In order to develop the Groove algorithm, it was necessary to establish realistic performance measurements from LP replay. Surprisingly, little real data exists on this beyond the various popular - and entirely qualitative - articles on the WWW extolling the advantages of analogue discs which are often hagiographic to say the least!
Here are our own test results from a first class deck playing various industry test discs, plus a few theoretical musings.
Measured results
Harmonic distortion on 1kHz tone (established to be -20dB below measured peak groove velocity): 0.44%, 2nd harmonic, and 0.2%, 3rd. (Outer grooves.) Worst case (5cm/s groove velocity, inner groves): 3% THD.
Wow and flutter: No sidebands greater than 60dB below signal. (Very much lower than THD!) This is an excellent result and much better than tape machines of the same epoch. (But see information about manufacturing faults.)
Frequency response: +/-1dB 40Hz to 15kHz (220pF cartridge loading). A remarkably good result considering the mechanical nature of the medium.
Stereo separation: >24dB at 1kHz. Decaying to -12dB at 10kHz. (On commercial recordings, it is often measured that LF decays as well; to about -12dB at 40Hz, but this is signal processing prior to the cutting amplifiers to reduce the LF vertical component.)
Rumble: Can be very low, but in fact much more a function of the disc (and therefore presumably the lathe) than the replay deck itself. John Lennon's Double Fantasy LP recorded -30dB below peak level rumble with a peak at 10Hz (and this is with a two-pole rumble-filter switched into circuit!).
Noise: On a very quiet, new LP the noise (post RIAA equalisation) peaks to about -40dB below maximum velocity when measured on a fast-acting peak-reading meter. A surprisingly poor result. However, the spectrum of the noise is benign. So, if the result is carefully de-clicked so that, even very low level "clicks and pops" aren't influencing the result, the noise peaks to about -60dB below peak. The practical dynamic range of an LP may therefore be taken to be about 66dB (by assuming that the RMS value of the noise is about 9dB below the peak value and that the RMS value of a peak sinewave signal is 3dB below maximum modulation level or 0dBFS on a peak-reading meter.1)
But what is the theoretical dynamic range? Records are a highly-standardised mechanical-medium, it is therefore posible to calculate the dynamic range.
The standardised groove geometry on an LP record is of a 2 thou* groove on a 5 thou spacing (the latter being based on 200 grooves per inch). As the diagram right illustrates, the absolute maximum modulation of a groove is ±1.5 thou which is equivalent to 76μm pk-pk modulation.
At the other end of the scale, the surface roughness of an LP is about 25nm, this being about 1⁄20th the wavelength of light2. So calculating we can say,
76μm⁄25nm = 3040 = 69dB
This is happily very close to the figure derived from measurement, the latter being inevitably a little higher due to mechanical noise in the tuntable bearings and electrical noise in the cartridge and preamplifier3. So we can be very confident of this figure. We can say that an LP record is roughly equivalent to a correctly dithered 13 bit-resolution digital-medium3.
* Thousandths of an inch, equivalent to the American term "mil" and equal to 25.4 μm.
Oscillogram and spectrogram of 300Hz at 5cm/sec, inner grooves. Distortion is predominantly 3%, 2nd harmonic
References and Notes
1. There are several issues with measuring noise levels with peak meters because you have to relate peak values to RMS values; the latter being the form in which noise voltages and levels are expressed.
We can assume the RMS value of Gaussian noise is 1 ⁄ 6th of the peak to peak value (see illustration right). In other words, the RMS equivalent is 1 ⁄ 3rd of the peak value (for example, as shown on the DAW peak meters). One third = -9.5dB. In other words, the noise voltage RMS value is 9.5dB below the peak value. However, we can't express this RMS equivalent with respect to 0dBFS because that is itself a peak value. We need to express it with respect to an RMS maximum value. Assuming a sinewave, the maximum RMS signal is 0.707 of (or 3dB below) 0dBFS.
Thus two things:
The RMS value of noise is 9.5dB that which reads on a fast, peak-reading meter.
The dynamic range of the system is 6.5dB below the meter readings. Thus, if the meters hover around -75dBFS, the dynamic-range is 81.5dB.
2. Versatile Stereophonic Pickup, J. Walton. Wireless World August 1961. (Presumably Walton was taking an average figure of 500nm for the λ of light; a reasonable assumption. Walton worked for Decca.)
3. In fact, the various mechanical "rumble" mechanisms and electrical interference (especially hum) are the principal contributors. The internal resistance of a typical moving-coil phono cartridge is 10 ohms. The maximum RMS output from the cartridge is about 5mV (that being 20dB above the nominal 0.5mV @ 5cm/s level).
The RMS thermal noise generated in 10Ω in a bandwidth of 20kHz is 57nV. So the ratio between maximum signal and noise voltage is,
5mV⁄57nV = 87,720 = 99dB.
This is very substantially (28 times!) better then than the theoretical maximum from the vinyl medium. The preamplifier will probably reduce this margin by 6dB but clearly the electronics are comfortably better than the mechanical limitations of the medium itself.
A similar figure for dynamic-range is possible from a moving-magnet type provided that the inductive nature of the cartridge is isolated (as it is in the PHLUX cartridge.) Here, the internal generator resistance is 100 times higher but, because of the square-root sign in the calculation for thermal noise, the noise is only 10 times higher. The output voltage is also about 10 times higher too, so the ratio of maximum-signal to noise is maintained.
3. At least it is at low frequencies. Groove excursion is limited by RIAA equalisation to a quarter of the 76μm pk-pk amplitude above 2kHz when the record is cut. Thus the resolution of the medium is reduced to 9 bit at high-frequencies. For more information concerning maximum recorded levels on an LP, see this page.
In order to develop the Groove algorithm, it was necessary to establish realistic performance measurements from LP replay. Surprisingly, little real data exists on this beyond the various popular - and entirely qualitative - articles on the WWW extolling the advantages of analogue discs which are often hagiographic to say the least!
But what is the theoretical dynamic range? Records are a highly-standardised mechanical-medium, it is therefore posible to calculate the dynamic range.
Apple Certified Developer. Stereo Lab, 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.