Maybe we need to step back and take a look at the processes used to both record music to a digital medium and to subsequently replay music stored on a digital medium.
(Let me just add a caveat that I am no expert in this field and, if there are any inaccuracies or errors, I would really appreciate being corrected)Some items to remember regarding the recording process:
- Music, as performed, consists of a complex and continuous analog signal
- Music, as digitally recorded, consists of a sequence of discrete numeric values
- The recording process must convert the continuous analog signal into a sequence of discrete numeric values
- The analog-to-digital conversion process achieves this by taking snapshots (samples) at precise time intervals
- The precision of the time intervals is critical to ensure that subsequent digital-to-analog conversion can be accurate
- The number of snapshot intervals (samples) per second is defined as the "sampling rate" (44,100 samples per second for CD)
- Each snapshot is then converted to a numeric value represented using a resolution of 16 bits (giving 65,536 possible values)
- The resulting series of values is then packaged into a standard format (defined by the Red Book standard) and written to disk
Some items to remember regarding the playback process:
- Music, as digitally recorded, consists of a sequence of discrete numeric values
- Music, as reproduced, consists of a complex and continuous analog signal
- The playback process must convert the sequence of discrete numeric values into a continuous analog signal
- The precision of the time intervals is critical to ensure that they "map" 100% to the sampling in the ADC process
- Various DAC technologies use different approaches to reconstitute the analog signal from the 16-bit snapshots
There are two major sources of inaccuracies in this end-to-end process:
- Errors due to data loss (aka bit drop-out)
- Errors due to differences in the timing between original sampling and reconstitution
Ignoring data loss errors as being "off-topic" in this thread, let's examine the issue of errors arising from differences in timing (ie "jitter").
Most graphic examples use a simple sine wave to show how the reconstitution of the continuous analog signal is achieved by "plotting" the discrete sample values against time, usually resulting in a series of "steps" going up and down. In these examples it is difficult to visualise why timing errors are likely to cause audible "nasties" apart from minor "ripples" in the "plotted" waveform. Music, however, does not consist of a single frequency sine wave of a fixed amplitude, but rather comprises a very complex mix of wave forms of multiple frequencies and multiple amplitudes all running along together. If you can picture this, then imagine a timing error that breaks the precise "mapping" back to the original, then the impact of "jitter" can also be understood - as the impact will vary across the multiple and parallel different waveforms.
In the early days of CD and the motto of "perfect sound, forever", most people believed that the only possible source of distortion in the digital record-store-replay process lay in data loss.
One of the earliest companies to recognise that timing inaccuracies (or "jitter") were another source of distortion was the UK-based Trichord Research (then run by Graham Fowler and Tom Evans). Trichord developed and marketed their Trichord Clock upgrade which basically replaced the stock crystal oscillator (or "clock") in a CDP with a high accuracy (typically 5ppm) crystal oscillator. Trichord continued to develop and evolve this concept through later iterations (Clock II, III and IV).
Back in the mid 1990's I had a Rotel RCD965BX CDP and decided to try out the Trichord Clock II variant available at the time.
The result was a clearly discernible improvement in detail retrieval and spatial cue reproduction (staging and imaging) plus a very clear improvement in clarity at both frequency extremes.
This approach of fitting a more accurate clock was Trichord's approach to "fixing at source" (aka "prevention").
Paul McGowan - now of PS Audio fame - developed a slightly different approach with the Genesis Digital Lens which fitted between the transport and the DAC and was designed to "eliminate jitter" (aka "cure").
Both approaches - prevention and cure - were effective in improving sound quality in the digital replay chain - as attested to by many a reviewer and consumer.
These improvements were all achieved though the reduction of jitter.
So, I would tend to argue against anyone saying that jitter induced errors are undetectable audibly!!
