Is 24;96 always better than 16;44?

[Kippyy]

I'm a newbie to this site and hope someone can help me understand what these sample rates really mean.

I've been reading posts here, but am still confused with the basics. I'm a Mac user, with all my music(mostly pop) ripped from CD to Apple Lossless in itunes. I use the ipod/itransport.

Questions:

1) Can I obtain sound off the CD's better than I have now encoded at 16;44?

2) If so, will I hear the difference?

3) If higher sample rates will sound better, do I have to re-rip all my CD's to achieve the improvement, or does the Max system convert my lossless files to something better?

4) Is 24;96 et al, mostly applicable for jazz/classical music?

 

Thanks

 

[PeterSt]

Hi Barry,

 

Many parameters come into play, and it is quite hard to explain it in layman's words. But here is a first go :

 

First of all it is my opinion that 16/44.1 is good enough to squeeze out the best you ever heard. I mean : this is relative to what most perceive or can perceive, and it related to the playback means to start with.

Of course you know about $80 DVD players vs. 10K CDPlayers, and whether the price difference is justified or not, there is a difference, and it comes from somewhere. With playback software this is similar, but in order to avoid the ever recurring jitter (from software) debate, let's stick to the CDPlayers for better understanding (or no need to understand :-).

 

Now, once you'd have heard that, say, best from 44.1, you can already wonder whether you need better. But for arguments sake, let's say Yes.

 

Now it becomes a tad more difficult ...

The first thing you should know is that at 44.1 samplerate, a 20050 Hz tone is a pure square. This is because of it being "digital". This is not good. Ok, you could say "but I don't hear a 22050 tone anyway", but let's then go to half of it : 11025. This is not a pure quare, but a 2 stepped square. Thus, less of a square, but still very squarish.

 

In between the lines, but very important, counts that anything which is a sqaure (or is squarish) will create harmonics. Squares just do that. This by itself creates an enormeous mess in the audible spectrum. I don't say this is directly audible, but it can be visualized anyway, and for theories this is not good.

Please notice that I am talking about squarish where the original is sine(ish), and it is just "digital" doing this to us.

 

When the sample rate is doubled, this squaryness because of digital is halved again. So this is good.

When it is quadrupuled, the same. So, again good. The more the sample rate is highered, the less digital plays its nasty role.

But keep in mind that I am talking about the *recorded* sample rate.

 

When the recorded sample rate is 44.1, there are tricks to get rid of nasty digital, and they come down to similar I talked about in the above : upsample the original. The same applies, but ... almost;

 

Besides it will create fake samples (think of interpolation like a sampled value of 10 and one of 20 where 15 is in the middle and forms the additional sample), there is something else going on, and this is a nasty thing by itself :

When D/A converters apply there means of internal working, they may apply 256 times oversampling. And indeed all the nasty digital is faded away, and the staircase like waves have become rather fluent of it. The nasty thing comes in when you'd see that some of the squaryness is actually intended. This is about square sounds by nature (trumpet, well resined violin sticks) and about dynamics (a hit on a drum rim). Those are flattened just the same along the process.

 

The latter is about oversampling (sigma-delta) D/A converters, and they just work like that. They can't do without oversampling.

Not oversampling D/A converters also exist, but for the real high sample rates (192KHz) they don't (not 100% true because I own one). But anyway for 96KHz they sure do, so, problem solved ?

 

Not really, because when no oversampling is applied, the earlier described problem of the mess in the audible spectrum plays its role again. Filters are needed to counteract that, and those filters by themselves "destroy" sound to a certain extend.

In the end it is one or the other, and both have their sure downsides.

 

Of course I was talking about two very different subjects, with a recorded higher sample rate opposed to upsampling, and with oversampling as most D/A converters do as an actual third.

It is my opinion that upsampling and oversampling never work out for the better, but it requires much more words to describe why for theories, and how it comes that for the net result it sounds better, including the mess I talked about (meaning : no filter to be applied because net it would destroy more than the mess influences sound).

 

With this as a very rough basis (already going too far I'm afraid), here are the answers to your questions. Ok, my answers :-)

 

1) No (assumed you have ripped into a lossless format). But of course the software matters, but let that go for now.

 

2) It is the other way around : you will hear the difference because it gets worse with upsampling (after some experience), and FIRST assumed to have the best playback means.

 

3) You'd never have to re-rip in any circumstance. Or the playback software may upsample in real time, or off line software will take your current rips for input.

 

4) As it seems (available assortment), mostly for Jazz. Don't forget to take DVD-A into account, because that too just can be ripped.

 

Peter

 

[Mike19]

Peter -

 

Could you repeat that please ? [just kidding]

 

[audioengr]

IMO in general, a native 24/96 track will sound better than the same track downsampled to 44.1.

 

Upsamplers that do his after the fact are all over the map. Hardware upsamplers generally suck IMO.

 

I like the Foobar 0.8.3 SRC plug-in. It does the upsampling on the fly and sounds great.

 

Even better are static upsamplers that re-write the files to 24/96: R8Brain and Adobe Audition.

 

Steve N.

Empirical Audio