PlanetZ forums - scopeusers.com

...hi to all i want know is preferred to set scope to work at 48khz 24 bit or 44.1 24 bit ? 48 khz use more bus pci? i know that if use 48 khz i the scope no need to resample tha file. if i use the 44.1 the system need to resample it to use. is real info or if i work at 44.1 . The at 48 24 bit is more better than 44.1 24 bit. i need to use48 ?
thanks...

u don`t have to care about the pci bus at all. 48 or 44, its less important.

since the final product will likely be at 44.1k, i say 44.1k.

the more data you have, the sharper the effects work will be : that's just science. The more you will use EQS, effects, etc ... the more you will then benefit from a high sample rate... Of course, you eventually will have something at 44.1 but for example 30% of 1000000000 has nothing to do with 30% of 1000. For that reason, even if you shift from 48 to 44.1 the original file contains more data. That's just an example of course ...

Will you ear the difference though ? Not sure. It depends on many things (human ears, gear, final monitoring,...).

Of course, there are many theories but science does not always match reality I am not sure you will find 1 thruth ...

I use to work at 48 but takes quality (preamp, mikes..) remain the most important things for sure ... really !

Jo

bill3107 wrote:...but for example 30% of 1000000000 has nothing to do with 30% of 1000 ...

sorry, but you are entirely wrong here - they even are identical
a perfect example to proove the nonsense of bit deepth without paying attention to the context

no matter how many decimals you use, it will never be any more accurate than 'almost one third with an infinite rest' 30,3333...
for a representation of a continuous signal the integrity of the relative value 'points' is important - not their absolute figures.

for (most) processing an increased precision is appreciated as rounding errors may add up terribly in loops, but once the processing has finished it doesn't matter at all to continue with the result stripped to a lower resolution.
Further processing will just start from a new point, but still keep the signal's integrity (if it's a quality processing)

since disk space isn't a concern today, there's of course no need to fall back to 16-bit continuously and for convenience one simply puts everything in a 24bit file.
Nevertheless under optimal conditions just 20 of those bits can be used to represent your signal, the rest is randomness.
Without a high precision studioclock don't even think of representing 20

cheers, Tom

i understand. But as sample rate gives you the number of samples "taken" per second - the higher the SR, the bigger files are - this means that there is more information as soon as you increase the sample rate. You are right, my example was not very good though ...

24 bits gives a bigger dynamic range which is clearly an improvement compared to 16 bits and increasing the sample rate is quite irrelevant but i thought (i mean thta seems obvious to me but may be i am wrong ?) that 48 or 96 was a good idea for processing stages as an algo applied to 48000 samples gives a better result than 44100 .... ??? Of course the difference is actually very very small so i admit it is quite useless (except audio+video project because of synchro)

But my ears are wide open .. Could you give me more information Tom ? thanks for your comment...

If anyone can hear the difference between 48k and 44.1 then you have bat ears. It really is a non-argument. i had a DAT that would only record at 48. Did it sound better than my CDs? No. Record at 24 bit and do the following:

1. Record at 44.1 if you're going to end up on CD.
2. Record at 48 if you're doing anything for television, although any Avid con convert CD to 48k anyway so even that is a non-issue, or if it's for DVD.
3. Use the extra time gained from not having to ponder over this crap by having a pint with your mates down the pub.

actually it were just those 'badly choosen' values of your example that made me look at this from a different angle - it made things look less abstract (at least for myself)

both the sample rate and the bit deepth are (probably) equally important in processing algorithms.
Filter design is significantly easier at higher sample rates - but you will have to meet the timing specs...
a high numeric precision is required for anything that runs in iterating loops which modify a certain item over and over again in every pass.

on the other hand that's just DSP processing 101 and no big secret, even for a mathematically humble person like me...
the real details are somewhat beyond my own skills - nevertheless I'm fascinated by the mathematical approaches to the subject

I also wouldn't bother about the 44.1 versus 48k difference, but choose what's required in the data exchange process (same with the other formats)

cheers, Tom

take a value f.e. 3333 dez. at 24 bit and convert to 16 bit;
hers the algo: x=y/256 supriesed?
now convert back to 24 bit x=y*256
result is still 3333 at 24 bit
you loose nothing!!!!
----------
thats why my 8bit akaisamples sound so sweet
and now have fun with 64bit vista
-----------
advatage of 48000 Hz to 44100:
you are able to record 24000 Hz to 22050;
if you are able to hear even the 22050 contact 'book of records'
----------
have to add there are differences for the code to work with;
depending on the quality of the developer this could make difference for adding effects;
the matharguments are only theroretical nature but sell . .


good vibes from vienna

tgstgs wrote:-----------
advatage of 48000 Hz to 44100:
you are able to record 24000 Hz to 22050;
if you are able to hear even the 22050 contact 'book of records'
----------

yes, even further, show me the loud speakers which reproduce any of that....

i agree .... there are limitations to theory. I wil think about that as i am used to 48 but that's just an habit. And wow if i use 44.1 instead of 48 i will have more DSP power !!!!! This is a good argument too

jo

http://digido.com/


here a guy explain well what happen with digital quantization and sample rate.

if you get enough dsp, always work/mix with the higher resolution and bit depht, and let the mastering studio do the dithering/sample rate conversion...
bit depht is faraway more important than changing sample rate from 44 to 48, that will be more significant if you go from 44 to 96 or 196kHz. On the other hand, Sample rate is important cause bigger sample rate increase the frequency range that is rendered (for the ADC-DAC stage mainly - not sure what about that in digital processing-if someone know...?)

read the text carefully!

1. hes seller, wants to make you buy ->

2. its simple math
take my example in above post and tell me im wrong . .
the so called noise distance is not relevant becouse scope is 32bit anyway,
once in digital domain to lower means lowers noise too; distance stays the same;

---------

for audiorecording 44100Hz is ok.
the analog circuit in front of adc is more important;

if you want to improof your recordings conzentrate on the preamp, cable, mics and the room your recording;

reminds me of the original formant forcing technix for clasical voise recording
http://www.studio96.de/story%20040806%2 ... mants.html

good vibes from vienna

tgstgs-you are exactly right.

2. its simple math
take my example in above post and tell me im wrong . .

sorry, i'm not truly involved in algorythm...
but if you do your algo with eg 16744447, you get as result "65407,99609375", now how do you write this number with 16 bit?
maybe you will write "65408"? now go back to 24bit:
"16744448"
...so it's happening some datas modification during the bit depht change...

just what i want to say is that it's sound better to make eg a fade on a parameter from a higher value as 16777215 (24bit) to zero than from 65535 to zero ... the sweeping will be smoother in the first case. Here the bit depht are really significant. I'm not an engineer but I can ear the difference when I drive my cem3378 filter cutoff with a CV comin from digital generator thru a 8bit DAC or thru a 12bit DAC

but we are not talking here about sample rate !

from my side, I record all at 44kHz... to get time to drink much more with friends..

I think that if everyone do recording and processing at higher quality then do dither only one time for the final stage of mixdown rather than doing recording/proc. at the quality of the final media, it's cause the result are better in the first case.

the second things is that you own for the future your original tracks in a quality that maybe some media will handle in some time.



best regards.

EDIT:-----------------------------------------------
see no model matching your earings
but who knows
maybe future technologie makes a hit out of the rounding error
max bits at max samplerate and send to the guy

good luck and
good vibes from vienna

attn pub friends and dsp savers, can any of you plz tell me why people got into trouble giving you the option to record to 96kHz? maybe they wanted to make music for dogs or sth. Oh, and by the way, I'm sure SACD buyers and sample converter developers appreciate your opinion.

I record everything at 48k. Since I usually track to an Alesis HD24, it makes sense because the HD24 is spot-on accurate at 48k but in theory has some (very minor) long-term drift at 44.1. And, years ago, when my ears were younger, I compared a bunch of sample rates using SunRize's Studio 16 on the Amiga. 48k seemed to sound just a little better than 44.1. And I did not hear a penalty when converting from 48k down to 44.1,

The big advantage though is that by recording at 48k and giving the talent 44.1, I am still in possession of the 'highest and best' copy at 48k, while the talent gets the more usable 44.1k.. With me having a copy with - theoretically - more information, at least the talent has an incentive to go through me for future work on the project instead of simply making a 1 to 1 copy of what they already have and giving the work to somebody else.

best,

john

Hi,

From a technical point of view quadratic transform used in FIR and IIR filters is correct if the sample frequency is far from the filter frequency and whe should not forget that shanon theorem was devised for telephony app ... you should look to the distorsion curve at hig frequency!

We can also think that, for stereo signal the sample frequency have an incidence on spacial quantisation. Another point is the slew rate or dv/dt of the sampled signal which is affected by resolution and sample frequency.

All of this is theorical as ultimately your signal will depend on the input ADC stage (for acquisition) and output's DAC and filter (for restitution). So the same numeric filel can sound differently on another machine (with different output stage).

so ... do as you like!

cheers

Michel

Processing greatly improves at higher bit- and samplerates.
There are even mastering engineers that upsample client files prior to processing and downsample to CD-quality later on.

With cheap ADC's it makes sense to record at high samplerates and then downsample later on via software, because good hardware filter-designs are very costly.
By recording at higher rates you move the nyquist frequency way up in ultra-sonic territory which will clean up the hearable range.

Great software filter-designs are rather cheap. You don't have to pay much for r8brain PRO i.e which easily surpaces any hardware filter.