When did studios begin 24bit recording ?

[Eric Natural]

I'm curious about digital studio technique and when did "high-def" begin - maybe call it second generation post 16 bit ?
Another way of asking the Q is how many of my(common pop & jazz) recordings have been up-sampled ? What year did 24 bit recordings get made ?
TIA

 

[voodooless]

My little AI pal says it was somewhere around the end of the 1990s.

But many analog recordings have been remastered in 24 bit, well older than 1990s. Not that the analog stuff really benefits from it…

 

[Svperstar]

My little AI pal says it was somewhere around the end of the 1990s.

But many analog recordings have been remastered in 24 bit, well older than 1990s. Not that the analog stuff really benefits from it…

Funny this popped up on this site, I literally just had someone on Discord a few weeks ago say that some cassette tapes from the 1990s needed to be ripped in 24-bit because a tape can be "too loud" for 16-bit. Not even kidding.

Don't think that is possible. I took the 24-bit rip of these tapes, converted them to 16-bit, then compared the longest track, everything was identical.

 

[DVDdoug]

Funny this popped up on this site, I literally just had someone on Discord a few weeks ago say that some cassette tapes from the 1990s needed to be ripped in 24-bit because a tape can be "too loud" for 16-bit. Not even kidding.

Don't think that is possible. I took the 24-bit rip of these tapes, converted them to 16-bit, then compared the longest track, everything was identical.

You are right!!!

0dBFS (zero decibels full-scale) is the highest you can "count to" with a given number of bits. The "numbers" in a 24-bit file are bigger but everything is scaled to the same loudness when you play (or record).

24-bits can to quieter before it drops to zero (dead digital silence). That means more dynamic range. The extra dynamic range is on the quiet side rather than the loud side. The dynamic range of cassettes limited by noise. 16-bits has more dynamic range than just about anything analog.

24-bit also has smaller digital "steps" (higher resolution). The DAC smooths everything so there are no "steps" in the analog output anyway but there is something called quantization noise which is lower at higher resolution. You can't normally hear it with 16 or 24-bits but if you make an 8-bit file, you'll hear it like a "fuzz" on top of the audio. (If you want to know what low resolution sounds like, try an 8-bit file.)

Floating point audio has a different 0dB reference (a value of 1.0). 32-bit floating point has a virtually unlimited range somewhere around -700dB to +700dB.
There are a few floating point digital recorders and audio interfaces and they can go over 0dB without Clipping, but there are still real-world physical voltage limits and I'd guess they go to +12 or +20dB. And they are still noise limited on the quiet side.

Since you don't want the DAC to clip when you play back, the final file shouldn't go over 0dB, even if you have a format that can go over.

 

[danadam]

I'm curious about digital studio technique and when did "high-def" begin - maybe call it second generation post 16 bit ?

Apparently DECCA had a 18-bit/48 kHz system in 1979

Decca 1

Welcome to the Nagra VPR-5 web site and some other old items of media technology.

www.vtoldboys.com

Sony manufactured 24-bit recorder in 1997:

The Extraordinary Sony PCM-3348HR Digital Multitrack Recorder

The Sony PCM-3348HR is a 48-channel 24-bit/16-bit digital audio recorder, using 1/2-inch tape in DASH (Digital Audio Stationary Head) and DASH PLUS formats.

www.vintagedigital.com.au

I can also find there dcs 900 (1989): https://www.vintagedigital.com.au/dcs-900/
and Prims Sound AD-124 (1992): https://www.vintagedigital.com.au/prism-sound-ad-124/

How popular they were in recording studios (or how soon they upgraded), I don't know.

 

[voodooless]

some cassette tapes from the 1990s needed to be ripped in 24-bit because a tape can be "too loud" for 16-bit

That just makes no sense. If someone is too loud, you just turn down the gain. The number of bits play no role.

 

[Ropeburn]

You are right!!!

0dBFS (zero decibels full-scale) is the highest you can "count to" with a given number of bits. The "numbers" in a 24-bit file are bigger but everything is scaled to the same loudness when you play (or record).

24-bits can to quieter before it drops to zero (dead digital silence). That means more dynamic range. The extra dynamic range is on the quiet side rather than the loud side. The dynamic range of cassettes limited by noise. 16-bits has more dynamic range than just about anything analog.

24-bit also has smaller digital "steps" (higher resolution). The DAC smooths everything so there are no "steps" in the analog output anyway but there is something called quantization noise which is lower at higher resolution. You can't normally hear it with 16 or 24-bits but if you make an 8-bit file, you'll hear it like a "fuzz" on top of the audio. (If you want to know what low resolution sounds like, try an 8-bit file.)

Floating point audio has a different 0dB reference (a value of 1.0). 32-bit floating point has a virtually unlimited range somewhere around -700dB to +700dB.
There are a few floating point digital recorders and audio interfaces and they can go over 0dB without Clipping, but there are still real-world physical voltage limits and I'd guess they go to +12 or +20dB. And they are still noise limited on the quiet side.

Since you don't want the DAC to clip when you play back, the final file shouldn't go over 0dB, even if you have a format that can go over.

24 bit does NOT have "smaller steps" as in higher resolution per dynamic unit. The upper 16 bits are "spaced" identically to 16 bit. The additional 8 below are simply offering more range before you reach logical silence. They don't raise the resolution of the upper 16.

That's also why conversion is so quick and easy computationally. You don't need to "stretch" and interpolate values, you simply ignore the lower 8 when converting 24 to 16. A bit more complicated in practice, but that's basically it. Likewise, from 16 to 24 you just declare the original 16 the upper bits and fill the bottom 8 with zeroes.

 

[sergeauckland]

You are right!!!

0dBFS (zero decibels full-scale) is the highest you can "count to" with a given number of bits. The "numbers" in a 24-bit file are bigger but everything is scaled to the same loudness when you play (or record).

24-bits can to quieter before it drops to zero (dead digital silence). That means more dynamic range. The extra dynamic range is on the quiet side rather than the loud side. The dynamic range of cassettes limited by noise. 16-bits has more dynamic range than just about anything analog.

24-bit also has smaller digital "steps" (higher resolution). The DAC smooths everything so there are no "steps" in the analog output anyway but there is something called quantization noise which is lower at higher resolution. You can't normally hear it with 16 or 24-bits but if you make an 8-bit file, you'll hear it like a "fuzz" on top of the audio. (If you want to know what low resolution sounds like, try an 8-bit file.)

Floating point audio has a different 0dB reference (a value of 1.0). 32-bit floating point has a virtually unlimited range somewhere around -700dB to +700dB.
There are a few floating point digital recorders and audio interfaces and they can go over 0dB without Clipping, but there are still real-world physical voltage limits and I'd guess they go to +12 or +20dB. And they are still noise limited on the quiet side.

Since you don't want the DAC to clip when you play back, the final file shouldn't go over 0dB, even if you have a format that can go over.

However, with noise-shaping, even an 8 bit recording can sound surprisingly good. Those who were at the Scalford show a fair few years ago may remember Pluto's demonstration of noise shaping. We took a good 24 bit recording and reduced the bits whilst increasing noise-shaping with very little change in sound quality and those who insisted that 16 bits wasn't enough were convinced we were pulling some sort of trick. At 8 bits, the noise level was audible in the quiet bits of a recording, a bit like a cassette tape without Dolby, but louder parts were reproduced pretty much normally. switching back to 24 bits then back to 8 bits was quite a revelation. Sadly, the switching took a few seconds, so wasn't instantaneous, but nevertheless, drew a few disbelieving comments from the 16 bits isn't enough crowd.

S.

 

[voodooless]

24 bit does NOT have "smaller steps" as in higher resolution per dynamic unit.

Yes it does. You sample the same voltages with more 256x more resolution.

You don't need to "stretch" and interpolate values, you simply ignore the lower 8 when converting 24 to 16.

You should really add dither, though.

 

[Ropeburn]

Yes it does. You sample the same voltages with more 256x more resolution.

1 bit still equals 6dB, which is why I wrote "per dynamic unit". You do not sample the 96dB range 16bit is good for logically in more and finer steps in 24bit. You sample 144dB range, with identical resolution per dB.

 

[voodooless]

1 bit still equals 6dB, which is why I wrote "per dynamic unit". You do not sample the 96dB range 16bit is good for logically in more and finer steps in 24bit. You sample 144dB range, with identical resolution per dB.

You’re confusing dynamic range with resolution. PCM uses linear amplitude steps, so adding bits reduces the step size (higher resolution), which lowers quantization noise and therefore increases dynamic range.

Switching your DAC from 24 to 16 bit doesn’t make it output a lower voltage, does it?

 

[Ropeburn]

You’re confusing dynamic range with resolution. PCM uses linear amplitude steps, so adding bits reduces the step size (higher resolution), which lowers quantization noise and therefore increases dynamic range.

Switching your DAC from 24 to 16 bit doesn’t make it output a lower voltage, does it?

By raising bitdepth you raise the available dynamic range, which is precisely why you do not raise resolution within any given dynamic "window", say 0 to -96. The resolution of 24bit between 0 and -96 is identical to 16bit, the additional 8 contain information below that.

 

[voodooless]

By raising bitdepth you raise the available dynamic range, which is precisely why you do not raise resolution within any given dynamic "window", say 0 to -96. The resolution of 24bit between 0 and -96 is identical to 16bit, the additional 8 contain information below that.

Let me reiterate:

Switching your DAC from 24 to 16 bit doesn’t make it output a lower voltage, does it?

And neither on the ADC side. Switching to 24 bit mode doesn’t not unlock more full scale voltage input, does it? The same voltage is quantized in 2^16 or 2^24 possible values.

 

[RandomEar]

By raising bitdepth you raise the available dynamic range, which is precisely why you do not raise resolution within any given dynamic "window", say 0 to -96. The resolution of 24bit between 0 and -96 is identical to 16bit, the additional 8 contain information below that.

Nope. 16 bit means 65536 distinct values between full scale and zero. 24 bit means about 16.7 Mio. distinct values between full scale and zero. You gain step resolution (smaller amplitude steps) and dynamic range, because the smallest difference from zero you can encode is getting smaller with increasing bit depth.

Since "full scale" is always the same value for any bit depth, your dynamic range is only limited by the smallest value different from zero you're able to encode.

 

[DVDdoug]

1 bit still equals 6dB, which is why I wrote "per dynamic unit".

You are confusing bits with "counts".

16-bit signed integers go from -32,768 to 32,767 and 24-bit signed integers go from -8,388,608 to 8,388,607.

So, let's say we have a 16-bit 0dB peak of 32,767 and we go down one step to 32766. My handy-dandy spreadsheet says that one step is -0.00027dB.

If we take away one bit of range/resolution we can only count to 16383 and that's -6db.

One step down from 8,388,607 is -0.0000010dB (a smaller dB step and a smaller voltage step).


...That's all "best case" near 0dB. At lower volumes the amplitude (or voltage) steps are the same but the calculated as dB is bigger (because it's a logarithmic ratio).

 

[Robin L]

My little AI pal says it was somewhere around the end of the 1990s.

But many analog recordings have been remastered in 24 bit, well older than 1990s. Not that the analog stuff really benefits from it…

I'd say analog stuff really does benefit. De-noising and otherwise digitally massaging older recordings is better when playing with 24 bits. There's a lot more headroom.

 

[tineared]

FWIW, Santana's The Swing of Delight was digitally recorded in 1980. It was a jazz fusion album featuring Herbie Hancock, Wayne Shorter, Ron Carter and Tony Williams
It was my first digital recording on vinyl

 

[DVDdoug]

De-noising and otherwise digitally massaging older recordings is better when playing with 24 bits. There's a lot more headroom.

Only if you lower the volume. With integer formats there is no headroom over 0dB. ...You could say there is more "footroom"

Or you CAN record/digitize lower to leave more headroom. Pros often record at -12 to -18dB, leaving plenty of headroom for unexpected peaks.

The good news is that most processing and mixing, etc., is done in 32-bit floating point, which for audio purposes is infinite headroom (or sometimes in 64-bit floating point). Audacity works internally at 32-bit float and REAPER works in 64-bit float

 

[Robin L]

Only if you lower the volume. With integer formats there is no headroom over 0dB. ...You could say there is more "footroom"

Or you CAN record/digitize lower to leave more headroom. Pros often record at -12 to -18dB, leaving plenty of headroom for unexpected peaks.

The good news is that most processing and mixing, etc., is done in 32-bit floating point, which for audio purposes is infinite headroom (or sometimes in 64-bit floating point). Audacity works internally at 32-bit float and REAPER works in 64-bit float

My bad. Yes, more "footroom might be more like it. Recently I recorded some children's assemblies off the mixer. Didn't really great fidelity but having all that headroom meant I could easily compress the result later. Usually recorded around - 12 or so.

 

[kemmler3D]

Another way to look at bit-depth for those heads that are spinning on "dynamic range" vs. "resolution"... it's kind of two ways of saying the same thing when it comes to digital.

Take an analog signal and digitize it. For now don't worry about how many bits are used. You only have so many "steps" to work with as you create samples, so there will be a random difference (error) between the original value and the digitized value, for each sample. You take the number that is closest to the real value, but the analog signal will rarely land exactly on the equivalent digital value.

You can now think of the digitized signal as the original analog value, plus a random error signal.

That should sound a bit familiar ... a constantly varying random signal? We call this noise. And in fact, that's what you get when you digitize something. And that's actually how it sounds when you listen to a 44.1khz+ signal but at a very low bit depth - just a really noisy recording.

If you only have a few steps, you get a ton of noise because you can't approximate the original signal very well. But what's really interesting is the original audio is pretty recognizable even if you only have like, 2-3 bits. 8 bits sounds a lot better than you'd expect and 12 bits is a little alarmingly good compared to what you'd guess.