A DSP test signal path

A DSP signal path test for studio processing

Experiments, DSP block schemas and this text by Ir. T. Verelst, copy righted but free for unchanged transfering of this whole page, and non-commerical applications (I.e. these signal paths may not just be turned into a commercial application). Discussions and work on these subjects are encouraged.

See also my previous page on a similar subject, with some practical technical data: http://www.theover.org/Prod . I'm working ona graphical editor for manipulating (Linux) Jack -audio connection+mix graphs (see the tcl wiki "Manipulating Jack/Ladspa Audio processing graphs from Bwise") When that is done, it will be possible to recreate and manipulate and combine signal graph (parts), until that time, I used tclscripts to start and connect the "jack-rack" programs and such. Of course you could recreate the graphs below by "starting jack-rack" and imitating the plugins and the settings, which is worth while, but getting it all right isn't easy. When/if the chain works, it's quite beautiful, and there appears to be a good match with a wide variety of recorded top grade materials, with an understandable (but not small or easy) parameter variation.

In short the signal path proposed here first of all o bring out a lot of the hidden signals of mainly the A-grade well known studio recording products like we all know from CDs and records/tapes. In my experience there's important musical information hidden in the FFT-transform of a lot of original recordings (including the studio processing).

This signal path brings out the main components I am aware of:

Averaged mid-frequency sound (in the most sensitive frequency range of the human hearing)
Medium low to low-mid frequency tube amplifier type waves (with compression)
Smoothed high frequency components (with pre-thoughts about mpeg encoding I suppose)
Various hidden low frequency and transient information in the above
Tube spoiling correction effects
provisions for making loud signals less annoying / dangerous

In many consumer audio products, deliberate and accidental damage has been done to what many people will still remember as the golden standards in record producing: in spite of the many home, hobby and so called pro studios that have emerged in the last decades, most "modern" products can't even come close to a good 50s juke-box sound, or even a crappy BBC4 satellite sound track from an original Pink Floyd record. The sizzles are wrong the basses either not there or horribly boomy, the mids scream at you for no apparent reason and the balance is probably aimed at a strange sound made for small 2-way speakers. Let alon the "atmosphere" of a record: even with decent (legal) CDs of top musical names, it isn't likely that you expensive HiHi system will actually sound right and not at all recognizable as "the latest ELO" or "country music on a good 8-track" even though a CD should be perfect. Better recent signal sources like itunes and (a favorite of mine) HDtracks can do good, but even at the top of the audio quality range (HD audio and Blurays) the digital mess is often disturbing, or someone "forgot" to set the Dolby to the right setting when digitizing.. That I'm not getting into here, though it is certainly possible (but not trivial) to correct these types of errors, too.

The whole signal path depicted below centers for a major part on the Linux "jamin" Fast Fourier Transform program (based on the FFTw3 library, running at 192 kHz, 32 bits float/channel), to provide low/low-mid, mid and very high frequency FFT filtering (clicking any of these reduced size images will show the original screendump size):