So I just put the 6H6 (V9) circuit back together and you know what? It is a pretty good noise limiter!
The unorthodox concept is called a starved shunt limiter. Basically they put a 4 Ohm resistor in series with the filaments (of the two triodes which are paralleled). The heater voltage is reduced to approximately 3 volts by 4-ohm series resistor R34. This has the effect of making the triode conduct as a shunt switch very sharply when a pulse above the average audio level occurs. Does it cause distortion on AM signals? Only on very high (super) modulation levels.
There is an incomplete explanation of how the limiter works here
https://www.radioexperimenter.us/hammarlund-super-pro/noise-limiter-fig.html which is taken directly out of the BC779 manual.
I will attempt to translate it to the SP400:
The control grids (pins 4 and 5) are connected to 0.05-mf capacitor C65, which is grounded at one end, and through 1-megohm (meg) resistor R32 to the most negative point (A) on the diode resistance. The cathode (pin
is connected through LIMITER OFF-ON switch SW3 to a point of higher potential (B) with respect to point (A), and the plates (pins 3 and 6) are connected to the point of highest potential (C) with respect to point (A). Thus, with the LIMITER OFF-ON switch at ON, the voltage drop between points B and C is the plate-to-cathode voltage of limiter Tube JAN-6N7.
b. Limiter Operation. (1) Development of bias. When a signal of constant amplitude is received, the average (that is, the d-c) voltage between B and C, which is also the limiter plate-to-cathode voltage, remains constant. Similarly, the voltage drop between A and G remains constant and charges capacitor C65 through R32. For a given signal level, the grids are held at a potential more negative than the cathode by the voltage drop across resistor R48. This voltage drop, which is applied as bias to the grid, and the plate-to-cathode voltage are so proportioned that the tube is biased to or beyond cut-off. Consequently, the limiter tube has no effect on the input signal.
(2) Noise reduction. Assume that a constant amplitude signal is received and that a very short noise pulse voltage of high amplitude is superimposed on the signal. The noise pulse causes an increased detector current and, hence, an increased voltage drop across the diode load resistance. As the plate and cathode are connected to points B and C, the plate-to-cathode voltage of the limiter is increased. Although there is also a corresponding increase of the negative voltage at A, the grid is not made more negative. Because of the 0.05-second time constant of R32-C65, capacitor C65 cannot charge and/or discharge rapidly enough to follow instantaneous voltage changes at point A. The cathode, however, becomes more negative because of the increased drop across resistors R30 and R31. If the noise pulse is a strong one, the cathode will be negative with respect to the grid. This positive bias in conjunction with the increased plate-to-cathode voltage causes current flow through the noise limiter tube. The limiter is then the equivalent of a low-resistance (approaching a short circuit) shunt from B to C on the detector load resistance. As this condition starts at the point where the noise pulse amplitude begins to exceed the signal amplitude, the peak of the noise pulse is clipped and the noise output cannot become much greater than the signal output. (With the noise amplitude held to a level only slightly greater than the signal, an operator can read signals that would be unreadable without the limiter.) Immediately following the pulse, the original bias voltage, across resistor R29 again biases the limiter, and normal reception is restored quickly. In practice the limiter is effective in reducing the effect of noise pulses, such as those produced by the ignition system of an automobile. If the amplitude of the noise pulse is lower than the signal amplitude, the limiter has no effect. In other words, the limiter can only prevent the noise output from rising appreciably above the signal output.
(3) Changes in signal amplitude. The limiter stage will affect signals, such as amplitude-modulated voice and music transmissions, which contain amplitude changes. Increases of amplitude above the average level cause the same action that is produced by noise pulses and, therefore, the limiter distorts the signal by clipping the peaks. Clipping and, hence, the distortion may increase considerably if the degree of modulation is high.
(4) Changes of average signal amplitude. The time constant of R32-C65 is long enough to prevent bias changes when noise pulses are received. The time constant is short enough, however, to enable automatic readjustment of the bias (voltage of C65) when the average amplitude of the signal changes. For example, if the level of the received signal increases or decreases because of fading, or if the receiver is tuned to a signal of given amplitude and retuned quickly to a stronger or to a weaker signal, the charge of C65 readjusts itself for proper operation under the new conditions.