To be a successful samplista you need to be patient and have great deal of concentration. Recording all the notes for several velocity layers may sound as a daunting task, but that's nothing compared to the editing of the thousands of samples you probably will have at the end of the day.Equipment
The thing that is more important than hardware is what
you record and how
you do it.
The first sample CD I bought (an Audio CD so I had to sample and program it myself) was recorded in 1989 using a Neumann USM stereo microphone and a Sony PCM 2500 DAT tape recorder in a rented studio belonging to the Swedish Broadcasting Company. Really simple. And it still sounds great. Sure there's only two sustained dynamic levels in addition to short notes (but most hardware samplers couldn't afford to load more than one instrument created from it during the 90's anyway) but the quality of the recording goes a long way at least to my ears. If it was recorded today, almost twenty years later, the difference would mainly be the number of dynamic layers and performances recorded.
Use a simple solution as possible with minimal noise that give you good levels. For instance, a solid state memory based stereo recorder that let you record 24-bit at 44.1 kHz and transfer them back to a computer over USB might serve you better than a clunky DAW.Noise
it the keyword. As signal path that is acceptable for a regular recording might not be working well for samples. Remember that samples are just small fragments of time played back again and again making details you otherwise wouldn't pick up very audible. Always do some test runs during the R&D phase of a project and edit some small test takes and map them in a sampler to get a better feeling for it.
The two main culprits is environmental noise and signal path noise. When recording make sure you have closed all windows, shut off any air-conditioning and silenced all equipment (like computers). This is why using a proper studio is a good idea. In the signal path, make sure you use good microphones (exactly what microphone you use depend on what you will be recording) and a great microphone pre-amp. Recommended pre-amps are Millennia, Avalon, Focusrite and Mackie Onyx, with the first two considered in a league of their own. Rent the best you can get if you can't afford to buy one.
Of course, after killing as much as noise in the signal chain as possible there's always the rumbling stomach of the player that need to be quieted before recording commence.Recording
In my opinion the best way is to make continuous recordings and takes as long they are well organized and follow a predictable protocol. What I mean by continuous take is to record several notes and whole scales with silence between them as an alternative to record each note into its own take or file. The recording session will run smother and it will make it simpler to sample yourself playing an instrument. It's actually quicker to split and label the individual notes at the editing stage than during the recording.
Let's pretend we will sample an acoustic guitar that has six strings and we want to sample four dynamics (p
* There will be one continuous take per string per dynamic, making the total number of takes 6 x 4 = 24.
* Begin with recording each string at a specific dynamic: 1st string p
, 2nd string p
, and so on...
* Structure each take the same way: Play each note in the scale, start by the lowest. Repeat notes and don't proceed to the next note in the scale until you're certain you've got at least one good one. Leave silence between notes.
* Name the file for each take along the lines of '1st-p.wav' for the 1st string p
take. If you are using a recording device make a short audio note at the beginning of each take to keep track of what the file should be named once transfered to a computer.Noise removal
Sometimes it's necessary to process the audio, i.e. to remove noise. That's preferable done on entire takes before any further editing to make the effect uniform. But beware that it might change amplitude and timbre of the recording. Also keep in mind the noise by definition being noise is more successfully removed from otherwise silent parts and lesser so from the actually sample data you're interested in creating a "pumping" compressed sound.Editing
The first step is to load each take into an application like Audacity and label each note. It may sound difficult but it's rather simple if you followed the same protocol when recording each take. Play back the take, click somewhere in front of the next note and hit Ctrl+B in Audacity to set a label. There's usually enough time to type a simple label without stopping the playback, use number pairs instead of actually note names. E.g. 1-1, 1-2, 2-1, 2-2, 2-3, and so on... with the first number indicating relative pitch in the scale and the second just being a counter.
The next step is to split the long takes into individual files based on your labeling. Create a new directory, like '1st-p/', for the take and select 'Export multi' in Audacity. The labels will be used as file names.
Open each file and trim the samples. You can use the 'Z' hot key in Audacity to snap selections to zero crossings. It's important to not cut the samples off to early. Often you can't really see the decay well on the waveform display, it's better to rely on your ears and a pair of headphones. Also make sure to not cut out the small buildup of one or two periods preceding the actual note.
If the waveform is off center and the body tending towards the top or bottom, then you might want to use DC Offset to correct it.
Loop (see this thread
for information about looping) and apply any compression, eq and pitch correction if you find it necessary. The lesser the better.
Normalizing has traditionally been the final step after you're done messing with the sample. This is an optional step not necessarily done with modern sample libraries with the possibly exception of drum kits.Mapping
When you have all samples edited, or at least a small subset to make a test, start building your virtual instrument by mapping them on the keyboard and into velocity layers. This is a sampler specific process depending on what you samples and how you want the final instrument to behave. With LinuxSampler use Gigedit to create new instruments out of your samples.Recommended readingAudio Sampling
by Sam McGuire and Roy Pritts, that run through all the basics, and Sample This
by Simon Cann and Klaus Rausch.