[Amps] Plate Load Calc.

Manfred Mornhinweg mmornhin at gmx.net
Wed Mar 7 08:56:08 EST 2007


Hi Steve,

> Thanks Manfred, that makes sense to me.  But then I wonder why fudge factors 
> are used at all?

I share your wondering! Actually, in some newer books they are not used. 
It seems that some author in the dark ages of electronics came up 
experimentally with these fudge factors, without understanding what 
really happens, and several generations of later authors copied these 
factors without stopping to think about it. It's very common to see 
rules of thumb, and even little mistakes, copied from one book to the 
next over many years!

 > Your example calculation doesn't appear any more
> complicated than the "fudged" handbook equations.  As a newby amp builder, I 
> am trying to understand the theory and calculations behind it, which sends 
> me to the handbooks. 

My suggestion: Do look into handbooks coming from a different tradition! 
If you use ONLY books printed by the ARRL, for example, you will see the 
same fudge factors and the same little mistakes in many books from 1930 
to 2000 and beyond! If instead you resort to books from the RSGB, or to 
German, French, Czech books (translated as necessary), you will learn 
completely different methods for calculating and doing things. Then you 
can choose the best of each. Japanese books are highly recommended, as 
they typically have a very logical approach. But it can be difficult 
getting translated versions!

 > I ran through your example calculation and came up
> with a K factor of 2.5.  The only mathematical difference between what you 
> displayed and what is in the handbooks is that you use RMS RF voltage, and 
> the handbook uses DC plate voltage.  So perhaps the K-factor is used since 
> DC plate volts is more easily known than the RMS voltage swing and 
> saturation voltage?

The ratio of effective DC voltage (less saturation) to RMS voltages is 
fixed: It's SQR(2), when running class B. So this would be known. What's 
not known is the saturation, because it depends on the specific tube, 
and the operating conditions. So, yes, the K factor represents an 
estimation of saturation voltage, and also includes the effects of 
different biasing according to operating class, which changes the 
conduction angle and thus the DC to RMS ratio. As a result, the K 
factors given in books make sense, but only as approximations in typical 
situations, and not as universally valid numbers! If you want real, 
correct numbers, you have to do the math for the situation on hand.

> So, can I infer that I should "design" for Class B SSB operation, and then 
> if I need to use the amp for Class C CW operation, just use more drive?

Not that simple! What really defines class C is that the tube conducts 
for LESS than half the time. So, first you need to make the bias more 
negative, to bring the tube into cut-off when there is no RF drive. And 
after that, yes, you should drive it very hard to improve efficiency!

The increased bias will make the DC to RMS ratio larger. The increased 
drive will make the ratio smaller. Depending on the ratio of drive and 
bias change, the K factor could end up higher or lower in class C, 
compared to class B! The most typical situation will be that the bias 
effect is larger than the drive effect, and so you end up with a 
slightly smaller DC to RMS ratio, resulting in a slightly larger K 
factor. If this is a problem for impedance matching of the plate to the 
antenna, you might want to reduce the plate voltage in class C 
operation, to keep the load impedance constant. Many factory-made amps 
do that! It allows tuning the amp in class C at low dissipation, then 
switching into class B (or more commonly AB) for linear operation, with 
unchanged tuning settings.

> My intended 'operating condition' and design is:
> - Three 813's in grounded-grid.

What a waste! These tubes work so nicely as pentodes, with grounded cathode!

> - CW and SSB on 160,80,40,20m

OK. Since you don't want to go above 20m, the high capacitance of these 
tubes becomes less of a problem.

> - Around 1000 watts output.
> - PI-EL output network.
> - 2700v supply.
> - TS-870 for exciter.

I would say that 2600 Ohm is a reasonable value for plate loading, 
allowing for some sag of the supply voltage. The parasitic capacitances 
will require a minimum network Q of about 12 at 20 meters, which is 
easily workable. I would try to design for a Q of 14 or 15 at 20m, and a 
bit lower on the lower bands.

If you wanted to work at higher frequencies, you would need a lower 
supply voltage to obtain a workable Q, because of the high plate 
capacitance of the 813. Otherwise you would end up with such a high 
loaded Q that the losses would be too large!

> Yes I admittedly have been too caught up in the calculations - but only in a 
> effort to understand what's going on.  I was struggling with the PI-EL 
> design in order to figure out what capacitors I have to obtain (lower values 
> are much easier to obtain than the higher values, and wanting 160m ability 
> in the amp doesn't help), and what kind of coils I need to start winding up. 
> I suppose with capacitors I can just find the largest values I can (say 
> 250pF tune, and 1500pF load) and pad or trim as needed.

Not really. A very important consideration is the minimum capacitance of 
the tune cap! A 250pF max cap will typically have 25pF minimum. This 
capacitance, combined with the tubes' capacitance and a little stray 
capacitance, already restricts you to a Q above 18 or so on 20m! This 
comes into the range where the coil losses rise to a dangerous level. If 
you manage to make a 20m coil with an unloaded Q of 300, you would get 
losses of 60 Watts or so in the coil! This can be a real problem. So, go 
for a tuning cap with the smallest possible minimum capacitance. 5pF is 
a good value to shoot at.

Using lower plate voltage reduces this problem, because the tube 
capacitance stays the same and the load resistance comes down sharply.

To reduce the range of the variable capacitors required, you can play 
with the Q. Using a lower Q at the lower bands reduces the requirements 
for maximum capacitance. It increases the coil size, but that's easy to 
handle, specially because the losses will be lower. At 160 and probably 
also on 80m you might need fixed caps in parallel, unless you can find 
or make variable caps with a very high ratio of minium to maximum 
capacitance. Parallel caps will reduce the matching range for antennas 
with non-optimal SWR. So, you might plan on a Q as low as 8 on 160m, and 
then see what fixed capacitors you might need to add.

>  And with coils, 
> wind them for the larger calculated values and cut them down as needed 
> during the on-board fine-tuning effort.

Yes. It's always easier to cut off the excess, than to add what's 
missing! :-)

I would use a separate tube-wound coil for 20m, and a tapped wire-wound 
coil for the lower bands.

Oh my, writing all this is tempting me to put my hands to old tube 
technology again! I have two pretty good 4CX1500B lying around... Hmmm, 
the two of them in parallel wouldn't be bad! Using water pipe to wind 
the tank coil, and a rewound pole pig in the power supply! Such a baby 
would be nice to bust the pirates that crowd the bands down here! :-)

73,
Manfred.

----------------------------
Visit my hobby website!
http://ludens.cl
----------------------------


More information about the Amps mailing list