> The maximum value is determined by stability criteria. The plate swing is
> divided in the potential divider consisting of Cgp and the impedance from
> grid to ground. This impedance includes Miller effect, the swamping
> resistor, the input Cgk, and the output impedance of the driver, modified
> by the length of interconnecting cable. If the ratio of this voltage to
> the plate swing is greater than gm times RL, it oscillates. For this, you
> must calculate with complex impedances.
>
> It could just happen (but Murphy Rules - OK?) that the driver you have
> with the length of cable you have that the Gods smile on you and you can
> get away with a higher load resistor than you'd expect.
>
> Some of the older books do quote the stability criteria for tubes - I
> don't have it here with me at work.
>
> The resistive portion is conveniently 12.5 or 50 or 200 ohms because of
> making transmission line transformers, if you have to have a low input
> VSWR.
>
> I use 200 ohms with a pair of un neutralised 4-250s, with no problems.
> However, you do need decent layout with grid and plate circuits well
> shielded.
>
> The input match can be a problem, because of the Cin of the tubes. This
> can be improved by using a low pass filter section to feed the tube grid,
> and making the output capacitor of the pi network LPF the tube input
> capacity. Of course, the resistor goes straight from grid to ground (or
> bias). The filter is designed as straightforward m=3 Butterworth.
>
> Another approach is to use a m=5 Butterworth i.e.
>
> -----L--------------L-----------------
> | | | |
> C C C R
> | | | |
>
> and to use the tube in place of the middle capacitor. This has the
> advantage that the middle capacitor is bigger, AND THE DANGER OF VHF
> PARASITICS!
> I wouldn't say that it is certain you'd get parasitics, but there's a
> high probability - and even Rich's low Q suppressors could have a problem
> curing them, since you might have enough cathode inductance for it to be a
> Colpitts.
>
> The m = 3 Butterworth filter has values (normalised) of
>
>
> ---------------L = 2-------------
> | |
> C=1 C=1
> | |
>
>
> and then the denormalised C = C/2*pi*F*R
> denormalised L = LR/2*pi*F
>
> It's a good idea to set the corner frequency, F, as high as possible. SO
> figure the value for R and the input capacity, and substitute back to get
> F, which is
> F= 1/2*pi*C*R
>
> F=30 MHz is where the filter is 3dB down, so that's a no-no.
>
> A Smith chart is useful here for deciding how far you can go, but it may
> be that you need to physically tune the input with a shunt L if the values
> are too awkward.
>
> Finally, if you need a higher terminating resistor than will be inherently
> stable, you have to neutralise. This isn't necessarily a problem, and you
> may even be able to put some NFB in. If the stability margin is likely to
> be at all marginal,
> neutralisation is likely to improve IMD because of the phase of the fed
> back voltage.
>
> 73
>
> Peter G3RZP
>
>
>
--
FAQ on WWW: http://www.contesting.com/ampfaq.html
Submissions: amps@contesting.com
Administrative requests: amps-REQUEST@contesting.com
Problems: owner-amps@contesting.com
Search: http://www.contesting.com/km9p/search.htm
|