32bit floating softens sound compared to fixed 24/32/48 bit

musurgio · Post by **musurgio** » Wed Sep 15, 2004 4:30 pm

What I have concluded after experimenting listening and comparing UAD-1 (32bit floating) Powercore (48bit fixed) Pulsar (32bit fixed) is that using Cubase 32bit floating too that floating alters the transient of the sound.
If you use too many UAd_1 COMPRESSORS ON YOUR MIX THEN YOU GET A MIX WITHOUT ANY PUNCH ! as opposed using Powercore's cmpressors.
Even 32bit floating recording onto Cubase reveals same defficiencies as opposed with using 24bit fixed recording.
Thats why I conclude Sawstudio has the best sound becuase it uses 24bit engine.
thats why VST24 has a better sound than Cubase 5.1 and SX.
Thats why SX sounds soft and nice...
Any thoughts ???
Regards,
Dimitrios

astroman · Post by **astroman** » Wed Sep 15, 2004 5:03 pm

there have been world class recordings made with 16 bit Protools...

imho you're completely on the wrong track - it just doesn't work that way.
Either you have a good processing or a bad one, regardless of the resolution.

It's a matter of fact that a high bit count can (and does) provide an enormous headroom which makes life easier when dealing with lots of tracks (say 48 and above), but the number format itself has no influence on the sound

It's all and entirely in the processing and each system you mention runs a different version of math library.

It's those differences that (may) affect the sound and of course the knowledge (and preferences) of each programmer.
Not a single product exists that was implemented by the same engineer in all 3 or 4 versions...

cheers, Tom

Post by **hubird** » Wed Sep 15, 2004 6:50 pm

Madonna's previous album was recorded in Paris with an old AudioWerk8 card, 16 bit.
The sound was reviewed as superb.

Stop the technicions.
Stop the bit hunters.
Stop the measurements.
Stop knowing better than what the leading soft houses offer.
Stop believing nonsense.
Stop stopping making music.
Get back to work.
Make a song.
Make music.
Eat food, not bits.
Think music, make art.
Hell, make MUSIIIIIIIIC.

Ah...!

[ This Message was edited by: hubird on 2004-09-15 19:55 ]

Liquid Len · Post by **Liquid Len** » Wed Sep 15, 2004 9:44 pm

On 2004-09-15 17:30, musurgio wrote:
What I have concluded after experimenting listening and comparing UAD-1 (32bit floating) Powercore (48bit fixed) Pulsar (32bit fixed) is that using Cubase 32bit floating too that floating alters the transient of the sound.

It's entirely possible that Steinberg's implementation of floating 32-bit in SX alters the sound. It used to be wrapped around their truetape saturation effect (in VST) so you would basically record a wet (processed) signal; if any of that processing is still included, it will likely alter your sound on its way in.
As Astroman has pointed out, it depends on how that particular algorithm is coded much more than how many bits. I can hear a *slight* difference in sound quality between steinberg's 16 and 24 bits, depending on the source material; now I generally record in 16 bits and create submixes in 24 bits. It all depends on where you want to draw a line between quality of sound and quality of song, but don't forget Sgt Pepper was recorded on a 4 track analog tape.

If you use too many UAd_1 COMPRESSORS ON YOUR MIX THEN YOU GET A MIX WITHOUT ANY PUNCH ! as opposed using Powercore's cmpressors.

I've run into this before, I concluded it was because I had too much processing in the mix and started focusing more on my source material and arrangements - if *everything* is highlighted, it's the same as if nothing is highlighted, and you can't boost frequencies that aren't there. Are you using the cubase mixdown function or recording live? I don't like the sound of a mix when it's rendered by Cubase - my ears are not infallible but it just doesn't seem to sound the same. Using Pulsar cards it's easy (and often more practical, especially if you don't use XTC mode) just to pipe the mix back into cubase and record it. That way, what you hear is what you get.

Even 32bit floating recording onto Cubase reveals same defficiencies as opposed with using 24bit fixed recording.
Thats why I conclude Sawstudio has the best sound becuase it uses 24bit engine.
thats why VST24 has a better sound than Cubase 5.1 and SX.
Thats why SX sounds soft and nice...

Sawstudio has a very different audio processing core than steinberg's, I remember when that company started out and they seemed to be a small startup company who had some very talented people at its core, who were programming in assembly instead of C++ to get the very best quality and efficiency out of the hardware, that sort of thing. If those algorithms are higher quality you'll get a more musically pleasing sound - cheap processing can definitely make your sounds muddier and less punchy.

[ This Message was edited by: Liquid Len on 2004-09-15 22:46 ]

Post by **garyb** » Thu Sep 16, 2004 12:25 am

i like sx's sound but to each their own.

btw-sgt pepper's 4 track was a 1/2" four track, a device of super high quality(equivalent to 16trk 2", considered the pinnacle of tape recorder technology)

medway · Post by **medway** » Thu Sep 16, 2004 1:10 am

SawStudio is programmed by one man, Bob Lentini.

samplaire · Post by **samplaire** » Thu Sep 16, 2004 4:15 pm

On 2004-09-15 19:50, hubird wrote:
Stop the technicions.
Stop the bit hunters.
Stop the measurements.
Stop knowing better than what the leading soft houses offer.
Stop believing nonsense.
Stop stopping making music.
Get back to work.
Make a song.
Make music.
Eat food, not bits.
Think music, make art.
Hell, make MUSIIIIIIIIC.

Ah...!

Huub, can I use your poetry as a lyrics to my next track? I'm not kidding, man

Post by **hubird** » Thu Sep 16, 2004 7:04 pm

that would be the ultimate use of those words

go ahead and show them bits kids that making sound is human work

cheers mate

samplaire · Post by **samplaire** » Fri Sep 17, 2004 6:16 am

Here You Go

firubbi · Post by **firubbi** » Sun Sep 26, 2004 10:12 am

Stop believing nonsense.
Stop stopping making music.
Get back to work.
Make a song.
Make music.
Eat food, not bits.
Think music, make art.
Hell, make MUSIIIIIIIIC.

...but does it help if we rec at 24/32bit and burn data on cd which is 16bit?
thanks

[ This Message was edited by: firubbi on 2004-09-26 11:16 ]

astroman · Post by **astroman** » Sun Sep 26, 2004 11:28 am

with his 'rhyme' Hubird puts music back into it's context again - opposed to the overproduced typical charts crap.

And his reference to the identically titled Madonna album is correct ('American Life' may apply as well, but I dunno).
Large parts (even of the vocals) were processed ny co-producer M. Ahmadzai on a 16 bit soundcard and later added to the big Protools system for final mix.
While the predecessor 'Ray of Light' is a total 'hifi' high end record the 'Music' album sounds much more pleasing (at least to my ears) and balanced due to those lofi elements.
It is a nice example of mixing (totally) different sound sources to a unit which finally scores on it's own - no need to be a Madonna fan to recognize this...

A great song will easily stand some recording flaws (a CD release of an antic reggae album by the Congos comes to my mind), so this is the wrong part to focus on during a production (imho).

32 bit (in SFP via VDAT) doesn't sound great because of the bit depth, but because the overall sound processing remains 'in the same context'.

24 bit is mostly about headroom, which yields a more detailed mixing result for the simple fact that you have a 'larger image' of sound to process.
If you 'spoil' a couple of bits in 16 bit data that's obviously much more significant than the same action on 24 bits.

For the final result (on consumer equipment) the increased precision of a 24 bit master will have absolutely no influence at all.

I guess that 75% of all CD players still feature a (effective) 12-14 bit converter at best, let alone clock quality and filter.
To make records sound good on crappy equipment one applies different tricks than increased resotution

Of course this would be another story if all home listening gear would feature 24 bit converters (and appropriate analog stages !) of the same quality as a Scope boards (or other quality brand) - but this simply isn't the case.
And it probably will never be as people seem quite happy with 128 kbit MP3s...

So you can make your mixing life easier with more headroom (and some increased resolution) in 24 bit (or 32 if that applies better or more convenient) and pick the right dithering process to a final 16 bit release.

But I'm absolutely convinced that even the most crappy 'downscaling' will have only minor influence on the final result, compared with other (possible) mixing errors...
Nevertheless one should try to do the job as good as it gets

cheers, Tom

firubbi · Post by **firubbi** » Tue Sep 28, 2004 10:26 am

Thanks Tom,
i'll jump to 24bit soon. should i set dither to 16bit while mastering back to 16bit for cd?
thanks

Grok · Post by **Grok** » Tue Sep 28, 2004 6:04 pm

On 2004-09-28 11:26, firubbi wrote:
Thanks Tom,
i'll jump to 24bit soon. should i set dither to 16bit while mastering back to 16bit for cd?
thanks

Yes.

Golden rule n°1:
-Always add dither when lowering the bit resolution.
Golden rule n°2:
-Apply dither only one time, at the end of the recording/mixing/mastering process, after all the others operations and just before lowering the bit resolution* (add dither to the 32 bit or 24 bit master, and then convert to 16 bit).

* this is done in one step with Samplitude

Grok · Post by **Grok** » Tue Sep 28, 2004 6:39 pm

Golden rule n°3:

With the Creamware cards, if you want to preserve the plain sound of the synths & FXes in the recorded audio files, you have to record audio files that have a 32 bit definition.

The Creamware's synths and FXes sounds outputs have a 30 bit resolution.
If you record these sounds in a 24 bit or a 16 bit audio file, you will truncate the audio signals coming from these synths and FXes outputs.

Truncating a digital PCM audio signal add harmonic distortion to the resulting audio. And this harmonic distortion is really not a euphonic harmonic distortion like these obtained with tubes processing! We don't want this distortion in our recordings, unless it is for specifical musical views (e.g.: bitcrushing is an FX that add this kind of distortion when used). With average monitoring, you may not perceive immediately this distortion when recording Creamware's audio signals in 24 bit audio files, but at the end the result will not be the same as if it is recorded in 32 bit files! And with recording these signals in 16 bit audio files, it will be worse than with 24 bit files.

This truncature harmonic distortion is auditively minimized with adding dither at the final conversion stage to 16 bit for audio CD(some dithers are better than others from this point of view).

Remembering the Golden rule n°2 (dither has to be applied only one time), you see that you can't apply dither to all your recorded 24 bit or 16 bit files to minimize the truncature harmonic distortion when recording the files from Creamware's synths and FXes outputs. Dither is a noise, and we don't want to multiply this noise, we want it to be unaudible.

That's why the best use with Creamware cards is to record and mix in 32 bit, with an audio engine that has no weaknesses with the support of these calculations. And with the adequate drivers communication between SFP and the audio soft (32 bit, not 24 bit, not 16 bit).

Best,
Grok

Post by **hubird** » Tue Sep 28, 2004 7:02 pm

On 2004-09-26 12:28, astroman wrote:
32 bit (in SFP via VDAT) doesn't sound great because of the bit depth, but because the overall sound processing remains 'in the same context'.
......
For the final result (on consumer equipment) the increased precision of a 24 bit master will have absolutely no influence at all.

Grok says recording Scope synths and effects at 32 bit is highly recommended.
I wanne believe that this is theoretical true, but Astroman suggest it's hardly relevant.
Does the fresh Levitating Fire-Breathing Pulsar Overlord stand with his former opinion?
I will test it by myself anyway, soon, but I'm curious to your opinion, Astro, if you don't mind

[ This Message was edited by: hubird on 2004-09-28 20:03 ]

Shayne White · Post by **Shayne White** » Tue Sep 28, 2004 7:08 pm

I took a really dynamic sound in MiniMax and recorded my noodling with it into a 16-bit file, a 24-bit file, and a 32-bit file.

The 24-bit file and the 32-bit file sounded indentical, and the 16-bit file sounded only *slightly* worse. I do all my recording at 24-bit and things sound great. Don't go crazy over searching for 32-bit software -- and it takes up tons more disk space.

That's my opinion.

Post by **hubird** » Tue Sep 28, 2004 7:42 pm

thanks Shayne for doing the test

It's what I suspected

Grok, if you think you get audible different results (with dithered 16 bit files), I offer you my server to put them on, for us to listen to

My A.D.A.M. nearfields will show every difference, specially concidering the 'air' part of the audio, my ears will be the only cul de sac

As this thread shows, I'm definitely not a chasing-for-every-bit type of guy, but if I get any prove that 32 bit recording of Scope sources is worth the difference, I would change to 32 bit, why not.
Something is not true untill the counterpart has be proven, if need be once

I doubt if that 'once' will happen though

[ This Message was edited by: hubird on 2004-09-28 20:44 ]

Grok · Post by **Grok** » Tue Sep 28, 2004 8:54 pm

One thing is unavoidable: truncating a audio PCM signal adds harmonic distortion. It's a simple, prooven, and theoricized fact. It maybe more (30 bit to 16 bit) or less (30 bit to 24 bit), but it's simply a fact (an inherent and basic digital PCM fact).

And, as we know, SFP doesn't run at 16 bit, in wich case there wouldn't be any problem to record in 16 bit or 24 bit files.

Then, as I said, when a distortion is minimal (30 bit to 24 bit), we may not perceive immediately the damage done (and there are chances for that, if the distortion is low and if our attention is focused on others parts of the job).
On one only audio 24 bit file, this truncature is very little distortion. In a 48 tracks mix with many furthers PCM digital processes, this cumulates into an audibly less transparent mix.

BR,
Grok

astroman · Post by **astroman** » Tue Sep 28, 2004 9:33 pm

Grok is of course right. The math transform of the various bit rates WILL include an unavoidable amount of distortion.
But as mentioned this is extremely low and can be safely ignored.

It doesn't even make sense to consider something like a 32-bit converter because the (also unavoidable) clock instability will cause a larger amount of 'distortion'.

And finally 'dithering' means nothing but to modulate the signal with a certain type of noise - with the trick being to pick the appropriate one. At least then any faint differences will be 'washed' away.

It's also worth to consider the type of music to be recorded - and it's audience.
Who's gonna listen to the (very) details of a synth in depth ?

If it's a typical 'hit' production, then the vocals are upfront and half of the audience will be more concerned with the sexual aura of the performer than with the instrumentation

On a dance track it's the impact that counts and a clean sound may not yield the best result in this context (as the 'distorted kick thread' shows).
12-bit drum samplers still score top prices for their 'special' sound.

I've had a good dose of car radio recently

and my (sarcastic) conclusion is that one can forget anything related to precision on pop vocals.
I really dunno why people spent 5 figure values on mic gear when the final result is almost synthezised in post production...

So finally all that's left is the recording of acoustic performances of instrumentalists and vocalists, sufficiently sophisticated that their unfiltered output can be enjoyed. In that context a very high precision makes sense.

Anything else that is processed up and down may be affected by one or the other item in the recording chain, but certainly not by the bit depth

In that sense: make music - if it sounds right, then it is right

my 2 cents, Tom

Shayne White · Post by **Shayne White** » Wed Sep 29, 2004 12:29 am

And with everybody listening to everything on MP3 anyway, a little 32-bit to 24-bit truncation isn't going to make any difference.