Note: This article has its roots in digital audio and video reproduction. However, it's quite applicable to many areas of data-acquisition and , especially, signal conditioning due to similarity of components used in these outwardly dissimilar disciplines. Indeed, DAQ boards may be tweaked with inexpensive, household items -- as I've tweaked this PC audio card -- for improved signal conditioning and data transfer.
Protecting CD/DVD players and ROM drives from external vibrations -- using aftermarket supports, spikes, stands, and home-made DIY solutions -- can reduce error rates and improve audio/video fidelity as well as data transfer.
When the era of domestic digital audio began in 1982/1983 with the launch of Compact Disc, another era was supposed to end. No -- not the demise of LP as such but audio's salvation from the effect of external vibration, which had dogged analog disc replay throughout its history. In its crudest form this sensitivity manifested itself as the groove-skipping ping that would result from an inadvertent nudge of the turntable or passing footfalls on a suspended floor. But this wasn't something to which CD was entirely immune.
Give any optical disc player a firm enough whack and it too will skip, albeit without the risk of damage to the disc. The real benefit of digital audio in this respect, or so believed, was that it would banish the lower-level, more subtle effects of acoustic feedback from the speakers and thereby spare us the hand-wringing over support tables, which were just becoming a major theme of audiophile life.
This turned out to be wishful thinking. Far from being immune to external vibration, optical players (a category which includes drives and burners) proved to be equally sensitive to it in their own way. Hence the plethora of support tables and other isolation devices -- some of which physically can't provide much isolation at all -- that still adorn the hi-fl accessories market.
Such products can elicit remarkable improvements in CD sound. Indeed some have marveled at the improvement in the clarity and organization of the sound elicited by placing a Stilipoints Component Stand under the system's CD player. Some feel this is not a change that could be called subtle. More than 20 years after CD's introduction, its susceptibility to external vibration still surprises many audiophiles, computer users and engineers.
The favored explanation for this occurrence has always focused (bad pun) on the transport mechanism. In the very worst case, vibration might cause reading errors that result in the player error concealing (interpolating) missing data. But this is a gross effect, not something that should afflict any stationary optical disc player in normal use.
At lower levels of vibration the data recovered from the disc should mostly be perfect once error correction has been applied, so changes in sound quality are presumed to occur because the player’s tracking and focusing servos must work harder than they would in perfectly quiescent operating conditions. Their increased current draw modulates the power supply voltage, which can elicit audible consequences elsewhere in the circuitry of the CD player.
In 2001, Ed Meitner of EMM Labs noted his pioneering use of deep cryogenic treatment -- his experiments into CD vibration effects: "If you put a CD player into an anechoic chamber, in front of a loudspeaker and sweep frequency, you will find a really vicious peak in the focus servo current around 800Hz, very high Q The disc resonates and the focus wants to follow it. This current demand modulates the power supply and generates jitter, which is influenced by the acoustic energy going to the CD player from the speakers. Since it’s around 800Hz, we have this problem with female voices. If you know someone who can sing in that frequency range, very loud, they can shut CD players down (because once focus is lost, the player can’t recover)."
But this isn't the only way in which vibration can affect optical disc players, as Meitner explained. Another thing that is influenced by mechanical vibration is the crystal oscillator (or clock). In the old days we got crystals in a metal can and that itself was very resonant. So we had the resonance of the chassis, the resonance of the board, and the resonance of the can. It’s very difficult to sort all that out.
Meitner notes: "I once made a recording from our jitter analyzer on a cassette while playing music in the room (where the CD player was operating). There were some situations where you could hear the words of the song within the jitter. Luckily, today we can get crystal oscillators that are embedded in a tiny ceramic chip. You can hit them with sledgehammer and it doesn't cause any jitter. Even printed circuit boards are affected by vibration because you have piezoelectric events through bending. That's why our two-channel converter any lo analog connections are not via a PCB trace but a wire."
Ever since jitter became a headline issue in digital audio roughly 15 years ago, it has been widely credited with being the root cause of why many CD players and other digital sources sound different. Yet there has been little attempt to quantify how much of it's generated when playing music in circumstances representative of normal use, rather than with synthetic signals on the test bench. Ed Meitner’s findings suggest that perhaps we've been measuring jitter incorrectly all this time, even with test signals, because equipment’s susceptibility to external vibration has not en stirred into the test procedure.
To try to establish whether this is indeed the case, let's reproduce Meitner's results, albeit in a listening room rather than in an anechoic chamber. The test setup: First generate a test signal comprising an 11.025kHz sine wave at an amplitude of -3dBFS, and burn it to CD-R for replay.
High amplitude, high frequency tones are the most sensitive test of jitter because the sidebands which appear on either side of the test tone are of higher amplitude than they would be with a lower frequency or amplitude, making them easier to detect. A spectral analysis of the test tone is shown in ill. 1, using the same vertical and horizontal scales as for the figures that follow. Zero on the frequency scale corresponds to 11.025kHz, which puts the test tone right in the middle of the spectrum. Frequencies of up to ±3kHz are accommodated on either side, so that the spectrum covers 8.025 to 14.025kHz overall.
Figure 2 shows what happens if some jitter is introduced to the test signal, in this case using computer simulation. Five frequencies of sinusoidal jitter (500Hz, 1kHz, 1.5kHz, 2kHz, and 2.5kHz) were applied to produce this spectrum, each at an amplitude of 50 picoseconds RMS, which is equivalent to half of one ten-thousandth of a millionth of a second. As one can see, the resulting sidebands appear at either side of the central peak, at spaces corresponding to the jitter frequencies and with an amplitude of about -115dBFS.
This confirms the sensitivity of this type of jitter measurement which, using 24-bit recording and a high degree of FFT averaging to suppress the noise floor can identify jitter components down to less than 10 picoseconds.
Measurements taken from three players -- a Meridian 508.24 CD player, Arcam DV89 DVD player, and the Meridian combined with an old Arcam Delta Black Box outboard DAC -- using a Lynx L22 sound card both to record the player’s right channel analog output playing the test disc and to generate a test signal that was replayed through a loudspeaker placed close to the player. Sound pressure level at the player was around 105dB (varying somewhat with frequency due to standing wave effects in the room), which is substantially higher than would usually be experienced in normal listening. Each player’s output was recorded with and without the loudspeaker operating, to see whether the vibrations from it had any effect on the measured jitter performance. Various signals were used to bombard the players, as will be described.
Let's begin by measuring the Meridian and Arcam players, using a linear swept sine wave of 30 seconds duration running from 750 to 850Hz, to see whether this would provoke the c800Hz disc resonance found by Meitner. The results are shown in Figs. 3 and 4, respectively, the light blue spectrum in the foreground representing the result when the loudspeaker was not operating and the red spectrum in the background showing the result when the speaker was active. In both cases the differences between the two spectra are minimal suggesting little if any effect from the loudspeaker. Sweeps from 125-500Hz and from 500Hz to 2kHz were also applied to the Meridian and Arcam players, again with little visible effect.
Using the old Black Box and board DAC with the Meridian was prompted by recollections that its Philips TDA1541 DAC chip was very sensitive to vibration. Indeed, Arcam went to the trouble of isolating it in the Black Box by placing it on an “island” within the main circuit board that is detached around most of its perimeter, like a moated castle.
Removing the Black Box's top plate and standing it on end atop the Meridian to expose it more fully to the loudspeaker gave the results shown in Figs. 5 and 6, for sweeps of 125 - 500Hz and 500Hz-1.5kHz, respectively. This time there are clear effects on the jitter spectrum, with peaks at ±518Hz. Exposing the Black Box to a steady tone at this frequency gave the spectrum of ill. 7, where the vibration-induced jitter side-bands are now of even higher amplitude because they are no longer being diluted across a swept frequency range.
Here is hard evidence that CD players -- and optical drives in general -- can indeed be affected by vibration. Let's put this in context. The effect was substantially reduced when the Black Box's top panel was replaced (ill. 8); the sound pressure level to which it was being exposed was much higher than would usually be the case in normal use; these jitter levels are way below experimentally determined audible thresh olds; and the Black Box is an old, out-moded design whose point of vibration weakness is almost certainly the defunct TDA1541.
It is a relief, of course, that modern optical-disc drives appear to have suppressed vibration-induced jitter effects to insignificant levels. But this makes it even more difficult to explain why vibration-reducing tweaks can still affect (improve) audio/video quality and data-transfer integrity. It remains somewhat of an unresolved mystery.
WHAT IS JITTER?
Digital audio operates by sampling the amplitude of the original waveform at precise time intervals during recording, and reproducing the sampled amplitudes at exactly the same time intervals on reply. Only if each sample is measured and expressed at precisely the right instant will the recording and replay processes capture and reconstruct the signal accurately. Any temporal errors -- termed jitter -- will change the wave shape and thereby introduce distortion.
The nature of this distortion—just as with analog timing errors in the form of wow and flutter—is generally frequency modulation, although there is a second manifestation of jitter that occurs with sonic digital converters which results in amplitude modulation. So unlike quantization error, which has no analog equivalent, jitter is something broadly familiar from the pre-digital era. Wow and flutter in analog tape recorders, the same in cutting lathes and turntables, and Doppler distortion in loudspeakers all have a broadly similar effect and are present at a level which is generally very much higher than the jitter in high quality digital audio equipment.
This suggests that the jitter in such equipment should he inaudible, and the few attempts to ascertain the subjective significance of jitter scientifically have reached that same conclusion. But anecdotal evidence indicates otherwise. Many audiophiles have experienced the changes in sound quality that result from fitting improved, lower-jitter master clocks to CD players, or have had the opportunity to experiment with different clocking provisions within the same equipment. Typically the improvements heard in these instances are obvious.
So what makes jitter audible at much lower levels than predicted? One of the key issues is likely to be signal -- dependence, where the pattern of jitter varies in accordance with the signal. This can potentially occur through a variety of mechanisms, outlined in the text, including equipment susceptibility to the effects of external vibration from loudspeakers.
EXAMPLES OF JITTER-REDUCTION METHODS
Also see: Digital-to-Analog Converter, Analog-to-Digital Converter, Digital Filter
Other Resources on this Topic:
Updated: Thursday, December 22, 2016 18:15 PST