The RIAA Replay Characteristic for Phono Stages, and Inverse RIAA Networks

Tavish Design phono stages are designed to implement the traditional RIAA replay curve. We do not implement the 1976 IEC Amendment which rolls off the replay response at 6dB/octave below 20Hz, and we do not implement the occasionally discussed “Neumann pole” at 50 kHz.

The 1976 IEC Amendment adds a 1st-order high-pass characteristic to the replay curve with a -3dB point at 20Hz. This adds rumble filtering to the playback response, but it does it in an intrusive and ineffective way. A first-order filter doesn’t give much attenuation to typical 5-15 Hz rumble frequencies, but it does affect in-band response by -3dB at 20Hz and -1dB at 40Hz. If a rumble filter is needed, we believe it is much better to implement it as a higher-order filter (such as a 3rd-order Butterworth filter with a 25Hz corner) which would more sharply attenuate rumble frequencies while also having less in-band impact.

There is a belief circulating that record cutting lathes incorporate an unofficial pole at 50 kHz to protect the cutting head from ultrasonic frequencies. While it is almost certainly true that some form of ultrasonic filtering would be required, it would likely be implemented as a more effective higher-order low-pass filter that would more sharply attenuate ultrasonic frequencies while having minimal impact on the in-band response at 20 kHz and below. The people who design record cutting lathes understand the RIAA characteristic and one has to believe that they would implement any ultrasonic filtering in a way that had minimal impact on the RIAA pre-emphasis. At any rate, the 50 kHz pole is not part of the RIAA standard.

Tavish Design calibrates its phono stages against an inverse RIAA network as described in Lipshitz and Jung. We have verified our network against the theoretical RIAA response to within ±0.08 dB. There is an additional measurement tolerance of ±0.05 dB when measuring the phono stage against the inverse RIAA network, so our RIAA measurements can be considered accurate to within approximately ±0.15 dB.