What concerns are there when converting 24bit 48khz to 16bit-44.1khz (CD quality)?

Question

My band is working with a somewhat pricey mastering engineer and we are wanting to keep the price down. For a price we're comfortable with, he will either provide us with hi-res masters (24bit, 48khz - the recording quality), or with CD-quality masters + DDP file (16bit-44.1khz).

My gut is to get the hi-res masters, buy Triumph for $79, and do the conversion and create the DDP file myself. However, the engineer said he does a separate mastering pass for each quality he does.

My question is: what might necessitate a different mastering pass for the CD-quality version? What benefit might there be to getting the mastering engineer to do the CD-quality version instead of just converting it myself from the hi-res version?

Answer 1

Dither, Noise Shaping, and Bit Quantization. These are the reasons for the separate mastering passes. All of which do not need to be considered when mixing/bouncing the audio in the native digital format it was converted to (24bit/48khz).

Also, it sounds like he will provide you with individual "HiRes" native files OR Will downsample for you and create a DDP file. (Which is a FINAL Copy including play order and silence between songs that you can usually upload directly to a CD cuplication service.)

If you are not making CDs or distributing digitally only (itunes?). You have no use for a ddp file. Although u may have use for the MEs tools and know how when it comes to downsampling.

EDIT: I suppose it is possible for the ME to EQ specifically for the lower quality (not likely!) to produce better translation at the format? Also he may handle level peaks differently when downsampling.

IF it were me, Id take the native hi res master and be about my business. Ive also been an audio engineer for years tho.

Answer 2

For going from 24 bit to 16 bit is relativity simple since it is evenly divisible, however you will still want to apply a dither to make up for the way the originally analog signal loses resolution. This allows simulating a wider set of distinct values than 16 bits would normally allow in exchange for some minimal noise. A great link about this was included by Takuya in the comments below.

With 48khz sampling, you can accurately capture waveforms of up to 24khz frequency. There may be some weird harmonic issues that occur during the conversion, but they would be outside the human hearing range.

It isn't truly a lossless experience since it isn't evenly divisible, but it shouldn't cause major issues either.