It inevitably causes distortion, since it makes the modulator tubes work into a non-linear load, plus it wastes hard and expensive-to-produce audio power, and I'm not sure it actually accomplishes anything. I once tried the original "ultramodulation" circuit that appeared in the 1956 issue of QST, using three 866-As. At first it looked good; the rf ammeter in the feedline kicked up higher with modulation, the plates of the PA tube (a 304TL) got brighter orange, and the DC plate current meter to the final kicked up slightly. I don't recall that I had a monitor scope at the time; I may have observed expanded positive peaks (actually, reduced negative peaks). BUT numerous on-air A/B comparisons yielded similar results; the signal sounded more raspy and distorted, but the audio didn't sound any louder or more piercing through the background noise. An analysis of the circuit explains why. The upward kick in plate current and rf feedline current, and brighter plate glow in the 304-TL were due to the upward kick in DC plate voltage, resulting from the portion of the rectified audio voltage that was not wasted in the power resistor. The main effect of the negative cycle rectification on the audio waveform itself was to add a component of even harmonic distortion. In effect, all that circuit does is to function as an inefficient form of controlled-carrier modulation with a strong even-harmonic distortion component added. A negative cycle recertification scheme that does work is to use a power resistor of the same DC resistance as the modulating impedance (plate voltage divided by plate current per Ohm's law). It doesn't have to be very large, maybe 10% of the modulator output power, or 5% of the DC input. Connect the cathode of the diode to the modulated B+ line, the anode of the diode to the resistor, and the other end of the resistor to ground. That way, during negative overmodulation peaks (when the PA plate is driven negative and stops conducting), the diode conducts and the resistor takes over to maintain a constant load across the secondary of the modulation transformer over the whole audio cycle, thus protecting the transformer from working into no-load (infinite impedance) during negative peaks that might exceed 100%. Since those peaks are likely to be of short duration and limited amplitude, there is no need for a large wire-wound resistor. In most cases a 20 to 50 watt resistor should be more than adequate even at the highest DC input power an AM ham transmitter might ever run. I don't use this circuit in any of my transmitters to-day, since the broadcast-type peak limiter pretty much eliminates over-modulation peaks, but back when I ran the 304-TL I didn't have any kind of peak limiter, so this circuit might have offered some protection to the transformer. Not having a monitor scope at the time, I went a step further and used an old globe shaped 866 (non-A) rectifier which has no shield over the rectifier filament. The blue glow was very prominent and served as an effective over-modulation indicator, since the 866 conducted when and only when the instantaneous plate voltage dropped negative to produce the blue glow. I used a small cardboard box to house the 866, spray painted the interior flat-black, and cut a window in the side of the box, making the blue glow more visible. I could not believe how lightly I had to drive the modulator to make the 866 flash blue, and actually operated it so that it always flashed a little on modulation peaks. I didn't have a monitor scope at the time and wasn't aware of the voice peaks that occurred at a fraction of the modulator plate current that the tube manual said the tubes were supposed to draw to produce half the audio power as the DC input power to the final. Of course I realised later on that the output power prescribed in the tube manual was for a sine-wave audio tone, and that the peak-to-average power for a voice waveform was far higher. Back then, most receivers were broad as the side of a barn and very few hams had panadaptors or spectrum analysers, so no-one ever complained that I was splattering even though the 866 flashed blue indicating over-modulation. I did blow a few modulation transformers back then, but that's a whole unrelated story that I have posted, in an unrelated topic. All that the "turbo connection" is, is to combine the two windings of the transformer to form a modulation autotransformer. Modulation autotransformers are not all that rare; they were used in several Wilcox transmitters, for example. Instead of a separate primary and secondary, the same winding is used as both. A simple mid-tapped winding would work, but it would produce a 2:1 step-down turns ratio, or 4:1 step-down impedance ratio. The autotransformer adds a section of additional turns to one side, out beyond the plate connection of the coil. That extended section of winding feeds the PA plate, producing more voltage swing than just half the plate-to-plate voltage as with a simple mid-tapped winding, thus producing less step down. The main precaution is that unless a separate modulation reactor and blocking capacitor is used, the modulator and PA must run off a common power supply when a modulation autotransformer is used. The turbo connection re-purposes the original secondary of the mod xfmr to serve as the turns extension to one side of the primary, converting it into an autotransformer. The turbo section must be phased correctly, to be additive and not subtractive. If the original transformer secondary has several taps, this allows the amount of additional turns for the turbo section to be chosen. Otherwise, you have what you have. The step-down may be greatly reduced, or the transformer may act as a step-up, not step-down. This allows a lot of positive peak head room, but may present too low a plate-to-plate impedance for a given pair of modulator tubes, requiring different tubes or maybe running a quad of the original types in push-pull parallel to successfully work into the lower p-p impedance. Also, since fully driving the modulator tubes may produce severe over-modulation in the negative direction, running back the audio level to 100% negative modulation may result in low modulator tube efficiency. As with any radical circuit modification, YMMV.