Bit depth question

[Max Arwood]

I'm talking 44K/32Bit

Since I have a bunch of frozen soft synth tracks, I was wondering about the bounce bit depth.  The frozen tracks are already @32bit according to what I have read. Wouldn't it be almost as fast to bounce all to 32, because all frozen tracks are already 32bit

This way the synth and other frozen tracks are merged and copied, no bit recomputation. I can't hear the difference, but I'm sure there are micro differences.

Any ideas?

[OutrageProductions]

Just remember that Sample Rate relates to frequency response and Bit Depth relates to dynamic range. Which one matters the most to you?

Most professionals and streaming/distribution services start from and derive their content from 48k/24b, so that is now the defacto standard. 

CD quality at 44.1k/16b is now passe.

[Will.]

2 hours ago, OutrageProductions said:

Most professionals and streaming/distribution services start from and derive their content from 48k/24b, so that is now the defacto standard. 

You'd be shocked at how in correct this is. Its what they want you to think. 

[John Vere]

Why would anyone use 44.1/16 ?  It’s a total holdover from the CD era.
As @OutrageProductionshas said everything in my world is now at 48/24 or 48/32. My new interface records at 32 and believe the rest will follow soon. 

I just uploaded 14 songs to a distributor DistroKid and they recommended 48/24 as minimum.  What the streaming services do with that is beyond my control but as we all know it’s how things are now done. 
I listened to them on Spotify, Apple, Deezer YouTube etc and they all sound the same to me on studio headphones.  If I can’t  hear any degrading then what else matters.  
 

There’s zero reason for 44.1 to even exist anymore. If I want a CD my burner converts the file  automatically.  

 

[Michael Richards]

is it higher or lower? I can’t see why any professional would go lower. I could be wrong.

 

[Michael Richards]

I agree, no reason for 44.1.

[msmcleod]

Just to give some perspective here.

The highest frequency a human can hear is around 20Khz if you're 21 or younger.  As you get older, this continues to drop.  I suspect the majority of users on this forum will struggle to hear anything over 15Khz.

44.1Hkz can reproduce frequencies up to 22Khz (so beyond human hearing). 16bit audio has a dynamic range of 96db across all of its frequencies. 

Compare this to tape: a maximum dynamic range 55db - with it's optimal  frequency reproduction between 120Hz and 800Khz (it starts to drop off either end of these frequencies).

Vinyl has a maximum dynamic range of 60db; frequency range is around 7hz to 50Khz (so better than a 96Khz sample rate).

However as far as dynamic range is concerned, vinyl and tape aren't even a match for 12 bit audio (72db), which a bunch of budget synths/samplers used in the 80's/90's.

Given that most tracks are of a single instrument, it's likely that the required dynamic range recorded on a single track is way less than 96db. 

Where things start to go wrong is when you mix in 16 bit - then you're dynamic range is significantly reduced.   However NO modern DAW mixes in 16 bit.  They all use at least 32 bit float or 64 bit float for mixing;  only going back to 16bit at the stereo master bus.

The advantage of recording in 24bit is you can record really quiet signals with no loss of dynamic range, and no perceived increase in noise. 

However, as long as you're recording signals at a decent level, there should be no issues using 44.1Khz/16bit for recording.... I mean, it's absolutely fine on CD.

Comparing 32 bit to 24 bit is problematic - 24 bit is an integer format;  32 bit / 64 bit are floating point formats.   Although you can store higher numbers in floating point (which is why it's great for mixing), you do lose precision at the extremes.

For a mixed down song, a 24 bit integer file and a 32 bit floating point file are equivalent in quality.

[Max Arwood]

I was thinking that frozen tracks are already 32 bit why not keep them at this resolution to the end? Will it bounce as fast not having to do the extra conversion? Or, does it convert to 32 or 64 bit and then mix the track signals to the final bus in that bit depth anyway, and then reconvert to the final bit depth? 

[Will.]

9 hours ago, msmcleod said:

Just to give some perspective here.

The highest frequency a human can hear is around 20Khz if you're 21 or younger.  As you get older, this continues to drop.  I suspect the majority of users on this forum will struggle to hear anything over 15Khz.

44.1Hkz can reproduce frequencies up to 22Khz (so beyond human hearing). 16bit audio has a dynamic range of 96db across all of its frequencies. 

Compare this to tape: a maximum dynamic range 55db - with it's optimal  frequency reproduction between 120Hz and 800Khz (it starts to drop off either end of these frequencies).

Vinyl has a maximum dynamic range of 60db; frequency range is around 7hz to 50Khz (so better than a 96Khz sample rate).

However as far as dynamic range is concerned, vinyl and tape aren't even a match for 12 bit audio (72db), which a bunch of budget synths/samplers used in the 80's/90's.

Given that most tracks are of a single instrument, it's likely that the required dynamic range recorded on a single track is way less than 96db. 

Where things start to go wrong is when you mix in 16 bit - then you're dynamic range is significantly reduced.   However NO modern DAW mixes in 16 bit.  They all use at least 32 bit float or 64 bit float for mixing;  only going back to 16bit at the stereo master bus.

The advantage of recording in 24bit is you can record really quiet signals with no loss of dynamic range, and no perceived increase in noise. 

However, as long as you're recording signals at a decent level, there should be no issues using 44.1Khz/16bit for recording.... I mean, it's absolutely fine on CD.

Comparing 32 bit to 24 bit is problematic - 24 bit is an integer format;  32 bit / 64 bit are floating point formats.   Although you can store higher numbers in floating point (which is why it's great for mixing), you do lose precision at the extremes.

For a mixed down song, a 24 bit integer file and a 32 bit floating point file are equivalent in quality.

Exactly 441/24 you will find most "professionals" use. I have first hand experience seeing this almost every second month and when I travel. When you ask why, the simply respond, its unnecessary. If 441/24/32 removes the noise floor in the digital universe, why should they record at 48/24? That the geniune answer you get. Spotify, Deeza, Itunes these platform dont care about noise floor, but only the K-system.

[Will.]

13 hours ago, John Vere said:

Why would anyone use 44.1/16 ?  It’s a total holdover from the CD era.
As @OutrageProductionshas said everything in my world is now at 48/24 or 48/32. My new interface records at 32 and believe the rest will follow soon. 

 

Why would you want to record at the bit rate and deph with built in analog to digital (AD) Converters lowering the noise floor for you? 

If you can record in 441/ 24 or 441/32 with the exact same recording as 48/24 why would you want to waste precious system resources? 

Keep in mind, recoring at 48/24 does not give you better quality then 441/24. What bring you the quality is your preamps of your interface.  Higher bit only creates higher headroom. 

The only reason and times "pro4fessional" studios use 48/24 is when they record an orchestra. I've got first hand experience with this and it is only because at 48/24 there is a higher headroom and lower noise floor within the digital universe - not because its better quality. Theres only a handful of studios that uses actual true analog gear today. So far, every high end and big name studio I have been to uses AD converters.

With my last work abroad, I asked the artist why are you renting a studio? He said: they've got "better equipment." We got to this studio and I immediately laughed in side. After we were done, I showed him the AD converters and the biggest USB hub I have ever seen. Most gear had a Midi to Usb converter or cable. Guess what happens in that setup - lower floor noise because of the analog to digital converters.

If you dont do music for movies that uses surround sound files, then recording at 48/24 seems a little unnecessary for everyday use. 

Edited October 28, 2023 by Will.

[John Vere]

Ok it’s more about being compatible on my system. I haven’t tried it yet with the Zoom but my Motu would simply crash when changing clock rates. So 48 solved the problem. 
When creating Tutorials I have to be able to have Cakewalk running and producing audio. Then Win Amp also playing Audio. Then the screen capture OBS recording both sources. As well as I might often have to use the internet to quickly grab a picture or web page screen shot. 
At 44.1 this is a train wreck. 
At 48 it all works.  
 

Back to the OP. I do all bouncing and stem exports at 32 with no dithering.  
Most audio interfaces are set at 24 and can’t be changed. For most people the majority of audio is recorded at 24. 

My new Zoom L8 can record at 32 which is a new thing in consumer gear as far as I can tell. I’m going to be using it from now on for my own original music because it will mean no dithering will be required at any step. 
If you read up on dithering it is yet another rabbit hole and you can obsess about it just like clock rate and bit depth.   It is only required when reducing bit depth so 32 solves this  little issue. 

As far as resources, storage is dirt cheap now. It’s the last thing I worry about , but that said I would never take the jump to 96/64.? This is not only for resources , but it is definitely outside the scope of reasoning in my work. 
I feel 48 is a happy balance in this day and age. I’ll leave 96 for those who think it matters. 

Edited October 28, 2023 by John Vere

[SloHand Solo]

If I'm using vst instruments only (no live recording), does the sample rate and frequency have any impact?  

I can go up to 64bit, but that seems like overkill.  So, 24bit or 32bit?

[John Vere]

VST instruments are audio tracks. So same thing. Lots of people upsample and claim it makes a huge difference.  

[azslow3]

10 hours ago, Will. said:

Why would you want to record at the bit rate and deph with built in analog to digital (AD) Converters lowering the noise floor for you? 

If you can record in 441/ 24 or 441/32 with the exact same recording as 48/24 why would you want to waste precious system resources? 

Keep in mind, recoring at 48/24 does not give you better quality then 441/24. What bring you the quality is your preamps of your interface.  Higher bit only creates higher headroom.

Many (most?) interfaces can record with lower latency under 48kHz (the number of extra samples is normally the same, but the sample "length" is smaller). Some interfaces can go even lower with 96kHz, but that is less common.

Up-sampling to 96kHz is "strait forward" (and almost all plug-ins are at least tested with 96, since that is "common" rate, unlike 88.2...).

3 hours ago, John Vere said:

My new Zoom L8 can record at 32 which is a new thing in consumer gear as far as I can tell. I’m going to be using it from now on for my own original music because it will mean no dithering will be required at any step.

In fact the opposite is true for (at least current) 32bit recording.

32bit was introduced to avoid gain staging, great for Field Recorders (when the level is unpredictable) and other similar situations (no wonder Zoom, as major player in field recorders, has adopted the technology). Note that fidelity is not top priority for field recorders...
What and exactly how they do that is "proprietary", but from all sources they are not using true 32bit ADC (all ADCs are for physical reasons fixed point), the output 32bit floating point is somehow composed from outputs of 2 different ADCs (in some devices they are 16bits...). Both ADCs are not used with "optimal" gain, the target is avoid digital clipping even in case an airplane will fly right over the microphone,  not to retain best fidelity.
And so there is quite some "math" in combining the result, which is in any case way more significant then any "dithering".

BTW when converting 24bit Fixed-point to 32bit Floating-point there is no dithering, the second has the same 24bit precision as the first one.

Note that the consequence of the last statement: saving the input from 24bit interface into 32bit file is just wasting space (extra 8bits in each sample will be constant).

45 minutes ago, John Vere said:

VST instruments are audio tracks. So same thing. Lots of people upsample and claim it makes a huge difference.  

It CAN make audible difference, VST dependent. That was discussed many times in this (and old Cakewalk) forum. With practical examples.

The reason is again pure technical, some mathematical algorithms used for audio processing or generation "produce" frequencies outside 20kHz. Even when working with a stream of samples which can't have such frequencies. So resulting "numbers" land somewhere in range of allowed numbers, generating aliased frequencies.
"Properly written" plug-ins are aware, they up-sample before and down-sample (with LPF) after such (or total) processing. But not all plug-ins are "properly written" (even so some of them are well known as "good").

But a difference between 48kHz and 96kHz in a file (physiologically) can't be spotted by a human. Such claims exist in the Internet, normally in High-end equipment forums/ads/blogs, sometimes even with tests... Note that hardware (including analog) is also a subject of "not properly made". So any such component in High-end chain will produce aliases in lower (audible) range. Such aliasing is so subtle, that it is not only declared as "the difference", but can be perceived as "a feature" (even so in reality that is just "a bug"...)  

---

So:

• record into 24bit (44.1/48/96kHz, depending from own preferences and required side-effects, 48kHz is a good balance).
• process in 64bit. 32bits are sufficient, but have 24bit precision. Every FX in a chain can "destroy" the last bit. After huge number of processing passes, that can "leak" into the result. So, practically that is impossible to perceive, but theoretically it is, and modern computers have power ?... Plug-ins are likely processing in 64bits in any case.
• up-sample for "buggy" plug-ins (or in general, for safety, if you have sufficient resources...)
• render/bounce/freeze into 32bit float. Only for the final result (so after maximizer) 24bit is a reasonable option. 64bits is an overkill for persistency
• do not bother about "dithering" options for anything above 16bit format, no equipment will be able to reproduce that (till you save intermediate files into 24bit, which as I have written is a bad idea).
• especially when working in 96kHz, put LPF before output to the interface (till you knows what your gear does with "High-end" frequencies...).

[JnTuneTech]

Not sure if you've noticed, but threads on this subject go on & on. In technical theory, what happens is well outlined by @msmcleod in this thread.

4 hours ago, John Vere said:

Ok it’s more about being compatible on my system. I haven’t tried it yet with the Zoom but my Motu would simply crash when changing clock rates. So 48 solved the problem. 

One example - and imagine how many variations there are with all the possible interfaces available, and to come. -Sadly, we never know until we try them specifically in practice.

4 hours ago, John Vere said:

Most audio interfaces are set at 24 and can’t be changed. For most people the majority of audio is recorded at 24. 

My new Zoom L8 can record at 32 which is a new thing in consumer gear as far as I can tell. I’m going to be using it from now on for my own original music because it will mean no dithering will be required at any step. 
If you read up on dithering it is yet another rabbit hole and you can obsess about it just like clock rate and bit depth.   It is only required when reducing bit depth so 32 solves this  little issue.

This also varies by many, many factors. The drivers for the various interfaces all handle the same connection & data conversion differently, and when you add that to the differences in hardware capability, using digital theory to predict it becomes troublesome. For instance, I would have to hear some identical recordings made with identical outboard gear & sonics to see if your 32-bit rated interface & driver actually produce what I might prefer over a 24-bit rated one in the end. -Avoiding any dithering may indeed be a good goal though.

The underlying conversions & best match choices in VST instruments is another of those rabbit holes. About all we can do is try each one, and find the best results with our ears, and of course monitoring system performance & graphs can help - sometimes.  -And the discussions will continue...  But it's sometimes interesting to hear some of the experiences from other folks out there. -Might even help me with the next audio interface choice I make, -who knows!

Edited October 28, 2023 by JnTuneTech
clarifications...

[reginaldStjohn]

I was confused about people claiming some interfaces recording at 32Bit since I have never seen an A/D manufacture that actually produces a 32 bit ADC at audio Sampling rates.

From Zoom's own website it says the Zoom L8 is a 24 bit interface.  If it allows you to record in 32 bit then it is converting t he 24 bits to some 32 bit format in the driver.  That is basically the same thing that all DAWs do . Am I missing something?

Edited October 28, 2023 by reginaldStjohn

[mettelus]

40 minutes ago, azslow3 said:

render/bounce/freeze into 32bit float. Only for the final result (so after maximizer) 24bit is a reasonable option. 64bits is an overkill for persistency

Poor OP is getting inundated with the perpetual discussion/debate, when this was all Max was asking. The practical side of this (beyond carrying precision through calculations) is that a signal cannot be digitally clipped inside the DAW. The downside is that a rogue plugin can spike pretty deep into the red on you, which is why a limiter on the Master bus is always prudent to protect your hearing.

*****

This is OFF TOPIC and more for John:

5 hours ago, John Vere said:

Ok it’s more about being compatible on my system. I haven’t tried it yet with the Zoom but my Motu would simply crash when changing clock rates. So 48 solved the problem. 
When creating Tutorials I have to be able to have Cakewalk running and producing audio. Then Win Amp also playing Audio. Then the screen capture OBS recording both sources. As well as I might often have to use the internet to quickly grab a picture or web page screen shot. 
At 44.1 this is a train wreck. 
At 48 it all works.  

I was testing out the PreSonus Revelator (condenser version), and got a pretty pronounced latency at one point (more annoying, but usable) from inadvertantly having the mic and CbB at different bit rates. I chuckled when I found that and thought that has to be one of the more impressive things I have seen in a long while... that mic was working like crazy to "trick" everything it was connected to to think it was working in their native bit rates. Even the CbB preferences window took like 30 seconds to open (with everything still connected), but PreSonus had no issues playing the game.

DISCLAIMER here... I would not recommend the condenser version to someone new (especially for recording voice with environmental noise and sibilants concerns), but have not gotten to test the dynamic version (yet). That mic is actually its own audio interface with its own ASIO (and OBS) drivers and counts as Dolby Atmos hardware.

[John Vere]

@reginaldStjohn The Zoom L8 is a bit confusing because it is 3 different devices. The manual is not exactly great and I had to test things myself to get answers. 
I will send an Email to  the company and ask just that because you have presented a good question. 
It wasn’t that long ago when I read that most consumer A/D converters are all 24 bit and therefore the reason most Audio Interfaces are only just that. 
To me it would be very misleading to say you are recording at 32 bit via a 24 bit A/D. ( or even16!!!)  Zoom was never the top of my list. So it wouldn’t surprise me. 

 If all it ever does is be my live mixer it was still the best option for that . Not to mention being able to record in multi track with out a computer. 
As an audio interface so far it’s outperforming my Motu M4.
I am totally aware of the limitations of using gear in these lower price ranges. I simply have never had the budget for the real stuff.  But the bang for the buck factor has improved ten fold or more since my first set up in the 80’s. Long live the bottom of the gear food chain. 

 

Edit Oct 29th. I now see that nowhere does Zoom claim that it is a 32 Bit device. A visit to the web site and looking at the specs for all of their Audio interfaces I saw only the UAC 232 ( see next post ) is listed as 32 bit floating. 

 

 

 

Edited October 29, 2023 by John Vere
Add update comment

[azslow3]

10 hours ago, reginaldStjohn said:

I was confused about people claiming some interfaces recording at 32Bit since I have never seen an A/D manufacture that actually produces a 32 bit ADC at audio Sampling rates.

From Zoom's own website it says the Zoom L8 is a 24 bit interface.  If it allows you to record in 32 bit then it is converting t he 24 bits to some 32 bit format in the driver.  That is basically the same thing that all DAWs do . Am I missing something?

In fact it seems like L8 is not 32bit recorder (unlike F2/3/6/8, M2/3/4 and UAC-232). So saving into 32bit from L8 make no sense...

[Glenn Stanton]

all it means is if you have a 24-bit IO and record at 32-bit - your lower LSByte is all zeros. when you add processing in 32-bit effects, then those bits may likely become populated. so it's just s wasted space if you only use 24-bit IO and plug-ins. and of course trying to playback 32-bit on 24-bit IO (or apps) means you're truncating the LSByte which may produce noise...