|
Post by jeromemason on Nov 20, 2017 21:49:41 GMT -6
Some of you guys are really smart with this and I'm not hip to making a session past 24 bit, but, when I get sessions to mix I always consolidate the active playlists and export then import that into my own session. Sounds crazy, but it saves me from having to deal with someone else' settings.
If the files are cut in 24 bit, does it do me any good to make these sessions 32 bit and if you wouldn't mind, just why?
Thanks!
Jerome
|
|
|
Post by NoFilterChuck on Nov 21, 2017 4:13:23 GMT -6
Plugin math. If the plugin was written with JUCE, the audio has been converted to either 32bit floats or 64-bit doubles depending on the DAW. You can look here if you're curious: github.com/WeAreROLI/JUCE/blob/4.3.1/modules/juce_audio_processors/processors/juce_AudioProcessor.cpp#L703Audio files are 24 bit because that's all the dynamic range we need to hear stuff. But computers are not designed with 24-bit memory sizes. they are designed with 32-bit-sized memory slots or 64-bit-sized memory slots. So, when your DAW loads up an audio file, it's copying those 24 bits per sample into a 32-bit-sized space of memory. so 8 bits are not being used. Or maybe they are, if they have a smart system to pack/unpack the samples into memory. things get weird when you need to read the 17th sample from the audio file, and your sample spans memory addresses. ie, the first 8 bits are at address 0x008667, and the 2nd 16 bits are at 0x008868. How do you keep track of which bits at a memory address to read from and which bits to ignore? Things get weird fast, so I doubt most system designers would do that. Perhaps @exponentialaudio would like to chime in. He's a plugin developer. If you're curious what plugins have been written with JUCE: juce.com/made-with-juce/users Superior Drummer 3 was written with Juce, as was UA Console and MAX/MSP. UVI as well as some EWQL plugins. Eventide's H9 Control App was written with JUCE. I may be mistaken, but I thought that PCM data is integer-based, and not floating point. If that's the case, then the DAW is converting that integer representation of samples that were recorded using a scale of −8,388,608 to 8,388,607 into a scale from -1.f to 1.f.
|
|
|
Post by iamasound on Nov 21, 2017 4:57:41 GMT -6
Sure, everybody knows that!
|
|
|
Post by kcatthedog on Nov 21, 2017 5:57:45 GMT -6
@nfc Impressive !
|
|
|
Post by lcr on Nov 21, 2017 6:41:28 GMT -6
What NoFilterChuck said
|
|
|
Post by ericbradley on Nov 21, 2017 7:13:43 GMT -6
I may be mistaken, but I thought that PCM data is integer-based, and not floating point. If that's the case, then the DAW is converting that integer representation of samples that were recorded using a scale of −8,388,608 to 8,388,607 into a scale from -1.f to 1.f. PCM can be either floating point or integer. The .wav format supports both 32-bit and 64-bit floating data.
|
|
|
Post by ericbradley on Nov 21, 2017 7:32:40 GMT -6
Some of you guys are really smart with this and I'm not hip to making a session past 24 bit, but, when I get sessions to mix I always consolidate the active playlists and export then import that into my own session. Sounds crazy, but it saves me from having to deal with someone else' settings. If the files are cut in 24 bit, does it do me any good to make these sessions 32 bit and if you wouldn't mind, just why? Thanks! Jerome The beauty of using 32-bit floating point when exporting is that you have a dynamic range of > 1500 dB. This means in practice that even an attenuated track will retain it's signal to noise ratio instead of being truncated. Or on the other side of the spectrum: A peaking signal will not clip onto the rendered file.
|
|
|
Post by mrholmes on Nov 21, 2017 9:55:48 GMT -6
If the files are cut in 24 bit, does it do me any good to make these sessions 32 bit and if you wouldn't mind, just why? Someone correct me if I am wrong but IMO you just ad 8 empty bits to it and that amkes no diffrence at all.
|
|
|
Post by Ward on Nov 21, 2017 11:02:40 GMT -6
Plugin math. If the plugin was written with JUCE, the audio has been converted to either 32bit floats or 64-bit doubles depending on the DAW. You can look here if you're curious: github.com/WeAreROLI/JUCE/blob/4.3.1/modules/juce_audio_processors/processors/juce_AudioProcessor.cpp#L703Audio files are 24 bit because that's all the dynamic range we need to hear stuff. But computers are not designed with 24-bit memory sizes. they are designed with 32-bit-sized memory slots or 64-bit-sized memory slots. So, when your DAW loads up an audio file, it's copying those 24 bits per sample into a 32-bit-sized space of memory. so 8 bits are not being used. Or maybe they are, if they have a smart system to pack/unpack the samples into memory. things get weird when you need to read the 17th sample from the audio file, and your sample spans memory addresses. ie, the first 8 bits are at address 0x008667, and the 2nd 16 bits are at 0x008868. How do you keep track of which bits at a memory address to read from and which bits to ignore? Things get weird fast, so I doubt most system designers would do that. Perhaps @exponentialaudio would like to chime in. He's a plugin developer. If you're curious what plugins have been written with JUCE: juce.com/made-with-juce/users Superior Drummer 3 was written with Juce, as was UA Console and MAX/MSP. UVI as well as some EWQL plugins. Eventide's H9 Control App was written with JUCE. I may be mistaken, but I thought that PCM data is integer-based, and not floating point. If that's the case, then the DAW is converting that integer representation of samples that were recorded using a scale of −8,388,608 to 8,388,607 into a scale from -1.f to 1.f. Pretty much, yes, in Layman's terms... if you don't want to get too technical. LOL
|
|
|
Post by jeromemason on Nov 21, 2017 23:31:15 GMT -6
I just experimented with a mix I was just sent. I mixed the whole song in 44.1/24 and then did a save copy in and changed the settings to 88.2/32 and I could swear there was a difference, a good difference. It didn't have any effect on my rig, in 44.1 my buffer only goes to 1028, in 88.2 it doubles so there was no performance issues, although I did have to bump up the disk buffer, that was all.
The difference I felt I could hear was my mix seemed to have more space between the notes kinda thing. If this was a phantom thing or my mind playing tricks on me then oh well. That surely wouldn't be the first or the worst scenario in which that happened. Once I engineered on a session and the producer wanted more bottom in the snare, I reached over and boosted with the 1073, he was like "stop right there" and I felt it was just right too. He walked out for a smoke and I looked over and the damn eq button was disengaged. Never said a word and didn't touch the bottom on the snare. He loved it on all 10 songs we cut that day lol.
|
|
|
Post by lcr on Nov 22, 2017 6:38:43 GMT -6
I just experimented with a mix I was just sent. I mixed the whole song in 44.1/24 and then did a save copy in and changed the settings to 88.2/32 and I could swear there was a difference, a good difference. It didn't have any effect on my rig, in 44.1 my buffer only goes to 1028, in 88.2 it doubles so there was no performance issues, although I did have to bump up the disk buffer, that was all. The difference I felt I could hear was my mix seemed to have more space between the notes kinda thing. If this was a phantom thing or my mind playing tricks on me then oh well. That surely wouldn't be the first or the worst scenario in which that happened. Once I engineered on a session and the producer wanted more bottom in the snare, I reached over and boosted with the 1073, he was like "stop right there" and I felt it was just right too. He walked out for a smoke and I looked over and the damn eq button was disengaged. Never said a word and didn't touch the bottom on the snare. He loved it on all 10 songs we cut that day lol. Not to dis your experiment, It very well could sound better but I wonder because many think plugins perform better at higher rates, if you started from scratch at the higher rate if the comparision would sound even better. This is an impossible comparision, because mixing the same thing twice isnt a fair comparision.. ongoing debate are A/B tests accurate? I do mix at 32/96K
|
|
ericn
Temp
Balance Engineer
Posts: 15,014
|
Post by ericn on Nov 22, 2017 8:35:57 GMT -6
Plugin math. If the plugin was written with JUCE, the audio has been converted to either 32bit floats or 64-bit doubles depending on the DAW. You can look here if you're curious: github.com/WeAreROLI/JUCE/blob/4.3.1/modules/juce_audio_processors/processors/juce_AudioProcessor.cpp#L703Audio files are 24 bit because that's all the dynamic range we need to hear stuff. But computers are not designed with 24-bit memory sizes. they are designed with 32-bit-sized memory slots or 64-bit-sized memory slots. So, when your DAW loads up an audio file, it's copying those 24 bits per sample into a 32-bit-sized space of memory. so 8 bits are not being used. Or maybe they are, if they have a smart system to pack/unpack the samples into memory. things get weird when you need to read the 17th sample from the audio file, and your sample spans memory addresses. ie, the first 8 bits are at address 0x008667, and the 2nd 16 bits are at 0x008868. How do you keep track of which bits at a memory address to read from and which bits to ignore? Things get weird fast, so I doubt most system designers would do that. Perhaps @exponentialaudio would like to chime in. He's a plugin developer. If you're curious what plugins have been written with JUCE: juce.com/made-with-juce/users Superior Drummer 3 was written with Juce, as was UA Console and MAX/MSP. UVI as well as some EWQL plugins. Eventide's H9 Control App was written with JUCE. I may be mistaken, but I thought that PCM data is integer-based, and not floating point. If that's the case, then the DAW is converting that integer representation of samples that were recorded using a scale of −8,388,608 to 8,388,607 into a scale from -1.f to 1.f. Sure Chuck, it's so simple when you bring math and science stuff to the table!😎 Do what works best for you and makes the clients happy !
|
|
|
Post by NoFilterChuck on Nov 22, 2017 11:48:55 GMT -6
My point was that the audio engine is gonna turn the 24-bit audio samples into 32-bit or 64-bit floats regardless.
|
|
|
Post by jeromemason on Nov 22, 2017 13:11:52 GMT -6
I just experimented with a mix I was just sent. I mixed the whole song in 44.1/24 and then did a save copy in and changed the settings to 88.2/32 and I could swear there was a difference, a good difference. It didn't have any effect on my rig, in 44.1 my buffer only goes to 1028, in 88.2 it doubles so there was no performance issues, although I did have to bump up the disk buffer, that was all. The difference I felt I could hear was my mix seemed to have more space between the notes kinda thing. If this was a phantom thing or my mind playing tricks on me then oh well. That surely wouldn't be the first or the worst scenario in which that happened. Once I engineered on a session and the producer wanted more bottom in the snare, I reached over and boosted with the 1073, he was like "stop right there" and I felt it was just right too. He walked out for a smoke and I looked over and the damn eq button was disengaged. Never said a word and didn't touch the bottom on the snare. He loved it on all 10 songs we cut that day lol. Not to dis your experiment, It very well could sound better but I wonder because many think plugins perform better at higher rates, if you started from scratch at the higher rate if the comparision would sound even better. This is an impossible comparision, because mixing the same thing twice isnt a fair comparision.. ongoing debate are A/B tests accurate? I do mix at 32/96K Well when I used to use NS10's I worked really hard to make mixes sound good on those speakers, when the mix sounded good on them I would flip to the good speakers and it sounded even better. Pretty sure if I'd started in 88.2/32 the mix would had sounded the same, just wouldn't had taken me as long to get there.
|
|
ericn
Temp
Balance Engineer
Posts: 15,014
|
Post by ericn on Nov 22, 2017 15:35:26 GMT -6
Not to dis your experiment, It very well could sound better but I wonder because many think plugins perform better at higher rates, if you started from scratch at the higher rate if the comparision would sound even better. This is an impossible comparision, because mixing the same thing twice isnt a fair comparision.. ongoing debate are A/B tests accurate? I do mix at 32/96K Well when I used to use NS10's I worked really hard to make mixes sound good on those speakers, when the mix sounded good on them I would flip to the good speakers and it sounded even better. Pretty sure if I'd started in 88.2/32 the mix would had sounded the same, just wouldn't had taken me as long to get there. Yeah I remember the first 192 /24 mix I did was a bunch of Opera pieces for broadcast, left a bad taste in my mouth because I was so proud of all the little detail I worked so hard to preserve, well between bringing it down to 44.1-16 and broadcast it disappeared! So I stuck around 48-24 to avoid that happening again!
|
|