a Buzz audio FAQ

[mute]

Buzz Audio FAQ v 0.2 by Mute

Answers by Cyanphase, Mute, Usr, Btdsys, LeeDC, Fuzzpilz, Oskari, and others

----------------------------------------------------------------------------------

Q: What is the internal processing quality of buzz?

A: Internal calculations in Buzz are done with 32-bit precision floating point numbers.
Although the calculations are done in 32-bit, the bit depth is dependant on your soundcard.
The samplerate can go all the way up to 96KHz or whatever your soundcard supports.

24 bit recording and playback is on many modern soundcards, such as the Creative Audigy 2,
M-Audio cards, Echo cards, (basically all ASIO cards available), etc.

Here are some examples of 16 bit soundcards:

* Creative Labs SB16, 32, AWE32/64, Ensoniq/AudioPCI 64, SB128, Live!, Audigy 1 (although the specs say different)
* (if not) all generic multimedia soundcards that you find pre-installed on your computer

Q: Tell me more about Buzz's 32-bit processing, what about these newer applications that say they are 64-bit, etc?

Single precision (32 bit) floating point numbers are less precise than double precision (64 bit), of course,
but I very much doubt that this causes anywhere near audible problems in realistic circumstances.

32bit is a floating point system, so the values can actually be greater than the theoratical maximum of 0db,
and also the values can represent decimal values and stuff so its meant to be more accurate and suchlike.
32bit PCM is the same kind of integer system as 16bit, but its much more precise.

The point of exporting at 24/32bits is that you can do further editing such as mastering your tracks in the
best quality possible, then only dither down to 16bits as the very final stage, this way you retain a more
accurate representation of the original signal, and dithering down can increase the theoretical noisefloor
of a 16bit signal etc., but u need a >16 bits signal to start with etc.... and stuff like that. (morex)

Q: Does X's "sound engine" sound better than Y's "sound engine"?

A: This is more or less audiosoftware's equal to an urban myth. But we'll explain it.

A "sound engine" is only meant to haul blocks of floats around between plugins. No processing except for the
occasional gain or pan (which is the same operation). Rhe only kind of processing that a _correct_ sound engine
is allowed to do is dithering in the last stage, and differences there should be _very_ unhearable, even more
so if you use a 24 bit output device.

If any host has its own sound, then that host is *broken*. The host should not do *anything* except shovel data
around between plugins, and some basic gain and pan stuff. No chorus or reverb or any other audio "sweetening"
should be done behind the users back in a professional audio host.

HOWEVER, some truth may enter this "sound engine" question when you are comparing applications like Sonar,
Cubase, etc. that all contain built-in mixers (which Buzz doesnt, due to it's Modular nature). In applications
with these types of mixers the sound is subject to the quality and "style" that the developers put into the EQ's
on the mixer channels. In Buzz though, there isn't any equalization being done unless you've added one to your
audio paths. In which case the EQ could range from a Voxengo GlissEQ or Waves Q10 to a lowly and basic Automaton EQ-7.

Q: What about the quality of Samplers (or trackers) in Buzz?

A: Samplers can use very different algorithms when it comes to pitching up or down. If you don't know what aliasing
is, then you simply have to believe that there are differences. An "audio mixing engine" (Buzz) does _not_ do
upshifting or downshifting, so this difference does not apply here.

samplers also tend to employ additional filters (beyond those used to cope with aliasing), there you have another
thing where they can sound different.

Q: What sample formats can buzz's wavetable load?

A: This deserves a 2 part answer:

* Buzz's Wavetable supports 16bit and 8bit PCM .WAV files at any sampling rate.
It also Loads XI (Fast Tracker Instrument) files, with multi-samples & loop points.

* Overloader "Componented Wavetable System" adds support for 32bit .WAV files (and possibly other formats).
These 32bit files can only be used by Buzz Generators (samplers and trackers) that support 32bit through Overloader.
At this time, the known plugins that support this are Fuzzpilz UnwieldyTracker and BTDSys Samplegrid 2,
which both support 32bit .wav files.

* Further information for Developers:

* The Comp. Wavetable is plugin based (found in Buzz\Gear\WTHelpers), it allows in 8-bit wavs (8bit loader),
16bit pcm wavs (gen wt handler), 24bit packed int wavs (gen wt handler), 32-bit pcm wavs (gen wt handler),
32-bit float wavs (gen wt handler), there is also a lpac loader which i never finished. If a developer wants
to add Componented Wavetable support for additional formats (mp3, flac, whatever) to be used in a machine
they are creating, they can do so by also writing an additional WTHelper plugin. You can find documentation for
this in Buzz\Help\Overloader\devmanuals\WTHelper.html, source code examples can be found in Buzz\Dev\
Plugins can also be written that will convert a format to another (such as Cyan's experimental LPAC loader),
so you could write a wthelper plugin to import a .ram file or whatever and convert it to
a 16bit 44.1 in the wavetable (etc).

* The bit depths themselves are only partially part of overloader, there is a wave entry flag that gets stored
in the bmx (that is used with oskari's permission) to flag a sample as being "extended", which means the sample
starts with 8 or so short ints (to keep it aligned to sound less harsh to trackers that dont know the new format),
with the first short int (if i remember) that specifies the format as being either 16bit,
24bit or 32-bit.

Q: Can I load other formats outside the wavetable?

A: Of course, many Buzz plug-ins and VSTS allow you to load many other formats (such as .mp3 or .sf2 (soundfont).
There are also harddisk plugins that allow you to load or "stream" files from your harddrive, bypassing the wavetable.
Such as Fuzzpilz INP and the rDev Continuous WaveReader.

Q: I was working with buzz several years before, but I stoped it because I found the sound somehow metalic and not so
full as I want.

A: when you use synths in buzz, they dont have any effects built in like you expect in other plugin formats such as VST.
Thats why they sound so dry and "metallic". Nearly all VSTs sound great when you pick a preset because they have chorus,
reverb, delay, etc. built into them. If you turn those effects in the VST off you'll find it sounds pretty thin too.
The idea in buzz is that you can just add these effects yourself, just connect your synth to some effects machines.

In many cases, Buzz synths (Kyrie Spectra, ld0d Jacinth, FSM Infector, etc.) are far more powerful and feature driven
than their VST counter parts (specially as far as synthesis goes). That is because we are geeks.

Q: What formats can Buzz export to?

A: Buzz's default recorder does 16bit 44.1Khz .wav files. You can record in many other formats and in many other methods
using Cyanphase's Overloader or anynumber of Buzz or VST plugins. Multitrack, 32bit, 96Khz .WAVs, .OGG, you name it.

Q: Why would I want to record in any other format than 16bit 44.1Khz? Isn't that the limit of an Audio-CD anyways?

(From LeeDC):

Just to explain to non-technical people, there is a reason to use more than 16 bits when you have to reduce to 16 anyway (for CD).
Each bit represents 6 dB of signal above the noise floor. If you normalize or raise the volume of a file, you aren't actually
adding intelligent data to it, just raising the level of the data that is there. That means that if you have a 16 bit wav
file at -12 dB the actual audio data is only represented by 14 bits. The others are simply 0. If you normalize this signal,
you aren't increasing the signal to noise ratio, just making the signal louder. So you still are only 14 bits (84 dB) above the noise floor.

Try this experiment. Create a sequence in buzz. Now turn the master volume down very low to -60 dB or more. record this.
Now turn the volume all the way up. Record this. Put the low level file into cool edit or some such program and normalize.
Now compare the files. The lower level file should be noisier and grainier than the other.

It's important to realize that when you do compression or limiting or eq'ing, lowering and then raising the volume is what
you're doing. If you have 24 or 32 bits, it's not going to matter if you lose a few of them during processing. On the other
hand, too much processing on 16 bit data can end in crudville.

EOF

-------------------------------

this can also be found @ http://www.plosive.net/buzz/txt/Buzz_Audio_FAQ.txt
_________________
http://www.lazytrap.com/

[mute]

someone had to do it, and i was bored. dont worry, it didnt take that much time -- alot of this resulted from buzzmachines msgboard .. so if u feel like you were quoted, you probably were.

have anything to add or fix?

ill edit (and spellcheck it... etc..) and format this eventually for buzzwiki, along with the midi faq i was working on before
_________________
http://www.lazytrap.com/

[JoaCHIP]

Internal quality

[nool]

very nice work mute
_________________
http://www.noolmusic.com
ReWorked the Entire Site - Spin Laden !

[white butttaaahh]

nice one mute

[fun-lovingjack]

This has to be the best thread ever.