Thank you guys.

Macodun · Aug 12, 2004, 01:19 AM

Well, after an intensive day and a half playing with the KX drivers, I am now about to un-install them.

My main reason is that 24/96 support for my Audigy 2 ZS doesn't seem to be implemented yet. Second reason; that Aqua skin really is nasty! lol And the other skins I've tried seem to be buggy in some way or another. And the third reason is that I'm way too dumb to really understand what's going on!

But I really do appreciate the effort you guys have put into this, and when you manage to give full support for the Audigy 2 ZS, (and perhaps make the documentation more comprehensible for thicko's like me), I will be happy to try it again.

So thanks again guys, you're doing a great job.

But, before I go, I'd like to respond to another poster who asked if 24/96 was any better.

Yes.

24 bit gives much more dynamic range, and therefore headroom, than 16bit; theoretically in excess of the range of human hearing. (Although, the hardware does reduce that greatly, but even so it's a vast improvement on 16 bit).

And people ask why use 96K (giving a recordable audio frequency of just under 48k) when the range of human hearing only goes up to 20k?

Well, actually, if you can hear a 20k sine wave you're lucky, my hearing struggles to reach 16k.

But that's only for a sine wave. Transients, which provide much of the positional information as well as "crispness" to a sound, have components greatly in excess of 20k, and they can be registered by the human hearing.

I've been involved in experiments where a source has had a 30k filter switched in and out, and even though the experiments were performed blind, one could always tell whether the filter was in or not. When it's in, the stereo imaging loses focus and the overall sound loses a degree of crispness.

So, again, yes, 24/96 is better. Good condenser mics connected to the Audigy via high quality pre-amps really show up the difference, especially when recording cymbals, xylophone, etc., and even acoustic guitar is audibly better.

Mac

**dj_stick** · Aug 12, 2004, 03:21 AM

the sample rate of the audio has nothing to do with the frequeny range, its is in relation to how many samples are processed per second, the higher the sample-rate the more accurate the re-production is
24/96 is in progress atm, but as most recordings are then sent to CD, which is 16bit and 44100hz, this is not the greatest of importance, unless you are keeping your audio digital and/or pressing to vinyl

there is a new skin (thread in the general discussion forum here) that should really help you understanding the mixer a little more

also look at the kX knowledgebase especially the getting started guide, to help you getting the basis of the program

TravelRec. · Aug 12, 2004, 06:31 AM

Quote:

Originally Posted by dj_stick

the sample rate of the audio has nothing to do with the frequeny range, its is in relation to how many samples are processed per second, the higher the sample-rate the more accurate the re-production is...

Hi dj!

That´s not correct! You can record nearly the half of frequency compared to the sample rate. This is a physicly law. And to avoid aliasing effects while A/D conversion, all circuits have analog or digital anti-aliasing filters that cut away all frequencies above 1/2 sample rate. So Macodun is right about that and the recordable frequency and freshness of sound depends directly on sample rate. The higher resolution of 24 bits gives you not really a better sound, but a greater headroom to avoid distortion with high input levels. All in all it´s cool to have 24/96 or 32float./96 for internal mixing / mastering, but for recording and playing 16/48 is good enough (my opinion).

Greetings!

TravelRec.

Tiger M · Aug 12, 2004, 12:59 PM

Quote:

You can record nearly the half of frequency compared to the sample rate.

Yep, Nyquist freqency, can't love it. Btw, the EMU 10k2 DSP still operates at the fixed 48KHz, and to achieve 96KHz resampling is used.

**dj_stick** · Aug 12, 2004, 01:00 PM

learn something new everyday

Tiger M · Aug 12, 2004, 01:04 PM

and more headroom can be achieved only with floating-point DSPs (at least they say so). 24bit conversion only gives you a more natural wave, even closer to the original analog wave

Macodun · Aug 13, 2004, 03:50 AM

Thanks DJ, I've had a quick look at those documents you mention, but as I'm trying to learn Cubase, and Fruity Loops, and what a VSTi is, etc., etc., my brain is starting to overload a wee bit at present.

As has been mentioned, the sampling rate is directly related to the sampled frequency (the Nyquist frequency) and the sample rate is twice the sampled frequency, (well, slightly more in practice, usually about 2.1 times the sampled frequency, otherwise there's no DC discrimination at exactly 2 X f.sample).

But I don't know if I agree that 24 has no more headroom than 16 bit Tiger.

16 bit gives 65,536 different levels (slightly over 96dB) (he says quicky checking his calculator! lol), whereas 24 bit gives 16,777,216 different levels, (slightly over 144 dB), so assuming both start at the same bottom level, then 24bit gives significantly more dynamic range, and hence headroom. (Except of course, the hardware can't handle much more than about 106/108 dB).

But, I suppose it all depends what you mean by headroom, and I 'spect I'm using the word wrong.

But you're right of course Tiger, 24 bit does give a much smoother conversion, much less quantising noise.

Mac

kuja · Aug 13, 2004, 10:52 AM

O/T:

How do you translate the number of values possible in a 16-bit integer into a dB dynamic range?

Macodun · Aug 13, 2004, 07:03 PM

As you know, an 8 bit word has 2^8 "levels", i.e. 256; a 16 bit word, 2^16, 65,536; and so on.

dB is just a way of expressing ratios logarithmically. And in the examples given, the first ratio is 256:1, and in the second, 65,536:1, in other words, the ratio is exactly equal to the number of "levels".

So to convert from a ratio to to dB's, you just apply the equation: 20 log ratio.

i.e. for 8 bit; 20 log 256 which equals: 48 dB (nearly, anyway)
for 16 bit; 20 log 65,536 which equals: 96 dB (and a wee bit).
and so on.

Note, the reason that 20 log is used instead of 10 log, is because we're talking about voltage levels, not power levels.

EDIT: Just to explain that a wee bit more.

Bels are used as a direct conversion from linear power ratios to logarithmic power ratios,

So, a 10W amp has 10 times the power of a 1W Amp, so it's 1 Bel more, 100W is 2 Bels more, 1000W, 3 Bels, and so on. (in other words, its just the log base 10).

A decibel is obviously just a tenth of a Bel, so just multiply Bels by 10.

But, that's for power ratios. When we talk about signals we're usually talking about voltage, and if you remember your Ohm's Law, Power is proportional to V squared, and of course, when converting to logs, squared becomes multiplying by 2.

So, we end up with: 2 x 10 x log (ratio),

i.e., 20 log 2^(number of bits).

Hope this helps.

Mac

Smurf43 · Jan 16, 2005, 07:50 AM

i think that burnt a brain cell

Nappylady · Jan 16, 2005, 02:16 PM

AH!!!!

I always wondered why sometimes it's 20 and sometimes it's 10. This should help me make more sense of it.

Thanks, Mac!

Aug 12, 2004, 01:19 AM	#1
Macodun DriverHeaven Newbie Join Date: Aug 2004 Posts: 12 Rep Power: 0	Thank you guys. Well, after an intensive day and a half playing with the KX drivers, I am now about to un-install them. My main reason is that 24/96 support for my Audigy 2 ZS doesn't seem to be implemented yet. Second reason; that Aqua skin really is nasty! lol And the other skins I've tried seem to be buggy in some way or another. And the third reason is that I'm way too dumb to really understand what's going on! But I really do appreciate the effort you guys have put into this, and when you manage to give full support for the Audigy 2 ZS, (and perhaps make the documentation more comprehensible for thicko's like me), I will be happy to try it again. So thanks again guys, you're doing a great job. But, before I go, I'd like to respond to another poster who asked if 24/96 was any better. Yes. 24 bit gives much more dynamic range, and therefore headroom, than 16bit; theoretically in excess of the range of human hearing. (Although, the hardware does reduce that greatly, but even so it's a vast improvement on 16 bit). And people ask why use 96K (giving a recordable audio frequency of just under 48k) when the range of human hearing only goes up to 20k? Well, actually, if you can hear a 20k sine wave you're lucky, my hearing struggles to reach 16k. But that's only for a sine wave. Transients, which provide much of the positional information as well as "crispness" to a sound, have components greatly in excess of 20k, and they can be registered by the human hearing. I've been involved in experiments where a source has had a 30k filter switched in and out, and even though the experiments were performed blind, one could always tell whether the filter was in or not. When it's in, the stereo imaging loses focus and the overall sound loses a degree of crispness. So, again, yes, 24/96 is better. Good condenser mics connected to the Audigy via high quality pre-amps really show up the difference, especially when recording cymbals, xylophone, etc., and even acoustic guitar is audibly better. Mac

Aug 12, 2004, 01:04 PM	#6
Tiger M kX user Join Date: Apr 2004 Posts: 851 Rep Power: 0	and more headroom can be achieved only with floating-point DSPs (at least they say so). 24bit conversion only gives you a more natural wave, even closer to the original analog wave __________________ Miss you, Steve... Last edited by Tiger M; Aug 13, 2004 at 05:55 AM.

Aug 13, 2004, 03:50 AM	Thread Starter #7
Macodun DriverHeaven Newbie Join Date: Aug 2004 Posts: 12 Rep Power: 0	Thanks DJ, I've had a quick look at those documents you mention, but as I'm trying to learn Cubase, and Fruity Loops, and what a VSTi is, etc., etc., my brain is starting to overload a wee bit at present. As has been mentioned, the sampling rate is directly related to the sampled frequency (the Nyquist frequency) and the sample rate is twice the sampled frequency, (well, slightly more in practice, usually about 2.1 times the sampled frequency, otherwise there's no DC discrimination at exactly 2 X f.sample). But I don't know if I agree that 24 has no more headroom than 16 bit Tiger. 16 bit gives 65,536 different levels (slightly over 96dB) (he says quicky checking his calculator! lol), whereas 24 bit gives 16,777,216 different levels, (slightly over 144 dB), so assuming both start at the same bottom level, then 24bit gives significantly more dynamic range, and hence headroom. (Except of course, the hardware can't handle much more than about 106/108 dB). But, I suppose it all depends what you mean by headroom, and I 'spect I'm using the word wrong. But you're right of course Tiger, 24 bit does give a much smoother conversion, much less quantising noise. Mac Last edited by Macodun; Aug 13, 2004 at 04:01 AM. Reason: spelling

Aug 13, 2004, 07:03 PM	Thread Starter #9
Macodun DriverHeaven Newbie Join Date: Aug 2004 Posts: 12 Rep Power: 0	As you know, an 8 bit word has 2^8 "levels", i.e. 256; a 16 bit word, 2^16, 65,536; and so on. dB is just a way of expressing ratios logarithmically. And in the examples given, the first ratio is 256:1, and in the second, 65,536:1, in other words, the ratio is exactly equal to the number of "levels". So to convert from a ratio to to dB's, you just apply the equation: 20 log ratio. i.e. for 8 bit; 20 log 256 which equals: 48 dB (nearly, anyway) for 16 bit; 20 log 65,536 which equals: 96 dB (and a wee bit). and so on. Note, the reason that 20 log is used instead of 10 log, is because we're talking about voltage levels, not power levels. EDIT: Just to explain that a wee bit more. Bels are used as a direct conversion from linear power ratios to logarithmic power ratios, So, a 10W amp has 10 times the power of a 1W Amp, so it's 1 Bel more, 100W is 2 Bels more, 1000W, 3 Bels, and so on. (in other words, its just the log base 10). A decibel is obviously just a tenth of a Bel, so just multiply Bels by 10. But, that's for power ratios. When we talk about signals we're usually talking about voltage, and if you remember your Ohm's Law, Power is proportional to V squared, and of course, when converting to logs, squared becomes multiplying by 2. So, we end up with: 2 x 10 x log (ratio), i.e., 20 log 2^(number of bits). Hope this helps. Mac Last edited by Macodun; Aug 14, 2004 at 05:25 PM. Reason: expansion o explanation

Jan 16, 2005, 07:50 AM	#10
Smurf43 DriverHeaven Newbie Join Date: Jan 2005 Posts: 7 Rep Power: 0	Ouch....... i think that burnt a brain cell

Aug 13, 2004, 10:52 AM	#8
kuja DriverHeaven Junior Member Join Date: May 2003 Posts: 45 Rep Power: 0	O/T: How do you translate the number of values possible in a 16-bit integer into a dB dynamic range?

Jan 16, 2005, 02:16 PM	#11
Nappylady DriverHeaven Addict Join Date: Jun 2003 Posts: 257 Rep Power: 0	AH!!!! I always wondered why sometimes it's 20 and sometimes it's 10. This should help me make more sense of it. Thanks, Mac!