[AMRadio] RE: negative cycle loading

[Donald Chester]

If a low-pass filter is used after the series-diode clipper, the filter 
itself will generate phase shift distortion, resulting in a tilt in the 
waveform. A simple pi section (constant-K) filter produces less phase shift, 
but the rolloff may not be sufficient to reduce the splatter to an 
acceptable level. An "M-derived" filter produces a better cutoff, but more 
phase shift.

It looks to me like Steve's (WA1QIX) circuit is a variation of the old 
series-diode limiter, which simply uses a diode in series with the 
modulation transformer secondary (or mod reactor). It has the addition of 
the "keep alive" circuit to allow adjustment the clipping point to slightly 
less than 100% to avoid overshoot, and maintains the correct load on the mod 
transformer during clipped peaks.

There is a major flaw in the old series diode/splatter filter method that is 
usually overlooked. It involves "slew rate" of the waveform following the 
clipper. I'll try to explain it non-mathematically. Beyond a certain 
frequency the low-pass filter, by definition, tries to limit the rate of 
change of the instantaneous voltage feeding the final. (The higher the 
frequency, the greater the rate of change of instantaneous voltage of any 
a.c. waveform and thus the modulated B+ voltage). The problem with the 
series clipper lies with the negative half of the audio cycle, during which 
the instantaneous voltage is decreasing towards zero.

Beyond a certain frequency, the modulated B+ as it exits the modulation 
transformer, is decreasing at a faster rate than the splatter filter will 
allow it to drop. In other words, the stored energy in the filter circuit 
tries to keep the instantaneous voltage at the filter more positive than the 
instantaneous voltage output from the mod transformer/clipper circuit. The 
result is that the clipping diode cuts off, allowing the mod transformer 
output voltage to drop precipitously, while the B+ line to the final is 
temporarily floating free, and as the energy stored in the filter section 
decays, the instantaneous voltage to the final is decreasing at a slower 
rate than the instantaneous voltage output from the mod xfmr. As the mod 
xfmr output voltage bottoms out at the crest of the negative peak, the B+ 
line voltage is still decreasing. As the instantaneous voltage from the mod 
xfmr begins to rise after passing the crest of the negative peak, a point is 
reached where the mod xfmr output voltage equals the B+ line voltage (which 
is still decreasing as the filter energy continues to decay). At that point 
the clipper diode turns back on, and the B+ line once again follows the mod 
xfmr output. The result of all this is that the modulated waveform at the B+ 
line to the final has a sharp peak at the crest of the negative peak, at a 
point somewhat less than 100% negative modulation, and the "pointy" negative 
peak is lopsided since the upswing of the peak is more rapid than the 
downswing; this represents a discontinuity of the rate of change of the 
waveform, exactly as occurs with overmodulation. Even though the negative 
modulation is limited to less than 100%, splatter still occurs at the point 
where the still-decaying B+ line voltage "crashes head-on" with the rising 
mod xfmr output voltage.