June 19, 2020
People say they prefer analog to digital recordings for many reasons. Analog’s assumed superiority comes from a misunderstanding of how digital audio works and how to improve the sound quality of digital audio. The most common analog format that is listed as superior is vinyl. Vinyl has a distinctive advantage as an analog format. It is conceptually easy to understand since the movement of the needle creates a movement of air (you can even hear the music just from the needle movements alone). Improving sound quality for vinyl is as simple as following the groove of the record better, improved vibration reduction and playback speed regulation. None of these benefits and tweaks are a panacea for analog audio. The case for digital audio will be made evident in this article. It will be clear why digital audio beats analog audio of any format in recording.
Many enthusiast purport vinyl is closer to the artistic intent. Throughout vinyl’s history it has been purely a mass-produced consumer standard, not a professional standard. Vinyl’s use in movie soundtracks and musical recordings was extremely short. The Fox Corporation when developing sound for “talkies”, always used optical sound. Warner Brothers movies used vinyl-based sound in the mid 1920s but, by the mid 1930s, Warner Brothers was using optical sound . According to the 1952 SMPTE’s Progress Report by the end of 1951 75% of original production recording for Hollywood was done on magnetic tape. Artists have not recorded directly on vinyl for a long time; thus, violating the artistic intent argument. The ideal music listening experience would be in the control room or a live performance listening to the artist. Magnetic recordings needed to be transferred to vinyl which caused a loss in quality. Vinyl also does not allow for a separation of playback and song composition. So, if you have a continuous piece that is longer than a single side, it has to be interrupted. Tracks are separated on vinyl records by periods of silence but, songs without silence separating them have become very common in the digital era. These fully integrated songs becomes inseparable on vinyl records because, their is no way to denote beginning or ending. This violates the artistic intent. Lossless digital downloads meet artistic intent requirements much better. Consider Mike Dean’s Album 4:20 which was mixed and mastered by the artist. The Album was delivered directly to the consumer at no point did the album need to change formats.
Vinyl has poor value for those who want to argue collecting it is valuable (ask millennials about their beanie baby collections). An example of this: if you bought the original 1973 Pink Floyd Darkside of the Moon (UK Edition) and never played it you could get between $1,700-$3,478.26 (USD). Other than considering that an unplayed record is a waste of money since you do not get to enjoy the music, but this is beside the point. Assuming this record cost $5 (an inflation adjusted version of cost of Tom Petty’s Hard Promises price dispute with his label MCA). looking at the Dow Jones in March 1973, it was at $951.01 while in March 2020 it was $21917.16. If you invested the money instead you would only get $115.23—in this case preserving the Pink Floyd album would have been a worthwhile investment. However, if you pick another popular album from the billboard charts in 1973 Houses Of The Holy (UK Edition) today it could net you between $100-$200. Houses Of The Holy is not a clear winner you can sell at the upper end and it would be worth it otherwise not so much. If you selected to hold on to Don’t Shoot Me I’m Only The Piano Player (UK Edition) you could get between $19-$35—in this case the stock market would have been a way better bet, sorry Elton. All prices were taken from Discog (accessed on April-May 2020—only mint or near mint where used). The only reason the number one album of 1973 The World Is a Ghetto by WAR was not used was because no mint or near mint version could be found for sale.
The following video illustrates some of the deficiencies of vinyl. Ironically, this video was about why vinyl so great, yet it is one of the best explainers of the format’s shortcomings.
TLDW (To Long Didn’t Watch)
- Requires lots of processing of the original recording before you can even start cutting
- To cut a vinyl record you need to limit the bandwidth that your music uses
- Limits the slew rate of your music to preserve the cutting head not your music quality
- “It limits you to what you actually can do”
- Forces you to put the bulk of your content in the midrange frequencies instead of having a full range composition
- Depending on your sonic decisions your playtime will be adversely affected (having more bass reduces your playback time and so does having a louder record)
- To fit on a record your music will need to take on the vinyl sound but not necessarily your mixed sound
Benefits of Digital Audio
Now finally: why should you use digital audio? The story of digital begins with the return to the direct vinyl recording in the late 60’s and early 70’s. Why, after forsaking recording to vinyl directly would direct disc mastering exist again? What problem does this solve? Direct disc recording fixes modulation noise! Modulation noise is unique to magnetic tape and it was very hard to measure.
Modulation Noise – A noise which exists only in the presence of a signal and is a function of the recorded amplitude of the signal. The signal undergoes modulation in both amplitude and frequency. These two types of modulation produce sidebands above and below the signal frequency.
Let’s Breakdown what this definition means. The first sentence means that modulation noise only occur sonce audio has been recorded to the magnetic tape. A tape that has no audio recorded will still have a slight hiss called surface noise, but then will gain additional noise from the modulation noise once a signal is recorded to the magnetic tape. A perfectly noiseless tape will-which doesn’t exist-still have modulation noise. This sentence also means that as the amplitude of the signal increases or decreases the characteristics of modulation noise changes. So, recording louder will reduce surface noise but, will cause an increase in modulation noise. The second sentence means that the frequency and amplitude of a signal is varied over time. In the case of our examples a pure tone with a constant frequency 440Hz that undergoes frequency modulation will now have a frequency that varies from 429Hz to 451Hz. In case with amplitude modulation the amplitude of the signal is varying over time from being multiplied by 1 and 0.98 (which is below human perception of 1.1 to 0.89 if the amplitude was 1). Amplitude modulation of two pure sine waves produce what are known as side bands which means now additional tones are produced at 380Hz (440Hz-60Hz) and 500Hz (440Hz+60Hz). These extreme examples are just with these parameters turned up. The modulation noise uses a combination of FM and AM modulation. The rate of frequency modulation mirrors a Mark Levinson No 515 turntable and the amplitude is modulated by randomly generated white noise to mimic uneven gain size.
Below are auditory examples for different types modulation
- Unmodulated Sinewave
- AM (Amplitude Modulated) Sine Wave
- FM (Frequency Modulated) Sine Wave
- A Sinewave with simulated Modulation Noise
- AM (Amplitude Modulated) Extreme Sine Wave
- FM (Frequency Modulated) Extreme Sine Wave
Frequency Modulation (FM) Example
Amplitude Modulation (AM) Example
Simulated Modulation Noise
AM Modulation Extreme Example
FM Modulation Extreme Example
The source of these errors came from a non-uniform distribution of iron oxide grain size on the magnetic tape. This is because, the iron powder cannot be ground to perfectly even sized grains. Magnetic tapes work by gluing a fine magnetic powder to the tape material. Powder application tends to be inconsistent causing level inconsistencies. This is very similar to gluing glitter to a sheet of paper. Some glitter always comes off. The frequency modulation comes from the imperfect speed regulation of the tape machine . It is noise generated at some frequency multiple of the incoming signal. Modulation noise destroys your low frequency response.
Nippon Columbia, or Denon, wanted to improve their recordings’ quality. In fact, according to Audio Restoration expert Thomas Fine, Denon, when pursuing digital, had one goal: “To produce recordings that were not compromised by the weaknesses of magnetic tape” . Early digital was impossible to edit so it had very limited use cases, but direct disc masters could not be edited either so most early digital recordings where done as backups to direct disc recordings. Denon’s using NHK’s (NHK is like Japanese BBC) early prototype can be heard on the Steve Marcus + Jiro Inagaki & Soul Media recording of “Something”. Honestly, the recording sounds fantastic, maybe a little limited on the high frequency but that’s the only fault I hear.
From what I can tell this was recorded on the prototype NHK system, so it had a sample rate of 30kHz and a nonlinear 13-bit depth, that was built in 1969. Sample rate is how often an incoming analog signal is measured per second. So a sample rate of 30kHz means that the incoming audio signal is measured 30,000 times per second. While bit depth is how many unique values the sampled amplitude can take on in this case 8192 unique values. However, unlike a CD the spacing between amplitude values is nonlinear by design in the NHK prototype. See additional reading for further explanation of digital sound.
Early digital audio was focused on providing the best audio to consumers. Soundstream–another early digital pioneer located in Salt Lake City, Utah–was used as a backup to a direct disc recording for a Virgil Fox session. Jerry Bruck, the recording engineer on the session, remembers receiving a call from Soundstream founder Thomas Stockham that the analog feed had a hum problem -80dB down from peak amplitude which was well into the noise floor of the best recorders of the late 70’s. The best tape recorders with signal to noise reduction could only produce 70dB of signal to noise so this hum was inaudible on tape. This shows you one of the unintended consequences of digital recordings was improved signal-to-noise ratios. As far as I can tell no one really thought that improving the signal-to-noise ratio was that important as modulation noise was already removed. To put this into perspective, using OSHA’s guidelines for noise exposure, no worker should be exposed to an average level over 85dBA for a whole day without hearing protection.
If we make your playback level 85dBA then subtract the 70dB signal to noise ratio (I know it is not A-weighted) this gives us the noise floor of 15dBA that is needed to hear this dynamic range during playback. This noise floor is quieter than a quiet study room (20dBA) and the North Rim of the Grand Canyon (30dBA) . In most playback cases, vinyl records and magnetic tape are effectively noiseless. Soundstream’s test recordings were good but had limited high frequency content due to only sampling at 37.5kHz (the true maximum frequency was around 15kHz for anti-aliasing purposes) with 16bits. This early recording allowed Soundstream to be co-booked for many more recordings. Telarc Records was impressed by these early Soundstream recordings but found the high frequencies limited and asked for an improvement which Soundstream agreed to do. Soundstream bumped the sample rate up to 50kHz.Telarc Records and Soundstream worked with band director Fred Fennell to record an assorted set of pieces. Band director Fred Fennell who directed the live band could only say the words “wow” when he heard the digital recording of the session . Telarc Records is famous for its bass drum sounds but, this is extremely difficult to cut into vinyl records but, trivial for digital recordings. If you own a copy of Telarc’s Stravinsky’s Firebird Suite & Borodin’s Polovtsian Dances on Vinyl and love its rich warm sound, then you agree with me that digital recordings sound excellent.
Now it’s time to discuss digital misconceptions. A common misconception – with sampling is you are losing information. This is not true. An example, let’s say we want to represent data from a single linear line and we need to pick a way to preserve information about the line. We can choose between two representations A or B which contains more information?
Representation A: A set of points
Representation B: An Equation
Y(x)=2x for x equal 0 to 23 with a step size of 1
Neither representation contains more information about the line data. Each can be used to create the other’s representation. Even if we took smaller and smaller steps both would remain interchangeable. This is the same with sampled audio. There is a lot of discussion about how digital audio cannot represent the complex overtones of an instrument because, they surpass the audio bandwidth of even the 192kHz. Can analog recordings represent these higher frequencies? Let’s say you are working on capturing an ensemble containing a full orchestra–a Buchla Music Easel and Minimoog Voyager (both are fully analog synthesizers)–directly to vinyl. You use Earthworks QTC50 omni microphones for orchestra and room microphones. While you record the soloist with a Neumann TLM103 you connect everything to a fully analog SSL XL-Desk. First, the Earthworks QTC50 begins to roll off at 50kHz and both analog synthesizers can produce harmonics past 20kHz but the Neumann TLM103 can only produce up to audible content up to 20kHz . Your nice SSL XL-Desk  has an upper frequency response of 40kHz which is less than your QTC50 so you’re actually rolling off the high end of the microphone. While if you recorded this signal digitally at 192kHz all of this would be preserved. All physically realizable devices have to have finite decaying bandwidths or else they would require infinite amounts of energy to power (look at definition of energy in signal processing and divergent infinite series to gain an understanding of this point or Technology Connection’s Videos ).
Digital audio does not sound worse than analog. Digital has many advantages over analog from degradation-free copying to longer playback times. You will not gain any benefits from playing audio back using analog media–at best it will sound the same. The collectors’ value of analog media is minimal and speculative at best on average. Analog audio limits artistic expression and can conflict with artistic intent. In particular, vinyl has been a purely consumer audio standard for decades and has not factored in music capture, reproduction or production professionally in a long time. If you enjoy maintaining vinyl and the comradery you gain from being in the analog community that is fine. But you will gain nothing in sound quality from analog music playback compared to digital music playback. You will also lose many of the benefits that digital provides like tracks seeking, glitch free playback, error correction, easy archiving and organizing.
- Xiph’s Digital Show and Tell Video
- Xiph’s 24/192 Music Downloads…and why they make no sense
- “Introduction to Signal Processing” by Sophocles J. Orfanidis
- Technology Connection’s “Nyquist-Shannon; The Backbone of Digital Sound”
- Technology Connection’s “Sound By Numbers: The Rise of Digital Sound”
1. “Motion Picture Sound Recording “ by John G. Frayne
2. “The 35mm Album Master Fad” by Thomas Fine presented at the 135th AES Convention
3. “Modulation Noise in Tape Recordings” by Robert Z. Langevin
4. “ The Dawn of Commercial Digital Recordings” by Thomas Fine
5. “An Historical Overview of the development of PCM/Digital Recording Technology at Denon” by Almon H. Clegg et al
6. “How loud is too loud?” by OSHA (https://www.osha.gov/SLTC/noisehearingconservation/loud.html) accessed June 7th 2020
7.Earthworks QTC50 specification (https://earthworksaudio.com/products/microphones/qtc-series/qtc50/) accessed June 7th 2020
8. Neumann TLM103 specifications (https://en-de.neumann.com/tlm-103#technical-data) accessed June 7th 2020
9. SSL XL-Desk Owner Manual accessed June 7th 2020