24 vs 16

[CoolCat]

can anyone post a couple "different" pics that shows the real difference between the sample/freq rates etc..?

i found the visual easier to understand. obviously some disagreement on that presentation.

 

[Farview]

The waveform at the output looks the same as the waveform at the input.

Other than that, those graphs just need twice as many horizontal lines.

 

[CoolCat]

so are you saying the graphs are ok, but the horiz lines and little dots aren't really correct?

how about this one...

 

[Farview]

The dots are correct. There really isn't a scale on the horizontal lines, so it's hard to tell.

The thing that is really wrong with all of the graphs that you posted is they give the impression that the reconstructed wave is stair-stepped. It isn't. The reconstructed wave looks just like the original. (with the exception of frequencies above the low pass filter and the nyquist frequency)

 

[CoolCat]

been reading more on the 16versus 24, web searches..
almost all the charts on the web show the "stairstep" to represent the new, smoother conversion versus the old. one did show a smoother output but that was after the internal filtering circuit (i think).

more you learn the less you realize you knew...

somewhat skeptical..
So you say all these companys and articles are wrong
for showing the "stairstep" instead of the reconstructed signal,

You think it's all just a marketing plot to present the data for better simplicity to the masses;
old product= bad and choppy...the good=new and smooth!!....= BUY MORE NEW WIDGETS NOW.

 

[SouthSIDE Glen]

A couple of points that need clarification here I think...

First, the resolution of the "stairstep" is not a result of word length/bit depth, that is a result/representation of sample rate. The narrower the width of the stairs, the higher the sample rate.

Second, the stairsteps are meant to be a graphical representation of digital *values*, not an actual representation of a digital "wave"; i.e. there are no actual vertical or horizontal "lines" between the digital values. In those graphs, it might be more illuminating to indicate each digital value as a point, with it's vertical position representing it's amplitude as measured in the 16- or 24-bit word value, and it's horizontal position representing it's place in the horizontal timeline, with the horizontal spacing between dots being determined by the sample rate (the higher the rate, the closer the horizontal spacing.) There are no actual amplitude "lines" between the dots because there are no actual values between them.

The difference is in the accuracy of the location of the points or dots as related to the original analog amplitude at that position.

G.

 

[Farview]

If you recorded a sine wave digitally and it didn't come out the other end a sine wave, it wouldn't do you any good to record anything digitally. In fact, CDs with test signals would be useless.

The stairstep thing is analogus how the data is stored, but it gets reconstructed as a smooth wave. It's not really marketing hype, it just isn't telling you the whole story.

 

[CoolCat]

thanks G and Farview.
found one last pic, i'll post...(don't want to use up to much of my allotted jpg space!

this shows the rebuilding and reconstructed. at least I'm getting the terminologys a bit better understood....

like all things you can go really deep into the science, eventually when my brain starts sizzling, and in parellel,
the marketing can predictably "trash the old" and "praise the new"...understandable.

 

[crosstudio]

If you recorded a sine wave digitally and it didn't come out the other end a sine wave, it wouldn't do you any good to record anything digitally. In fact, CDs with test signals would be useless.

The stairstep thing is analogus how the data is stored, but it gets reconstructed as a smooth wave. It's not really marketing hype, it just isn't telling you the whole story.

dude you are missing the point. the stairstep shows you how much of the original signal will be present in the recorded signal.

what gets "recorded" are the stairsteps that are represented in the graphic. what gets heard as a result of the Digital-to-Analog converters is the sine wave that you speak of... BUT and this is a HUGE BUT... the later recorded wave is not the original wave. it is the D/A's interpretation of the stairstep representation of the original wave. if it were an exact duplicate of the original wave, i would be recording in 8bit/11khz and save myself a heap of disk space.

oh, and as of this week, I now record in 24/88.2 instead of 24/48.

Now that I can archive my tracks to NAS and also to DVD, i don't have as much of a concern about the size of the files. I'm only concerned about my HD being able to push 88.2.

 

[littledog]

Cross -

I'm sort of in the same boat, but I still record at 24b/44.1k

With me, I'm just archiving to DVD, and only the 4.2GB size. So I still hate having sessions that are 20-30GB, because I hate backing up to 7 or 8 DVD's, instead of 3-4.

I guess I'm just lazy, and don't feel like my audio suffers all that much at 44.1

 

[Farview]

if it were an exact duplicate of the original wave, i would be recording in 8bit/11khz and save myself a heap of disk space.

If your original wave was under 5k and only had 40db of dynamic range (and you don't mind a high noise floor) it would be the original wave.
The wave that gets recorded still gets reconstructed properly. Anything above the nyquist frequency will be filtered out before the wave gets recorded.

The stairstep thing never happens. The storage medium doesn't conncect the dots, the DA does. Your waves are not stair-stepped in any way.

 

[crosstudio]

littledog:
i don't have sessions that span 7 or 8 DVDs. although i play in a band, i'm essentially a one-man band (the black Dave Grohl... well... the broke Dave Grohl) and i record 1 song at a time.

the longest/largest track count song that I have (real drums, real congos, real timbales) is a 8 minute DC Go-Go song in 24/48 that is 2gb.

incidently, the one song i've done in 88.2 thus far is 4:27 long and thus far takes up 1gb for 13 tracks and i haven't even laid the vocals yet.

of course, it would be much smaller if i looped the repeating audio parts instead of just recording from MIDI to audio.

I'll let you know in a couple of weeks whether I can hear the difference. I'm not subjective with new toys, and 88.2 makes my old toys seem new.

 

[NYMorningstar]

littledog:
i don't have sessions that span 7 or 8 DVDs. although i play in a band, i'm essentially a one-man band (the black Dave Grohl... well... the broke Dave Grohl) and i record 1 song at a time.

the longest/largest track count song that I have (real drums, real congos, real timbales) is a 8 minute DC Go-Go song in 24/48 that is 2gb.

incidently, the one song i've done in 88.2 thus far is 4:27 long and thus far takes up 1gb for 13 tracks and i haven't even laid the vocals yet.

of course, it would be much smaller if i looped the repeating audio parts instead of just recording from MIDI to audio.

I'll let you know in a couple of weeks whether I can hear the difference. I'm not subjective with new toys, and 88.2 makes my old toys seem new.

Why were you recording in 24/48 and not 24/44.1? When you dither down to a cd format of 16/44.1, do you change sample rate first or do you do it all at the same time? Is there any additional noise introduced when you have to do the additional step?

 

[CoolCat]

The stairstep thing never happens. The storage medium doesn't conncect the dots, the DA does. Your waves are not stair-stepped in any way.

i think i get it a bit more?
with G's explanation and Farviews comment, both agreeing...i think.

so as G said there isn't an actual "digital wave", its just a way to represent the A/D c.

in other words your saying I wouldn't see a "choppy stairstep digital wave" on a oscilliscope, it doesn't exist?

thats were i was off, i guess.
I thought this was a actual "stairstep digital wave form" captured from an oscilliscope screen.
Taking the signal coming off the output of the first stage converter before filtering or the "mathematical rounding off" before the final DtoA took place.

so your both saying the real output (oscope screen) of the AD converter would be a flat "pulsed line"?
O's and 1's with no "digital stairstep waveform" visual. is this correct?

the Digital Waveform...
it is a pretty good representation to sell something though.
little tiny smooth bumps for the good stuff and HUGE CHOPPY caveman like lines for the old stuff.

right up there with DIGITAL READY HEADPHONES?


edit: apologize for the caveman comment.

 

[Farview]

Taking the signal coming off the output of the first stage converter before filtering or the "mathematical rounding off" before the final DtoA took place.

I don't understand the sentence. The filtering is analog filtering on the way to the A/D. The D/A doesn't filter anything because it was all filtered in input.

so your both saying the real output (oscope screen) of the AD converter would be a flat "pulsed line"?
O's and 1's with no "digital stairstep waveform" visual. is this correct?

You are correct.

the Digital Waveform...
it is a pretty good representation to sell something though.
little tiny smooth bumps for the good stuff and HUGE CHOPPY caveman like lines for the old stuff.

right up there with DIGITAL READY HEADPHONES?


edit: apologize for the caveman comment.

It is misleading.

 

[crosstudio]

Why were you recording in 24/48 and not 24/44.1? When you dither down to a cd format of 16/44.1, do you change sample rate first or do you do it all at the same time? Is there any additional noise introduced when you have to do the additional step?

i was recording in 48khz so that the extra 10% worth of information will help me get the most accurate interpretation of the sound and have more to work with during all the mix processing (EQ, compression, autotune, delay, reverb, etc...).

SEKD RedRoaster [i.e. Samplitude Master] converts the stereo mix to 16/44.1 just before the CD burn process. It also gives you the option to save the individual stereo mixes to 16/44.1 WAV files.

 

[Farview]

The extra 10% with 48k gets negated during SRC. in fact, doing it this way guarantees that your audio will go through two stages of anti-aliasing filters instead of just one. (the one on input that gets rid of everything above 24k, and the one during SRC that gets rid of everything above 22.05k)

BTW, the difference between 22.05k and 24k is less than one musical whole step. (the difference between a C note and a D note, for example)

 

[bennychico11]

to add to that, I would think the extra filtering on the high frequencies from converting would diminish them at a faster slope...although I guess they would be minimally noticed (maybe a couple dB?)

from mshilarious' graphs here
near the top of the human hearing range, with 44.1kHz the slope starts around 17kHz and by the time you reach 20kHz you've lost 16-17dB. With 48kHz the slope starts later and you only lose about 2-3dB at 20kHz

That's enough for me to go with 48kHz....but then again, everything I do remains at 48kHz anyway.

 

[Farview]

The musical difference between 17k and 20k is almost 1 1/2 steps. You ears are probably attenuating harder than that anyway.

Obviously, if your target rate is 48k, then that's what you should record at.

 

[Chibi Nappa]

In decimal, the analogy would be, which number is more precise:

1 or 1.00000000?

In this case we know the numbers are exactly equal.

Oh jeez. Scientists and engineers all over the world are groaning in agony right now.

1 doesn't even gaurantee the accuracy of the "1's" position. It could be anything greater than 0.5 and less than 1.5

1.00000000 gaurantees accuracy to 7 decimal places. Huge difference.



Nothing to do with audio, but don't ever say that 1 is exactly equal to 1.00000000 around a scientist or engineer! You'll give him a heart attack!