A little sample explanation

[Theo Verelst]

Without fun discourses about studio signal path design, acoustic pre-treatment of sounds instrument tuning, and more interesting subjects worthy of discussion a little explanation on terms and a wide and inaccurate description of a main subject concerning playing back samples.

 

The difficult signal path issues that make for interesting digital synthesis include two major subject I'll not really go in to, namely the idea that a set of audio samples on a row can have a certain knowledge associated with them. So the "sample" of lets say 44,100 stereo piano note 16 bits "samples" could be made such that you know about the signal that was sampled from a real piano didn't contain any frequencies above 22050 Hertz. You could also have knowledge about the signal in the piano sample, interpreted as the collection of subsequent, equidistant short aperture samples taken from the continuous electronic sound signal, such that you'd know you've put a certain compression effect on the piano mike signal, and some eq. Or maybe you know there's a piano enclosure ringing effect in the shortly called sound sample (again interpreted as say a file with signal values on a row, corresponding with the progression of the time variable in linear sense) corresponding to a certain wood resonance of a certain size.

 

The other thing besides "signal knowledge" narrowing down the interpretation of a certain sound bite is that there can be intended signal components built in popularly put "the piano sample" which can be either filtered out by certain digital signal processing, or brought to the front by the same. It can be that the processing to deal with intended sound elements is in normal language "linear" according to say standard college physics interpretation, but non-linear, non-stationary, infinite order or non proper initial state filter are also possible to mess with sequences of digital samples passing through a piece of DSP software acting in congruence with an electronic (time and amplitude discrete) signal filter, like an equalizer, volume control, etc etc.

 

Here, I only want to touch upon the main meaning of samples as digital recordings, in a sampler, effect, PC audio program or otherwise. A sample in it's briefest interpretation would be single, usually digital, digtized value of an electronic signal, nowadays usually in at least 16 bits accuracy (about 65534 possible amplitude discrete steps put in a signed digital variable). Of course, like ona CD usually, one "sample" of for instance 24 bits follows the previous, and this process of sampling in the Analog to Digital Converter goes on for some time, usually resulting in a file on a computer of a certain size, containing some number of seconds at least of the "sampled" signal, which originally was electronic.

 

Now, the stream of signal samples coming out of the ADC can be called a sampled signal, just like a file where you store them can be called "a sample". Only slightly confusing!

 

The main thing about this whole "virtual" signal path and "quasi" tape recording and virtual tape is that playing back the samples (in the sense of a little recording) isn't as perfect as musicians would like it to be. In spite of Digital to Analog Converters being specified as highly accurate in certain electronic test and measurement equipment dimensions, all available DACs, be it on your computer mother board, your phone or a specific box you use in the studio to play back digital signals such that you can put them on your mixer and amplifier, no matter which type of conversion you use and no matter how esoteric certain specs sound, there are going to be major errors in you electronics signal when playing back any type of digital "sampled" signal source. Be it software creating virtual instrument "waves", a DAT recording playing back through some DAC chip, your studio ADC/DAC unit on you computer, or the built in DAC (always at the end of the digital signal path putting out an electronic signal at your output bus) in a digital musical instrument, the DAC does certain things pretty inaccurate for human perception.

 

Mathematically, theoretically speaking that's because the way to create an analogue, "normal" electronic signal from a digital "sampled" signal is more involved than a DAC chip will ever do. And moreover, theoretically a perfect reconstruction of the electronic signal that the ADC had recorded coming from the output of the DAC would require "infinite" time.. Normally, nothing much can be done about the distortion that inevitably comes from the very imperfect DAC "reconstruction filter" (sometimes called oversampling, anti-aliasing and more named technical aspects of the practical chips). When consider a certain sub-class of digital signals, and doing a lot of work, it is possible to significantly improve the result of the analog signal coming out of the DAC, but it's hard and seldom is done.

 

TV