48kHz vs. 44.1kHz Sample Rate @ 24 Bits

Mshil...

I use a dbx Quantum I, I'm not sure what's inside so finding a data sheet is out. I just have to go with what I see and that there is no change in the filtering from 44.1 to 48. If there is, then its a small one.

Bob the Mod Guy
 
Bob's Mods said:
Does anyone know, in those tape based studios, when those began to roll off the high end?
Click to expand...
It seems to me that the high end started to roll off about 16-17k. It has been a long time, so I could be wrong.
 
Farview said:
It seems to me that the high end started to roll off about 16-17k. It has been a long time, so I could be wrong.
Click to expand...

I think it also depends on the tape speed. 15 IPS will give a fat bass response, but the high frequencies start to roll off. 30 IPS will get flat response to 20k, but everything moves up one octave. The bass drops out.


sl
 
I keep thinking that the roll off at 44.1 may be no better than the roll off of tape based recording FWIW. The belly aching over roll off at 44.1 may be undeserved if it matches that of a tape recorder.
 
( conjecture )

Say top rate src was available would the conversion redeuce the noise floor, and contribute to a crisper sound?
 
nonovice said:
Say top rate src was available would the conversion redeuce the noise floor, and contribute to a crisper sound?
Click to expand...

The noise floor (dynamic range) is more effected by bit depth, in this case 24 bit, rather than sample rate. Higher sample rates get you better accuracy and better plugin resolution. There are more bits because there are more sample words per second. More bits overall provide for more precise math which is what mixing in the box and plugins are all about. The value of these additional sample words is very dependant upon the quality of your convertors which is another topic altogether. This may not mean much to you if your happy with your mixes and plug in performance at 44.1.
Its my understanding that DVD audio at 96 kHz sounds killer and hands down beats 44.1 kHz. If your system is set up from the get go to do DVD audio then I guess there is a marked improvement in sonics. If you track at 96 and the final product is 44.1 it may not be worth the effort unless its going to end up in DVD audio at some later time.
Audio may move up to 24/96 someday which I'm guessing is a big jump in quality from 16/44.1. Anything less may only have a marginal improvement in sonics depending on material and track count.
Do two demos, one at 44 and the other at 48 or 96 and compare. Do two or three different style demos this way and compare. Is there some meaningful improvement in your mixes or plugin performance? Is the quality maintained when you downsample? Is any improvement in sonics a result of pushing the low pass filter farther out to 30kHz from 17.5kHz or from just plain more bits to do math or both? Results would be expected to vary widely because of different gear and user finesse. Is 24/96 where we need to be for that next leep in performance?
Gear now is topping out 192kHz but the majority of home recordist record at 44.1 anyway. Gear manufacturers provide a variety of sample rates for us to choose from yet we continue to work at 44.1. Are they trying to tell us something we're not getting or are they just selling it cause we're buying and don't know any better? Who needs 192 anyway? If higher rates don't buy us anything and we are stuck at 44.1 for the very long haul, whats the sense in all this technology just for the sake of technology?
Sorry, I've been ramblin on.

Bob the Mod Guy
 
OK here is my test

The test tune was the demostration program on my trusty ancient Yamaha YPP-15 piano (c. 1990). It's a crap keyboard patch, but the performance is exactly the same, and the harpsichord sound had plenty of high-frequency content.

It was run direct into a Presonus M80 for about 30dB of gain, then into an Alesis AI3 converter. The three files are 44.1, 48, and 48 converted to 44.1 using Wavelab's SRC.

The results were rather unusual, as I don't have a theory to explain them, I'll just post the links (24 bit .wav files):

44.1

48

44.1 SRC
 
After listening to those three, I could certainly hear differences:

44.1 native sample sounded best

48 native sample was next in line

48 to 44 sounded the hoakiest

There seemed to be an additional element of cheeze factor in the native 48kHz sample. The down converted sample sounded worst. Voxengo's R8brain convertor does a good job. You may wish to try it.

Bob
 
Bob's Mods said:
After listening to those three, I could certainly hear differences:

44.1 native sample sounded best

48 native sample was next in line

48 to 44 sounded the hoakiest

There seemed to be an additional element of cheeze factor in the native 48kHz sample. The down converted sample sounded worst. Voxengo's R8brain convertor does a good job. You may wish to try it.

Bob
Click to expand...

What was odd about the samples to me was an FFT between 44.1 and 48 was very dramatically different at ALL frequencies. Doesn't make sense. I figured, well I guess the converter just doesn't have a very good implementation, that's why I tried the SRC. The FFT between 44.1 and the SRC is nearly identical :confused: They sound a little different to me, but I can't identify them in a blind test. 48 stands out though, and to me is more accurate to the instrument--I have the pleasure of listening to it live! The attack of both 44.1 samples is too soft in comparison.

Again, the only theory I have to describe such an obvious difference is converter quality, but that wouldn't explain why the SRC ended up with practically the same file. It is highly unlikely both converter and SRC would make identical errors.

I'll have to repeat the test with different converters, maybe next week.
 
From a theory point of view, the 48 kHz sample should be more accurate. Maybe you should trust your ears rather than a computer plot in trying to determine fidelity.
The 44.1 sample actually sounded abit more pleasing to me than the 48 sample. The 48 sample recreated that ditty in all its cheezy glory.

Bob
 
Bob's Mods said:
From a theory point of view, the 48 kHz sample should be more accurate. Maybe you should trust your ears rather than a computer plot in trying to determine fidelity.
The 44.1 sample actually sounded abit more pleasing to me than the 48 sample. The 48 sample recreated that ditty in all its cheezy glory.

Bob
Click to expand...

I refuse to reject measurement of audio; it is a critical tool. If we are going to ignore our eyes, let's be consistent and discard all the LEDs and VU meters in our studio--but that is not for me.

Anyway, after spending more time in analysis, I noted that Wavelab has a flaw in its FFT graph that causes two waveforms at different sample rates to display incorrectly in comparison--one wave, in this case, is shift by a half-step, which embarassingly enough, wasn't obvious until I cranked the resolution way up.

As a workaround, I converted the 44.1 file to 48, this time using the higher quality bundled plug rather than Wavelab's stock tool. The theory being that the high frequencies that are attenuated in the 44.1 original aren't going to be restored by an SRC upsample.

And look here:

Graph

Thus I believe I've proven what we set out to look for--at least for the AI3, the high frequency attenuation is measurable and it's audible--whether or not one prefers it :)
 
It appears your graph is displaying high freq info above 18.5 kHz. This is your upsample demo at the 48Khz sample rate? It is interesting. Maybe in a mix I would expect data to be there but on a single track?

Maybe the effect is similar to software compression (zip files) where certain bits can be dropped to reduce the size of the file. When the file is returned to full size it becomes normal again. The common wisdom has it that you can down sample but you can't upsample.

Bob
 
I looked at another 48kHz sample and that one appears to roll off sharply at 20k. The roll off may occur at a lower knee depending upon the audio content. Just how important is the non audible portion of the mix to the overall character of it's sonics?

Bob
 
Bob's Mods said:
Maybe the effect is similar to software compression (zip files) where certain bits can be dropped to reduce the size of the file. When the file is returned to full size it becomes normal again. The common wisdom has it that you can down sample but you can't upsample.

Bob
Click to expand...

Yes, exactly right. Those frequencies that were attenuated at 44.1 are gone, I only upsampled because that was the only way I could graph 44.1 against 48.

As for audibility, this is audible range. I can't directly hear a sine wave of frequencies that high, but both of us must be able to sense them, because we both heard on audible difference in the files. The portion of the FFT below 18.5 showed no material difference, so I left it out for clarity.

As for the signal vs. noise issue, these charts are just above the converter's rated SNR of 100dB, but the behavior of the sample rates is practically identical to the published specs of the converter chip (attenuation, but no other distortion), just a bit larger in magnitude.
 
I think maybe we are just splitting hairs. My guess is the next big jump in performance is 24/96 DVD audio. Anything else is more or less chump change.

For me the audio range is what I can actually hear. Not the harmonic content above. But yes, that stuff does influence the character of what you hear in your hearing range. Last time I checked my hearing some years ago, my hearing maxed out at 15k. Its probably somewhat less now. The limits of your hearing decays with age. Same thing for your voice too, its range decays with age. Have you watched the BeeGees try and sing their stuff on PBS? You can see them strain to make the highs that in a younger day were probably a piece a cake.

Bob the Mod Guy
 
You must log in or register to reply here.
Back
Top