Sound: Vinyl Records Vs. Digital Files

All music is beautiful
~Billy Strayhorn

A Vinyl Record
A Vinyl Record (source).

I apologize to everyone who looked for my article earlier today. I was delayed in writing it last night. I think I need to start working ahead and building a buffer.

sunstreetreviews asks

Audiophile friends tell me that vinyl records have a much richer sound than MP3 files. Is that really the case, and if so, why?

In principle, there’s no reason a digital file should produce lower-quality sound than a vinyl record. In reality, tests are inconclusive. Why this is the case has to do with how sound works, and how sound is stored on a computer file.

Some Sound About Sound

Sound is a wave of pressure in the air. The frequency of the wave determines the pitch. Let’s unpack this statement. Imagine that I hold a string between my two hands, and I move my hands so that the string wiggles. This is very analogous to sound. In fact, the wiggling of strings on string instruments is what causes the sound they make. The more wiggles on the string, the higher the pitch of the “note.”

Low Vs. High Frequency Notes
The waveform on the left represents a low-pitch note, while the waveform on the right represents a high-pitch note.

As more complicated sound is made (for instance if I speak), the strength of the pressure changes to control the volume and make a pattern (like a word). The result looks something like this:

A single note becomes loud and soft and loud and soft
A single pitch becomes loud, soft, loud, and then soft again.

The combination waves like this is called beating. In a real song, there will be many notes at once, and the note will change as the song goes on, so the change of the higher-frequency wiggles will change.

Sampling

Now that we understand a little bit about how sound works, we can talk about how it’s stored on a computer. Because the computer can only hold a finite amount of information, some of the waveform is necessarily lost. The computer keeps track of the strength of the pressure at evenly spaced points in time. The computer also doesn’t have perfect ability to keep track of the strength of the pressure of the wave at any given time. It estimates the pressure. How good the estimation is is called the resolution. The number of times the computer measures the strength of the pressure wave is the sampling rate.

How a waveform is discretized
The original sound wave (left) is discretized with a low sampling rate (middle) and a high sampling rate (right). The large blue dots indicate that the computer is not quite sure what the strength of the sound wave is.

Vinyl Records

Sound is recorded on a vinyl record by copying the shape of the sound wave onto a groove on the record. The shape of the groove mimics the shape of the pressure wave, just like the plots I’ve shown you above. The resolution is given by the recording device’s precision when it cuts the groove. Our machines aren’t perfect and our recording devices aren’t perfectly precise. Similarly, the sampling rate is determined by the width of the needle that cuts the groove. The needle tip cuts a groove the same width as the tip, and we lose all information about how things change under the needle. So, the wider the needle, the  lower the sampling rate. (I am glossing over some details, many of which I don’t understand.)

A needle cutting a vinyl record
When cutting the groove in a vinyl record, the cutting needle flattens out everything underneath it, and so its thickness determines the “sampling rate” while the precision with which the machine can cut determines the resolution.

This is a bit surprising. We don’t usually think of our world as “discrete” or “digital” in the same way that computer information is. Whether or not space itself has this property is open to debate. However, on a larger scale our world is made of atoms, and these are the “pixels” of the universe.

The Comparison

A vinyl record has a very good resolution and a very high sampling rate. It would take up a lot of computer memory to get the same sound quality. However, we ourselves can only hear so well. Once the sampling rate is high enough, and the resolution good enough, we can’t tell the difference if the sampling rate is higher or the resolution is better.

Furthermore, the human ear hears some notes better than others. By not recording those notes, we can reduce the size of an audio file without any obvious change in audio quality.

“Lossless” audio formats, like .wav, use a high enough sampling rate and a good enough resolution, that we can’t tell the difference from a vinyl record. However, to conserve space on hard drives, we often compromise between audio quality and memory usage. “Lossy” audio formats, like .mp3 use various tricks, like the one described above, to reduce the file size. In the end, whether or not a digital file has a lower-quality sound than a vinyl record depends on the person who produced the file. For more information, see Gregor Robinson’s comment below. Thanks for the correction, Gregor.

Further Reading

Skeptoid has an article discussing the comparison between vinyl records and digital files.

Related articles, courtesy of Zemanta:

Questions? Comments? Hatemail?

As always, if you have any questions, comments, complaints, or insults, please let us know in the comments!

January 9, 2013, EDIT:

It seems I’ve made some serious errors when discussing the technicalities of music recording, both digital and vinyl. This is not my area of expertise and I apparently did not do nearly enough research. I apologize to everyone I misinformed earlier this week, I should have corrected the erroneous information now. Thanks to Gregor Robinson and Jonathan Griffitts for correcting me. Those of you who want to know what changed, check their comments. I will do my research more thoroughly the next time I explore a subject I don’t know well. Thanks for reading, everyone.

9 thoughts on “Sound: Vinyl Records Vs. Digital Files

  1. There is actually a big difference between lossy formats and lossless digital audio formats (which may be in fact compressed; see FLAC). It’s not just that the sample rate is lower. Lossy compression makes use of interpolation and other tricks to approximate the original lossless digital signal, whereas lossless compression makes use of inherent redundancy of the digital information. Some signals are more redundant than others (http://en.wikipedia.org/wiki/Information_theory).

    The MP3 codec, for instance, is a lossy format that relies on a model of the human auditory system to estimate portions of the waveform that are difficult for the human ear to notice. The MP3 encoding algorithm identifies the presence of frequencies that are probably “masked” by the presence of others — and just plucks them out of the file like they never existed. It turns out that formulating accurate models of psychoacoustics is difficult, and people vary, which makes this process imperfect. So there’s no surprise that many people hear significant difference with such codecs. The difference is exaggerated at low sample rates.

    As an experiment, try converting a WAV or other lossless format into pairs of files of similar sampling sampling rates (be sure to disable VBR — how to do this depends on your encoder). Make one an MP3 and the other a WAV. See what happens when you vary the bitrate. Listen to the sea-alien sounds that start to happen when you start cranking down the MP3 bitrate.

  2. This information is not actually correct: “Sound is recorded on a vinyl record by copying the shape of the sound wave onto a groove on the record. The stronger the pressure, the deeper the groove.”

    The sound information is actually recorded by motion on the other axis, wiggling the groove to the left and right (that is, radial motion relative to the center of the record). The groove actually looks like a tiny graph of the sound waveform, big enough to be seen with minimal magnification (well known to oldsters like me who have handled a lot of vinyl records). On monaural records that’s all there is, but on stereo recordings the depth of the groove actually records stereo information, specifically it represents the difference between the two channels.

    I do a lot of work involving signal processing and the relationship between sampling rates and signal bandwidth. Therefore I’m really tempted to launch into a nerdy clarification of the relationship between sampling rate and signal bandwidth, and just what is meant by “resolution”, but I’ll spare you.

    1. Thanks for the correction, Jonathan. I’ll update the post. I think many might be interested in your explanation, so please do correct me on sampling rate and signal bandwidth.

      I intentionally glossed over the trade-off between frequency and time resolution given by Fourier transform. I hope to discuss this in another post. If you think this is a good idea, I will use next Sunday’s post to explain.

Comments are closed.