Every part is generated in code from dimensions and profiles: no model files. Module notes carry their sources at the foot.
What you hear is the machine’s own chain. The record and replay equalisation, the bias, the tape’s saturation and hiss, the replay gap’s loss and the head bump all act on the track. So do wow and flutter, noise reduction and the generations, and a limiter catches what is left.
Scale and drawing
Tape widths are exactly proportional, 0.15 to 0.25 to 1 to 2 inch, but all drawn far larger than life against the deck. Real tape is about 0.05 mm thick and the ribbon here is drawn roughly two centimetres thick, so it reads.
The recorded flux pattern is a visualisation, not a real waveform. It is drawn on both faces near the head run so it reads from the front, even though the oxide faces the heads.
The drawn pattern's period is the recorded wavelength magnified two hundred times. At 15 ips a 10 kHz tone records 38.1 micrometres, which draws as 7.62 mm, so the pattern crowds towards the gap as the tape slows.
The record and replay gaps, the coil turns and the oxide layer are drawn thicker than life so the section reads. The cassette's gap is drawn as one line across its lane stack rather than four.
Models, not measurements
Every number here is a model, not a measurement of a particular machine. The site’s own sentence is inherited: learn the direction and size of an effect, never the digits. The generation figures, 3 dB at the first bounce, 7.8 dB at the fifth and 0.97 dB off 10 kHz per pass, are that model unchanged.
There is no bias frequency anywhere on this page. The specification names none and nine years of papers have never asked for one, so the bias is drawn and described and never numbered.
Wow and flutter depths are illustrative and exaggerated, so the wobble is audible within eight bars. The vari-speed reaches an octave either way so the speed trick can be heard, which is further than a studio machine’s trim.
The delay figure uses a nominal two-inch record to replay spacing, chosen so 15 ips gives the 133 ms the site’s chapter quotes. The teaching point is the relationship, not the constant.
The bias positions are a three-way model of a continuous adjustment, and the hysteresis is drawn schematically.
Meter ballistics and reel-pack capacity are illustrative.
Scope
No named machines. The specification names no tape machine models, no consoles, no microphones and no processors. This is a generic instrument of its era.
The head-block physics is scaffolding for the examined consequences. Bias, azimuth, sel-sync and the head names have never been asked for in nine years of papers, while hiss, saturation, restricted high frequencies, bouncing and vari-speed have been asked for every year.
“Head gap” is used two ways in the literature and on this site: the magnetic gap in a head, and the spacing between the record and replay heads. This page says record gap, replay gap and head spacing.
The section caps are on the head and tape parts only, which is why a cut shell shows no cap face.