can safely dissipate is always lower than the rated power. This is not any sort of de-rating for safety or improved life - it is instead an essential part of designing the circuit, to keep the device operating AT (not below) its maximum safe internal temperature, which should result in a reasonable lifetime. If you truly derate it from this point, so that the silicon runs at significantly less than 150 degrees Celsius (that's the typical silicon temperature allowed for plastic-cased devices), additional (speak: essentially
unlimited) lifetime will result.

Vapor cooling isn't practical with solid state devices, because between water boiling temperature and the device's internal absolutely maximum temperature you have just 50 degrees Celsius difference, and this temperature difference must push all the heat from the silicon to the water. On its path, that heat will find the internal thermal resistance from the silicon to the case, more resistance from the case to the copper block, more resistance through the copper into its fins, more resistance from the fins to the water, and then some, as the water is stirred around only by the vapor bubbles. If you allow one half of this 50 degree gradient to the whole path from the device's case to the water, and the other half just to the path from the silicon to the case, then the maximum allowable power dissipation will be only ONE FIFTH the rated power! This is because the rated power is based on having 125 degrees from silicon to case, and you are allowing only 25.

It's a bad bargain to use transistors in such an inefficient way, so forget vapor cooling.

When using water cooling, you have basically the same thermal resistances, except that the resistance from the fins into the water is much lower, if you pump the water by fast enough to create a lot of turbulence. But the greatest benefit of water cooling over vapor is that your cooling medium might be just 30 degrees Celsius warm, instead of 100 degrees. This gives you much more than twice the temperature difference from the silicon to the cooling medium, and thus allows dissipating much more than twice as much power, from the same devices, mounted the same way on the same copper block!

So there is no point in vapor cooling for solid state devices, as long as we keep using silicon. When semiconductors come around that can work safely at higher temperature (maybe silicon carbide?), vapor cooling would become more attractive.

Vapor cooling comes into its own when your active device can work so hot that the difference between almost cool water and boiling water is much smaller than the difference between boiling water and the allowable core temperature of your device. This IS the case with many big vacuum tubes, but NOT with solid state devices. When you have a lot of heat to remove, like hundreds of kilowatts, bringing in a small amount of water, and moving away a large volume of steam (which is lightweight and moves easily in a pipe) can be easier than bringing in A LOT of cold water, and removing that lot of water after it has warmed up.

For my amp I'm using water cooling. Should I say "of course"? Maybe. 
Silent operation was just one consideration. Another is that I just couldn't find a way to move heat from my silicon to air, at the rate needed, for a very simple reason: The thermal conductivity of copper is finite! To dissipate a certain amount of heat, without making the air VERY hot, you need A LOT of air. You can't make the air very hot, because the heat sink must stay just lukewarm, in order to have enough thermal gradient between the heat sink and the silicon, to move all that heat over all those thermal resistances along the way. So, if you need to evacuate a lot of heat, but not make the air more than lukewarm, you need A HUGE LOT of air, and for that you need a large heatsink, in addition to powerful fans. And a large heatsink inevitably has a long path from the transistors to the average location in the fins! And that long path has too much thermal resistance, even if it's made from copper.

One solution is using a great many small transistors, each of them located directly on a fin of the heat sink, not on the baseplate 10cm away. But that creates a difficult construction, and RF-wise it's not so easy to have hundreds of paths, each up to 0.02 wavelengths long or even more. Specially not at very low impedance.

So, it's much easier to use a watercooled copper block. The path length from silicon to water is just 2cm or so, and the water can be kept cooler, in average, than the air in a heat sink. The contact surface between fins and water can be very much smaller than between fins and air, and this allows using the small cooling block, with the favourable short thermal path lengths.

Anyway, the kind of class AB amplifiers hams are mostly using (regardless of whether it's solid state or empty state), and its associated low efficiency, should be a thing of the past soon. As LDMOSFETs for UHF and microwaves become more capable, a high efficiency, low cost switching amplifier for HF is moving closer. The end result would be a legal limit amp smaller than a shoebox, with its power supply built in, of course, that can run key-down all day long putting 1500 watts into the antenna, while dissipating only about 100 watts, less than what a normal HF radio dissipated while running 100 watts into the antenna. Heat dissipation issues are far less important in such an amp.

Years ago I was experimenting in this line, but didn't get much beyond 40 meters with good efficiency and low distortion, simply because the FETs I used were too slow. So I decided to develop my low cost, watercooled conventional class AB MOSFET amplifier first, and perhaps later on, if my head still works and any ham radio activity is left, develop a switchmode amplifier.

Without any doubt, even when some day 95% of all hams are using highly efficient solid state parametric synthesis amplifiers, there will be some hams left saying it can't be done, and that tubes are best...

That's not to say that tubes don't have their advantages. If a 70 year old ham has a 55 year old junkbox, well filled with quality parts covering all history from the development of the first tubes up to 1970 or so, and this 70 year old ham is very comfortable with tube technology, but has a hard time learning new solid state tricks, I won't blame him the slightest bit for building a tube type amplifier, even in 2012! But for a young ham who hasn't much of a junk box yet, it would be rather stupid to try and buy all the antique parts he needs, often at collector's prices, to build himself an amplifier to an antique design. 
He is much better off buying modern, low cost parts, and building a modern amplifier.

I'm in between, so I think I understand both extremes! ;-)


Visit my hobby homepage!
Amps mailing list
Amps at contesting.com

More information about the Amps mailing list