Long, long ago I started a series of articles for TastyFresh on sampling. I promised a piece on one-shots, a piece on loops and a final installment on multisampling. I'm here to finally make good on part III.
The good news is that the basics are not that overwhelming. In my experience the truly complicated aspects of multisampling are either software specific or well beyond my needs. For this article I will focus on the basics concepts behind multisampling, practical reasons for a DJ or producer to build an instrument and a number of working examples. What I won't be doing is helping you plumb the depths of articulation switching and conditional re-triggering. These concepts are valuable and important but also more appropriate to sound design or instrument modeling than making a banging mix.
Throw The Map In The River. . .No, actually, don't do that.
The first two ideas you need to understand are key mapping and velocity mapping. Start with a mental picture of a piano keyboard. Each key sounds a different note. Any given key also sounds different depending on how hard it is pressed. To create a digital replica of a piano recordings of each key played at every possible volume (AKA "velocity," assuming the key is moving faster for louder notes) would be required. The recordings would then need to be organized so that exactly the right ones could be recalled to duplicate a live performance on the physical piano being replicated.
This organization of sounds is mapping. The standard technology for communication between electronic instruments and software is MIDI. In MIDI each note (each key on a keyboard) has a number and note velocities have numerical values from 0-127. Thus, notes played in any octave can have 1536 different combinations of note number and velocity. If one wanted to exactly model one octave of a piano (not counting use of pedals, etc) you would need 1536 samples. As you will see there is nothing stopping any of us from doing this, except that it is completely impractical. The important thing to understand is that this array of 1536 unique values (combinations of note and velocity) per octave is how all assignments of notes to samples are specified. Any sample or group of samples can be assigned to any combination of those values. But how?
But It Doesn't Look Like A Map
PolyPhontics is a Macintosh utility for building multisample instruments in the Sound Font format. Sound Fonts have been around for about fifteen years and were originally developed for use with Sound Blaster brand PC sound cards. The format continues to be actively developed by E-mu, Ensoniq and Creative Labs (all sister companies) but musicians mostly know it as a widely supported standard format for sampler and synth patches. Most full featured samplers, multisample players (NI Kompact, et al) and many sample based synths support Sound Fonts. Thus, if you have any doubts about where your multisample instrument will be used you might want to consider making it a Sound Font.
PolyPhontics allows you to take a group of samples and map them keys and velocities to build a virtual instrument. This is essentially the same as the basic instrument editing done in a full featured sampler like Kontakt, Mach5, Independence or (Ableton) Sampler. The first difference is that PolyPhontics is an editing utility not a sampler or plug-in itself. You will need to take the instrument you create to a Sound Font compatible synth or plugin to use it. The other is that PolyPhontics is decidedly no-frills compared to the flagship samplers listed above. I purchased it primarily to create multisampled instruments to use in Logic Express 7's EXSP sample player (Logic Express 8 includes a full version of EXS).
Here is our first view of the anatomy of a multisampled instrument. This is the PolyPhontics Document Editor window from an instrument of Gregorian chant samples I created for a song called Pope Gregory.
Technically, in Sound Font terminology this isn't an instrument. It is a "bank." The subgroups of sounds listed on the left of the window are the instruments included in this bank. Shortly after doing the work shown above I learned (the hard way) that support for multiple instrument Sound Font banks is not consistent. Some of my Sound Font compatible plug-ins would recognize only one instrument of a multi-instrument bank. It turns out that, for me anyway, making banks with only one instrument was the way to go. Below is the Document Editor view of the Gregory instrument I actually used for the song.
Listen to Sample 1
The most important thing to understand about these PolyPhontics windows is the right side of the Samples pane. The Low, Base and High columns represent the key mappings of the samples included in the instrument. The numbers are MIDI note values. Low and High are the highest and lowest notes that will sound the sample in question. Base is the note that sounds the sample at its original speed and pitch. All notes in the range between High and Low other than Base will play the sample sped up or slowed down so as to sound the appropriate number of semitones higher or lower.
If you are working with drums your choice of base note can be arbitrary. Basically you can just map your samples where ever you find convenient. In the case of my Gregorian chant samples I needed to be more careful. I needed to know what notes were being sung when each sample was triggered at each pitch. Thankfully the sample library I got the samples from included starting pitch in the file names, making it easy to pick logical base notes for each one and help me find samples likely to harmonize with each other.
To use a sound in PolyPhontics it must be added to the PolyPhontics' Library. From there you can start grouping samples into instruments and editing specific properties for each sample. It is important to remember that in this case you are not actually editing the sample. You are editing how the sample is played. No changes are made to sound file itself. Here is the Sample Editor page for a snare sample I included in a Sound Font instrument of my most used drum sounds. 
As you might expect the keyboard across the top allows you to visualize the sample's key mappings. The Velocity Range sliders do the same for velocity mapping. It is also possible to set looping, envelope, panning and reverb from this page. Your mileage may vary on how/if some of these settings are handled by different Sound Font players.
If you are assembling a relatively simple instrument and or portability is critically important to you Sound Fonts (and utilities like PolyPhontics) have a lot to recommend them. There are also many reasons to turn to a more full featured sampling instrument.
Ableton, the company that gave the world Ableton Live, released Sampler during Live 6's run. Sadly, it isn't available in any standard plug-in formats (VSTi, DirectX, AU) but if you have already invested in Live its kind of a no-brainer. It is a fraction of the cost of some other samplers and it is masterfully optimized for use in Live. It does not come with a 20GB sample library like some of the competition but 19 of those gigabytes are probably stuff you don't want. Besides that, you can't buy Kontakt or Mach5 as a digital download. Last I checked you couldn't even download a demo of MOTU Mach5, which I have heard is fantastic to use. But, isn't boxed software just so 20th century?
Make sure you read that last line with a really bored, world-weary tone in your voice, even though it is true.
Sampler will import many foreign sample formats, including Sound Fonts and Kontakt. However, it assumes that you have come to Live and Sampler to stay and does not provide easy exports. Launching Sampler is as easy as any of Live's other built in instruments. You just drag it's icon out of Live's browser and onto a MIDI track. Feeding it a patch, sample or virtual instrument works the same way.
Here is a Live 7 window with a standard bass patch loaded up in Sampler. Sampler's mapping pane is nicely placed between the device area and the mixer.
If you are familiar with Live's Racks feature you should recognize the bars representing each sample's key range. Changing them is as easy as dragging the bars one way or the other. Ableton could have made this easier, but the software would have needed to read your mind and I'm told their lawyers warned them off of it.
Remember how you had to go to different places PolyPhontics to see velocity and key mapping information, and could only see velocity information for one sample at a time? Sampler provides this information in a much nicer way. In the picture above you can see a series of bass samples. None of them overlap keys. Click the Vel button next to the Key button to reveal the velocity map for this instrument.
This instrument uses no velocity layering and all samples are set to respond to all velocities of their mapped notes. This screen shot only gives the bottom half of the velocity range, but you can trust me.
Here is the lower half of the key map for one of my drum kits. You can see that most of the sounds do not overlap keys, but three high hat sounds overlap exactly.
I could have done this just to make the sound bigger. The hits are from the same source and the same session so they sound quite a lot alike, but not identical. Besides just being louder, layering them makes the sound fuller and more complex. It also gains a noisy quality that I like. But I wanted something else as well. I wanted more variation over the velocity range than I could get with any one hat step. One click on the Vel button reveals what I did.
The ranges of the three hat steps overlap, but only at higher velocities. The three hat step samples are layered so that all three will sound at velocities over 100. Numbers 9 and 11 will sound at velocities from 76 to 99. Velocities 75 and below only trigger the sample 5K_HH_11_step.wav. This is what they sound like, from loudest (highest velocity) to softest.
Listen to Sample 2
I didn't demonstrate it, but I mentioned that before PolyPhontics will let you work with a sample it has to be added to the utility's Library. Ableton also makes this easier. Adding a sound to my drum kit is as easy as dragging the file into the list of samples in the mapping pane.
Looking at Live's device window reveals the settings we saw in the PolyPhontics sample editor and much, much more.
On the sample tab you can set root key, detuning, looping and sustain options, among others. Sampler's other tabs include a volume envelope and synthesis style filters, oscillators, and a multistage pitch envelope.
As you can see there is a lot to work with here. Besides being able to quickly and easily do any sample mappings that strike your fancy, Sampler gives you modulation routing and LFOs to play with. I love less complex samplers like the ones I wrote about in part 1 of this series but I am a sucker for this stuff too. Like everything else, it's all about the right tool for the job.
This concludes my series on sampling. I know plenty of you actually know more about these matters than I do, but I still hope it has gotten people thinking about what can be done. With tools like this at our disposal there is no excuse for sounding stale.