In Issue 192, J.I. Agnew began his series on the topic of signal levels, and the technical considerations in getting them recorded onto disk. The series continues here with a look at magnetic tape recording.
After early disk recording came tape recording. (Reel-to-reel tape and cassette tape operate on the same principles and share the same units of measurement and reference.) Information is stored by magnetizing ferric oxide (aka rust) particles that are rapidly moving across the front of a tape head. The "fluxivity" reference of 0 dB (known as reference fluxivity) corresponds to 185 nWb/m, as established by Ampex when they first commercialized the technology. As with dBu and velocity, once again, this is neither a maximum nor a minimum, but just a convenient reference.
Signals can go much lower than that, up until they are buried under tape hiss. Different tape types, tape machines, tape widths, tape speeds and operating conditions result in wildly different noise floors. The lower limit is, as usual, left undefined.
The upper limit would theoretically be the point of complete magnetic saturation. This would be the point where all magnetic domains on the tape would be aligned in the exact same orientation (north/south poles facing the same way), where further magnetization would simply not be possible. In practice, this point is rarely reached, because tape stops behaving in a linear fashion long before it reaches hard magnetic saturation. Different tape types have different headroom capabilities and different types of music have different linearity requirements (tape compression sounds pleasing and is considered desirable in rock and pop music, as evidenced by the sheer number of software emulations of this effect, for use by studios that do not have access to tape machines), so the upper limit is a very fluid concept and varies with frequency as well. A typical range between the noise floor and usable levels for practical recording purposes is around 78 dB, for 1/4-inch tape running at 15 ips, without using any kind of noise reduction systems.
Then came digital. Originally described in the 1930s by Bell Labs employee Claude Shannon, it would take several more decades until the commercial application of digital sampling in sound recording was implemented. In the digital world, we have the concept of dBFS, which stands for dB Full Scale. In this concept, the 0 dBFS reference is also the maximum signal level possible to store, which is a hard ceiling. Nothing can go above 0 dBFS, so all other dBFS values have a negative value. In Shannon-based PCM (pulse code modulation), it is possible to theoretically calculate the minimum signal level that can be represented, as a function of the quantization (bit resolution). The dynamic range for 16-bit PCM is 96 dB and for 24-bit PCM it is 144 dB. However, these figures are indeed theoretical, as in practical ADC (analog-to-digital converter) and DAC (digital-to analog converter) units, the available dynamic range is lower, sometimes even significantly so. The Compact Disc format uses 16-bit PCM exclusively, and while on the digital side of things, signals from -96 dBFS to 0 dBFS can be represented in binary code, the electrical output of most CD players do not allow for a 96 dB dynamic range to be achieved.
Digital recording, therefore, has both an upper and a lower limit, with the hard upper limit also acting as the reference. This is also the root of the modern "loudness" issue, the tendency to over-compress the audio signal. Since the digital domain imposes the same hard upper limit of peak signal values for everyone and everything, regardless of the equipment used, the only way to make your music sound louder than other music is by intentionally restricting its dynamic range. The idea is that since you cannot have a peak that is louder than anyone else, you need to reduce your peak-to-average ratio, to increase apparent loudness, But once you start going down that road, it is a race to the bottom of sound quality. Once you succeed in sounding louder than the previous loudest record by restricting the peak-to-average ratio, the next person wanting to be louder than anyone else would just need to restrict the peak-to-average ratio some more, and even more, and so on, until there is absolutely nothing left of the dynamics in the music.
Different releases of ABBA's 1980 song "Super Trouper" show different levels of loudness compared to the original 1980 release. Courtesy of Wikimedia Commons/Kosmosi.
Much of that has nothing to do with sound and music, but is driven by marketing statistics. If a record label demands higher apparent loudness because it is known to sell more records, this is based on statistics of sales as related to loudness. Radio stations have a long tradition of using specialized equipment to "condition" the signal, with an aim of increasing loudness to improve their broadcasting range without an increase of their legal power limit, which directly translates to advertising revenue.
In the days before digital, the race for the loudest album was already on, already prompted by jukeboxes with fixed level controls playing different records one after another. The loudest ones would stand better chances of making a rowdy crowd of youngsters shut up and listen. Maybe some of them would even remember which song it was and would go buy it the next day. And it actually probably did work that way. But with the vinyl disk medium not having a hard upper ceiling for peak levels, there were many ways to make it sound louder without sacrificing the dynamic range, and sounding loud and good was definitely better than just sounding loud.
So, creative engineering came to the forefront: pitch and groove-depth control systems to permit higher levels on disk while automatically preventing one groove from cutting into the next; injecting hydrogen and then helium into the cutter head to reduce the thermal time constant and extract a few extra dB out of it before blowing it up. Improvements in tape formulations and in tape machines in general allowed music to get louder on tape.
But where things can get really complicated is when we are trying to transfer acoustic phenomena with a defined lower limit and no upper limit into electrical signals with no lower or upper limits, to mechanical or magnetic media with only nominal references and no clearly defined upper or lower limits, and with vastly different overload characteristics and problems, before finally transferring that to digital with a hard upper limit and a clearly defined theoretical lower limit which is rarely, if ever, achieved in practice, then back to an electrical signal and oftentimes (especially now with the resurgence in vinyl records and the constant flood of reissues) back to the mechanical medium, before returning once again to the electrical domain, just prior to exiting back into the real world as acoustic phenomena that we can actually listen to. This is what you are listening to. It is an eternal attempt at translating and scaling dynamics and signal levels to best represent a balance between artistic vision and economic viability.
Header image courtesy of Pexels.com/Nikita Korchagin.