[CQ-Contest] Bandpass filter

[Tom W8JI]

> Now this gets very interesting !

Sometimes things get useful when the information is wider.

> A hunderd watts measured at any BPF stopband may destroy the BPF .. be
> it ICE, Dunestar, W3NQN or any other.
> W3NQN, if done with real good capacitors may be able to take 100W on
> stopband for some moments on the stopband.

<<snip>>

> The stopband energy becomes heat. More than 95% of it goes to heat.
> The passband energy goes through the filter .. a part of it; say 5% or
> so would be transferred to heat. 95% goes to the output connector.

I disagree with that, Jukka. Traditional filters neither route things to a 
ground nor dissipate a significant portion of unwanted energy as heat inside 
the filter.  Most of the heat comes from fundamental energy, and can be 
aggravated by high circulating currents. This is why hi-Q tank circuits get 
much hotter than the same components in lower loaded -Q circuits. It isn't 
harmonics heating them.

The mechanism primarily involved in reduction of unwanted feedthrough is 
reflection of energy. The system supplying unwanted signal is misterminated. 
In PA stages, we have to be very careful of filter appearance as reflected 
through the system and appearing at the output device, because it can alter 
termination and waveshape. This is why input filters on GG amplifiers should 
be low-pass or band-pass with a design that makes low impedance at odd 
harmonics at the cathode.  Efficiency can dramatically be reduced if the 
filter reflects a high impedance at the second harmonic at the cathode, 
because it rounds switching waveforms. High impedances at odd harmonics can 
enhance efficiency.

I find incorrect information about stubs in some Internet articles. I find 
articles saying distance from PA to stub, or antenna to stub, makes no 
difference for harmonic suppression because the system is matched. The 
system really is NOT matched at the out-of-band frequency. It cannot be 
matched (without a diplexer), because the mismatch mostly stops the unwanted 
signal. Optimum distance from PA to stub really depends on the PA 
characteristic, and the change can be quite dramatic. For example, a pi- 
L-network PA has better harmonic suppression with a low impedance load at 
the harmonic frequency shunting the network output, while a traditional pi 
has better suppression with a high impedance termination. The stub actually 
has a different optimum location depending on PA filter type, and even on 
layout inside the PA.

In this case power couples through mutual coupling in elements, and filter 
location (distance from antenna) and filter type can change what happens in 
the antenna! This is because the filter (or radio) misterminates the line 
from the 20-meter element on 15-meters. There is no doubt most of the 
suppression is from reflection, and the 20M system simply not absorbing 15M 
energy that is potentially available.

If he wanted to find the absolute maximum energy available, he would run the 
20M element through a T network tuner into a dummy load. Then, when the T is 
tuned for maximum energy into a load, it would be pretty close to the 
maximum possible power level. That would almost certainly be much higher 
than the power he measured into 50-ohms. The real power into his receiver 
could be less than he measured, or more, depending on what the 20M element 
of the antenna likes to see for optimum 15M transfer.

Since the 20M bandpass filter mismatches the 20M element and transmission 
line from that element, the location of the filter can greatly impact 
suppression. The 15 meter signal on the 20M feedline has a very high SWR, so 
moving the filter and the impedance of the filter on 15 greatly affects 
match of the 20M element into the system for 15M energy. No so with a 
broadband dummy load.

While this is long, the main point is location of a filter in the line, as 
well as the characteristics of the source (antenna) and load (filter or 
receiver) will greatly affect harmonic suppression. It will also affect 
power rating of the filter for out-of-band signals.

The same is true for stub placement.

I am not saying a filter cannot "blow up" from the application of 
out-of-band energy, but thinking the components inside dissipate the 
unwanted power formerly measured into a load is absolutely incorrect.

73 Tom