New data on the sampling-rate argument.
- Thread starter Where She Wept
- Start date
paladin shredder
Member
even though there isn't an audible difference for anything higher than 44.1, i think that while tracking you do have more headroom the higher the sample rate. i've also heard that if you mix down something recorded at a higher sample rate to 44.1, that your mixes are louder because of the headroom you had during tracking. this could be way off so correct me if i'm wrong.
tgs
Elder
I see this argument with "more headroom" thrown around so much that it even starts to bug me.
Well first and foremost you can't go over 0dBfs. So there is no way that it could be louder.
What you're getting is better dynamic range since the S/N rate will be lower. However, I can surely say that since I switched from recording on a 4-track cassette recorder, I have NEVER had a problem with dynamic range in my recording format, even back when I was recording on open reel. I've had hiss and noise in all their forms, but it came from various effect units and cheap mic preamps.
Since I switched to digital (with the old blackface ADATs) I have never recorded with anything other than 44.1 or 48kHz and I have never felt the need for it.
So ok, if you need a really extreme dynamic range and extremely low S/N ratio, then go for that, but if you have a system that actually is able to take advantage of it and produce no noise whatsoever, then well...I guess you're wearing gold diapers.
LSD-Studio
HCAF crusher
the opposite is the case,
by using higher samplerates you're as well adding more high-freq noise etc.
you're actually REDUCING the dynamic range
Yeah, dynamic range comes from increasing the bit resolution/depth (from 16 to 24), cuz there's 8 more bits in the digital "words" with which to represent the level of audio, thus increasing the precision and the amount of values you can represent between the minimum and the maximum (think of it as adding decimal places to a number)
tgs
Elder
Aaron Smith
Envisage Audio
That article was a very interesting read. I've hardly even considered upping the sample rate in the majority of the work I've done. I mixed an EP at 96k once, just because that was the way it came in, and I couldn't tell any difference at all...and I had to downsample for mastering, since I don't have the 96k Finalizer.
Doesn't digital processing supposedly sound better at higher sample rates? And if so, I would be interested in having this same kind of test set up, but instead of merely playing back a CD-quality source against a "high-quality" source, also playing back the sources after applying some degree of digital EQ- one with 44.1 kHz processing and the other with 192 kHz.
Am I just totally wrong though? Is it actually higher bit rates that make digital processing sound better? I can't remember right now.
Doesn't digital processing supposedly sound better at higher sample rates? And if so, I would be interested in having this same kind of test set up, but instead of merely playing back a CD-quality source against a "high-quality" source, also playing back the sources after applying some degree of digital EQ- one with 44.1 kHz processing and the other with 192 kHz.
Am I just totally wrong though? Is it actually higher bit rates that make digital processing sound better? I can't remember right now.
Torniojaws
They call me Juha
And in technical style:
16-bit has 2^16 = 65 536 possible waveform values ("steps in stairs")
24-bit has 2^24 = 16 777 216 possible waveform values
An analog signal has an infinite bit depth, or in other words it draws a smooth curve regardless of "zoom level" so to speak.
Now a digital signal has a finite level of bit depth, which you define when recording. It allows you to pick how small the stair steps are in height (sample rate defines how "long" the stair steps are), as you see from this pic. Analog signal is on top, and the digital model of the form is below:
More steps = more accurate digital model of the analog signal.
Omega_Void
Your Eventual Destroyer
Quote from the article linked above:
"Women, who were involved in less than 10 percent of the trials, did relatively poorly, getting just 37.5-percent right."
Any result which deviates significantly from the 50/50 "coin-flip" distribution indicates that the subjects did hear a difference in the audio! And women have a different neurological hookup for hearing than men, they process input from both ears the same, which men don't. Surely it's the assumption that the high-sample-rate audio would not sound worse than CD which is skewing that result...
An assumption which may not be justified.
"SACD ... has lower real resolution than the CD-system in the highest octave.
"SACD is in the high frequency range quite mediocre, even compared to a good CD-system one-bit DAC, and of course clearly inferior to a CD-player with a real multi-bit converter."
Quoted from this article.
If we want the truth about sample rates we need this sort of test to be performed using the same format at different sample rates. The article referenced above is relevant to consumers of audio product, less so to the people who are engineering the mix, IMHO.
"Women, who were involved in less than 10 percent of the trials, did relatively poorly, getting just 37.5-percent right."
Any result which deviates significantly from the 50/50 "coin-flip" distribution indicates that the subjects did hear a difference in the audio! And women have a different neurological hookup for hearing than men, they process input from both ears the same, which men don't. Surely it's the assumption that the high-sample-rate audio would not sound worse than CD which is skewing that result...
An assumption which may not be justified.
"SACD ... has lower real resolution than the CD-system in the highest octave.
"SACD is in the high frequency range quite mediocre, even compared to a good CD-system one-bit DAC, and of course clearly inferior to a CD-player with a real multi-bit converter."
Quoted from this article.
If we want the truth about sample rates we need this sort of test to be performed using the same format at different sample rates. The article referenced above is relevant to consumers of audio product, less so to the people who are engineering the mix, IMHO.
so compressing at a higher sample rate: would this mean the frequencies at inaudible ranges would affect the compression of audible frequencies? ex. a transient at an inaudible frequency (if they exist) would cause compression of an audible freq. not experiencing a transient at that point?...
Omega_Void
Your Eventual Destroyer
Yes - unless the particular compressor you use filters the control signal. Some give you the option, some do it automatically without telling, some don't do it at all.
Hehe.. This should elaborate a little bit.
"The difference between the Neve 1073 EQ and the 1073SE EQ is in the fact that the 1073 up-samples to either 176.4 kHz (for 44.1 or 88.2 kHz sessions) or 192 kHz (for 48 kHz or 96 kHz sessions). Up-sampling is great for EQs, because it ensures that the EQ curves in the plug-in match the EQ curves in the modeled analog EQ. As you probably already know, sampling rate is proportional to the amount of DSP needed in a plug-in. Additionally, some DSP is needed to do the up-sampling and then the down-sampling back to the session's sampling rate."
This article was interesting and I think what they are doing is proving that a new playback format is not necessary. But it does nothing to show whether or not recording your individual tracks recorded at higher rates matters or not. I personally have settled at 48 K.
What really matters here is not dynamic range. How many of us use more than 10db of dynamic range anyway? What matters is the Nyquist frequency. This is where the digital low pass filter is placed. The Nyquist frequency is a sharp cut off that is at half the sample rate. 44khz has a Nyquist freq of 22khz. 96khz has its cut off at 48 khz.
Well our hearing stops at about 20-22 khz anyway so what is the difference? Many of you know that when you turn on a high pass or low pass filter at any frequency, you send a phase distortion ripple down the line. So if you make a sharp 40db per octave slope at 22khz, you send distortion down the line to say 16khz. I'm guessing at the exact numbers because I cant remember the exact specs. But this is the point.
So the higher up you go, the less you interfere with the human audible range. So lets go back now. When we are recording our individual cymbal tracks, does it matter if we are recording phase distortions into our tracks that are now permanently embedded into the sound? I think it does.
This is not the same as what they were testing for in that article. They are taking a recording that is already mixed and mastered at 96k for example and then doing a DA/Ad conversion and outputting it at 44.1 At this point, what the engineer did during recording makes no difference.
Now we need the same kind of rigor applied to to a test where the entire process of tracking, mixing and mastering is done at different rates. I suspect that it will make a larger difference than what was found here.
By the way, there is something else worth mentioning about 96k. I have found that for metal, it sounds worse than 48k. I feel 48 k is substantially wider than 44.1 but when I do projects at 96, this sample rate imparts a high-end emphasis that ends up making the recording sound bass-light.
We interpret sound contextually. When we hear a recording that has a lot of treble, we often mistake it for being bass-light. And when we hear a recording where bass has been added, we often call it dull sounding. I have found that these 96k recordings allow such a wide high frequency extension that they give the recordings a bass-light sound. Not so cool for metal. 48k is my favorite. And I always do analog mastering where I do a DA/AD conversion so there is never a sample rate converter applied.
Colin
What really matters here is not dynamic range. How many of us use more than 10db of dynamic range anyway? What matters is the Nyquist frequency. This is where the digital low pass filter is placed. The Nyquist frequency is a sharp cut off that is at half the sample rate. 44khz has a Nyquist freq of 22khz. 96khz has its cut off at 48 khz.
Well our hearing stops at about 20-22 khz anyway so what is the difference? Many of you know that when you turn on a high pass or low pass filter at any frequency, you send a phase distortion ripple down the line. So if you make a sharp 40db per octave slope at 22khz, you send distortion down the line to say 16khz. I'm guessing at the exact numbers because I cant remember the exact specs. But this is the point.
So the higher up you go, the less you interfere with the human audible range. So lets go back now. When we are recording our individual cymbal tracks, does it matter if we are recording phase distortions into our tracks that are now permanently embedded into the sound? I think it does.
This is not the same as what they were testing for in that article. They are taking a recording that is already mixed and mastered at 96k for example and then doing a DA/Ad conversion and outputting it at 44.1 At this point, what the engineer did during recording makes no difference.
Now we need the same kind of rigor applied to to a test where the entire process of tracking, mixing and mastering is done at different rates. I suspect that it will make a larger difference than what was found here.
By the way, there is something else worth mentioning about 96k. I have found that for metal, it sounds worse than 48k. I feel 48 k is substantially wider than 44.1 but when I do projects at 96, this sample rate imparts a high-end emphasis that ends up making the recording sound bass-light.
We interpret sound contextually. When we hear a recording that has a lot of treble, we often mistake it for being bass-light. And when we hear a recording where bass has been added, we often call it dull sounding. I have found that these 96k recordings allow such a wide high frequency extension that they give the recordings a bass-light sound. Not so cool for metal. 48k is my favorite. And I always do analog mastering where I do a DA/AD conversion so there is never a sample rate converter applied.
Colin
AudioGeekZine
arsehole know-it-all
let's add some expert (and some more amateur) opinions to the discussion
96khz...What's it good for?
Just the first 7 pages though, after that it got hijacked and never got back on topic.
96khz...What's it good for?
Just the first 7 pages though, after that it got hijacked and never got back on topic.
Similar threads
