“I use Digi001 / ProTools LE on a Power Mac 9500 with a G3/300 upgrade card. When bouncing 48 kHz ProTools session to disk as a 44.1 kHz AIFF file, a three-minute song takes about 45 minutes (on a system like mine) to convert at “tweakhead” quality level. The computer is working so hard that no other task can run. What is actually taking place that requires so much time and processing power?”
First, while this is two days in a row of questions from Digi001 users (and no, they were not from the same person) both questions are pertinent to many other systems so don’t think we are going off the deep end with the 001 – it’s just a coincidence.
I‘m never quite sure how literally I should interpret this type of question. Surely you don’t want to know what is _really_ going on in there. Whether or not no other task can run while your computer converts audio files is questionable. Digi simply locks you out of being able to do any other task because they want all of the processing power available to use for the job at hand. Other tasks could happen simultaneously, but you’d pay for it in the amount of time it took to do the conversion.
The simple answer to the question is that sample rate conversion is extremely difficult to do well. In the case of converting from a higher rate to a lower one you are asking a program to make decisions about throwing away pieces of your audio data without changing the way it sounds to you. This is a complicated process that requires a lot of CPU cycles to achieve. You can’t just slow the clock down because that changes the pitch. Samples have to be removed. Of the 48 thousand samples taken per second, which 3900 are the expendable ones? Even that question dramatically oversimplifies what is taking place. Critically listening to the output an any audio pitch processing device will tell you that we’re still pretty far away from being able to manipulate audio like this without sonic consequence.
Which brings me to my next point, and is the question you did not ask. What about the quality of sample rate conversions? There are some engineers who believe there is no sample rate conversion algorithm anywhere that sounds as good as simply converting the signal back to analog and resampling it at the desired rate (assuming it is through excellent A/D and D/A converters). Eight or so years ago this was pretty universally agreed upon. This is exactly why so many people still prefer to record at 44.1 kHz even though they have faster sampling rates available to them. Nowadays conversion schemes are much, much better (especially in high end products) and with the growing popularity of higher resolution recording systems sample rate conversion is something that is increasingly ‘tolerated.’ Nevertheless it is still a good idea to check your work. Compare input to output and make sure you aren’t losing too much. Try alternate methods and see if they sound any better to you. Maybe you can save time and end up with quality that’s just as good or better.