The Great 96 kHz Debate....all invited!

[chessrock]

I don't see anything wrong with the initial test of doing 88.2, copying and down-sampling to 44.1.

Since you're downsampling like this, there would be no need for anit-aliasing filters and the like, am I correct?

This way, at least you're comparing the very basic concept of higher sampling versus lower sampling. It may not be perfect, but it would still be very important and relevent in theory.

 

[mshilarious]

how many of us are using Lavry-level gear anyway? i'm thinking about a Delta 1010, motu, or something else similarly "prosumer". decent for prosumer, but certainly nowhere near the level of Lavry's converters, etc.

Exactly. What everybody should test is their own converters. Practically, that is all that matters.

However, we shouldn't mistake the result of our individual tests for the conclusive argument of whether higher sample rates are better, since we are limited by the performance of our gear.


I don't have 96kHz converters anyway

 

[mshilarious]

I don't see anything wrong with the initial test of doing 88.2, copying and down-sampling to 44.1.

Since you're downsampling like this, there would be no need for anit-aliasing filters and the like, am I correct?

This way, at least you're comparing the very basic concept of higher sampling versus lower sampling. It may not be perfect, but it would still be very important and relevent in theory.

That could be. Either Lavry or Katz (I forget which) stated that downsampling routines have been good enough for many years to not cause a loss in quality.

The only hitch I can think of is you are still dealing with two different clock rates for the D/A conversion . . .

 

[Reggie]

Chess: That's not really what I want to test. I want someone to test a recording at 88.2 and then converted to 44.1 versus a recording of the same source at the same time recording at 44.1 originally. I don't have a way to split signal to two separate soundcards or I would try it myself. With me?
I can't argue with the fact that 88.2 sounds better than 44.1.

 

[mrface2112]

<<Since you're downsampling like this, there would be no need for anit-aliasing filters and the like, am I correct?>>

actually, according to Katz (page 222, got the book right here)....:

"The downsampling is accomplished with a digital circuit called a decimator, which is a form of divider or sample rate converter, and which must contain a filter at half the sample rate to eliminate aliases, requiring a 22.05kHz cutoff at a 44.1kHz SR. This filter must be designed without compromise or it will affect the sound. Some manufacturers concentrate on transient response, others on phase response, ripple, linearity, or freedom from aliasing. But getting all of these characteristics are important, and getting it right is expensive--precision construction requires more math, and math requires labor ane parts (size of the integrated circuit die). This the filters in a typical compact disc player or in the converter chips used in most of today's gear is mathematically compromised."


so....in short, yes, even an "even number" downsampling from 88.2 to 44.1 will require an anti-alias filter....and will affect the sound.


cheers,
wade

 

[mrface2112]

<<That's not really what I want to test. I want someone to test a recording at 88.2 and then converted to 44.1 versus a recording of the same source at the same time recording at 44.1 originally. I don't have a way to split signal to two separate soundcards or I would try it myself. With me? >>


Reggie--

That's what i've been talking about this whole time. that's exactly the test i'd like to run, but i also don't have a way of getting the signal split into 2 separate interfaces, etc.

in theory, there should be a difference between the 88.2->44.1 file and the native 44.1 file. the question is whether the detriments from resampling from 88.2->44.1 are "worse" than a native 44.1.

a further test of this would be going from 96->44.1 and comparing it to a native 44.1. and then a further test would be to apply signal processing to the files before you resampled them, and an even further test would be to do so over an entire mix and see how "bad" the "buildup" becomes--if at all.


cheers,
wade

 

[mshilarious]

in theory, there should be a difference between the 88.2->44.1 file and the native 44.1 file.

See that's what I don't agree with. A theoretically perfect converter would only differ between 88.2 and 44.1 in ultrasonic frequencies. Once you discard those frequencies and downsample, it should be identical to a perfect 44.1 sample.

Now, is the implementation of sampling and downsampling imperfect enough to cause a benefit for one or the other? I think that is what we are trying to test.

 

[chessrock]

Chess: That's not really what I want to test. I want someone to test a recording at 88.2 and then converted to 44.1 versus a recording of the same source at the same time recording at 44.1 originally. I don't have a way to split signal to two separate soundcards or I would try it myself. With me?

Gotcha.

Okay, I got it ...

Take some material that's already been tracked at 96.1.

Then re-record each track, individually, twice, at the two different sampling rates that you want to compare ... then do your down-sampling on the higher-res tracks, remix and compare.

There would obviously be two generations of conversion going on, but at least whatever degredation you have would apply to both samples, so you'd still have a valid comparison. Besides, people who record digitally and mix outside the box go through multiple conversions anyway, so it's not like it's highly unusual.

 

[Bob's Mods]

Some points I would like to make here are, a less analytical approach should be all thats neccessary on your own system. You've been recording with it for some time and have a feel for how it responds. Record a few songs at the higher rate and the difference should be apparent. Its not the difference between gold and lead but there should be enough of change to realize something has changed.

I wish to clarify another thing. The word length is constant no matter what frequency you record at, its 24 bit. Its just that there are more 24 bit words per second at the higher rate. Kinda like more folds in an accordian.

I believe its really in the mixing stage and addition of plugins where the difference is apparant. I've noticed that if I record one track and one only, I cannot hear the difference between 44 and 96. When math needs to done for mixing and plugs, you can hear the difference. Tracks stand out better. Complex waveforms are produced when you mix more than one track that must be mathamatically summed. Its not happening in the same way as your old analog mixer board, it done with math. So, if you've got to stuff lets say, 5 tracks together (mix, sum) they are all time aligned in increments of 44,100 per second. Each sample must be summed to together, one per each track, in this case 5, to derive a new 24 bit word that is an approximation of that sum action. Five tracks, each recorded at 24 bit must be added together to create one final mix track that represents all five. The more tracks you have, the more compromises the math has to make to resolve them. If you have more samples, 96,000 per second rather than the 44,100, the math becomes more precise in its ability to resolve the complexities of stuffing mulitple 24 bit words into one. Its really about total bit count at 96 k that create more accurate math and less rounding errors.

Looking at this from a purely bandwidth/filter view is only one section of this puzzle. More (in band) data words per second make the math of mixing and plugs more accurate as the complex waveforms generated by mixing are more accurately resolved. Hence, your individual tracks stand out better.

One last point, the complexity of the math of 96 vs 88 is mostly overdone. The machine just works harder, thats all. If the proper formula is applied, the result should be true. Any rounding differences should inaudible.

Bob

 

[Stefan Elmblad]

The only reason 96kHz sounds better is that the ugliness of digital artifarcts at the upper frequencies, just below the maximum Hertz threshold (Nyqvist theory) are moved higher up the frequency spectrum, and as a result can't be heard. When you eventually dither or mix down to 44.1 it comes back again. And to underline this fact you can just upsample a 44.1/48 file to 96kHz and enjoy the "much better audio", even though it has nothing to do with the recording quality itself.

So, it's a waste of time and disk space. There's no need to record at 96kHz EVEN IF the end result is going to end up on a 96Khz album, which would be highly unlikely. That would be either a .wav album only playable on PC's or the Alesis CD24 format which is for the masterlink only. Just upsample at the mastering stage if you wanna go that route.

Otherwise, forget it.

 

[regebro]

And to underline this fact you can just upsample a 44.1/48 file to 96kHz and enjoy the "much better audio", even though it has nothing to do with the recording quality itself.

I just need to get my new soundcard in to test this really quickly!

 

[Bob's Mods]

Stef,

I found when I downsampled and dithered to 16 bit, the enhanced definition I attained at 96k was almost wholely retained. I will be doing more tracking at 96 in the months to come. I will then have more of a comparision with my earlier 44dot1 stuff.

It seems for all the sound cards and gear that advertise 2496 compatible, few are tracking and mixing at 96 kHz. And now, 192? Where are we going with this?

A friend at work, who is an audiphile, set up a surround sound DVD. He said he picked up FleetwoodMac's remaster'd Rumours work and couldn't believe the difference in playback quality and level of depth to which he could hear each instrument. He stated there was no comparison to the regular 44dot1 CD. I am suspect the higher sample rate of the DVD format contributes to this. If it did not, then the standard 44dot1 rate would have been maintained. Neil Young is a BIG fan of the DVD audio format too. If I remember, there may be an article posted on the web with him on this very subject. There must be something to the higher sample rate. I have never heard DVD audio as so I have no opinion on its fidelity.

Bob

 

[Cloneboy Studio]

So, it's a waste of time and disk space. There's no need to record at 96kHz EVEN IF the end result is going to end up on a 96Khz album, which would be highly unlikely. That would be either a .wav album only playable on PC's or the Alesis CD24 format which is for the masterlink only. Just upsample at the mastering stage if you wanna go that route.

I disagree.

Reason: take this example--if you have 24 tracks at 96khz the artifacts are less apparent on ALL 24 TRACKS! This will eliminate a ton of unwanted "junk" in your high end when you mix down.

It's the same principle as analog recording and keeping noise levels down--because that little bit of noise per track adds up when you have 8, 16 or 24 tracks into a major problem.

Same idea.

 

[mshilarious]

A friend at work, who is an audiphile, set up a surround sound DVD. He said he picked up FleetwoodMac's remaster'd Rumours work and couldn't believe the difference in playback quality and level of depth to which he could hear each instrument. He stated there was no comparison to the regular 44dot1 CD.

He is comparing a remastered surround sound version with a CD version that was probably remastered in the '80s. How is that a controlled test of bit depth and sample rate?

 

[Bob's Mods]

I disagree.

Reason: take this example--if you have 24 tracks at 96khz the artifacts are less apparent on ALL 24 TRACKS! This will eliminate a ton of unwanted "junk" in your high end when you mix down.

It's the same principle as analog recording and keeping noise levels down--because that little bit of noise per track adds up when you have 8, 16 or 24 tracks into a major problem.

Excellent point!

msh...you are correct. It's not a controlled test.

Bob

 

[mcolling]

Excellent point!

msh...you are correct. It's not a controlled test.

Bob

oooooo- bob's mods is on the run!!!!

looks like the 96 k boys are going pull this one off!!!!

 

[Stefan Elmblad]

I disagree.

Reason: take this example--if you have 24 tracks at 96khz the artifacts are less apparent on ALL 24 TRACKS! This will eliminate a ton of unwanted "junk" in your high end when you mix down.

It's the same principle as analog recording and keeping noise levels down--because that little bit of noise per track adds up when you have 8, 16 or 24 tracks into a major problem.

Same idea.

Not 100% true, cause the noise we are talking about isn't actually recorded on the tracks, they are a product of the samplerate.

 

[mshilarious]

oooooo- bob's mods is on the run!!!!

looks like the 96 k boys are going pull this one off!!!!

Umm, I'm an anti-96Ker

That is to say I doubt there is any audible improvement due to the use of the higher sample rate alone. I also believe it would be very, very difficult for a consumer to design a well-controlled test to prove it either way.

 

[Bob's Mods]

I find it kind a strange really. Marketers seem to think we want/need/demand this capability, yet many users are reserved about the use of, and belief in it as an option.

It actually falls into line with many of my other experiences in the home recording phenomenon. And that is that great recording comes from the sum of many small tweaks that add up to a quality recording (artist ability aside). Its just part and parcel of the bigger picture. There is no one thing that helps make a technically good recording, its the sum of all parts, and this is one part.

Bob

 

[Blue Bear Sound]

On An Aside...

Not saying this applies to people in this post necessarily, but it's funny that many of the people arguing the most about using higher resolutions are the ones who would get far better results if they just improved their arranging/tracking/mixing skills, or simply used better converters!

If someone can't get an amazing sound out of 24/48 with good converters, then the problem is NOT digital resolution.....