Cut and Thrust: RIAA LP Equalization Page 3
The origin of the "Neumann 3.18µs pole" appears to be anecdotal. Certainly, the notion doesn't withstand scrutiny of Neumann's SAL 74B cutting amplifier, which, together with the SX 74 cutting head, is the most popular combination in use today. As I first saw Swedish mastering engineer Goran Finnberg remark in a post on the ARSCLIST forum (footnote 13), the E (equalization) card of the SAL 74/74B has as its initial circuit element, immediately following the Mod input, a Sallen-Key active unity-gain second-order low-pass filter constructed around a BC212B/BC182B transistor pair. This has a corner frequency of 49.9kHz and a Q of 0.72, making its alignment almost that of a maximally flat Butterworth filter (footnote 14).
Taking into account an additional RC network prior to the following op-amp, which has a corner frequency of 482kHz, this filter's SPICE-simulated amplitude response is as shown in fig.12. It is less than 0.1dB at 20kHz, compared to the +0.64dB of the "Neumann correction" at this frequency. So the "correction" actually adds to in-band amplitude error rather than reducing it. It does somewhat reduce phase distortion, but, as already remarked, this is almost certainly inaudible in any case.
Fig.12 Amplitude vs frequency response of the input filter on the E-board of the Neumann SAL 74/74B cutting amplifier. This is quite different from that assumed by the "Neumann correction" in fig.6.
In other wordsat least for records cut using the SAL 74/74Bthe first-order "Neumann correction" makes matters a little worse rather than better. A modified correction regime is not the answer, since this entire approach ignores numerous realities about phase and amplitude error in the record-cutting process and elsewhere that, taken together, indicate that no fixed deemphasis correction can ever be justified.
For instance, the Neumann SAL 74/SX 74 uses a feedback system to control the motion of the cutting stylus, but this doesn't operate over the entire audible frequency range. It is fully active only to 14kHz, above which it is rolled off. An optional shelve-down filter, enabled or disabled by connecting or disconnecting circuit-board pins, is provided to control the cutting level through the last half-octave of the audible frequency range. High-frequency phase and amplitude response will alter substantially according to how this filter is set. Worse, the SX 74 cutter has a mechanical resonance at around 22kHz, above which the signal level cut on the disc falls off rapidly. This inevitably introduces additional amplitude and phase errors. Other makes of cutterand earlier Neumannswill behave differently, both electrically and mechanically, complicating matters still further.
Still more amplitude and phase errors are almost certainly present in the source recording. Large-diaphragm studio microphones often begin rolling off before 20kHz, again introducing much larger in-band phase and amplitude errors than Neumann's RIAA preemphasis modification. These may be added to by the inherent rolloff of analog tape machines and, particularly in older recordings, by electronic rolloffs, some of them due to the use of transformer coupling.
At the replay end are still more sources of amplitude and phase error, caused by electrical interaction between the pickup cartridge and its load, and as a result of resonance between the effective tip mass of the cartridge and the elasticity of the record vinyl, which can occur anywhere between 15 and 50kHz. Plus, there may be nontrivial phase errors introduced by the system amplification, as there may be by the loudspeakers. All of this makes the high-frequency amplitude and phase performance of the complete recording and replay chain extremely complex, and impossible to predictand thus to correct with a fixed filter.
Research conducted by Ortofon in the early 1980s, which led to the introduction of its Ortophase cartridges, concluded that high-frequency phase effects are audibly significant in record replay, but that for optimum results, a particular balance needs to be struck between amplitude and phase responses (footnote 15). So what's really needed, if anything, is not fixed changes to RIAA deemphasis, but a variable HF phase corrector analogous to the low-frequency equivalent that Peter Craven and Michael Gerzon suggested for undoing the highly variable phase errors that also occur at that end of the frequency range (footnote 16) (although, when it comes to phase distortion, low-pass filters are less culpable than high-pass filters).
To be practicable, though, such an adjustable phase corrector would probably have to be digitalwhich doesn't quite fit the bill for analog disc replay. The best advice has to be to leave well enough alone. The RIAA replay curve does not need tweaking.
I am grateful to Sean Davies for his comments about the realities of disc cutting, and for supplying copies of Neumann SAL 74B circuit schematics.
Footnote 12: John Atkinson, darTZeel NHB-18NS review, Stereophile, Vol.30 No.6, June 2007.
Footnote 13: Link.
Footnote 14: This use of a second- rather than a first-order filter makes eminently good sense. A single ultrasonic pole would merely flatten off the ultrasonic part of the RIAA preemphasis curve, whereas a second-order filter introduces a 6dB/octave rolloff beyond the audioband. As I understand it, Neumann had encountered problems with radio-frequency interference when it exported cutting equipment to the US, where some was used well above ground level in high-rise buildings equipped with roof transmitters. In such circumstances, a mere flattening of the record EQ ultrasonic response would not have provided adequate RFI protection.
Footnote 15: Torben Groth and Frits Nygaard, "A Method for Optimizing Sound Quality in Moving Coil Cartridges," Preprint 1866, Audio Engineering Society 71st Convention, March 1982 (available via www.aes.org).
Footnote 16: Peter Craven and Michael Gerzon, "Practical Adaptive Room and Loudspeaker Equaliser for Hi-Fi Use," AES UK DSP Conference, March 1992 (available via www.aes.org).