|
Exporting from Ableton (pg. 5)
|
View this Thread in Original format
| dannib |
| Cyrus is right. if you have to ask that kind of question in the first place, you don't understand digital audio principles. |
|
|
| Kismet7 |
| quote: | Originally posted by thecYrus
no, he's not lazy. that's just his answer to every noob who asks this question. |
Actually, its pretty much a standard, you wont find many mastering engineers telling producers to mix to 0dbfs on a DAW, unless they are nubs of course. |
|
|
| dannib |
Kismet. you have to ask yourselves why some mastering engineers say that. Its because if they didn't, noobs like you would be clipping their master buss all over the place. Its for precaution, nothing else.
As i originally said. As long as the master buss doesn't clip it is fine. You are better off mixing at a higher level, expecially higher then -18db lol. this is even more important if you are routing certain tracks out to be processed by analogue devices. Read up on signal to noise ratio. |
|
|
| Kismet7 |
| quote: | Originally posted by dannib
Kismet. you have to ask yourselves why some mastering engineers say that. Its because if they didn't, noobs like you would be clipping their master buss all over the place. Its for precaution, nothing else.
As i originally said. As long as the master buss doesn't clip it is fine. You are better off mixing at a higher level, expecially higher then -18db lol. this is even more important if you are routing certain tracks out to be processed by analogue devices. Read up on signal to noise ratio. |
yep im a "noob", yet probablly better than your mixing skills. If you want we can have a mixing contest.
We'll send it to professional mastering engineer asking who has a better mix. All this talk is becoming nonsense, when real world results probably dont reflect much of what you have to say. So lets see who really knows more eh? And this offer goes to anyone debating in this thread. |
|
|
| echosystm |
| quote: | Originally posted by dannib
Why can't he just digitally turn the mix down before he applies any additive processing? |
ok this is going to get complicated. below is my understanding of the topic.
if your song is getting mastered 100% in software, you are best off rendering as hot as possible. otherwise you will need to unnecessarily bring the level up and lose quality. that much is obvious.
however... if the sound is going through a DAC and we are using hardware, we have a different set of problems. first off... headroom is ONLY an issue if we plan to BOOST frequencies in an eq or some other effect BEFORE we compress the signal (as is usually the case). by compressing the signal and not adding any makeup gain, we have effectively created headroom. thus, if we were doing eq boosts AFTER compression, headroom is irrelevant. if we are not doing any signal boosting, we would be better off always having the signal as hot as possible, so that we don't need to use as much makeup gain on the compressors and bring up the noise floor.
firstly, digitally attenuating the volume of the mix, prior to the DAC, is not a good idea. to adjust volume in the digital domain, bits are truncated. so, the more you turn the signal down, the closer you will come to hitting the floor of the dynamic range. in other words, by reducing the volume, you lose quality. with 32bit float, it's not really a big issue, but for 24bit fixed point it might be - lower dynamic range.
instead, you would be better off allowing the signal to go through the DAC without touching the volume, then use a PASSIVE attenuator prior to the other mastering fx hardware. this is better than using an amplified circuit to reduce the volume, as passive attenuation won't affect the sound as much. obviously you don't have the bit truncation problem either, as there has been no digital attenuation.
now... if we had 6db headroom on our mix, we would have a signal peaking at -6db coming out of the DAC. in the DA conversion, we have applied the SNR of the DAC to the signal - a noise floor is created. throughout the mastering process, we would need to bring the signal up AT LEAST 6db, therefore we have raised the noise floor by at least 6db too.
i do not understand why you would want to do this, when the signal could have simply been attenuated beforehand. likewise, if no major EQ boosts are done, prior to compression, you have brought up the noise floor for absolutely no reason.
so, in light of the above, i can only hypothesise two reasons for leaving headroom:
1. to make sure noobs don't have the signal clipping.
2. because the amplification circuits in the hardware distort at very high or very low levels of amplification (the extremities i was talking about earlier). if this is the case, we would obviously be better off amplifying the signal by 3db rather than 1db.
hopefully someone can point out the flaws in my logic, or validate my conclusions.
edit: all of the above is obviously in the context of the song being mixed on a computer... |
|
|
| Subtle |
Maybe if you are using an analog mixer or something you should export at -6db
Could that be what mr. Babicz meant ? |
|
|
| Kismet7 |
| quote: | Originally posted by Subtle
Maybe if you are using an analog mixer or something you should export at -6db
Could that be what mr. Babicz meant ? |
Nope, the exact opposite. If your using analogue mixer you can mix hot. |
|
|
| Subtle |
| quote: | Originally posted by Kismet7
Nope, the exact opposite. If your using analogue mixer you can mix hot. |
Mix yeah, but export ? |
|
|
| Kismet7 |
| quote: | Originally posted by Subtle
Mix yeah, but export ? |
Whatever you've mixed is what you are going to be exporting. |
|
|
| kitphillips |
| quote: | Originally posted by echosystm
from my understanding, it needs to be a multiple of the CPUs word length, otherwise it is just a waste. the word length is the amount of bits that can be moved in one chunk. eg. one 32bit float is two words... a 24bit float would be 1.5 words. 4 bits would be wasted, so there's no reason to do this; 32bits would have the same overhead, but obviously offers more precision.
in a fixed point number, the decimal position is fixed. in a floating point number, the decimal position can be anywhere. the benefit of floating point numbers is that they can express a far greater range of numbers, as the decimal place shifts. the benefit of this to audio is that each sample can have a greater range of values. therefore, it is more accurate than using a fixed point number. the dynamic range of 24bit fixed is something like ~130db, whereas 32bit float is 150db+.
in order to allow the decimal point to float, we have to use some bits to store where it is. for example, a 32bit float number uses 1 bit for the sign, 8 bits for the decimal position and 23 bits for the actual digits themselves. i assume the reason that 16bit float isn't used is because the overhead of storing the decimal position and the extra cpu load of working with floating numbers outweighs the benefit, over a more simple fixed number. |
Interesting info dude, thanks;) |
|
|
| DJ RANN |
For some reason there has been so much misinformation about the optimum/best/preferred/etc. mixing levels.
Just because Robert Babicz prefers it that way doesn't make it perfect or even right. It's probably the preference for him because if he receives a clipped mix (even minute) there's all he can do about it, so to be on the safe side he says -6dbfs (i.e. he can add gain if it's too low without too much added noise but can't take away clipping distortion once it's there etc.). The other reason is that he doesn't do one track at a time, so having that headroom to balance several tracks to form an album is a big deal for him.
Also, you have to realize that the reasoning is for commercial issues - it's a very safe margin because he probably has to act like he can't trust the levels on the project he being given, so it's a safe margin of error or him to work with.
Even though Echosystems reasons are right, there is a direct scientific reasons (suggested mainly by Mr Katz) for the -3dbfs mix levels.....but I disagree with that anyway....
A purely digital signal (no samples) with a value of 0 dBFS can still clip when converted to analog, due to intersample peaks created by the DAC.
Also, depending on the way your equipment is calibrated or metered, if a full scale square wave has an RMS value of 0dbfs, then a pure sine wave at the same calibration could have a a level of up to +3dbfs RMS (and therefore clip), so the theory goes that if you lower your final mix peak level to -3db, then there is no danger of this.
My problem(s) with these is that firstly, we work nearly all in enclosed systems using the same daw to compose as we do to master (and even if we use another machine/host to master) it's already an offline file), meaning the only DAC going on is output the end result to our speakers. Hardly any of us have the signal leave our DAW systm and go in to analogue to be converted
the chances of having a pure sine wave in your track which is exactly 45 degrees out of phase are basically non existant (unless you're writing test tones and calling it music).
Basically IMO, a fraction below odbfs is the perfect level, and lower isn't making the most of your available dynamic range and SNR. |
|
|
| echosystm |
| quote: | Originally posted by DJ RANN
A purely digital signal (no samples) with a value of 0 dBFS can still clip when converted to analog, due to intersample peaks created by the DAC. |
interesting! but if this was such an issue, wouldn't it be a problem in playback (like after it is mastered to 0db) too? |
|
|
|
|
