Hopefully you will find useful what we have to say about this very important subject to master.
Blog post: https://www.bluecataudio.com/Blog/tip-o ... tal-audio/
The Truth and Nothing But the Truth About Latency
-
Peter - Blue Cat Audio Peter - Blue Cat Audio https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=560426 - KVRist
- 34 posts since 14 Apr, 2022
We hear and read so many wrong things about latency that we decided to write a detailed article about it.
Hopefully you will find useful what we have to say about this very important subject to master.
Blog post: https://www.bluecataudio.com/Blog/tip-o ... tal-audio/
Hopefully you will find useful what we have to say about this very important subject to master.
Blog post: https://www.bluecataudio.com/Blog/tip-o ... tal-audio/
-
- KVRer
- 2 posts since 6 Apr, 2023
I finally read the article today after bookmarking it a week ago.
First, thank you for doing it, hopefully others will find it useful.
Next, let me add this, I tell students to hold out their hand and look at their five fingers:
The 'Pinky' is Acoustic
The 'Ring' is Analog
The 'Middle' is Digital
The 'Index' is Analog (again, and...)
The 'Thumb' is Acoustic (again / finally)
This is the journey that sound takes from the source to the destination, from a Mouth to an Ear.
What gets forgotten is that Inside of a traditional 'Recording Studio', there are two Acoustic environments, one for the Microphones and one for the Speakers.
The 'Studio/Live Room' and the 'Control Room'.
Before any electronics are deployed, the Recording Engineer places the sources within the room because after the sound of the sources themselves, the room itself will have the greatest affect on what is being recorded. Only after that physical acoustic placement does she chose Microphones and determine Microphone Placement.
Obviously a single Vocalist with a Single-Mic is different than a Piano Stereo-Mic'd, or a Drum Kit Multi-Mic'd, or a Full Band with GoBos (go-betweens) and Many-Mics.
Once more than one Mic is deployed, then even before a single XLR cable is plugged into them, a Tape Measure should be brought in to determine the distances between each of the Sources and each of the Microphones. As pointed out in this article, these distances can be multiplied against the Speed-of-Sound and converted into Milliseconds of ACOUSTIC LATENCY.
This Acoustic Latency is best addressed by physically moving the Microphones and/or Sources themselves in order to (at least attempt to) have have each Acoustic Sound Wave arrive at each Microphone Capsule at exactly the same time. Zero Milliseconds of DIFFERENCE in the latency between them.
This can be virtually impossible of course, and there are many different techniques, but the process is similar in the physical space to what the PDC as described in the article is doing in the digital space. You could figure out the farthest Microphone Capsule in physical distance and move all other Mics to equal that same distance. So one Snare Hit would enter two or twenty Microphones at precisely the same moment.
But say in the case of Drums, you'd often have Close Mics, Overhead Mics, and Room Mics all placed a say three Different Distances from say the Snare and/or Kick. The Snare-Top & Snare-Bottom or Kick-In & Kick-Out Mics could be perfectly measured against each other (and their Polarities adjusted), and the Stereo Spaced-Pair Overhead Mics Left & Right could be perfectly measured against themselves, same with a Stereo Pair of Room Mics, but How do you physically space these Three Concentric Rings against each other? They will receive the Sound Wave at different times, or Acoustic Latencies.
Now, you plug in all the XLRs, you adjust Pre-Amps, maybe tweak EQs, maybe Gates, maybe Compressors, all "on the way in", and then you hit Record either on a Tape Machine or on a Computer attached to an Interface.
The next question: Did anything from the Microphone Capsule to the Record/Conversion incur any additional Latency. If so, wouldn't That actually be ANALOG LATENCY as opposed to the ACOUSTIC LATENCY we just discussed and as was described in this video?
I've always heard that "Analog has no Latency". That once a signal is analog, it's electrons are moving through circuits at the Speed of Light rather than the Speed of Sound. Is this true?
There is one particular latency that I'd point out that often gets forgotten in conversations about Audio Interface Recording Latency:
The problem is real and easy to understand, a Microphone is plugged into an Audio Interface's preamps (rather than an Analog Console's), then an Artist is given a pair of Headphones. They sing... and then... they hear themselves... with a performance destroying delay.
What's happening (as you obviously know but others may not) is that:
Sound leaves the Artist's Mouth,
the Microphone (Transducer) converts that Acoustic Energy into 'Mic-Level Analog' Audio,
the Interface's Preamp amplify that signal to 'Line-Level Analog' Audio (still in realtime),
but then the Interface's A-to-D Chips convert the signal from Analog to Digital Audio (at a particular Bit-Depth and Sample Rate, sat 24-Bit/96kHz),
that signal travels Digitally through the Computer, probably through Drivers, and into DAW software,
within the DAW (and/or the Interface's Driver Software) that signal may or may not be processed with Computer-Native or Outboard-DSP,
then the signal is sent back out of the DAW, back through the Computer,
back into the Interface's D-to-A Chips which convert the signal from Digital back into Analog,
then (back in realtime) the 'Line-Level Analog' Audio is amplified by a Headphone Amplifier to Headphone Level,
and Finally back to the Artists Ears
The problem here is that no matter how low the Latency of that A-to-D / D-back to-A roundtrip journey, the Artist is also listening to the same source through another path: Directly from their own mouth, through the bones in their skull, into their ears. And they've had a while to get to know how that sounds.
So Acoustic Latency, and Roundtrip Latency are definitely well known issues, and there are plenty of counter measure, such as nearly Every modern Interface has a way to route the signal from their Microphone Pre-Amplifier directly to their Headphone Amplifier, in analog realtime and bypassing the roundtrip latency altogether.
But the Latency that gets forgotten, that is very similar to the Digital Roundtrip yet has existed since the dawn of Tape Based Recording, long before A-to-D & D-to-A Conversion. The Latency that went by another name, yet caused the development Two Different approaches to and categories of Recording Studio Consoles, both 'Split' and later 'Inline'.....
is Tape Delay.
Tape Delay survives today as a creative effect, and for good reason. Digital Delays just aren't the same.
But the Tape Delay that drove every Recording Engineer and Artist crazy, and is directly comparable to the Digital Roundtrip Latency "problem" of today, is the latency caused by the space between the Record Head and the Playback Head on a Recording Tape Machine. The standard issue 24-Track 2-inch Multi-Track Tape Machine in most Recording Studios had a 'Three Head' design: an Erase Head, a Record Head, and a Playback Head (later sync/re-pro).
So if a track or a dozen was 'armed', when you hit the Red Button, the Erase Head would magnetize all the iron on the tape for that track back to zero, the Record Head would lay down whatever signal you were sending it, and the Playback Head would pick that signal back up and send it back to you.....but after some delay.
For example: if the gap between the R & P heads was 1.5 inches, and the machine was run at 15IPS (inches per second), you would have a 100ms delay time. Double the speed to 30IPS, 50ms. Halve the gap between heads to .75 inches, 25ms....maybe.
But 25 milliseconds is still too much latency to perform with, so other measures were devised. Namely, all the equipment in the Recording Studio had to be DOUBLED.
There had to be two separate listening paths, one for the Artists and one for the Producer/Engineers.
Artists and Musicians needed to Hear a signal Before it went to tape, in order to 'stay in the pocket' of their performance.
Producers and Engineers needed to hear the signal After it was recorded on tape, in order to know what they got, how the overall song was building, and what else they still needed to get.
Consoles were developed that 'Split' the say 24-channel inputs on the Left half of the board for inputs that were tapped into for Artist Mixes and then sent to the 24-Track Tape Machine, and then on the Right half of the board another 24-channel strips were used for the Returns from the Tape Machine in order to build the 'Rough Mix'.
In about 1975 Harrison introduced 'In-Line' Consoles that took that same dual-path concept but put the two paths into every channel strip, so that say a 48-Channel console that fit in the above Control Room actually had 48-Inputs AND 48-Tape Returns on a 1-to-1 basis, and could add a second 24-Track tape Machine doubling everything in about the same console footprint. And at the bottom of each channel strip was often Two sets of Faders and Pans, that ran to two different Stereo Busses, which after the Recording Session during the Mixdown Session could be joined together providing 96-Channels to the 2-Track Mix (48-Tracks from Tape, plus Midi Synths & Drum Machines, multiple Parallel Processing options).
How did I get this far down the rabbit hole?
Well, many of the biggest studios here in Los Angeles at their Zenith commissioned Solid State Logic XL-9096k Inline Consoles for roughly a Million Dollars. These had 96-Channels, 48 to the left, 48 to the right. On mix down they had over 200-Channels feeding the Left & Right 2-Track. If you rolled your chair out of the 'Sweet Spot' all the way to Channel-1...all the way down to Channel-96, the journey alone could take you ten whole seconds, and how many milliseconds it would take for sound to reach you from the other side of the room...who knows. But we're back to Acoustic Latency issues.
Now, I say ALL OF THAT to say this...
I know from experience that Many Many Analog channels can be combined, In Parallel, not only in realtime, but to great effect. Ask Michael Brauer.
So what you described in the article as Analog Latency, I'm not sure exists on the Ring or Index Fingers. The Pinky, Middle, and Thumb...definitely.
However, while I've obviously thought quite a bit about these 5 Fingers or Phases of sound, as well as the Conversions between each, I haven't spent as much time contemplating the Speed / Latency implications within each stage of a sound's lifecycle.
This article combined with my pedantic nature provided me with that opportunity, so Thank You!
G.
First, thank you for doing it, hopefully others will find it useful.
Next, let me add this, I tell students to hold out their hand and look at their five fingers:
The 'Pinky' is Acoustic
The 'Ring' is Analog
The 'Middle' is Digital
The 'Index' is Analog (again, and...)
The 'Thumb' is Acoustic (again / finally)
This is the journey that sound takes from the source to the destination, from a Mouth to an Ear.
What gets forgotten is that Inside of a traditional 'Recording Studio', there are two Acoustic environments, one for the Microphones and one for the Speakers.
The 'Studio/Live Room' and the 'Control Room'.
Before any electronics are deployed, the Recording Engineer places the sources within the room because after the sound of the sources themselves, the room itself will have the greatest affect on what is being recorded. Only after that physical acoustic placement does she chose Microphones and determine Microphone Placement.
Obviously a single Vocalist with a Single-Mic is different than a Piano Stereo-Mic'd, or a Drum Kit Multi-Mic'd, or a Full Band with GoBos (go-betweens) and Many-Mics.
Once more than one Mic is deployed, then even before a single XLR cable is plugged into them, a Tape Measure should be brought in to determine the distances between each of the Sources and each of the Microphones. As pointed out in this article, these distances can be multiplied against the Speed-of-Sound and converted into Milliseconds of ACOUSTIC LATENCY.
This Acoustic Latency is best addressed by physically moving the Microphones and/or Sources themselves in order to (at least attempt to) have have each Acoustic Sound Wave arrive at each Microphone Capsule at exactly the same time. Zero Milliseconds of DIFFERENCE in the latency between them.
This can be virtually impossible of course, and there are many different techniques, but the process is similar in the physical space to what the PDC as described in the article is doing in the digital space. You could figure out the farthest Microphone Capsule in physical distance and move all other Mics to equal that same distance. So one Snare Hit would enter two or twenty Microphones at precisely the same moment.
But say in the case of Drums, you'd often have Close Mics, Overhead Mics, and Room Mics all placed a say three Different Distances from say the Snare and/or Kick. The Snare-Top & Snare-Bottom or Kick-In & Kick-Out Mics could be perfectly measured against each other (and their Polarities adjusted), and the Stereo Spaced-Pair Overhead Mics Left & Right could be perfectly measured against themselves, same with a Stereo Pair of Room Mics, but How do you physically space these Three Concentric Rings against each other? They will receive the Sound Wave at different times, or Acoustic Latencies.
Now, you plug in all the XLRs, you adjust Pre-Amps, maybe tweak EQs, maybe Gates, maybe Compressors, all "on the way in", and then you hit Record either on a Tape Machine or on a Computer attached to an Interface.
The next question: Did anything from the Microphone Capsule to the Record/Conversion incur any additional Latency. If so, wouldn't That actually be ANALOG LATENCY as opposed to the ACOUSTIC LATENCY we just discussed and as was described in this video?
I've always heard that "Analog has no Latency". That once a signal is analog, it's electrons are moving through circuits at the Speed of Light rather than the Speed of Sound. Is this true?
There is one particular latency that I'd point out that often gets forgotten in conversations about Audio Interface Recording Latency:
The problem is real and easy to understand, a Microphone is plugged into an Audio Interface's preamps (rather than an Analog Console's), then an Artist is given a pair of Headphones. They sing... and then... they hear themselves... with a performance destroying delay.
What's happening (as you obviously know but others may not) is that:
Sound leaves the Artist's Mouth,
the Microphone (Transducer) converts that Acoustic Energy into 'Mic-Level Analog' Audio,
the Interface's Preamp amplify that signal to 'Line-Level Analog' Audio (still in realtime),
but then the Interface's A-to-D Chips convert the signal from Analog to Digital Audio (at a particular Bit-Depth and Sample Rate, sat 24-Bit/96kHz),
that signal travels Digitally through the Computer, probably through Drivers, and into DAW software,
within the DAW (and/or the Interface's Driver Software) that signal may or may not be processed with Computer-Native or Outboard-DSP,
then the signal is sent back out of the DAW, back through the Computer,
back into the Interface's D-to-A Chips which convert the signal from Digital back into Analog,
then (back in realtime) the 'Line-Level Analog' Audio is amplified by a Headphone Amplifier to Headphone Level,
and Finally back to the Artists Ears
The problem here is that no matter how low the Latency of that A-to-D / D-back to-A roundtrip journey, the Artist is also listening to the same source through another path: Directly from their own mouth, through the bones in their skull, into their ears. And they've had a while to get to know how that sounds.
So Acoustic Latency, and Roundtrip Latency are definitely well known issues, and there are plenty of counter measure, such as nearly Every modern Interface has a way to route the signal from their Microphone Pre-Amplifier directly to their Headphone Amplifier, in analog realtime and bypassing the roundtrip latency altogether.
But the Latency that gets forgotten, that is very similar to the Digital Roundtrip yet has existed since the dawn of Tape Based Recording, long before A-to-D & D-to-A Conversion. The Latency that went by another name, yet caused the development Two Different approaches to and categories of Recording Studio Consoles, both 'Split' and later 'Inline'.....
is Tape Delay.
Tape Delay survives today as a creative effect, and for good reason. Digital Delays just aren't the same.
But the Tape Delay that drove every Recording Engineer and Artist crazy, and is directly comparable to the Digital Roundtrip Latency "problem" of today, is the latency caused by the space between the Record Head and the Playback Head on a Recording Tape Machine. The standard issue 24-Track 2-inch Multi-Track Tape Machine in most Recording Studios had a 'Three Head' design: an Erase Head, a Record Head, and a Playback Head (later sync/re-pro).
So if a track or a dozen was 'armed', when you hit the Red Button, the Erase Head would magnetize all the iron on the tape for that track back to zero, the Record Head would lay down whatever signal you were sending it, and the Playback Head would pick that signal back up and send it back to you.....but after some delay.
For example: if the gap between the R & P heads was 1.5 inches, and the machine was run at 15IPS (inches per second), you would have a 100ms delay time. Double the speed to 30IPS, 50ms. Halve the gap between heads to .75 inches, 25ms....maybe.
But 25 milliseconds is still too much latency to perform with, so other measures were devised. Namely, all the equipment in the Recording Studio had to be DOUBLED.
There had to be two separate listening paths, one for the Artists and one for the Producer/Engineers.
Artists and Musicians needed to Hear a signal Before it went to tape, in order to 'stay in the pocket' of their performance.
Producers and Engineers needed to hear the signal After it was recorded on tape, in order to know what they got, how the overall song was building, and what else they still needed to get.
Consoles were developed that 'Split' the say 24-channel inputs on the Left half of the board for inputs that were tapped into for Artist Mixes and then sent to the 24-Track Tape Machine, and then on the Right half of the board another 24-channel strips were used for the Returns from the Tape Machine in order to build the 'Rough Mix'.
In about 1975 Harrison introduced 'In-Line' Consoles that took that same dual-path concept but put the two paths into every channel strip, so that say a 48-Channel console that fit in the above Control Room actually had 48-Inputs AND 48-Tape Returns on a 1-to-1 basis, and could add a second 24-Track tape Machine doubling everything in about the same console footprint. And at the bottom of each channel strip was often Two sets of Faders and Pans, that ran to two different Stereo Busses, which after the Recording Session during the Mixdown Session could be joined together providing 96-Channels to the 2-Track Mix (48-Tracks from Tape, plus Midi Synths & Drum Machines, multiple Parallel Processing options).
How did I get this far down the rabbit hole?
Well, many of the biggest studios here in Los Angeles at their Zenith commissioned Solid State Logic XL-9096k Inline Consoles for roughly a Million Dollars. These had 96-Channels, 48 to the left, 48 to the right. On mix down they had over 200-Channels feeding the Left & Right 2-Track. If you rolled your chair out of the 'Sweet Spot' all the way to Channel-1...all the way down to Channel-96, the journey alone could take you ten whole seconds, and how many milliseconds it would take for sound to reach you from the other side of the room...who knows. But we're back to Acoustic Latency issues.
Now, I say ALL OF THAT to say this...
I know from experience that Many Many Analog channels can be combined, In Parallel, not only in realtime, but to great effect. Ask Michael Brauer.
So what you described in the article as Analog Latency, I'm not sure exists on the Ring or Index Fingers. The Pinky, Middle, and Thumb...definitely.
However, while I've obviously thought quite a bit about these 5 Fingers or Phases of sound, as well as the Conversions between each, I haven't spent as much time contemplating the Speed / Latency implications within each stage of a sound's lifecycle.
This article combined with my pedantic nature provided me with that opportunity, so Thank You!
G.
-
Blue Cat Audio Blue Cat Audio https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=39981 - KVRAF
- 6335 posts since 8 Sep, 2004 from Paris (France)
You are indeed right, what I called "analog" should have probably been named "acoustic", because that's what I meant. I'll probably update the article to make sure it is not confusing.
The latency in analog cables is totally negligible (speed of light vs speed of sound), unless your cable is several thousands kilometers long
The latency in analog cables is totally negligible (speed of light vs speed of sound), unless your cable is several thousands kilometers long