88.2k vs. 96k?

Hi IBBIAZ,

I'm sure you could find a lot of nonsense answers to this question. Many of the answers to this question are displays of superiority rather than useful information.

The results depend on how you downsample, what type of filtering you do and whether or not the samples are "competently prepared" (are already anti-aliased) and so on. Because of this, results will vary somewhat. Whether the results really vary enough to worry about is also an issue --- and I doubt that they do. If anyone claims they do, I would challenge them to a double-blind listening test.

Regards,
Dave Clark

i doubt anyone could tell the difference between a 88.2->44.1 and a 96->44.1 conversion reliably. as frank says, the math is cleaner with 88.1, although the difference between the two probably wouldn't show up at half the sample rate if floating point math is used for downsampling.

my next question would be why are you recording at a high sample rate to begin with - if you are holding out for 96k to become a standard then record at 96k. if you just wanted to record at a high sample rate, you are going to lose that information anyway. i would actually avoid downsampling if possible, since it's an unnecessary process if all you ever plan to do is 44.1. the same is not true for bit depth, but that's a flame war for another time.

I assume it's some sort of multi-tap filter algorithm? If so, even though the code might always be complex, the taps should be falling directly on even sample boundaries, and one would think you could optimize that case and not have to create some sort of floating point interpolation based array and just look at raw source samples and not do any interpolation, as long as the taps were integral.

ORIGINAL: DaveClark


In the resamplers I wrote, all the math is dirty!

Regards,
Dave Clark

Do you mean Paris Hilton dirty or rounding errors dirty ?

ORIGINAL: DaveClark

Hi Frank,

I'll have to do some research and get back to you on that one... Rounding errors might explain a lot about her, so maybe these are actually the same thing.

Regards,
Dave Clark

You have a point, some parts of her are not as round as I like to see. Her brain also appears to produce some math and logic anomalies at times. Random values can be good or bad...

jacktheexcynic


i doubt anyone could tell the difference between a 88.2->44.1 and a 96->44.1 conversion reliably. as frank says, the math is cleaner with 88.1, although the difference between the two probably wouldn't show up at half the sample rate if floating point math is used for downsampling.

my next question would be why are you recording at a high sample rate to begin with - if you are holding out for 96k to become a standard then record at 96k. if you just wanted to record at a high sample rate, you are going to lose that information anyway. i would actually avoid downsampling if possible, since it's an unnecessary process if all you ever plan to do is 44.1. the same is not true for bit depth, but that's a flame war for another time.

Sorry to resurrect this old thread (could have started another one!  ), but I'm STILL fretting about 88.2 vs. 96 after seven years of using SONAR.  I can't really detect a sonic difference with my converters (Apogee AD/DA16X), and my computer can handle either rate just fine.  I'm just starting a new project and my brain brought up the issue again.
 
I've been using 24/88.2 all these years with no problems--no problems when converting to CD-quality, no problem with disk space or computer resources.  But what keeps bugging me is the possibility that 24/96 could be the consumer standard some day.  Is that day near?  CD's look like they're on the way out.  What will happen to all the stuff recorded at lower rates?  Will the upcoming Audio DVD players be able to play any rate?  Trend-seers, please respond!

Searched, read dozens of threads on this subject here and on other forums.    Not enough said on this specifically.  Have the Katz Book.  Full of facts.  I need interpretation of facts.

READ THIS  and be sure to watch the video in the OP.... it takes about 1 hour... but it's educational...and on this topic.

If you are recording at 88.2 it would probably be best to go ahead and bump up to 96.  That is a future standard, possibly, tho as stated elsewhere it seems that MP3 is now standard.  Unless, of course, you are doing video.

Converting rates doesn't seem to hurt these days (your computer does more math than that recording), but 96/24 bit is probably safest bet for high end recording.  That said, I use 44.1 at home and the studio.

@

Best advice of all.  Thanks, skull!

Hey SeveredVesper Have a look at this link:

http://musikality.net/tut...pth-decibels-loudness/

If you look at the 2nd diagram down (a digitally sampled waveform) the vertical lines represent the sample rate ie the number of times per second samples are taken. Have a look at the third diagram down now. (digital sampling ocurrs in discrete quantized steps as shown above) Look at the horizontal lines. They go across to the vertical axis of the graph. The bit depth is the number used to describe the number of possible levels that can be on that vertical axis. As you can see from that very crude diagram there are only around 10 or 11 levels to describe the levels on that vertical axis. Any sample has to land on one of those. If the voltage level of the waveform being sampled lies in between, then it is rounded either up to or down to the nearest level. Pretty crude. The bit depth in that case might be 3 bits or just a little higher. 16 Bit however represents a possible 65,536 levels on that vertical axis. (between -1 and +1) 24 bit depth represents 16,777,216 levels (between -1 and +1) hence it is much finer. While the sample rate determines the resolution of the horizontal axis of that waveform the bit depth represents the resolution of the vertical axis.

This is a basic answer only but for every sample that is taken vertically (ie on the horizontal axis) a word with a certain number of bits is used to describe what level the waveform is on the vertical axis or (amplitude axis)

The topic of this thread however is related to the number of vertical samples per second not bit depth. And people are asking whether you need to sample a wave either 88,200 times a second or 96,000 times a second and to be truthful there is no audible squat of difference. In fact it can be shown that really any sampling rate above 50,000 times a second is not really necessary. There have been some very interesting tests done (Mix Magazine) where very high quality analog signals have been put through a 44.1Khz and 16 Bit digital signal chain and even some of the best engineers in the world could not pick the slightest difference between the analog signal chain and the digital one. What to you make of that! Sort of blows this stupid 88.1K vs 96K argument out of the water.

But 24 bit recording does offer some good things and a very low digital noise floor is one of them. This means lower level recording inside your DAW which is a good thing.