CEO of SPL Hermann Gier, who you can watch in the video above, has kindly shared the story of the brand new SPL Machine Head tape saturation plugin with us. Read on to find out how Machine Head began and how SPL and SoundArt surpassed the finest digital tape saturator ever built.
The original Machine Head Tape Saturator hardware
The creative nucleus of the project was the Jam Productions recording studio, which Kai Lukas has been running as a sound engineer since the 1990s. The sonic advantages of the 24-track 2-inch Lyrec TR-533 tape machine awakened in Kai the vision and desire to make this sound available in the digital world.
At that time there was nothing comparable, so Kai Lukas teamed up with Frank Hartmann and they started from scratch. First of all, intensive measurements of the Lyrec were carried out with many trials and even more listening tests until a model of the algorithm finally emerged.
The development was carried out by Kai, a graduate engineer in communications engineering specializing in digital audio signal processing, and Frank, a graduate engineer in communications engineering specializing in digital audio and video signal processing and chip design. The two joined forces and founded the engineering office SoundArt.
The Holy tape saturation code
The same algorithm in the brand-new Machine Head plugin was developed in the mid-90s as a plug-in under C++ with floating point calculation. It ran on Motorola 56002 fixed-point DSPs the original Machine Head.
Careful scaling of the algorithm for the fixed-point implementation was necessary to prevent internal clipping. At that time, the DSP assembler code was developed by comparing the processing result of the hardware with the processing result of the floating-point reference code, using music as a test signal with properties that ensured the occurrence of all operating states of the algorithm.
The correspondence between the two processing results was the quality criterion for the DSP implementation. The plug-in based on the floating point reference implementation from the mid-90s now had to be made to run again after 28 years on a Windows 2000 workstation with a Pentium Pro processor. CD backups of the source code directories were used to run the reference source code in a current development environment. It was possible to achieve a bit-identical audio result compared to the reference implementation on the Windows 2000 workstation.
A tape saturator plugin for the modern age
Brining Machine Head into modern age required adapting its original code to the C++ functions available today, whereby the results of the audio processing were also bit-perfect compared to the reference version at the time. The Machine Head hardware only supported sampling rates of 44.1 or 48 kHz.
In today’s production environment, support for higher sampling rates is required. Therefore, adjustments had to be made for higher sampling rates without changing the sound of the original filters with all their characteristics. For this purpose, SoundArt 2024 developed a new filter approximation system.
Thanks to the S/P-DIF standard, it was possible to carry out comparative measurements between the hardware machine head, the floating point reference code and the new plug-in
The ultimate version of Machine Head
When the original algorithm was implemented in 1996/1997, there were limitations in terms of computing power. This has changed. SoundArt has therefore decided to use its current
knowledge and the improved code components to create an ultimate version of the Machine Head. With today’s computing power, the filter banks could be improved and the positive characteristics of the tape machine could be imitated more precisely. This has further refined the sound.
The new function in the “Ultimate” version is the Low Frequency Adjust. This parameter is used when calibrating tape machines to equalize the frequency response by compensating for the increase in low frequencies caused by head mirror resonances depending on the selected tape speed.
The “Ultimate” version allows independent adjustment of the low-frequency response from an ideally calibrated compensation with Low Frequency Adjust to 0.0 through to over- or under-compensation, which may well be justified from a sonic or artistic point of view.
While the “Original” version only offers integer steps for setting the High Frequency Adjust from -6dB to 6dB, the “Ultimate” version has a 10-fold finer parameter resolution. In addition, the resolution of the drive parameter has been doubled compared to the “original” version.The optimizations to the algorithm in the “Ultimate” version improve the sonic result in terms of openness, clarity and spatiality.