PlanetZ forums - scopeusers.com

downsampling is a bad idea in any case imho(unless it's for effect).

If a project were to be put onto CD, would it be best to record it all in 44.1k or 96k?

I can see the arguments for both, as downsampling from 96k may be more 'representative' as it involves averaging...

On 2004-10-05 14:27, garyb wrote:
downsampling is a bad idea in any case imho(unless it's for effect).

This the reason why this thread has got me baffled. Why not just try to get the best sound possible at 16/44.1?

I'm not knocking the technical aspects of the whole thing(it is a general discussion thread after all), it is very interesting indeed. I just question the why. Unless you are all looking to the near future when CDs WILL be at those high sampling rates?

First of all it's a personal pleasure !
With 24 or 32 bits ( using VDAT doesn't hurt either ) I record what I heard playing synths ( i.e taking an accurate picture )

With oversampling I pull the mist, the dust out of the picture.

If you record at 96 / 32 then downsample to 44.1 then dither to 16 bits you have a more faithful picture of the sound you originally heard than directly recording at 44.1 / 16.

But I find the oversampling part more potent than the bit part because I discovered unsuspected depth and shine in lot of presets.

I was suprised though to hear a difference from a 32 bit to 24 bit dithering, I thought 32 bit was overkill but ... no.

However I must say I took some very sharp sounds with delay and the sounds lost contours, with rounder, duller sounds the difference is less obvious.

I tried recording at 48 Khz and putting filters at 20 Khz + but the results wasn't satisfactory for me.

I'am sad I can't try it on all my synths and most of the modular because of the DSP shortage but I'am really hooked, so much I hardly imagine playing at 44.1 anymore.

Paul:

I record at 24 / 44.1. The 24 bits give me a substantial increase in dynamics. This gives me more maneuvring room. For example, I don't feel compelled to keep input levels quite as "hot" as I used to. Clipping therefore is less of a problem now than it was when I recorded at 16 bit.

So for me the reason to go 24 bit was purely quantitative, not qualitative. I don't really care whether 24 bits "sounds better" or not. I'll leave that hocus pocus up to the magicians. (JUST TEASING.)

Incidentally I didn't make my 24 bit move until hard drives became cheap enough (i.e. this past year).

From my experience, the Minimax sounds different (and more pleasant) at 96 kHz than at 44,1 kHz. There is a little perceived difference...

Downsampling is an issue, all downsampling converters are not equals, the free Voxengo R8brain being one of the best, if not the best one in any platform (remember: Alexeï Vaneev, the Voxengo dev, is a king!). The Voxengo R8brain with its "Reusable royalty-free converter DLL with explained API"...

<font size=-1>[ This Message was edited by: Grok on 2004-10-06 09:31 ]</font>

So what diference is more significant?
24-bits @ 44.1 kHz, (only increase the bitdepth)or 16-bits @ 96kHz (only increase the samplerate).

If you increase both settings it will apearently sound better for sure, but the problem is that 24/96 wil use about 3.3 times more DSP in comparision to 16/44.1, and that's a lot if someone like me only has 6 DSPs

Another thing is, I still have the old STS3000 which is only reading and recording 16-bits samples maximum (no 24 or 32 bits), And I use a lot of samples in most projects. I'll have the STS5000 next month, so untill then I have to stick to 16-bit recordings, too bad.
When I get 12 DSPs this week and the STS5000, I think I'd rather go for 24-bits ("only" 50% more DSP load) instead of going to 96 kHz (220% more DSP load)

Good choice, or not?

Darce

My sts3000 has 16, 24 & 32 bit options....

On 2004-10-06 10:03, ChampionSound wrote:
So what diference is more significant?
24-bits @ 44.1 kHz, (only increase the bitdepth)or 16-bits @ 96kHz (only increase the samplerate).

44.1/16 compared to 96/16 - the sample has just over twice the resolution using 96/16 comapred to 44.1/16

44.1/16 compared to 44.1/24 - the sample has 256 (maybe 128) times the resolution using 44.1/24 compared to 44.1/16

@ wayne

oops! Mine too!
I'm embarrased.... When I read the sts-3000 manual before I ever used a STS sampler, It said that it only can handle up to 16-bits, but I never checked it before myself, my bad... But anyway, the pdf manual is wrong about that.
Thanks anyway, wayne!

@ cleanbluesky

My thougths exactly...

_________________
Pulsar 2 user (incl. Luna I/O box)

<font size=-1>[ This Message was edited by: ChampionSound on 2004-10-06 10:56 ]</font>

On 2004-10-06 10:41, cleanbluesky wrote:
...
44.1/16 compared to 96/16 - the sample has just over twice the resolution using 96/16 comapred to 44.1/16

44.1/16 compared to 44.1/24 - the sample has 256 (maybe 128) times the resolution using 44.1/24 compared to 44.1/16

you cannot compare the loudness and the frequency domain

someone with enough math skills might calculate the various scenarios and determine the respective levels of distortion.

yet it would be of little practical use because he doesn't know how each type of distortion is perceived...

the greatest fidelity is not always acoustically most pleasing - anyone remembers the complaints about CWA's stock devices, eqs, compressors... ?

cheers, Tom

Tom is right.

Bits measure amplitude plus sign (positive or negative voltage).

Sample rate determines how many "slices" of the alternating current are taken.

Increasing bits lets you record a wider range of amplitudes, i.e. a wider dynamic range. This also means an increase in the dynamic range between noise floor and clipping.

Increasing sample rate lets you record a wider range of frequencies.

As Tom pointed out, an increase in bits will not allow you to capture higher frequency signals. An increase in sample rate will not give you any more dynamic range.

Tom: What are the complaints about CWA's stock devices? I'd like to know because I find the EQs and compressors quite musical and transparent. (Except when you push an EQ past +6 db boost, which is heresy in my world anyway).

Cheers,

Johann

For an overdose of geek talk I suggest you head there : http://www.hydrogenaudio.org/forums/index.php?

Thanks for the Voxengo's link Grok, I was using Sound Forge tools that may not be the best around, more tests ... duh !

ChampionSound

You don't save DSP power by working in 16 bit, because in fact you do not work in 16bit. As soon as a 16 bit signal is feed into SP, empty bits will be added below the 16th bit.

Immanuel, I guess you're right, that makes totally sense!
What was I thinking?

On 2004-10-06 13:57, blazesboylan wrote:
... Tom: What are the complaints about CWA's stock devices? I'd like to know because I find the EQs and compressors quite musical and transparent. ...

doesn't sound good - no character - boring... if memory serves

it's kind of amusing: people keep asking for ultra precise 'the best' reproduction technique and literally flood the sources with all kinds of distortions right afterwards.

a tube adds harmonics, something that wasn't there in the first place.
From the fidelity viewpoint it spoils the signal - yet it's cool because it's hot (the heating wire) - it makes a warm sound because of the dark orange glow - it's desired because it's expensive...

we're talking about quality differences a magnitude below a regular tube's distortion - or (to use another example) people shell out $$ for a multi gig piano lib... and the thing cannot even do faithful piano 'string' resonance.

Yet noone bothers, because it's considered unimportant (or is it just unknown ?) - but that effect is acoustically at least 10 times more prominent than the differences between 44 and 96 k

cheers, Tom

I know what you mean,
But can you simply compare a tubesound (which adds something extra and which is a pleasant sound indeed) with trying to get the best fidelity at the end of your recording chain? (96k)
I think if people want to have that tubesound, they should use a tube for that extra edge on their existing sound. But wouldn't it be nice if the tubesound itself could be captured at high quality, so that the tubesound (which distorts the original signal in a way) , and all the other distortions that suppose to be in the signal chain, can be reproduced as good as possible at the end?

Sorry if I haven't explained it clearly

Maybe fidelity should be treated with more respect.

If people would work on the source material more, capturing that great performance, and spend less time making sampled digital pianos sound "warm", then I, for one, would bitch less about contemporary music and how dull it sounds.

When you capture a live piano, there's so much air and colour and all that fun stuff in the instrument, mics, preamps, etc, that some transparency is necessary if you're going to process the sound at all -- because you don't want to muddy the recording!

(This is not an argument for sample rates, mind you. I just think that the stock CWA EQs and compressors are great.)

Cheers,

Johann