hi all
1. I wonder how many asio2 24bit in- and out-channels you could load with either 44.1khz or 96khz. would be nice to have a few testers like in the masterverb test thread.
I've 30 dsp's and i wasn't able to load 24 ASIO2 IN's and OUT's with 96khz... got something like "biggest module does not fit..."
2. Till today i always recorded at 44.1khz 24bit. now an international producer (american) told me that i should use at least 96khz. and in the usa, the common samplerate is 192khz, he says.
another mastering engineer here in switzerland told me that it doesn't make sense because the rounding error while converting from 96khz to 44.1 in the mastering progress is worse than record everything from the beginning at 44.1khz. What do you guys think about this? How do you record your analog stuff like drums, guitars...
greez
roman
<font size=-1>[ This Message was edited by: katano on 2006-08-21 06:54 ]</font>
the samplerate (test) thread
well, the swiss folks seem to have a practical viewpoint and don't just believe in 'bigger is better' hype.
I agree with your fellow engineer, but honestly, I never touched that type of gear anyway...
to fight aliasing artifacts 96k is indeed an improvement over 44.1k (from the 'technical' facts), anything above doesn't improve things anymore.
on the other hand the afforementioned downsampling (a very math intense process) will always have it's negative influence.
In 'standard' applications (simple code in regular software) it's more or less certain that it outweighs the 'less aliasing' benefits.
Regarding dedicated, hi-quality digital format converters (in software) I'm not entirely sure, but that stuff is pretty expensive...
It has been mentioned frequently that a dedicted wordclock will significantly improve 44.1k signals, if all is properly (and stable!) synced from 1 source.
I'd even suspect that my old A16's 18bit converters outperform cheapo 192k gear - it has a mighty impressive clock circuit... but that could be pure phantasy as well
Finally I'm convinced that professional 192k gear sounds amazing nevertheless.
There's most likely analog technology at it's best inside... so finally it's not the samplerate that makes the nice sound, but the solid craftmenship behind such a unit - imho.
afaik in the video/DVD business 192k is indeed standard, but that has few to do with 'musical' quality.
cheers, tom
I agree with your fellow engineer, but honestly, I never touched that type of gear anyway...
to fight aliasing artifacts 96k is indeed an improvement over 44.1k (from the 'technical' facts), anything above doesn't improve things anymore.
on the other hand the afforementioned downsampling (a very math intense process) will always have it's negative influence.
In 'standard' applications (simple code in regular software) it's more or less certain that it outweighs the 'less aliasing' benefits.
Regarding dedicated, hi-quality digital format converters (in software) I'm not entirely sure, but that stuff is pretty expensive...
It has been mentioned frequently that a dedicted wordclock will significantly improve 44.1k signals, if all is properly (and stable!) synced from 1 source.
I'd even suspect that my old A16's 18bit converters outperform cheapo 192k gear - it has a mighty impressive clock circuit... but that could be pure phantasy as well
Finally I'm convinced that professional 192k gear sounds amazing nevertheless.
There's most likely analog technology at it's best inside... so finally it's not the samplerate that makes the nice sound, but the solid craftmenship behind such a unit - imho.
afaik in the video/DVD business 192k is indeed standard, but that has few to do with 'musical' quality.
cheers, tom
In recording, you will hear no difference. Processing or synthesis, that's where the main gain in higher samplerates lies.
Here's an A/B, of a simple saw>LPF modular synth. The 96kHz recording was resampled to 44.1kHz by SX's basic resampling feature. Needles to say which sample is at which sample rate, I suppose.
Here's an A/B, of a simple saw>LPF modular synth. The 96kHz recording was resampled to 44.1kHz by SX's basic resampling feature. Needles to say which sample is at which sample rate, I suppose.
more has been done with less
https://soundcloud.com/at0m-studio
https://soundcloud.com/at0m-studio
@astroman: what do you mean with "I never touched that type of gear anyway..."?
@at0m: is this a test?
both samples are 44.1... i hear artifacts in the a.wav so i guess this is the converted one... what about a better converter like waves L2?
and I guess it'll be definitively a better result when converting 88.2 to 44.1 because of less rounding errors.
so, what would you guys recommend for recording? I use a tfpro16 preamp and an A16 Ultra via Z-Link...
@at0m: is this a test?
both samples are 44.1... i hear artifacts in the a.wav so i guess this is the converted one... what about a better converter like waves L2?and I guess it'll be definitively a better result when converting 88.2 to 44.1 because of less rounding errors.
so, what would you guys recommend for recording? I use a tfpro16 preamp and an A16 Ultra via Z-Link...
Waves L2? Not sure what use that would be for this, although it does have a dithering algorithm for changing bit depth...
Personally I don't see the point in using >44.1khz right now, for the following reasons:
1. CPU's are just not fast enough to deal with processing lots of channels at high sample rates with good algorithms
2. PCI buses and motherboard bussing in general simply do not have the bandwidth to ferry all this data around reliably (in fact many systems struggle just with 44.1)
3. there are virtually no good algorithms for native CPU's anyway... putting up the sample-rate just increases the CPU load without sounding any better
I think there's more value in investing in good quality AD/DA stages (Apogee, Lavry, Lucid etc) and decent outboard. I just got a (cheap!) Apogee AD-1000, it sounds incredible to me, and only 44.1k, 20 bit (oh noes, it's not 24/192, what am I going to do!)
Personally I don't see the point in using >44.1khz right now, for the following reasons:
1. CPU's are just not fast enough to deal with processing lots of channels at high sample rates with good algorithms
2. PCI buses and motherboard bussing in general simply do not have the bandwidth to ferry all this data around reliably (in fact many systems struggle just with 44.1)
3. there are virtually no good algorithms for native CPU's anyway... putting up the sample-rate just increases the CPU load without sounding any better
I think there's more value in investing in good quality AD/DA stages (Apogee, Lavry, Lucid etc) and decent outboard. I just got a (cheap!) Apogee AD-1000, it sounds incredible to me, and only 44.1k, 20 bit (oh noes, it's not 24/192, what am I going to do!)
Which one is which, I hear a huge difference between the two...On 2006-08-21 07:31, at0m wrote:
In recording, you will hear no difference. Processing or synthesis, that's where the main gain in higher samplerates lies.
Here's an A/B, of a simple saw>LPF modular synth. The 96kHz recording was resampled to 44.1kHz by SX's basic resampling feature. Needles to say which sample is at which sample rate, I suppose.
(a) sounds much sharper, I don`t know if in a better way, but just much more aggressive...
(b) is more smooth, damped.....
Which is which....
Did you do something wrong, I don`t expect such an difference between 44,1 and 96knhz...
How did you downsample th 96khz sound to 44,1?
With a good interpolation?
I don't own, I don't plan to own, I cannot even afford itOn 2006-08-21 08:18, katano wrote:
@astroman: what do you mean with "I never touched that type of gear anyway..."?
it was a joke, as I consider myself quite familiar with the technical side, but don't want to claim ultimate knowledge - which would require hands-on experience...
you missunderstood his intention - it's the synth engine that produces the 'artifacts'.@at0m: ...what about a better converter like waves L2?...
Oscillator A ran at 44.1k, Oscillator B at 96K
here you're totally on the wrong trackand I guess it'll be definitively a better result when converting 88.2 to 44.1 because of less rounding errors...
the case exists in reality, when Craig Anderton found in SOS tests that soundcard manufacturers just playback at twice the rate to fulfill the number...
This produces nothing but artifacts, but possibly those are experienced as an extra airy touch due to their hf nature
up and downsampling are complicated filter operations due to the nonlinear character of waveforms - it doesn't matter what relation the factor has to the original.
cheers, tom
ok, i see... at0m, can you confirm? if it is like astro said, i don't like these synth generated artifacts, i prefer the b.wav 
i still didn't get the converter stuff. it must, matematically, be much easier to convert a 88.2 to a 44.1 than a 96.0 to a 44.1 because you can "simple" devide it by 2. Man, i can't be that wrong
i still didn't get the converter stuff. it must, matematically, be much easier to convert a 88.2 to a 44.1 than a 96.0 to a 44.1 because you can "simple" devide it by 2. Man, i can't be that wrong
you areOn 2006-08-21 08:46, katano wrote:
...i still didn't get the converter stuff. it must, matematically, be much easier to convert a 88.2 to a 44.1 than a 96.0 to a 44.1 because you can "simple" devide it by 2. Man, i can't be that wrong
say if you sample a sine (you know that familiar picture)...
you can easily increase the horizontal steps, that is the time axis (sample rate), as it's linear.
but the corresponding points on the signal's amplitude curve don't share this attribute, you have to recalculate them properly according to the geometric rules.
the 'cheaters' just calculate the average between 2 points and hence the curve is distorted - not that much that you'd run out of the room, but it is...
cheers, tom
<font size=-1>[ This Message was edited by: astroman on 2006-08-21 08:57 ]</font>
trueOn 2006-08-21 08:54, garyb wrote:
nope, doesn't work that way. you can do the research if you want, but going from 88.2 to 44.1 is not just a matter of dividing by two. there are just as big rounding errors and artifacts in that process as in going from 96 to 44.1.
with 88,2 you would have twice as much values for every single amplitude
if you have lets say for example 1 and 3 for (88,2khz) you could add those 2 and then divide through 2
(1+3)/2=2 as the new amplitude value for both in 44,1
But if you have 1 and 2 as amplitude values for 88,2, you have a problem because you cannot save floating values on a fixed point system.... ( it would be 1,5)
So you have to decie if you take a 1 or 2 for this value and so you do some kind of interpolation, which is done by integrating the other amplitude values in front and behind the calculating ones...
Rounding is just a bad way and not a real interpolation...
quality depends on algorithm...
sorry for my "germanized" english.
<font size=-1>[ This Message was edited by: hifiboom on 2006-08-21 09:08 ]</font>
<font size=-1>[ This Message was edited by: hifiboom on 2006-08-21 13:04 ]</font>
that's right, "rounding" is just an easy handle for the idea.
there are plenty of suckers recording at 96k and filling hard drives like a jelly donut factory, and getting no better final results to cd.
how american of your associate. figuring that if he just hit it hard enough and threw enough numbers at it, it would have to be better.
there are plenty of suckers recording at 96k and filling hard drives like a jelly donut factory, and getting no better final results to cd.
how american of your associate. figuring that if he just hit it hard enough and threw enough numbers at it, it would have to be better.
if you want to "look at this sound", you can just view a 1280x1024 TFT at a resolution of 1024x768.On 2006-08-21 09:16, garyb wrote:
that's right, "rounding" is just an easy handle for the idea.
there are plenty of suckers recording at 96k and filling hard drives like a jelly donut factory, and getting no better final results to cd.
how american of your associate. figuring that if he just hit it hard enough and threw enough numbers at it, it would have to be better.
Everything seems smeared, and it does look like shit...
Even the best interpolation can`help out of this....
ATM a good 44,1 setup is much better than an "interpolation again and again" setup...
If you like to go 96khz, you have to go completly that way....
Recordings at 96khz with high quality converters( not just that one that interpolate internal to 96K), samples at 96khz, mixing at 96khz and master as Super-Audio CD or DVD-Audio....
Other constelations are mostly "pseudo-quality"...
ture words...On 2006-08-21 09:30, tgstgs wrote:
i record at 44 when end in cd and 48 when end in dvd (16bit)
a good microphon and its right position bring more than all bit and Hz
good vibes
in many ways this "96Khz improvement" is just a big marketing hype....
erm, we just go up to 192Khz...
"much better!!!", of course...
Next step will be 3xx khz and more that bring you THE 4th dimension in soundquality...
numbers sell quite well....