The recently published National Recording Preservation Plan from the Library of Congress includes a recommendation to "encourage scientific and technical research leading to the development of new technologies to recover, reformat, and preserve audio recording media". Although at first sight this passage seems to refer to high-tech projects such as IRENE, there may be other, more modest ways to advance audio preservation technology. Here is an example.
WNYC's NEH-funded project has made freely available over 700 hours of unique, vintage audio newly dubbed from WNYC discs and tapes housed at the New York City Municipal Archives. The project also accumulated a lot of ancillary information ("technical metadata") about issues ranging from condition of the original items to what choices were made during the reformatting process. (This is typical for a project of this ilk: technicians and archivists have a tendency to document processes fastidiously.) It is generally standard practice to keep this kind of information within the internal confines of the project. But what if other technicians and archivists could learn from it?
For example, consider stylus size selection, one of the disc-playback procedures most shrouded in mystery. Until the 1950s, with the wide adoption of "microgroove," disc grooves varied widely in size. It is generally known that groove size grew smaller over time, but the rate varied greatly from company to company. Although the vast majority of styli manufactured today are for the vinyl-era microgroove (still in use today), a handful of specialized vendors manufacture disc styli in a wide range of sizes and shapes, with tips ranging from 0.6 thousands of an inch (mils) to 4.0 mils or more. Since production runs are extremely small, the styli are quite expensive.
What stylus size should today's transfer technician use to play back vintage discs? Most of the available literature gives broad and often contradictory advice, and even a recent draft document by the scientific-leaning Audio Engineering Society (AES) starts with the vague assertion that "the correct stylus is the best-sounding one". Although the document does not define what "good sound" means in this case, there is no question that the sound extracted from a disc can change dramatically depending on stylus size and shape. In this example, note how much cleaner the second pass sounds, with a stylus only two ten thousandths of an inch (or 0.2 mils) smaller:
Faced with potentially dozens of choices, transfer technicians use their ears daily to pick one stylus for its "best sound," and many of them document their choices. What do their choices tell us?
As an example, we ran some simple statistical analyses on a subset of the data from WNYC's NEH project. First, a bit of background: the discs in this project are variously called "transcription discs," "lacquers," or (misleadingly) "acetates". Their materials and principles are still used in modern vinyl manufacturing. But radio transcription discs are different from commercial shellac "78" discs of the same time period, from their physical size (sixteen inches in diameter, as opposed to ten or twelve) and rotational speed (33.3 RPM, unlike the 78 RPM of commercial discs) to having slightly smaller grooves.
The first figure (click on the figures to see larger versions) shows the overall distribution of stylus sizes for 901 sides of the NEH project. Note that the most popular size was 2.8 mils. This does not conform to most recommendations for transcription discs in the available literature, which suggests using styli from 2.0 to 2.3 mils. Our two main transfer technicians did not arrive at this size with any preconceptions, but rather the bias developed over time: most discs just seemed to "sound better" with a 2.8 mil stylus.
Although measuring grooves was not part of the normal workflow of this project, for this study we measured the grooves of a small random sample of discs covering a range of years before 1960, and found them to be all in the 4 to 5 mil range edge to edge. Since the groove walls are at 90° with each other, basic geometry tells us that a 2.8 mil stylus would be riding quite high in the groove. Is this stylus choice the result of a particularity of our collection? Is it due to our transfer technicians' preferences? Forgive me if I give the standard reply: "More work needs to be done" comparing our collection to other broadcast collections.
The project's workflow was such that discs which sounded unsatisfactory with any stylus were set aside by the two main technicians for transfer by another technician with a wider range of styli (and more time!), one of which was chosen blindly. The stylus distribution on these 76 of these "bad shape" sides is quite different (next chart); in this case, 2.0 mil was the most popular size. This could obviously be a case of technician's bias; but interestingly, there were no significant stylus choice differences between the two main technicians. So a potential conclusion is that, in our collection, discs seemingly in worse shape simply sound better with a 2.0 stylus.
Is there a change in stylus size over time? Not a clear one, according to the next figure, which charts average stylus size per year (bubble size represents number of sides). The discs range from 1935 to 1963; stylus size remains fairly even until a precipitous drop in the late 1950s with the advent of microgroove. This may make sense from a practical standpoint: a radio station would be most concerned with compatibility across its machines and with other studios, so changing recording characteristics may be counterproductive until a new, well-established standard arrives —in this case, microgroove. But of course, this applies only to our collection, at least for now; more work needs to be done.
Just for fun, and as a contrasting comparison, take a look at the next chart, which uses data collected (and generously shared) by the University of California Santa Barbara as part of its collaborative project with the Library of Congress, the National Jukebox Project. The subset of data in this chart includes more than 3200 sides of commercial shellac ("78") Columbia discs covering the "acoustic era" of disc recording, from 1900 to 1925. In this case, you can see a clear downward trend, which corresponds to what most of the literature suggests for this kind of discs. However, it is interesting to see how gradual the trend is, and how big the difference is in just twenty years' time.
It is always dangerous to extrapolate statistics too far from their source, but think how useful such data could be to technicians and collection managers. When starting an in-house reformatting lab, this kind of information could help them decide on what sizes of expensive stylus may cover most of their disc holdings. If they know the type of discs and the years of manufacture, they could stock up on certain stylus sizes to safeguard against possible future shortages. And moreover, publishing this data could contribute to a slim body of evidence-based knowledge in audio archiving, at a time when direct training is becoming more difficult in today's isolated archives.
Here is a practical example. Shortly after running these numbers, I came across a glass-based lacquer disc which was badly peeling —I knew that I would only get one or two passes. Based on the literature, I may have chosen a 2.0 mil stylus, but our data seemed to clearly point that, given one choice, it should be a 2.8 mil. The disc was challenging but the audio ended up sounding satisfactory. Perhaps a 2.0 mil stylus would have also sounded fine, but I somehow felt like I was standing on more solid ground when I made the choice.
There are potentially many other applications for this type of study. Analyzing playback equalization curves, correlations between external symptoms and audio artifacts, elliptical vs. conical styli, stylus size vs. groove size... much of this data is already kept in institutions and vendor projects, and only needs simple analysis. Hopefully, this kind of reporting will be more common in reformatting projects, and even requested by grant-giving institutions.
Numbers cannot replace the wisdom —or ears— of the elders. The aforementioned AES draft document states that "in most cases, the judgement will be made aurally by the engineer who makes the transfer." Although human ears should almost always be the ultimate arbiter, there is no reason not to help that decision making with some historical perspective, backed by facts. The data is there, waiting to be released and be useful.
Thanks to John Passmore, Nathan Coy and UCSB's David Seubert for their assistance with this article.