You may consider the following as condescending, but be assured that it is not. It is just in the sole purpose to retrieve the files in the best possible conditions, and that in our common goal: what is best for the track.
Actually, it is more an ideal statement, if your tracks do not comply with all of the criteria, it will still be possible to deal with them. No hard feeling though.
No MP3, no compressed format of any kind. Expected file types are ideally WAV or AIFF both preferably in 32 bits float (16 bits are also accepted, but frankly speaking, if not too late, set your sequencer’s project on 24 or 32 float; all sequencers can do it for quite some time now, along with the setting up of the audio interface on 24 bits).
About the sampling rate, it’s up to you. I would say it is useless to go for 192 KHz, even worse, it wouldn’t give justice to the true sound, especially in the harmonics. A rate of 44.1 KHz is more than enough (appropriate?) for the human ear without generating too much distortion (especially in the top part of the spectre). Avoid as much as possible 48KHz and 96 KHz as mathematically inappropriate with a CD format in mind. Prefer a 88.2 KHz (or a multiple of 44.1 KHz).
Since 88.2KHz is the double of the target CD format (44.1 KHz), there will be less approximation errors during the downscaling.
Let me insist on one point: no sampling rate conversion! From recording to mastering, it is better not to alter it, regardless its value.
Mixing and only mixing!
When the mixing is done, do not apply any compressor limiter or any other modifier on the Master BUS, no normalisation either. The file should only be the result of the summing of your rendered tracks, with the only exception of artistic purpose. In that case, send me the two versions: one without anything on the Master BUS , the other one with the desired effect.
Ideally export the mix with medium RMS levels, not exceeding -16 DbFS.
Same rule, no normalisation. Preferably, use the tracks faders (if you have access to them), do not use master bus fader to lower the level.
Check that your export does not include peaks above -3 DbFS (especially in the case of a 16 bits export). Ideally -6 DbFS; if below, even better. This way, I have more headroom to work with.
Should you have access to the mix, or mixing yourself, avoid to overdue the compression as much as possible in order to let me deal with it. Over-compressing is harder to correct than the other way around. On that note, I am not a big fan of loud masters, but, if requested, I can do it nonetheless.
It may sound obvious, but should some voice takes include some sibilants, try to use a de-esser directly on the track. It can be dealt with during the mastering but it will then be applied to the mixing as a whole, affecting the other instruments (in the affected frequency range anyway), so better solve the problem at its source.