jazzi wrote:Hello everyone,
I have a recording done in 24 bit resolution and 192khz sampling rate. I would like to learn the best practices for reducing this recording down to 16 bits and 44.1khz. I plan to do this conversion as the very last step in the process, unless it should be done sooner for a reason I can't think of.
The best practice is, first of all to work on a copy of the file and not the only one you have. Then do all your editing first, then when you are done, FIRST save to a new file (using "Save as..." in the File menu, not "Save") as a 32-bit Floating Point PCM file (either WAV or AIFF). That way if you are not content with any of the reduction process later you can start over with the final processed result at full original quality. Don't use 24-bit because, when you apply edits GoldWave processes your edits in 32-bit Floating Point and store them to a temporary 32-bit Floating Point PCM file, so any edits you make will be 32-bit floating point data and not 24-bit Integer.
Once you've done that, then I believe you should first resample to 44100 Hz. GoldWave does a pretty good job at this, if you are using the latest version (of either v5 or v6), but if you wanted the very best quality you'd want to use a third party program like r8brain by Voxengo (even the free version does a very high quality resample) or something. But I'm just going to stick to telling you how you would do this in GoldWave.
Then the next step would be to apply a 16-bit noise shaped dither, ideally a triangular noise shape. I'm not positive how to do this without a dither plugin, but I know that dither is mixing a very low level (probably somewhere around -90 dB) of random noise in with the file to mask the aliasing effects (quantization noise) of the bit-depth reduction. The noise shaping is probably just making it so that the higher random noise frequencies are louder than then lower ones, since they are less audible to the human hearing, but yet still achieve the masking of the quantization noise. Using a VST Plugin would be the best way, as I don't know how to create a triangle noise shaped random noise manually.
Then finally after doing that, then you'd simply just save the file as a 16-bit PCM format (WAV, AIFF, or some other, or a lossless format like FLAC). Make sure you click "Save as..." and not "Save" so that you don't lose your original 32-bit Floating Point PCM "master" file that I told you to save earlier!
jazzi wrote:Also, I would like to know how the "maximize volume" feature works in Goldwave. If for example the loudest data is at -7dB, what does the volume maximizer do? Does it multiply the amplitude of all data points by the same constant, or is the volume level mapped in another way?
I have some engineering background so feel free to speak technical or refer me to some articles if you like.
Thank you all for your help.
It takes the value you enter into what you want to maximize it to (in the "maximize" effect window), it converts that number to a linear value, then it calculates the linear value of the loudest sample in the file, and then it divides the two. The result of that will become the multiplier that will be applied to every single sample, it's not mapped any other way (you'd use the dynamics effect for that
).
So if you entered 0db in the desired level box, and the loudest peak was 0.5, then GoldWave would first convert the 0db to linear: (10^(1/20)) ^ "dB value you entered" (in this example "0") = 1
Then it would divide that by the loudest peak level 0.5, so: 1/0.5 = 2
So "2" will be the multiplier to which every sample is multiplied! Every sample will be multiplied by 2!
Really quite simple but I felt like explaining it in the geekiest way I could just for fun.
Hope this helps answer your questions, have a good day!