Well, lets see.... If you have a filter with a center frequency of 1 MHz
and a Q of .5 then the cut off frequency would be 2 MHz and lowest
frequency would be Zero by the definition of Q= Center Frequency/ Bandwidth.
The Idea is to have sufficient circulating current thru the output
capacitor to maintain as close as possible a sinusoidal wave form voltage
during an entire cycle at the cathode. How close it is to a sinusoid
depends on the Q. It just depends on if you are matching or trying to
eliminate the third order component of the polynomial to a high degree. I'd
guess if you completely eliminated the third harmonic you would eliminate
the IMD as well since they are created by the same nonlinear characteristic
of the VI curve.
In doing your model place a diode at the output of your filter. Submit
2 tones at the input and do a Fourier transform of the output to see what
IMD and Harmonic products are produced.
To some degree a low pass filter should reduce the IMD somewhat due to
the fact that harmonics reflected back to the exciter. In that case the
exciter are reduced in amplitude and it does not see a nonlinear load
during each RF cycle, but that does not necessarily mean that the voltage
waveform at the cathode is sinusoidal.
73
Bill wa4lav
At 03:26 AM 5/6/2003 -0700, 2 wrote:
> >> There are resonant pi networks and low pass filter pi networks......
> >> These are resonant. If you perform a frequency plot of the output vs the
> >> input you will
> >> get a small signal up to the point of resonance ( the low pass part of the
> >> frequency plot) and then a peak at resonance that produces a much larger
> >> signal and as you increase frequency further the output continues to drop.
> >> The magnitude of this peak compared with that at lower frequencies is
> >> related to the Q of the network.If your pi network in your tube output
> >> transceiver was simply a low pass filter you could not peak the output
> >> power when tuning it.
> >>
> >Yes, quite right on the output PI network, Bill. However, if you plug the
> >values for Q=1 into a circuit simulator (I just ran them), the frequency
> >response of a tuned input PI looks a lot like a classic chebyshev
> >low-pass filter with a ripple of about 0.2dB (values taken from
> >Orr's Radio Handbook, K8RA 3CX800A7 amp, 14 MHz tuned
> >input). For Q=2 (taken from my Drake L-7 schematic), the frequency
> >response is starting to look more like a resonant PI (2 dB ripple).
> >Not sure why Pittenger used Q=1 (seems like I hear people generally
> >recommending Q=2). But if a 3 pole Q=1 resonant PI network
> >provides enough energy storage to be useful as a tuned input, but
> >only has 0.2dB frequency response ripple, then why wouldn't a 5
> >pole low-pass with the same 0.2dB ripple provide some useful
> >storage? Perhaps I need to break out the Spice simulator and run
> >some half cosine pulses thru said 0.2 ripple low-pass to find out.
> >
> >BTW, just for the record I am defining Q=1 to be Xc1=Xc2=50 ohms.
> >Not sure if that is a correct definition when the Q gets this low.
> >
>** This simplified method of determining Q is the one used in Eimac's
>'Care and Feeding', however, the actual Q is seemingly a bit higher.
>Theory and analysis have their place but the goal is that the SWR must be
>low enough to prevent the transceiver's PA-protect circuit from
>throttling back. With some radios, this means a SWR of under 1.2 to 1.
>Seemingly the fastest way to get there is to start with Xc1= 15 - 25
>ohms, and make the other two reactances adjustable
>
>cheers, Mike,
> >
> >
> >
>
>
>- R. L. Measures, a.k.a. Rich..., 805.386.3734, AG6K,
>www.vcnet.com/measures.
>end
|