How much of a difference does a high-end audio interface really make?

[mixyguy2]

That's not true at all.

We'll agree to disagree...

I've seen these discussions go on ad nauseam on other sites like Gearslutz and the "more is better" people are never changing, so exiting this one. To each their own!

[VELLTONE MUSIC]

I guess there is reason most devices to work with standard 24/48 ...same as speed limit inside the city...less noise,less problems.
It's a frame and compromise ratio of size/quality - mix made with internal 32bit/96khz process,exported to 32bit float is super vivid and dynamic,but 10-20 times bigger,then same mix exported to 24bit/48khz loss a lot of dynamics and color,but small size...why streaming platforms even use mp3 format,is it because of the quality???
Audible difference even via 24/96 sound card,i just wonder how much better will be the listening experience, if i can listen my 32bit float mix,the way it is inside the daw,without converting to low 24 bit resolution and loosing all the beauty i work to create.
In the beginning of third millennium - 50-60 years old stereo format,which i'm pretty sure somebody during 90's said let's try 24/48 cinematic format,16/41 isn't good enough ,now again 32bit/96khz is debatable...that's not serious for discussion.
Personal experience,not math data and theories.
Cheers

[stoopicus]

That's not true at all.

We'll agree to disagree...

I've seen these discussions go on ad nauseam on other sites like Gearslutz and the "more is better" people are never changing, so exiting this one. To each their own!

I agree completely with you in terms of recorded instruments in the absolute sense that going higher will not make the recording sound better.

I do think audio rate matters in terms of ensuring the Nyquist frequency is as far above 20kHz as possible. However oversampling can achieve this almost as effectively as a higher sampling rate.

Despite having a 96kHz interface in my mixer I still use 48 myself - but at the same time I do recognize that it can, in fact, matter.

[whyterabbyt]

mix made with internal 32bit/96khz process,exported to 32bit float is super vivid and dynamic,but 10-20 times bigger,then same mix exported to 24bit/48khz

Your maths is suspect. (As is the notion that the ratio might be variable across different files, as you seem to be implying)

If you go from a 24 bit word to a 32-bit word, and change no other factor, your storage requirements are increased by a third (32/24 = 1.33)
If you go from a 48Khz sample rate to a 96Khz sample rate, and change no other factor, your storage requirements are doubled (96/48 = 2)
Combine those, and your storage requirement is multiplied by 2.6666

A recording which would need 10Mb of space when recorded at 24b/48KHz will take up slightly less than 26.7Mb when recorded at 32b/96KHz

So no, not '10-20 times bigger', not even close, they're not even 3 times bigger.

You'd only just pass 10 times bigger (10.667) if you recorded at 64 bit/192Khz or 32bit/384Khz.

[Ploki]

I guess there is reason most devices to work with standard 24/48 ...same as speed limit inside the city...less noise,less problems.
It's a frame and compromise ratio of size/quality - mix made with internal 32bit/96khz process,exported to 32bit float is super vivid and dynamic,but 10-20 times bigger,then same mix exported to 24bit/48khz loss a lot of dynamics and color,but small size...why streaming platforms even use mp3 format,is it because of the quality???
Audible difference even via 24/96 sound card,i just wonder how much better will be the listening experience, if i can listen my 32bit float mix,the way it is inside the daw,without converting to low 24 bit resolution and loosing all the beauty i work to create.
In the beginning of third millennium - 50-60 years old stereo format,which i'm pretty sure somebody during 90's said let's try 24/48 cinematic format,16/41 isn't good enough ,now again 32bit/96khz is debatable...that's not serious for discussion.
Personal experience,not math data and theories.
Cheers

As whyterabbyt pointed out, your math is off.
But so is your history. 48k was a thing in pro audio by the end of the 80s already.

Sampling rates were chosen for practical reasons tied to technical limitations of mostly video gear of that era.

In any case, i explained why 32bit FP isn’t any more “detailed” (even if you believe bit depth means detail) but you choose to ignore it so i’m not sure what’s the point here

[Sonic Illusions]

Babies are born with bat ears already

[koalaboy]

Not sure if you're serious, but 24bit is fine for studio.
it has 144dB of dynamic range, and analog paths are limited by physics to about 120dB, for now.
Recording 32bit would mean you would 48dB of thermal noise to recordings.

32bit floating point is effectively 23bit mantissa, 1bit sign, and 8bit exponent. Effective resolution is 24bit, scaled. Single sample cannot have more than 24bits of resolution with 32bit FP.

Honestly, I'll believe the folks that build the hardware: https://www.sounddevices.com/32-bit-flo ... explained/

So for 33% more storage space compared to 24-bit files, the dynamic range captured goes from 144 dB up to, essentially, infinite (over 1500 dB). But more importantly, audio signals above 0 dBFS are preserved in the file, rendering clipped audio a thing of the past.

Now perhaps this is all 'semantics', but the reality is that 32bit FP is extremely useful in recording (and mixing) because it removes analog limitations. Part of the whole reason for moving ITB is to allow greater flexibility.

[Ploki]

Not sure if you're serious, but 24bit is fine for studio.
it has 144dB of dynamic range, and analog paths are limited by physics to about 120dB, for now.
Recording 32bit would mean you would 48dB of thermal noise to recordings.

32bit floating point is effectively 23bit mantissa, 1bit sign, and 8bit exponent. Effective resolution is 24bit, scaled. Single sample cannot have more than 24bits of resolution with 32bit FP.

Honestly, I'll believe the folks that build the hardware: https://www.sounddevices.com/32-bit-flo ... explained/

So for 33% more storage space compared to 24-bit files, the dynamic range captured goes from 144 dB up to, essentially, infinite (over 1500 dB). But more importantly, audio signals above 0 dBFS are preserved in the file, rendering clipped audio a thing of the past.

Now perhaps this is all 'semantics', but the reality is that 32bit FP is extremely useful in recording (and mixing) because it removes analog limitations. Part of the whole reason for moving ITB is to allow greater flexibility.

SoundDevices makes field recording devices, designed to be used on location recording for (usually) film/tv/media.
I said 24bit is fine for studio.
32bit is useful when location recording because you have less predictable environment, but it doesn't come free anyway, it's basically several parallel preamp stages+ADCs that the system switches between when it detects a clip, then combines everything in a single 32bit FP file. Noise floor bounces around everytime that happens. It's far from perfect and it's a contingency.
You don't want noise floor and preamps doing some auto-shit in the studio, because it's unnecessary and more of a hinderance than practically useful.

You keep on believing, I'll keep on learning.

I agree completely with you in terms of recorded instruments in the absolute sense that going higher will not make the recording sound better.

I do think audio rate matters in terms of ensuring the Nyquist frequency is as far above 20kHz as possible. However oversampling can achieve this almost as effectively as a higher sampling rate.

Despite having a 96kHz interface in my mixer I still use 48 myself - but at the same time I do recognize that it can, in fact, matter.

Why "almost" as effectively? what's less effective with oversampling?'

I use the + to mean the more recent UFXII hardware that had similar upgrades to the UFX+ and UFXIII from what I recall. You're correct that it's not an official SKU, but there are hardware differences.

Recent UFX (II and III) use different chips than UFX+ and early UFX II, because something happened with availability of AKM chips. It wasn't an "upgrade", it was a "sidegrade" because they were forced to replace the chips. It wasn't a choice.

[VELLTONE MUSIC]

I guess there is reason most devices to work with standard 24/48 ...same as speed limit inside the city...less noise,less problems.
It's a frame and compromise ratio of size/quality - mix made with internal 32bit/96khz process,exported to 32bit float is super vivid and dynamic,but 10-20 times bigger,then same mix exported to 24bit/48khz loss a lot of dynamics and color,but small size...why streaming platforms even use mp3 format,is it because of the quality???
Audible difference even via 24/96 sound card,i just wonder how much better will be the listening experience, if i can listen my 32bit float mix,the way it is inside the daw,without converting to low 24 bit resolution and loosing all the beauty i work to create.
In the beginning of third millennium - 50-60 years old stereo format,which i'm pretty sure somebody during 90's said let's try 24/48 cinematic format,16/41 isn't good enough ,now again 32bit/96khz is debatable...that's not serious for discussion.
Personal experience,not math data and theories.
Cheers

As whyterabbyt pointed out, your math is off.
But so is your history. 48k was a thing in pro audio by the end of the 80s already.

Sampling rates were chosen for practical reasons tied to technical limitations of mostly video gear of that era.

In any case, i explained why 32bit FP isn’t any more “detailed” (even if you believe bit depth means detail) but you choose to ignore it so i’m not sure what’s the point here

You mistaken 32bit fixed point and float point i guess.
Fixed point is small size.
Float point takes huge amount of space - 3-4 min audio is around 400-500 mb.
Export something and you'll see.
I have basic internet education and understanding of digital stuff,just have enormous time of listening and shaping the sound,so i care about the difference of sound quality and i was very nicely surprised when heard same track exported as 32bit float
I use 32bit/96khz inside the cakewalk,just because this is maximum my sound card allow and used to export to 24/48.
The curiosity makes me export in different formats and was really amazed how audible is the difference.
Now the audiophile in me can't rid of that feeling,that 24bit is too much of compromise compared to 32bit float...:):)
It is an important topic to me,so yes high end interface,format and player matters...
The debate now is like i am talking about a song i like a lot and you telling me,this song is not good at all
From internet:
32-bit fixed point is fundamentally different from 32-bit floating point because fixed-point numbers always have the decimal point in the same position, limiting their dynamic range, while floating-point numbers use an exponent to position the decimal dynamically, allowing for a much wider range of values and greater flexibility, especially when dealing with large dynamic ranges in audio and other applications.
Cheers

[BertKoor]

You mistaken 32bit fixed point and float point i guess.
Fixed point is small size.
Float point takes huge amount of space - 3-4 min audio is around 400-500 mb.
Export something and you'll see.

No, that can't be right. The size of 32bit integers and 32bit floats is exactly the same: 32 bits, which is four bytes.

A 32bit float is in essence a 24bit number with 8 bits to let the point float.

No analog gear goes beyond what can be captured by a 20 bit integer.

But, whatever. Bigger must be better, I guess. Until enough is enough.

[vurt]

I have basic internet education and understanding of digital stuff,

then maybe listen to the people with actual qualifications and jobs in the field

[Sonic Illusions]

3 minutes, 32-bit, 512 kHz, stereo:
Uncompressed (WAV, AIFF etc.)Bit Rate: 32768 kbps File Size – Decimal (1kB = 1000 bytes):
737.28 MB (megabytes) File Size – Binary (1kB = 1024 bytes): 703.125 MiB (mebibytes)

[Etienne1973]

From internet:
32-bit fixed point is fundamentally different from 32-bit floating point because fixed-point numbers always have the decimal point in the same position, limiting their dynamic range, while floating-point numbers use an exponent to position the decimal dynamically, allowing for a much wider range of values and greater flexibility, especially when dealing with large dynamic ranges in audio and other applications.
Cheers

With my human ear, I can no longer detect any difference in dynamics between 24-bit and higher bit resolution audio. At most, I notice a quieter noise floor during signal pauses audio >24-bit.

[stoopicus]

I agree completely with you in terms of recorded instruments in the absolute sense that going higher will not make the recording sound better.

I do think audio rate matters in terms of ensuring the Nyquist frequency is as far above 20kHz as possible. However oversampling can achieve this almost as effectively as a higher sampling rate.

Despite having a 96kHz interface in my mixer I still use 48 myself - but at the same time I do recognize that it can, in fact, matter.

Why "almost" as effectively? what's less effective with oversampling?'

With oversampling - my understanding anyway, I haven't actually implemented it myself, but I have studied it (Pirkle, Oppenheim & Schafer, etc) - basically you either pad the samples out by alternating with zeros or interpolated values for the added samples, apply the effect and the steep antialiasing filter, and then downsample back.

With a higher sample rate, you have actual samples there and not interpolated ones and get to skip the additional filtering as well (and resulting phase effects, etc).

Like I said, it's a slight difference.

[Ploki]

With oversampling - my understanding anyway, I haven't actually implemented it myself, but I have studied it (Pirkle, Oppenheim & Schafer, etc) - basically you either pad the samples out by alternating with zeros or interpolated values for the added samples, apply the effect and the steep antialiasing filter, and then downsample back.

With a higher sample rate, you have actual samples there and not interpolated ones and get to skip the additional filtering as well (and resulting phase effects, etc).

Like I said, it's a slight difference.

you only get to skip filtering if the delivery format is 88.2/96k, else, you will, at one point have to apply filtering.
I'm a proponent of DAW-level oversampling so you only need to filter once per channel strip for example, not every plugin consecutively. But then again, you only ever oversample compressors and saturators so a lot of overhead for little benefit

So, if "actual samples" are more or less noise, since there's rarely any useful signal above 20k, what's the effective difference between interpolated and recorded samples?