Sujet Sommation analogique ou numerique ?
- 1 395 réponses
- 110 participants
- 132 435 vues
- 120 followers
Afficher le sujet de la discussion
J'aime bien ces arguments imparables.
Moi j'entends, si t'entends pas va te laver les oreilles.
Souvent les mêmes que lorsqu'un ingé son chevronné chambre son stagiaire parcequ'il n'entend aucune différence ntre une carte son firewire et une PCI (situation vécue !
).
Anonyme
9677
scare
3151
Squatteur·euse d’AF
Membre depuis 19 ans
(j'ai le boulier qui gratte 
)
Tambour RLRLRLRLRLRLRLRLRLRLRLRLRLRL ....... pschhhhhhhhhhhhhhhiiii !!
And the winner is ....... ????
Voici la réponse de BLA :
Hi Ephrem--
In one sense, you're oversimplifying the issue. In another sense, you're over-complicating the issue. You're applying a theoretical ideal to a computer's ability to add without generating noise.
If the process is the same (same DSP functions performed by same software on different computers), that means the logic is the same. Therefore the deltas (the level relationships) between the DSP-generated harmonics (jitter) and the fundamentals could be so slight that an auto-nulling experiment would be unable to produce any data. In this case it would be best to measure with a spectrum analyzer. I don't think the human ear would be able to detect levels of harmonics that are -80 or -120dB from an auto-nulling experiment, yet a spectral analysis would be able to show it.
And you confess by your own hand that you can hear the difference between console summing and in the box summing. If you insist on using auto-nulling, the better approach would be to mult the signals so that analog summed and digital summed signals can be inverted and compared.
Yes, absolutely it's audible and measurable.
Jitter can be large enough to be audible but small enough not to corrupt a CPU process. Personally, I can hear variations in signal caused by jitter levels as low as 3 picoseconds. That's far, far, lower than anything that would cause a computer system to become unstable.
Yes, and we're right back to the original problem--the low level non-linearities caused by the bounce process have no mathematical relationship to the harmonic structure of the audio.
A 2vrms signal from a DAC running at 0dB FS will cause some line input stage non-linearity, but it's all small-signal stuff--2nd and 3rd order THD caused by the physical geometry of the on-substrate transistors. That's a far cry from the non-linear behavior caused by a bounce, and if you compare the two you'll certainly be able to hear the difference.
And by reducing DAC output level you're reducing output bit-rate. You want to make your DAC do as much work as possible without going over, and then allow your analog stage to do whatever work needed to bring signal up to +15, +18, whatever level you work at.
The argument I'm trying to make is that offline bounces don't sound as good as other methods of making a two track mix. Every method gives non-linear artifacts. But analog summing gives the most musical results in spite of whatever non-linear behavior may exist, and the best way to get that result is to run your DAC at 0dB center pan, and handle volume and pans in the analog domain.
Listen we can debate theories all day long. The best thing to do if you really want to know is to experiment with the sonic differences between a bounce, a two track digital summed master (made by running out of the interface master and recording a stereo pair of the master mix), and a two track mix made by analog summing. Then decide for yourself what works.
Matt
Pas encore eu le temps de lire
Bonne journée !
)Tambour RLRLRLRLRLRLRLRLRLRLRLRLRLRL ....... pschhhhhhhhhhhhhhhiiii !!
And the winner is ....... ????
Voici la réponse de BLA :
Hi Ephrem--
In one sense, you're oversimplifying the issue. In another sense, you're over-complicating the issue. You're applying a theoretical ideal to a computer's ability to add without generating noise.
Citation de Scare :
That would means one offline bounce made with two (or three) different computers (CPU) give out different results ? Because of harmonic signal generated by two kind of DSP would be different. After reading your sentence i've gone to the studio, there are three computers there. I've loaded 16 tracks on each, made the offline bounce. Afterwards i've done a nul phase tests between files pairs : full silence !
If the process is the same (same DSP functions performed by same software on different computers), that means the logic is the same. Therefore the deltas (the level relationships) between the DSP-generated harmonics (jitter) and the fundamentals could be so slight that an auto-nulling experiment would be unable to produce any data. In this case it would be best to measure with a spectrum analyzer. I don't think the human ear would be able to detect levels of harmonics that are -80 or -120dB from an auto-nulling experiment, yet a spectral analysis would be able to show it.
And you confess by your own hand that you can hear the difference between console summing and in the box summing. If you insist on using auto-nulling, the better approach would be to mult the signals so that analog summed and digital summed signals can be inverted and compared.
Citation de Scare :
Do you mean that any signal jitter caused by CPU processing is written on the file during the offline bounce, and be revealed during Bounce listening ?
Yes, absolutely it's audible and measurable.
Citation de Scare :
Also, I would say if "CPU jitter" is big enough the generate errors, most of the time the computer would crash, OS and software as well. no ?
Jitter can be large enough to be audible but small enough not to corrupt a CPU process. Personally, I can hear variations in signal caused by jitter levels as low as 3 picoseconds. That's far, far, lower than anything that would cause a computer system to become unstable.
Citation de Scare :
To be honest, i do not understand the second reason. If you speak about digital summing, there is no DAC converter in the offline bounce process, you can make it without sound card into the computer.
Yes, and we're right back to the original problem--the low level non-linearities caused by the bounce process have no mathematical relationship to the harmonic structure of the audio.
Citation de Scare :
And in analog world, by keeping all DAC's at full-scale output, means close to 0 dBFS, you come closer to achieving line input stage distortions, non more from my experience.
A 2vrms signal from a DAC running at 0dB FS will cause some line input stage non-linearity, but it's all small-signal stuff--2nd and 3rd order THD caused by the physical geometry of the on-substrate transistors. That's a far cry from the non-linear behavior caused by a bounce, and if you compare the two you'll certainly be able to hear the difference.
Citation de Scare :
On the contrary, I use to reduce DAC output to achieve clean analog summing and match analog signal levels. Please clarify this reason.
And by reducing DAC output level you're reducing output bit-rate. You want to make your DAC do as much work as possible without going over, and then allow your analog stage to do whatever work needed to bring signal up to +15, +18, whatever level you work at.
The argument I'm trying to make is that offline bounces don't sound as good as other methods of making a two track mix. Every method gives non-linear artifacts. But analog summing gives the most musical results in spite of whatever non-linear behavior may exist, and the best way to get that result is to run your DAC at 0dB center pan, and handle volume and pans in the analog domain.
Listen we can debate theories all day long. The best thing to do if you really want to know is to experiment with the sonic differences between a bounce, a two track digital summed master (made by running out of the interface master and recording a stereo pair of the master mix), and a two track mix made by analog summing. Then decide for yourself what works.
Matt
Pas encore eu le temps de lire
Bonne journée !
Construction du nouveau studio
Visitez le site THD STUDIO / page FB THD Studio
Dossier : Conversion analogique-numérique (pour les courageux)
Anonyme
9677
Anonyme
66280
Citation :
Personally, I can hear variations in signal caused by jitter levels as low as 3 picoseconds.
J'aime bien ces arguments imparables.
Moi j'entends, si t'entends pas va te laver les oreilles.
Souvent les mêmes que lorsqu'un ingé son chevronné chambre son stagiaire parcequ'il n'entend aucune différence ntre une carte son firewire et une PCI (situation vécue !
).Rickton
656
Posteur·euse AFfolé·e
Membre depuis 16 ans
Anonyme
The cat
7112
Je poste, donc je suis
Membre depuis 21 ans
Anonyme
5668
AlienVisits
190
Posteur·euse AFfiné·e
Membre depuis 21 ans
J'avoue que les explications de BLA ne m'ont pas non plus convaincu, c'est assez confus de leur part je le reconnais. 
__________________________________________________________________________
STUDIO D'ENREGISTREMENT A MARSEILLE
MIXAGE EN LIGNE
MASTERING EN LIGNE
nonconforme
8888
Je poste, donc je suis
Membre depuis 21 ans
Salut à tous,
content de voir ce sujet remonter même s'il a pas mal dérivé de l'idée initiale.
J'ai découvert un nouvel argument idiophile tout à fait rafraichissant : un jitter qui serait induit par le calcul d'un CPU passerait dans la bande audio (c'est bon ?).
Là on est en pleine science fiction.
Enfin, tout ça a commencé quand Apogee s'est mis à vendre du câble low-jitter...
Blague à part, je me permets de vous recommander la lecture du dernier SOS, pages 108-118, qui traite de la question des horloges externes dans une installation audionumérique. J'espère que les conclusions de l'article feront fermer leur grande bouche à toute une série de charlatans.
Dans le cas de la sommation, de toute évidence, ce sont les défauts audio de l'analogique (quoi que, avoir du headroom et une joli distorsion avant saturation n'est sans doute pas un défaut) et son ergonomie qui DOIVENT suffire à fair préférer cette solution.
Pourquoi inventer des concept ridicules qui font hurler de rire les spécialistes du numérique ?
Pour les horloges, qui sont une CONSEQUENCE de la nécessité de synchroniser différents convertos A/N (les cartes d'acquisition sont synchrones), il a fallut inventer une mythologie car, pour un ingé son qui ne comprend rien au numérique (presque pléonasme), il n'était pas imaginable qu'un appareil aussi cher qu'une horloge n'amène pas un gain audible (ça sonne MIEUX). Alors que, selon la théorie du signal, l'horloge inter d'un appareil est TOUJOURS préférable qu'une externe, qui se traine aux fréquences audio.
content de voir ce sujet remonter même s'il a pas mal dérivé de l'idée initiale.
J'ai découvert un nouvel argument idiophile tout à fait rafraichissant : un jitter qui serait induit par le calcul d'un CPU passerait dans la bande audio (c'est bon ?).
Là on est en pleine science fiction.
Enfin, tout ça a commencé quand Apogee s'est mis à vendre du câble low-jitter...
Blague à part, je me permets de vous recommander la lecture du dernier SOS, pages 108-118, qui traite de la question des horloges externes dans une installation audionumérique. J'espère que les conclusions de l'article feront fermer leur grande bouche à toute une série de charlatans.
Dans le cas de la sommation, de toute évidence, ce sont les défauts audio de l'analogique (quoi que, avoir du headroom et une joli distorsion avant saturation n'est sans doute pas un défaut) et son ergonomie qui DOIVENT suffire à fair préférer cette solution.
Pourquoi inventer des concept ridicules qui font hurler de rire les spécialistes du numérique ?
Pour les horloges, qui sont une CONSEQUENCE de la nécessité de synchroniser différents convertos A/N (les cartes d'acquisition sont synchrones), il a fallut inventer une mythologie car, pour un ingé son qui ne comprend rien au numérique (presque pléonasme), il n'était pas imaginable qu'un appareil aussi cher qu'une horloge n'amène pas un gain audible (ça sonne MIEUX). Alors que, selon la théorie du signal, l'horloge inter d'un appareil est TOUJOURS préférable qu'une externe, qui se traine aux fréquences audio.
Affiliation : Dirigeant Fondateur d'Orosys - Two notes Audio Engineering
[ Dernière édition du message le 10/06/2010 à 14:48:11 ]
- < Liste des sujets
- Charte