I’ve long been a fan of William Basinski’s works with tape loops and have admired his techniques so I set out to understand the nature of his work with analog gear and to try and replicate it in my DAW with plugins.
Original Recording Equipment
William Basinski used reel-to-reel tape decks when making his disintegration loops. Specifically, they were likely originally recorded on Norelco Continental reel-to-reel recorders and then from a Revox G36 onto a CD.
In the notes for his album “The Clocktower At The Beach (1979),” he mentions “Having the Norelco Continental reel-to-reel decks was instrumental in so much of my early work. With four speeds, these stalwart workhorses helped me to discover hidden, unheard sounds. My magic machines. This piece was one of my earliest drone pieces. Each of my pieces is a singular experiment that starts with some kind of spark, but with ‘Clocktower’ I was most encouraged to finally pursue long-form works that create space and distort time.”
This statement says that there were four speeds so this could be the Norelco Continental 401 Model 3534 reel-to-reel tape deck which features four playback and recording speeds: 15/16 inches per second (I.P.S.), 1 7/8 I.P.S., 3 3/4 I.P.S., and 7 1/2 I.P.S.
Also in an interview with SOS Magazine, he mentions the Norelco Continental: “I got some junk tape decks, these big old Philips or Norelco Continental 40lb portables. You could slow the Norelco way down — seven‑and‑a‑half IPS, three and three‑quarters and then one and seven‑eighths. I just started experimenting with changing speeds and seeing what kind of frequencies I could pull out of these drones I recorded, from the TVs to the refrigerator to everything.”
The fact that he mentions they are 40 lbs., indicates these must have been the Norelco Continental 401 Model 3534 reel-to-reel tape decks. Also popular during the time was the Norelco Continental 101 Model EL3586 which had only the one speed of 1 7/8 I.P.S.
You can also see an Ampex 456 in this image from the SOS article. This is a studio reel-to-reel recorder and Ampex 456 tape is famously associated with “sticky shed syndrome,” a degradation issue where the tape’s binder breaks down over time, especially in humid conditions, causing the oxide layer to flake off the polyester base.
Indeed this is the same type of issue associated with the tape used on the Norelco Continental machines and which caused the tape degradation found on Basinski’s Degradation Loops.
Also from the SOS article, it mentions he was a fan of Brian Eno and Robert Fripp’s “Frippertronics” setup and that he built his own version of that: “I started getting some newer quarter‑inch tape decks. I had the two flatbeds to loop with, mixing into one of the Sonys or something. I did a lot of that, and then short‑wave radio experiments with loops. I wanted a Mellotron, so I was trying to make one with these little string snippets. Slowing them down and stuff, you would find this huge well of melancholy. And it was like, Oh, OK, here we go.”
In a Frippertronics type of setup, there are two reel-to-reel machines side-by-side with one recording and the other playing. As the tape goes from one to another there is a delay determined by the length of tape between the two machines, and the original tape is slowly overwritten causing the sound to fade slowly over time. The sounds repeat, overlap, and slowly fade over time.
He was also recording shortwave radio broadcasts around the time before the disintegration loops were made and he also mentioned recording the piano performance of a friend, but the loops used for the disintegration loops were described by him as “the tapes were from the ’70s, and the loops were from the ’80s.”
The Degradation Mechanism
- He transferred these tape loops that were physically from the 70’s but had sounds from the 80’s to a CD using a Revox G36.
- These loops he transferred were originally recorded on Norelco Continental tape machines.
In the SOS interview, he said about this time in 2001:
“At the time, Basinski’s old tape loops had almost become decorative, having been left hanging for some years on a dead tree branch he’d found in a parking lot. “I would group the ones I wanted together with little safety pins on another side.” He picked off six of these loops, each roughly between 12 and 16 inches in length, suspended them from a boom stand, and then randomly selected one. “That was what became ‘dlp 1.1’,” he says. “I was like, Oh my God. This is so grave. This is such an amazing melody.” All of the tapes contained chance snatches of audio from a CBS radio station broadcasting from the top of the Empire State Building, slowed to three and three quarter IPS. At this point, Basinski’s main tape recorder was a Revox G36 that he’d picked up in a nearby house clearance in Brooklyn in the late ’90s. “It was this townhouse in the Hasidic section of Brooklyn about a mile from my house. And the man was a very interesting guy. He had worked for Bell Telephone since, like, the ’30s. He had all of these old headsets and microphones. I got a portable cutting lathe and this Revox, which looked kind of ratty.” With the Revox, Basinski’s technique involved using D‑sized batteries to spool the tapes around and maintain tension. “They’re heavy,” he says, “and you’ve gotta keep the tapes tight, or they’ll get chewed up in the machine.”
So the actual audio was from a CBS radio station broadcasting from the Empire State Building and slowed to 3 3/4 I.P.S. The station was likely WCBS or WNEW which both broadcast from there during the 80’s. He said they were slowed to 3 3/4 I.P.S. and if they were recorded with the Norelco Continental 401 originally, the highest speed would have been 7 1/2 I.P.S. so slowing to 3 3/4 I.P.S. would mean a reduction by half, so these recordings were at half speed of normal. Also you can see in the image of his Revox recorder that the loop is wrapped around D batteries to keep tension and allow it to loop.
- Audio Source: 1980’s CBS radio station WCBS or WNEW from Empire State Building.
- Speed: Half-speed (half of normal speed).
- Length: Physically 12 to 16 inches in length, so with an original speed of 7 1/2 I.P.S. these would have been 1.6 to 2.13 seconds long, but slowed by half these are 3.2 to 4.26 seconds long as played (in loops) on the disintegration loops album. DLP 1.1 is a 12 inch piece of tape so that loop would be 3.2 seconds long.
- Tape Type: Likely type I or II (Chromium) or specific for the recorder. The Norelco Continental 401 could accept type I or II tapes and special formulations also.
- Tape Speed: 7 1/2 I.P.S. as recorded originally on the Norelco Continental, and during the degradation the speed on the Revox G36 was 3 3/4 I.P.S.
Basinski further explains, “Listening to the woozy sample repeating in the first loop, he decided to create a counter‑melody to it using the in‑built arpeggiator in his Octave Plateau Voyetra 8 synthesizer, leaving it to play live and unsynchronised along with the main theme.
“I got a cool, randomly arpeggiating sort of French horn sound dialled in,” he recalls, “and it was going really nice. I didn’t know what was going to happen, so I started recording. And then, I went to make a coffee, and came back and started realising, something’s changing, what’s going on? And then I started to see what was going on and hear what was going on.”
- Additional Audio: French horn sounding arpeggio repeating from an Octave Plateau Voyetra 8 synth was used on DLP 1.1 and DLP 2.1 but not on the others. This was a counter melody playing asynchronously along side of the loop.
- Recording of the Counter Melody: The french horn of the Voyetra was being recorded onto the tape going through the Revox G36 because you can hear it degrade in the later stages.
Basinski looked at the moving tape and could see that it was shedding the iron‑oxide coating from its plastic backing. “I thought, Check the levels… check the levels. Gotta make sure, y’know. Let’s just see what happens, because this is a one‑time thing.”
- Levels: Compensation was applied (manually) by Basinski to maintain the volume level as the tape degraded.
Adding to the haunting effect was the outboard Lexicon reverb that Basinski had used to treat the sounds. As the notes cracked and slowly disappeared, the Lexicon eerily filled in the gaps. “You can hear it more because the dropouts seemed to happen at the end, where the decay on the notes was. The strongest notes seem to hold on the longest. Kind of like people’s memories and in dementia and things like that. So fascinating.”
- Reverb: An outboard Lexicon reverb was used to treat the samples before going into the CD recorder. Although the exact Lexicon version is not mentioned, it could be a Lexicon PCM60 which came out in 1984 and was widely used.
After recording the first loop for just over an hour, Basinski moved on to the second one, creating ‘dlp 2.1’ and so on. Over two days he produced the six pieces that were to make up the four albums of The Disintegration Loops. For ‘dlp 2.1’, Basinski had similarly added a sequence on the Voyetra 8, although he chose not to add any overdubs on the later loops. “The second one started to flake out faster,” he says, “and I realised, Listen, don’t do counter melodies here. “Some of them were so perfect and eternal that I knew they didn’t need anything. After that, it was just the loops by themselves. I realised, Stay out of the way. This is not your business. You just do your work as an engineer here.”
Understanding the Mechanism of Tape Degradation
Dividing the different aspects of tape degradation, I came up with these basic types:
- Hiss
- Harshness
- Motor Noise
- Crosstalk
- Wow/Flutter
- Dropouts
- Saturation
As described by William Basinski in the interview with SOS, “Basinski looked at the moving tape and could see that it was shedding the iron‑oxide coating from its plastic backing.”
This is typical “sticky shed syndrome” of tape which is a breakdown of the binder that holds the magnetic particles to the tape, causing dropouts. So we don’t really need to know what kind of tape was used but that it was from the 1970’s and suffered this stick shed degradation as described. This is what caused the change over time of the recordings of the disintegration loops and gave that deteriorating character to the recordings during the transcription process performed by Basinski.
Given that the main mechanism is sticky shed syndrome, when degradation of this nature occurs, it is likely that the parameters described have distinct changes over the same time period as the disintegration loops were transcribed.
Listening to the DLP 1.1 recording, I was able to describe each of the parameters during the process and create the following graph:
Further categorizing the nature of each parameter, I constructed a table of properties that would change over time to match what was actually heard on DLP 1.1:
- Hiss – linear increase over time
- Harshness – subtle then increasing with a S-curve in the latter portion followed by steep increase.
- Motor Noise – steady or slight rise over time
- Crosstalk – linear increase over time with sporadic increases during latter stages
- Wow/Flutter – linear increase over time with sporadic increases during latter stages
- Dropouts – subtle then increasing with a S-curve in the latter portion followed by steep increase.
- Saturation – stable or decreasing slightly over time
These follow the physical properties of degradation as described in various sources so they make sense from both a listening perspective and a scientific one.
Translating to a DAW
In the DAW, the routing of the loops and effects were made to mimic the recordings of the disintegration loops and processes.
The chain for the disintegrating loop looks like this:
Sample (half speed) –> Level adjust (VUMTdeluxe) –>Tape plugin (SketchCassette II) –> Reverb plugin (Bettermaker BM60) –> EQ (Claro) –> Level adjust (VUMTdeluxe).
To add the counter melody using a Voyetra 8, I did not run this through the tape degradation as it would not have been when Basinski recorded it. He mentioned that it ran along side of the tape and both were presumably then recorded to CD.
Sample
For the sample, I used an orchestral piece and slowed it by half using Serato Sample. I then repeated this in a loop over and over for the track.
Tape
The tape effect is easily the most important piece of this, so I tried a few different ones before settling on the RC-20 Retro Color plugin by XLN Audio or the Sketch Cassette II plugin by AberrentDSP.
I initially tried Reels by Audiothing, but could not get the dropouts to become severe enough to replicate DLP 1.1. Sketch Cassette II and RC-20 both seemed to have better modeling of the late-stage dropouts.
I’ll use one or the other but not both, and here are the settings I used:
Sketch Cassette
In Sketch Cassette, I chose a type II standard tape and started out at the “used” age. Wow, flutter, and age I increased using automation (see red arrows in the image below) over the length of my track. Most important was the dropouts which I increased in depth and intensity over the length of the track. While I did not increase the hiss, it could be an item that adds complexity to the recording and may be something I want to automate as well.
RC-20
In XLN Audio RC-20, I chose to automate certain items (see red arrows in the image below).
The Revox G36 was one of the last decks to use all tube amplification, so I chose the distortion in RC-20 to be from the tube pair. Wow is typically between 0.5 and 1 Hz, while flutter is from 10 to 20 Hz. In the wobble section, these settings were 100% wow and 11% flutter with wow frequency of 0.5 Hz and flutter at 10 Hz.
On the chromium tape that was likely used on the Norelco Continental 401, the roll off of the highs would be around 3.2 kHz because although the 401 rolls off around 8 kHz, the chromium tape on the 401 would be around 3.2 kHz because the 401 could not handle the higher bias of the chromium tape and would therefore roll off faster. On the low end, it would not be so prevalent and would be rolling off around 80 Hz. These are the points where I set the EQ on the RC-20.
Width I turned down a bit to simuate the crosstalk in the tape. 7.5 tape noise seemed to fit so I chose that and had it duck and follow appropriately. I played around with this later and adjusted further.
The magnetic section is for simulating the tape so during the track I turned up the dropouts and at full, it really seemed to simulate exactly what I could hear in Basinski’s DLP 1.1. Wear was at 100% while flutter was at 11% and the flutter frequency was 20 Hz in this section.
During the track I would increase the wobble slightly and the magnetic section more, while heavily increasing the dropouts toward the later stages of the track to simulate the degradation, and this seemed to work really well. I could reduce the width a bit to increase the sense of crosstalk, but it may not be necessary.
I favor the RC-20 because
- There are fewer controls to automate to get the disintegrating tape sound.
- The dropouts seem more aligned with the DLP 1.1 track of Basinski’s.
- There is more control over the parameters.
- Sketch Cassette II is made for cassette simulation and this is reel-to-reel and RC-20 is able to simulate the reel-to-reel tape better in my opinion.
Reverb
The BM60 reverb plugin by Bettermaker was chosen as it replicates the Lexicon PCM60. To tune it to be exactly like the disintegration loops, first I sampled the DLP 1.1 recording and copied the reverb using Chameleon. Then I compared it to settings of the BM60 plugin until I got it about right (almost exactly actually).
Counter Melody
To simulate the counter melody used on DLP 1.1 and 12.1, I used the UVI PX V8 that is a simulation of the Voyetra that Basinski used. He said he had a french horn sound dialed in and that is program 24 on the Voyetra and on the PX V8 plugin. Using this, I can get almost the exact same sound.
Routing-wise, this has to go through the tape degradation as it sounds like it was experiencing dropouts along with the tape in the DLP 1.1 track.
The main notes are A and A# repeated from what I can hear and what I could see in Notegrabber.
Although the sound of the french horn is not exact, it is close enough for me in this exercise because the goal here is not to duplicate the track but to understand how it was made.
Further Analysis of DLP 1.1
Here is a plot of the amplitude vs. time (blue) and the spectrogram (orange). Looking at the amplitude, you can see it declines until near the end where it increases and this is the sound of the tape dropouts and re-entries. So it is not amplitude-compensated throughout but is turned up a bit more at the end.
Zooming in, you can see the individual loops that are about 6 or 7 seconds long. This is the length of the loop with the Voyetra playing on it, but the underlying loop is about half that which is about right.
Final Outcome
Choosing a loop of my own, I constructed a project that would have automation for the tape parameters described above and that gradually degraded the sound. Rather than go for an hour in length, I chose a shorter version so you would be able to see it and hear it in a shorter time to get the gist of how I did it and what it sounds like.
Towards the end of the track, I cut up the loop into some smaller pieces to simulate it breaking down into nothingness.
Intent and Direction
So the intent of this all was not to make carbon copies of William Basinski’s great works, but rather to give insight into how to simulate such a thing so that other, different things can be done with this.
I can envision using very different loop material, different loop lengths, different amounts of tape speed (slow or fast), not degrading it at all, looping in rounds of different materials, re-integrating loops (it’s been done before though), and an endless amount of other ideas.
All I ask here is to please take this at face value and don’t use it to make exact duplicates of other’s work, but rather seek to understand how to do these things and then do your own inventive art.