Live Music Archive
Librivox Free Audio
Metropolitan Museum
Cleveland Museum of Art
Internet Arcade
Console Living Room
Books to Borrow
Open Library
TV News
Understanding 9/11View Post [edit]
| Poster: | NoiseCollector | Date: | Feb 1, 2008 12:24pm |
| Forum: | noisecollector | Subject: | The Art of NoiseCollecting: More Sound Sources and Techniques for Audio Sampling |
I recently mentioned mp3's and my disdain for them in most cases. Let's look at some techniques I have used over the years to to try and make them usable in a mix and then we can start looking at unusual sources. Again we are assuming you already read the other article or have a basic knowledge of navigating the web and using sound software (more on that later... ) so we can presume you know where to find the mp3's that everyone else can find already. i must reiterate that not only am I not using spell check, but I am not condoning using copyrighted media for any purpose other than allowed in fair use doctrine, get permission is possible and keep it legal.
An mp3 is a compressed sound file. Some musicians and others familiar with audio terminology may equate compression with audio compression like soft knee or noise gate effects produced by audio compressor hardware and software. In audio, compression is evening out volume levels and raising the overall volume level. In files it refers to making it smaller. Technically but in layman's terms, mp3's remove every other band of frequecies (depending on sample rate) and squashed the remaining bands into a smaller footprint, the missing data is estimated or stretched to resemble the original but we all hear the sqwishy metallic sound up to around 128kps at least. It's much more obvious to those with aged ears on a screen via grapshic visualizatios. Trying to mix a low quality mp3 in a mix of wav quality tracks is not fun, trust me.
The first thing that pops into your head is probably, well convert it to wav format then. Many audio software editing suites don't allow using mp3 tracks like my archiic seqencer so I convert to wav initially and see what I am left with. Not all softwaere is created equal and even the same software will give different results without changing any parameters. Ever convert a wav to mp3 and there's a pop in it? Ever burn a cd and there's a skip or error? Well, what do you do? Delete and redo it, right? Same for file conversion, got a cool mp3 sample and need it in a higher quality format for editing, mixing and processing but it sounds like crap when you hit render/save? Look at your settings, make sure maximum quality is selected try again. Still sounds like crap, try a lower sampling rate wav like 22k, hey it works. Try converting that, if needed, to 44.1k and it may work or just use it as 225k if possible. There is thing called aliasing caused when different sampling rates interact with each other that I won't touch on at this time. Just understand that different numbers work better in different ratios when working with sampling rates. Numbers that evenly divide into each other and factors thereof are alledgedly less affected by the swishy metallic demons from another dimension. Let's look at another way that gets you the sound without doing any conversion other than analog-digital, which considering the source isn't going to matter much as far as additional noise or distortion is concerned.
Open up your audio mixer's recording controls. Since we are looking at sound optimized for sampling, we'll use the mono mix recording source. If what you need is stereo, use the stereo one but we can make things stereo later. In most cases the content we will find on mp3's are voice and sounds that we can pan the way we want in a mix or in the ram of our keyboard however we choose. So the mono mix is selected, now unless you want some wicked feedback make sure you mute out unwanted stuff on the PLAYBACK controls as this is now your recording mixer. Keep the levels as high as possible but it's digital and clipping is not our friendly warm fuzz anymore. Mute out your CD player, built in synth, phone, mic, line, aux and whatever else you may have. On my mixer I need to turn the recording level on the mono mix down almost all the way and inch up the wav audio on the PLAYBACK mixer. For some reason the levels are way out of proportion on my setup, adjust yours accordingly. Oh yeah, did I mention we are going to record the mp3 as a wav file? The resulting wav file will be a copy but by keeping all those levels down and non essential sources muted, we can minimize additional noise and distortion and preserve the sound. Try not to EQ it live, You can always remove a frequecy with an EQ with decent results but increasing the volume of frequecies adds distortion, if there's not enough bass, cut down the mids and a little bit of treble. Don't crank the bass or you will get mud, again trust me. So converting that mp3 can be as easy as hitting record on one software button and play on the other. Snip off the silence and any other unwanted sound and hit save. You can always edit and process sounds, it's better to capture them as crisp and unprocessed as possible, you might even want to keep some nosie or tape hiss for noise reduction down the road.
Going forward we will be dealing with wav files. If you use aiff or soundfonts or any other format, simply insert your usable quality sample format in place of the word wav. So now we have the capabilty to get a usable wav from an mp3. Let's find some mp3's! Of course archive.org has tons of them, Freesound and CCMixter as well. Odeo has them, some news websites have them, lots of "books on tape", podcast and research projects are out there. College websites, historical societies, and more have googles of terabytes of mp3 files. This opens up endless possiblities for your audio toolbox. I have hundreds of CD's packed with stuff I'll never listen to... weird huh? So now that you have billions of 0's and 1's on shiny little pieces of plastic and crammed into your smoking hard drive, how do you find them when you need them? good question.
Catalogging and indexing your collection can be a daunting task if you don't start from the beginning. Let's start with the harddrive, I put all the multisamples there so my software sampler can access them in real time. I have folders in folders in folders of samples broken down by general types like: instruments, machinery, people, etc. I then break them down into subcategories like electronic, world, synth, etc. Pretty cool huh? No. I wish I had put them all in one folder and made the file names like this:
synth_moog_mini_C3
world_chinese_zither_A#4
Where the name would indicate everything those silly folders do and allow my software sampler to access the same route on every program I create. There are too many to do this and it makes downloading large numbers of files difficult but there are software programs that will run bulk processes on files with incremental naming I believe. If I had the chance to start over I would. Keeping track of where you got the sample can save you a ton of grief down the road if someone claims copyright ownership. Anyway, I burn the bulk of stuff like zip folders full of samples I might use later and mp3's of speeches and historical stuff and delete them off my drive. A simple numerical labelling system and a spreadsheet might help you keep track of what's what and where, I just write enough on the disk to hopefully jog my memory as to what all is on there. Yeah, I know...
When it comes to storage, file size may be a concern. For instance my Alesis synth can hold an 8MB flash ram card for samples. You may be thinking that's not a lot and you are right. Even some top dollar samples only hold 32MB. I supplement my Alesis's sampling capability with a sampling program called Mellosoftron which is only limited by my processor speed, mempory and drive space. So I use them to compliment each other. For playing large samples that require maximum fidelity and can be easily sequenced so that I can record them without supporting tracks playing (or else they get into the mono mix). The way my system is set up (cheap) is the standard soundcard that has one in and out cannot record me playing along with my percussion track and recording the sampler track (no render?) without the persussion track bleeding into the mix. If I can't hear it I cannot play along so it's either burn the track to cd and play it on the stereo while recording the sampler or a neat trick I figured out that may benefit you if you have a similar situation. Pan the sampler output audio hard left (again using mono samples this won't matter) and pan all other tracks hard right. Now I can play my software sampler along with my audio tracks in my sequencer and either only record the right channel or record in stereo and delete the duplicate drum track. Pan the sampler track to where it belongs and presto. The other sampling option is the Alesis. So for samples that I can use a lower sampling rate like 33k or 22k with smaller sample sizes (synths, percussion, etc) the Alesis is perfect especially if I want to do cool stereo effects and other processing in the onboard sound processor that would take forever with a mouse. Now that we mentioned lower sample rates some audiophiles are retracting and putting up defensive ear wax as we speak. Let's look at lower sample rates and how they can be used, believe it or not.
When is comes to audio fidelity lower bit rates and sample rates are universally interpreted as crappy sound. In the case of bitrate you are correct, except in the extreme exception of 1 bit recorders at a sample rate of 9billion terrawatts or whatever they run at. Let's keep it simple and concentrate on 8 bits and up. Although 8 bit sounds have a cult following and give that glitching video game feel, for the most part they suck. Stick with 16 bit. If you are into the 24 bit and 32 bit and 96k and all that, go for it. Your music will inspire insects and small rodents the world over, the humans won't hear much of a difference, trust me. So now I have these managable but perfectly capable 16 bit sound files, they must be CD quality right? Not necessarily so. CD's are recorded at 44.1 thousand cycles per second. That's a lot of digital snapshots of sound taken every second, lowering this results in less dynamic range and loss of frequencies starting at the high end. In order to produce the highest frequency we can here you need twice the sample rate as that frequency. To hear an 14k whistle you need at least 28,000 (28k) samples per second to capture the highest freqency and acuurately reproduce the sound. An exception to the rule and another source of samples is the high speed cassette. Some higher end track recorders that used cassettes back in the day tried to overcome the limitation of the tape's frequency deficiency by recording the tape at double speed. This gave the tape and analog equivalent of more samples per second. Since more tape was used to record the same amount of sound, more of the sound was captured. When played back there was less top end loss. So my buddy hands me an old 6 track hi speed multi-track cassette around the turn of the century that sat in a box until I figured out how to get the sounds out of it. I only have a regular tape player so what could I do. If I record it and then run time stretch or pitch shift processing it will sound like darth vader in the shower right? The solution was simple, record it at 16bit 22k in wav format. Convert the wav from 22k to 44.1k with good software that preserves true pitch in relation to doubling the speed. The result? The 8 minute long barry white on xanax meditation becomes a 4 minute vocal track with all frequencies in tact. The 4 minute long recording of an elephant on acid falling through a forest of power lines becomes a 2 minute guitar solo. Problem solved and pandora's box of cassette 4 track master tapes is open...
Before I get into more searching techniques I want to touch on software briefly. I am a freeware user, sonic spot, hitsquad, kvraudio, etc. Chances are if it's free and does something sonically or musically I have tried it, and if it worked I kept it. As I mentioned before I use Mellowsoftron for sampling, it is shareware not freeware so I paid for a key to unlock it. For recording I started with Voyetra Digital Orchestrator last century and have upgraded along the way due to the fact that I have hundreds of half started songs in the exclusive format and it does all I need. It does not host VST effects or instruments but has 1000 midi tracks and as many audio tracks as my pc can handle. Without the bells and whistles of modern software and plugins, this allows me to run as many as 35 tracks of CD quality audio on a $300 tower with a little bit more memory than it came with in 2004 when I got it. Not bad for $75 upgrade to Record Producer. To use VST's I use freeware like Minihost and Wavosaur, Minihost is great for playing and recording synth tracks and renders wavs easily to be imported and synced in Record Producer. I use Cool Edit 96 (yes 96) for sample editing, I also use yamaha's free editor and might even use awave to set sample loop points. I have read that wavoasaur can do this and that would be great to help limit my arsenal of various programs. Other great ones are VAZ, a virtual synth you cna play via midi and the popular and even more versatile SynthEdit which allows you to create your own synth patches and even export your own custom VST instruments and sample players (my next venture someday). I also use Granulab a freeware granular synth with some midi control capabilities and a few other subtractive and additive synth programs with various sample export and track rendering functions to create samples and synth tracks alike. A hard to find but great piece of freeware is Leon's Free Mastering software, google it. Aside from all the MIDI and other audio programs that I won't get into, I use about allthe AnalogX stuff as well for sample creation, loopsetting, tuning and other sampling tricks. Most stuff can be done bein Cool Edit 96 which can still be found online, if you know how to find stuff.
So we need to know how to find stuff. I touched on this topic previously and now that we have a wider range of target sources from which we can create usable samples let's find some. We have looked for sounds and found a lot but how to find the hidden sounds is where we will set ourselves apart from the other noise collectors by overturning that proverbial rock and finding that rare gem underneath. First let's start with subtraction, we searched for a rare instrument and keep getting a school sports team in the results, there hundreds of hits in the search but we need the sound today. Use subtraction in the googling. Again I use google... intill I exhaust all attempts then I try other engines. In google searches by putting "-" in front of a word you want excluded from the search... a minus sign in front of mp3 will remove all search results with mp3 in them. It does this without reasoning so only use specific words you estimate will not be associated with what you need. Try removing more words, this can narrow down results considerably. This is great for the web but sometimes we need a general search for websites that contain wav or mp3 or aiff files that we want but are not searchable because it is not in the search engine keywords. Maybe it's mispelled, try that. Is it in another language? If you are trying to find it in English, try it in another language (google). Great words to search that might get you a sound are:
excerpt
listen
hear
demo
These will lead you to a few places including the commmercial sites, most of which at least allow previews. You cannot download them, if only there was a way to record live sound, like a mono mix slider somewhere...hmmm. Remember you need permission to do anything outside of the fair use doctrine (google it) but many give free samples and lower samplerate (which might work!) version for trying. Great way to find vintage or vinyl sounds for that lofi break. Another return in the search results will be online stores that sell books and cd's, you may know of one yourself. There are always excerpts and snippets and clips to sample before buying. News broadcasts and blogs will come up as well and much more if you use other keywords in conjunction and maybe even some creative subtraction work and you never know.
There are also search terms to avoid:
free
sample
download
And in some cases wav and mp3
"Free" and "sample" without intense keyword addition and subtraction will give you tons of disinformation of spamic proportions as will to a lesser extent download. Searching for "wav" without clever use of quotes and other keywords will return words with those letters and other nonsense if not filtered and you can imagine how many times mp3 is splashed on a page that has nothing to do with mp3's at all, let alone contain any. I might search in quotes things like "conga.wav" or "conga.mp3" or "conga.aiff" if I were looking for something specific, you never know. Found some doozies that way but found some keepers as well.
So you wanted a conga hit and found a darn loop.... why not make your own hits from those loops? For percussion you only need as little as one hit per percussion element in the loop. Avoiding the reverb trail of preceding notes or other instrments is dificult. If you get the attack from one hit and the decay from another you can either cross fade some crazy approximation or mimic the release, sustain and decay with your sampler or software. For persussion this is not usually prodcutive but can make interesting timbers. Maybe it's a melodic piece but you can only get snippets of the notes without other instruments or chord notes polute the sound. Try these things and see what you can come up with. There is a VST plugin called sniper, I believe. It's like an EQ cuttoff machine on steroids. It can literally remove specific frequencies like a suregeon removing a piece of shrapnel from your sound. Some noise reduction software will let you remove or keep noise or sound and can be used if yours allows you to create a noise profile for removing. Select an isolated sound of what you are trying to remove if possible and use that as the template for the processing. It will sounds strange but you might be able to use the results, maybe layered with a substitute sound or simply process it with effects if possible. Find the smallest snippets and save them for looping to create custom wav forms if nothing else. I try to be audio environmentally friendly and not waste any scrap sounds. If the source is stereo your sound might be biffered nicely in the stereo field and isolated enough to use. Perhaps a phase cancelling or vocal removing karaoke plugin might do it. Don't give up, at least you have an appoximation of the sound and know how to describe it and start searching again or try to simulate or synthesize it. For instance just the attack and sustain of an instrument might be the only difference between it and a common one that you can use your sampler or synth filters to emulate. Think about it.
