[BUG] Modelling inconsistencies

[ampetrosillo]

The tape modelling seems a bit off on some things. It seems as if the hysteresis and distortion have been well modeled, whereas the spectral effects and some other things do not work as expected. Analysis has been conducted in DDMF Plugin Doctor and in Cockos Reaper with the use of a sine sweep generator and Voxengo SPAN.

Bias: although it is hard to understand how strong a bias signal is being applied (the scale in ChowTape is from 0 to 1), bias settings over 0.50 (which I would assume to be 0dB) lead to a progressive decrease of distortion and increase bandwidth, whereas real tape (and tape machines) actually show a substantial reduction in the treble response for increasing levels of bias (and an increase when underbiasing). Also, distortion goes down up to a certain point, but then skyrockets over a certain value of overbiasing. For instance, the RTM SM900 tape specifications show decreasing THD up to +3dB bias (actually this changes according to tape speed) and then a sharp increase for bias signals stronger than that.

Sources:
http://www.jrfmagnetics.com/Nortronics_pro/nortronics_silver/pdfs/Nortronics_Silvr_pp1-9.pdf
https://sepeaaudio.com/content/RTM_SM900_Premium_High_Output_Studio_Audio_Recording_Reel_Tape.pdf

Tape thickness, gap width, tape speed:

These parameters do not appear to be modelled appropriately. Tape thickness shows massive treble loss for normal oxide thicknesses (eg. 17.5 microns). Here is a plot of Tape Chow set at 30ips, 3-micron gap width, 17.5-micron tape thickness, everything else at default, with a 20-second sine sweep at -24dB (presumably low enough not to incur excessive linear analysis artifacts due to distortion) and Voxengo SPAN set to show MAX, 1/12 octave smoothing, no slope.

Gap width also appears to be modelled correctly in the treble region, but with doubled frequencies in the bass region, with head bumps appearing at twice the expected frequency. Head bumps are more pronounced at lower speeds, but even at 3.75 with common gap widths (50-U, 50 microinches = 1.75 microns) the head bump, which is expected at no higher frequencies than 50Hz, peaks at well over 100Hz.