[Amps] Tubes, transistors, and 'abuse'

[Manfred Mornhinweg]

Steve,

> It would also be interesting to get Manfred's views on the Hardrock 50:
> 
> https://docs.google.com/viewer?a=v&pid=sites&srcid=ZGVmYXVsdGRvbWFpbnxoYXJkcm9jazUwYmV0YXxneDoxOGMwZTJiMDIzYjZmOWE

Oh boy, I get no rest! :-)

> This one uses four RD16HHF1's at 12 volts in a parallel/push-pull 
> arrangement. Similar design for drain DC voltage and RF output.
> 
> 5 watts in/50 watts out at a slightly lower price than the hfpacker.

Okay. I just had a look at the documentation.

First thing: The RD16HHF1 indeed seems to be the best choice for 
amplifiers of this kind. It's made for the purpose, designed to work 
from 12V, has the source at the case, is available, and quite 
inexpensive. Several weeks ago I placed an order for a few of them, to 
use as drivers in my own project, but they still haven't arrived 
(customs processing in Chile takes forever and a day), so I have no 
hands-on experience with them yet.

Several comments come to mind while looking through the docs:

The output transformer has a 1:4 turns ratio. At the nominal 50W output 
that means 12.5V RMS across the primary, so that's 6.25V at each drain, 
or 8.84V peak at each drain. One would think that this is a somewhat 
poor utilization of the 12-16V supply voltage, resulting in poor 
efficiency - but I don't know what the actual saturation characteristics 
of this MOSFET are! Maybe it can't do much better.

Reducing the turns ratio to 1:3 would need 11.8V peak at each drain, 
which clearly would require more than 12V at the supply. It might be OK 
with a 14V or higher supply, but not with 12V. So, unless one would want 
to use a transformer with two primary and seven secondary turns, the 1:4 
ratio is a necessary evil. And 7 secondary turns would probably prove 
unworkable due to delays, leakage, etc.

With the 1:4 trafo, the maximum theoretical efficiency for pure class B 
operation would be 57.8% at 12V, and only 43.4% at 16V, instead of the 
theoretical 78.5% for a perfect class B amplifier. Of course when 
operating in class AB the efficiency will always be lower, and 
additional imperfections of practical components reduce efficiency even 
further. So we cannot expect this amplifier to be a model of high 
efficiency. I would expect 50% at 12V and 38% or so at 16V, on a "good" 
band, and slightly less on a "bad" one, if the feed arrangement provided 
perfect drain-drain coupling (which it doesn't).

Instead this amplifier should at least be acceptably linear, given the 
relatively small voltage excursions at the drains, and the use of some 
negative feedback.

Since the single turn primary of the output transformer doesn't allow a 
true center tap, a bifiliar feed choke is required, and is indeed 
present. The bifiliar feed choke used in this design has a single turn 
on each side, on a relatively small core having the optimal shape. From 
the point of view of leakage inductance this is pretty much the best one 
can do. Only strap winding would be better. But the transistors are 
placed pretty far apart, introducing significant stray inductance into 
the drains-choke-bypass-sources circuit. It's hard for me to tell how 
much, but looking at the board layout and the choke's construction, my 
educated guess would be around 50nH per side. Given this amp's drain 
load of only 1.6 ohm per side, a maximum stray reactance of no more than 
0.3 ohm or so would be required for good clean functioning of this 
choke. 50nH has 0.3 ohm reactance at close to 1MHz, and from there up 
its performance starts falling apart! So I would expect this amplifier 
to exhibit somewhat recognizable drain-to-drain coupling only on 160 
meters, falling apart on the higher bands. From 40 meters up or so the 
coupled feed arrangement doesn't work, so we are back to the usual 
problem of these amps, with lower-than-expected efficiency and 
linearity, and horrible drain voltage waveforms.

A better, very tight board layout would improve this slightly, but there 
is just no way to get it right on the higher part of the HF spectrum. We 
have to live with this problem and accept the compromised performance, 
as we have been doing since solid state broadband HF power amplifiers 
first started.

The thermal aspects: The specs are 12 to 16V input, 10 to 12A current 
drain, 50W nominal output. That's a power input between 120 and 192W. 
After discounting a few watts needed for control circuitry, let's assume 
105 to 180W input power to the drains. If in all cases we get 50W output 
(which is unlikely, it's probably more on low bands and less on high 
ones), this would equate to an efficiency between 48% and 28%. This 
seems plausible, with the actual values probably falling closer to the 
middle of this range. The worst-case dissipation in the four transistors 
should be 130W, while typical dissipation would be well under 100W, in 
key-down testing. Let's take the worst case of 130W, which is 32.5W in 
each transistor.

The RD16HHF1 has an internal thermal resistance of 2.2 K/W and a maximum 
junction temperature of 150°C. It doesn't need electrical insulation to 
the heatsink, so the thermal resistance of the interface should be 
around 0.1 K/W. So the heatsink mounting surface just under each 
transistor can be allowed to heat up to 75°C while 32.5W of heat are 
being extracted from each transistor. With the heatsink used in this 
amplifier, whose data I can only estimate because I don't have its exact 
specs, brick-on-the-key operation for a sufficiently long time will 
likely overheat the transistors, while normal ham operation at the usual 
duty cycles should fall into the green range. If anybody wants to 
transmit an RTTY bulleting using this amplifier, I would suggest to 
place a fan on the heatsink. Even running very slowly and quietly at 
reduced voltage it should prevent overheating.

I don't like the way the designer paralleled the FETs very much. 
Resistive loading of each gate would have been safer than series 
inductors. But as stated above, I don't yet have working experience with 
these particular FETs. Maybe they have enough internal gate damping to 
remain stable in that circuit.

The single turn per side on the bifiliar feed choke can look suspicious. 
So I calculated how things are. It turns out that at 1.8MHz and 50W this 
core operates at 13.4 millitesla flux density (or 134 gauss, for old 
fashioned people). That's a little higher than usually recommended as 
rule of thumb, but not terribly so. I would expect this core to run 
warm, but not excessively hot.

So, overall, I think this amplifier is pretty decent. It does suffer 
from the usual problems of solid state broadband push-pull linear 
amplifiers, but that lies in the nature of the thing, not in any 
specific problems of this particular design. To avoid unnecessary loss 
and heating, it would be a good idea to operate this amp from the lowest 
supply voltage that produces normal output and good IMD. Very likely 
that's much closer to 12V than to 16V, and possibly even a tad lower 
than 12V.

I didn't look at the control circuitry or the low pass filters. I 
suppose you all are mostly interested in the amplifier section proper.

Manfred




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