Neumann's Last Pole


Neumann's disc cutting system with lathe and electronics rack - circa 1968

The RIAA Standard

As specified in the standard, the transfer function of the RIAA equalisation on the record side is given by,

H(jω) = [(1 + jωT1) (1 + jωT3)] / (1 + jωT2)

where T1 = 3180µS, T2 = 318µS , and T3 = 75µS

From which it's fairly easy to see that H(jω) ⇒ ∞ as ω ⇒ ∞.

This is clearly not possible as no amplifier can ever provide infinite gain. And it wouldn't be desirable anyway because an equaliser with such a transfer characteristic would amplify any spurious supersonic signals and noise thereby putting the recording cutter amplifier, coils, chisel and lacquer under unnecessary stress.

This may seem something of an oversight. However, when the RIAA equalization standard was written, it was just assumed that the lathe manufacturer would include some kind of bandwidth limiting filter to define the upper skirt of the passband, which is exactly what they did.

It was suggested by Wright in The Tube Preamp Cookbook (1995) that Neumann cutting lathes incorporated this bandwidth defining filter by means of an extra, high-frequency pole at 3.18µS (50 kHz) in the cutting equaliser. The existence of this extra pole has become something of an urban myth.

Dubbing this extra time-constant, the Neumann Pole, Wright argued that, because the filter was single-pole and with a relatively "soft" roll-off, its effect was significant at audible frequencies. (It would cause about 1dB loss at 20kHz). Moreover, because Neumann lathes are by far and away the most common, Wright argued that it would be prudent to add a complementary zero in the playback electronics; thereby to restore the overall record-replay transfer function to unity.

The theory of this is fine...... However, as you'll see, no such single-pole filter exists in Neumann cutting amplifiers (or probably those of any other lathe manufacturer)!

The illustration below is a simplified schematic of the SE66 recording equaliser; part of the VG 66 disc recording system electronics rack.


Simplified schematic of the Neumann SE66 recording equaliser in the VG66 disc recording electronics system

In the SE66, the inverse RIAA characteristic is accomplished by the interaction of capacitors C1 and C2 with the other circuit impedances and op-amp' gain. Essentially, as the reactance of C1 falls, the impedance of the input leg of the virtual-earth amplifier falls and the closed-loop gain rises from the 20Hz to about 500Hz (the start of the mid-frequency flat-section). After this point, the gain is held steady by the 18k resistor. For frequencies above this, the reactance of C2 falls so that C2 shunts more of the feedback voltage to ground and less to the virtual-earth summing point at the negative input of the op-amp'. This causes the closed-loop gain to rise from about 2kHz until the circuit "runs out of HF gain" at some unspecified frequency.

Above the passband, the Sallen & Key, active filter formed around the BC107B emitter-follower rolls the high-frequency signal; its breakpoint being 33kHz. The loss due to this active filter is less than -0.5dB at 20kHz.

In other words: there is no 3.18µS (50 kHz) Neumann Pole in sight!

Interestingly, this circuit is heavily dependent on the choice and performance of the operational amplifier at its heart. Essentially, as stated, the T configuration in the feedback limb of the amplifier causes the amplifier to have a rising HF gain until it operates - to all intents and purposes - open-loop. The last pole (the Neumann Pole if you must) is thereby defined, not as a tidy time-constant at all, but by the dominant pole of the operational amplifier compensation.

The choice of operational amplifier in the SE66 was Bob Widlar's paradigm-changing μA709. This was a fine component for its time, but it's an op-amp' with an arthritic slew-rate by today's standards (0.25V/µS). The circuit response therefore actually rises to a peak just outside the passband before falling away due to the op-amp compensation and the effect of the following filter. The overall response is illustrated in the figure below. Note how the active low-pass filter has no effect on response until well above 20kHz.

The SAB-74B

Interestingly, the circuit for the SAB-74B programme equaliser in the later Neumann VMS 70 lathe reverses the circuit elements, so that the supersonic filter (this time built around a complementary-Darlington transistor pair and with a turnover at 50kHz) precedes the op-amp-based 75µS pre-emphasis stage which is itself separate from the last stage where the 3180µS and 318µS time constants are implemented in a further op-amp circuit.

This reversal of the equaliser circuit topology in the later Neumann design was an odd decision because headroom is compromised in such an arrangement compared with the earlier SE66 equaliser as a result of the pre-emphasis stage carrying the full signal, rather than one in which the bass frequencies have already been removed. Nonetheless, the SAB-74B is said to be the most widely used equaliser in the cutting of contemporary records (Self 2010), so its performance is of great interest. Does it implement the Neumann pole?

The answer is a definitive, no. Despite its apparent differences, the SAB-74B equaliser relies once again on the op-amp to provide rising HF gain way beyond 20kHz. The SAB-74B benefits from a more modern and better op-amp': the LF356 FET op-amp with a slew-rate of 12V/µS and a DC gain of 106dB. (Much better therefore than the µA709 with a slew-rate of 0.25µS/V and an open-loop gain of 93dB.)

In fact, the characteristics of the LF356 op-amp would cause the gain of the SAB 74B to peak well outside of the audio band (100kHz) were it not for the presence of the up-stream supersonic filter which limits the peak to 50kHz. But, being an active filter with a reasonable Q, there is no appreciable effect upon the amplitude response at 20kHz as the figure below illustrates.


Frequency response of the Neumann SAB-74B amplifier with and without the supersonic, active filter. Note the total absence of the Neumann Pole

Conclusion

The Neumann Inverse-RIAA filters are interesting circuits which turn out to be rather dependent on op-amp' type and performance.

Nevertheless, it can be stated categorically that Neumann disc lathes did not introduce an extra, single-pole into the overall transfer function at 3.18µS (50 kHz); or indeed at any other frequency. Despite the inclusion of supersonic, low-pass filters, the equalisers adhere to the 6dB/octave rising response demanded by the RIAA standard at, and beyond, 20kHz. (For a discussion of phase-response, see the Appendix.)

For these reasons, Stereo Sauce software RIAA equalisation does not include the infamous Neumann pole in the transfer-function of the equaliser. Circuitry was never included to implement this in Neumann lathe electronics and the inclusion of a zero in the playback RIAA de-emphasis at 3.18µS can only worsen the overall frequency-response.

The Pspatial Audio Groove Sleuth Preamplifier similarly does not implement a zero at 3.18µS when the RIAA option is included.

Other names given for this pole in the replay transfer-function are 4th Pole or (incomprehensibly in the circumstances), Enhanced RIAA or eRIAA.

REFERENCES

Wright A. (1995), The Tube Preamp Cookbook , Vacuum State Electronics.
Blauert, J. and Laws, P. (1978) Group Delay Distortions in Electroacoustical Systems Journal of the Acoustical Society of America Volume 63, Number 5, pp. 1478-1483 (May)
Self, D. (2010) Small Signal Audio Design. Elsevier


Appendix - the Phantasy of Phase

Of course, there will be those that say, "Ahhh, but what about the phase response? The active, supersonic low-pass filter must be introducing untold havoc which the "Neumann Zero" helps to compensate....."

To some degree, this is correct: a zero above 20kHz on playback does introduce some phase-advance which compensates for some phase-lag introduced by the supersonic filter during recording. However, the phase-distortion of any active filter is best expressed in terms of group delay (the relative delay a filter imposes on various frequencies, or more technically ∂φ/∂f.) The group delay for band-limiting filter in the SAB-74B is illustrated below. It shows that the delay variation is <700nS; equivalent to the delay imposed by 0.2mm as sound travels in air and a result some 3000 times less than the threshold of audibility for group-delay at high-frequencies established as by Blauert and Laws (1978).


Group delay response of the supersonic filter in Neumann's SAB-74B cutting amplifier


Links

Home page

For all support issues, go here.

For Pspatial Audio sales, email: sales@pspatialaudio.com



© Pspatial Audio 2015 - 2016. All rights reserved. Apple Certified Developer. Stereo Sauce, Aria 51, Aria 20, Head Space, Groove Sleuth, iLOOP and FRANCINSTIEN T-Sym are trademarks of Pspatial Audio. FRANCINSTIEN and Bride of FRANCINSTIEN (BoF) are trademarks of Phaedrus Audio. Macintosh and the Mac logo are trademarks of Apple Computer, Inc.