I am trying to be on my good behavior and not question someone as respected as "Self". While i would suggest that anyone actually interested in that approach to build it and compare it to even the bare RIAA stage...
Not to add a veer on top of a veer.. but looking at that circuit, reminds me of an old trick I used back in the late '70s. back then I sold a Tape Noise Reduction kit, and a common problem when recording companded record playback onto tape with the tape's reduced LF response, is that very LF energy from warped or not centered records, would get encoded into the envelope of the compressed signal, but when lost on the tape, this missing envelope information would cause a phantom very LF modulation in it's absence to the expanded playback.
The solution I came up with, was a sliding HPF, added to the encode compressor, where the pole frequency shifted with the gain, and therefore signal level hitting the compressor. At high level/low gain the compressor had full LF bandwidth, since warp energy was insignificant in the envelope then. At low level/high compressor gain the LF roll off was more severe, to keep warp energy out of the envelope when it was most likely to cause errors.
My actual circuit was much simpler than this long description. A pole was formed by an input capacitor, and shunt resistor connected to the overall inverted compressor output. This looks similar to Self's impedance trick but instead of running a fixed inverting gain, my inverted gain and therefore effective impedance varies with signal level, giving me high impedance at low gain, to lower impedance at high gain.
Does anybody really miss the LF issues with vinyl?