Fidelity question 2: Digital 192, 96, 48 anyone counting?

[cybermooks]

I know that theoretically 192khz on newer equipment should produce better signal fidelity. Same with 96khz, but to use it is costly on some devices as you lose half the inputs to reach that level. I've settled on 24/48 as the best compromise. Now of course the 24 bit parameter is far more important than the khz, and any of these take you beyond the threshold of human hearing.

 

So the question is this: is anyone regularly using these high rates, and can you really hear any difference over 24/48?

[Yoozer]

Originally posted by cybermooks Now of course the 24 bit parameter is far more important than the khz,

You're right.

and any of these take you beyond the threshold of human hearing.

You're not entirely right. Even 44khz is beyond the treshold - what matters is that it's all about the aliasing

 

I'm not going any further than 16/44, but mainly because only my soundcard can handle 24/96. RGC Z3ta+ didn't sound shabby at all at that, though - packed a lot of power, almost as much as the Q.

[funkstar]

Originally posted by Yoozer You're not entirely right. Even 44khz is beyond the treshold - what matters is that it's all about the aliasing

44KHz samples means that it's only just above most peoples hearing limit - Nyquist's Theorem gives us a maximal frequency of 22KHz in the sample, and normal people tend to max out around 15-20 KHz.

[ChetSinger]

 

Originally posted by cybermooks So the question is this: is anyone regularly using these high rates, and can you really hear any difference over 24/48?

 

I used to record at 24/96. Then I started working with pre-recorded 44.1k material, and discovered that my 44-year-old ears sure can't hear the difference between 44k and 96k.

 

So I now record everything at 44.1k. But I agree that bit depth is important. So I record at 24bit, and keep all effects and mixdowns in my DAW software at 32-bit floats.

 

Chet

[cybermooks]

Right that's what I was thinking of; so the question is whether there's any value to these higher rates.

Ok if it is not hearing but aliasing, when does that factor out, i.e. what level?

[cybermooks]

Yes but until you get to 44.1 it would seem to be better to work with higher rates and then dither, i.e. in Logic.

 

True enough CDs are 16/44.1 although new formats promise 24/48, so it might be wise to use that in working anyway.

 

There have always been those LP audiophiles who claimed that there was "something missing" from the CD sound (which the rest of us might attribute to the noise or digital coldness). However some of the same people are now satisfied with 24/48. There was a group or two who objected to CD sound but found 24/48 fine. So my assumption is that yes, it is likely that some people can hear the difference between 16/44.1 and 24/48. However, I can't believe anyone can hear anything different higher than that.

 

On the other hand in terms of resolution consider the following analogy. Digital pictures still have far less resolution than film. In a 4X5 picture, or on a computer screen at say 1024X768 you can't see the difference. But if you go to 8X10 or 20X24 you can see the difference if the pixel count is low. Information has been digitized but it is not lossless; information is still lost as opposed to film. Or when something is scanned it is digitized; if the resolution isn't too great it will "interpolate," i.e. attempt to add missing information to the closest pixel.

 

I'm not sure how this translates into audio terms, but theoretically digitizing must similarly lead to some "loss;" because you are always sampling, at whatever rate, rather than replicating. The question is whether it is audible or not or whether makes any difference.

 

On this basis it would seem that good old analog is 100% reproduction, at least from a master, if only the noise were possible to get rid of. There was a lot of progress on this in tape from Dolby and DBX, but kind of fell off when things went digital.

 

All of this is a propos recording your masterpiece. You don't want it to suddenly sound like {censored} in twenty years because information is missing, unless that audio can somehow be interpolated as well.

[cybermooks]

Agree as far as storage goes, unless you use stone or something. But I'm not sure in what sense you mean analog "colors" the sound. I know there are a lot of methods, outboard and plug-in to restore "analog warmth" to digital sound, although I'm not sure you can actually restore anything that's lost but rather color it with an effect.

[MP3Chuck]

While the 22,500Hz upper limit of 44.1kHz recording may be about the highest frequency people can hear, it would still be a good idea to use 96kHz (or higher) if it's an option. Instruments, most notably the higher frequency things like cymbals, can create harmonics that reach over the 22,500Hz mark, and all those frequencies come back as low frequency noise when you're using a lower sampling rate.

 

I'm assuming (I havn't tested) that dithering down to 44.1 as your last step alleviates this low-frequency noise since the process would actually remove the frequencies greater than 22,500Hz.

[Boom]

The only problem using sample rates higher than 44.1k is when you have to convert them to down to 44.1 for a CD, 48k doesnt divide evenly into 44.1 and your sound quality degrades as a result. I guess the only way to go would be 88.2, or re-record it via analog (it's not as evil as some people think, converting to analog and back to digital, if you have good quiet gear).

 

When I first started out I used 48k and when I converted them in Soundforge to 44.1 it sounded harsh, I had much better results starting at 44.1 and that's what I still use and will continue to use until something like DVD audio or SACD is the standard.

[cybermooks]

Begs the question though why do you have to use 96 for that instead of 48? Why wouldn't 48 be adequate for other than cats and dogs? That would give you 24 and enough headroom I would think, unless some of those harmonics are even higher? Sure 96 will give you more headroom, but do you really need it given equipment, space, etc. considerations? I have some devices that will do 96 but at a cost of half the track count, i.e. with an ADAT connection you get 4 tracks instead of 8, whereas with 48 you don't lose anything.

Then if 96 why not 192? Does anyone know what the practical highest range of these harmonics are?

 

The only other possibility, and a very vague one, is that somehow sounds of higher rates affect us in other ways, i.e. physically rather than audibly, although most of us find that true of low frequencies not high ones.

 

Finally the spec on a lot of audio gear is 20hz to 20khz, the old flat 20/20; so who can hear it anyway through their stereo?

[ChetSinger]

 

Originally posted by MP3Chuck While the 22,500Hz upper limit of 44.1kHz recording may be about the highest frequency people can hear, it would still be a good idea to use 96kHz (or higher) if it's an option. Instruments, most notably the higher frequency things like cymbals, can create harmonics that reach over the 22,500Hz mark, and all those frequencies come back as low frequency noise when you're using a lower sampling rate.

 

This isn't really a problem anymore. The input anti-aliasing filters on modern converters are so good that harmonics above 22.5kHz are stripped before being digitized.

 

 

Originally posted by Boom The only problem using sample rates higher than 44.1k is when you have to convert them to down to 44.1 for a CD, 48k doesnt divide evenly into 44.1 and your sound quality degrades as a result. I guess the only way to go would be 88.2, or re-record it via analog (it's not as evil as some people think, converting to analog and back to digital, if you have good quiet gear). When I first started out I used 48k and when I converted them in Soundforge to 44.1 it sounded harsh, I had much better results starting at 44.1 and that's what I still use and will continue to use until something like DVD audio or SACD is the standard.

 

If the sample rate conversion is audible, it's because Soundforge is using a less-than-ideal sample rate conversion algorithm. I use Adobe Audition to do sample rate conversion, often from 96kHz to 44.1kHz, and I haven't heard any artifacts.

 

Chet

[cybermooks]

Ditto. The dithering algorithms in Logic work fine too.

 

I still also wonder about the question of degradation in AD/DA conversion multiple times.

 

The other factor to keep in mind here is the conversion capabilities of most of your equipment. While some newer synths are 24/whatever most of everything else is old 16 bit, if not less on real old stuff. In addition, apart from newer 24 sample libraries most of the samples out there are 16 too. So chances are a large part of your source material is already below the threshold your recording at.

 

For example, I have a Korg A1 effects unit that still sounds great. It also came with the "digital out" of its day, a 16 bit s/pdif connection. If I can record the analog out at 24/48 I should theoretically have better results than the "digital" I/O.