CompuServe Thread

John: I didn't know that you were so knowledgeable about sound. I have a very basic question for you: Background: As you may know I have a video post operation here and have primarily involved myself in simple projects like industrials and other low end work. But, recently since acquiring betacam editing I've taken on some more broadcast level productions for ESPN, PPV, and possible other networks like Discovery etc. On a recent project I found the audio editing quite over my head and we sent the audio clips out to a sound studio for the music, sweatening of voices and effects. The work was done on a MAC and transfer was done via a small DAT recorder I rented for the job. I got the tape back and lip synced it back up to the master and while very well done creatively I discovered some of the voice overs were laid down out of phase and the job must now be sent back for corrective action. Now for my question: What type of equipment ( boards, DAT, software ) would I need to look into to be able to do this type of thing on the PC? What type of costs are involved? Would I need a multitrack DAT or is a two track good enough. I have a 4 track TASCAM that works fair but not as good as the DAT. What are the differences between the portable DAT machines and the front loaders like the Alesis as far as quality is concerned. Do you see a need for time code capability if I am going to use the DAT just to dump to the PC from field recordings and not use it as a source feeder in edit. How long a musical run can one sustain at one playing off the computer, practically speaking ie minutes per megabyte at CD quality? Is there software that can effect up the sound like echo, reverb, waveform slicing and EQ? How many tracks does one get to work with? How difficult is it to maintain phase integrety between the left and right channels? Many questions, I know, but take your time. I just wanted to get a brief general overview of the state of the art as it relates to my business. As you might gather, dealing with outside service bureaus always makes me nervous.

But John, wavetable cards ARE second-rate approximations! It's just that FM is THIRD-rate! <g> Seriously, even the best wavetable cards have, what, 4MB of sound samples? Since there are 128 instruments in the General MIDI spec, plus the drum kit, and since each instrument should optimally be sampled in stereo (or, better yet, full Dolby Pro-Logic surround!) at 44.1kHz or more with 16 bits or more (24 would be better), with each note sampled separately rather than one that is transposed by changing the playback rate, not to mention sampling the same notes different times to handle different velocities (a piano or trumpet or violin note played loudly just doesn't sound the same as the same instrument playing the same note softly and then amplified to match the loudness), well, 4MB just isn't anywhere NEAR enough room! Most wavetable cards compromise by using mono samples, one note per octave, and one velocity sample per note. Some even do tricks like using one saxophone sample and transposing it up or down octaves to fake soprano, alto, tenor, and baritone saxes, or doing similar tricks with a viola sample to fake violin, cello, and contrabass (generally the boards with a lot less than 4MB of sample ROM). Other shortcuts including simulating "phat" sounds such as ensembles or honky- tonk piano by using multiple generators playing pre-existing samples with slight detuning (layering violins to simulate string ensemble, or a trumpet and a trombone to simulate brass ensemble, or pianos to simulate a honky-tonk piano, etc.) rather than sampling ensembles directly. This saves sampling room at the cost of reducing polyphony dramatically when such voices are in use: for instance, a brass ensemble made by layering two trumpets and a trombone would require three generators per note, so playing a four-note chord would take up 12 of the available generators, leaving precious few for the rest of the music at that point (remember that, for instance, a realistic piano arpeggio requires using up the generators for as long as the notes are decaying, not just as long as the "keys" are "down"). On the other hand, the best of the current line of Yamaha Clavinova CLPs sample one note for each group of three adjacent semitones (for instance, C4 to handle B3, C4, and C#4/Db4), in 16-bit stereo, with 9 samples each for ppp, pp, p, mp, mf, f, ff, fff, and sfz, with Yamaha's proprietary AWM (Advanced Wave Modulation) interpolating the other 118 velocities (for instance, if you play a note at a velocity 75% of the way between f and ff, AWM will take the samples at f and ff and interpolate them on the fly to generate a new pseudo-sample wherein the f sample has 25% and the ff sample has 75% influence over the final pseudo-sample). It can afford to do this because the CLPs only have a few instrument sounds (three or so pianos, an electronic piano or two, a strings, a harpsichord, a Clavi sound, a pipe organ, a celesta, and a vibraphone, if I remember right), so it wouldn't take huge amounts of ROM to store such high quality samples — even so, only a few of those are stereo-sampled even on the top-of-the-line model. Let's face it: if you want professional quality MIDI sound, you don't use the output of even the best wavetable card on the market. You get a Proteus or Yamaha TG-500 or Roland Sound Canvas or some such, which you can plug right into the MIDI Out of any sound card on the market all the way back to the original Sound Blaster! But MIDI Renderer has a CD-ROM full of samples. Even if we assume that the combined program code for the DOS, Windows, and OS/2 versions of the programs themselves take up a grand total of 150MB (they would have to be pretty darn big programs for what they do — 50MB would be more likely), that would leave 500MB from a 650MB CD to store samples! That's over a HUNDRED TIMES the sample base of even the BEST wavetable cards! Even taking into account that the MIDI Renderer has 200 instruments instead of just 128 (I think, therefore, that it is fully compatible with the GS superset of General MIDI, and not just General MIDI itself, or maybe they count the individual drums in the drum kit separately), there's room here for the samples to be MUCH more sophisticated than ANY wavetable card can handle! Theoretically, given moderate compression, they could let most of the General MIDI instrument sounds have the quality of sampling that Yamaha gives the piano sound on its very best Clavinova CLP! Or even go to 24-bit sampling! We already know they go to 100kHz, which is over twice what CD-ROM does (44.1kHz)! Also, there is no need for shortcuts since there is so much room! There are no tone generators to run out of, so you can have UNLIMITED polyphony! No note stops sounding before its time just because a tone generator was needed for a new note! Even so, there is one thing that samples just can't do, no matter how good they are. They can't dynamically change in character while a note is playing. For instance, a person playing a brass or wind instrument can blow harder or change the embrouchure (shape of the lips and mouth) to alter the sound mid-note — the "wailing trumpet" effect the best of the jazz legends are known for. String players can produce analogous effects by altering the bowing mid-note. These effects aren't just a matter of volume changing — if it were so, that could be simulated easily enough. But the actual TONALITY of the sound changes! Samples just can't duplicate that, since they are basically recordings. If one were to try to switch between samplings in the middle of a note, there would be an audible discontinuity just as there would be if one tried to switch between recordings in the middle of a note. This is why FM and similar synthesizing techniques still have a place. MIDI specs provide for such opcodes as "Breath Control," "Channel Aftertouch," and "Per-Note Aftertouch." Sampling systems, from wavetables to Roland Sound Canvas to Proteus to Yamaha TG-500 to the MIDI Renderer, cannot respond to such opcodes realistically. They are usually ignored by samplers. But true synthesizers can. FM especially can be very dynamic in its tonality. But FM isn't as realistic as sampling. This is why the Yamaha VR-1 is making such a splash right now. Although it's monophonic and limited to one instrument playing one note at a time, it's still considered a major breakthrough because it can give the flexibility (including dynamic tonality control) of FM with the realism of sampling, by using computer models of the actual sound generation processes of actual instruments, rather than simple recordings of them.