96 kHZ

[appleyardrules]

i tried to search it but the search wasnt working.
on my firepod
i recorded a song in 96 Khz rather than 41.4 or whatever
what are the benefits of this. and also, when i saved the wave, it was slow.
im generally a newbie kinda. so can someone just explain the whole difference between 96 kHZ and others and explain how to make 96kHZ recordings NOT be slow?

THANKS!

[asulger]

Originally Posted by appleyardrules

i tried to search it but the search wasnt working.
on my firepod
i recorded a song in 96 Khz rather than 41.4 or whatever
what are the benefits of this. and also, when i saved the wave, it was slow.
im generally a newbie kinda. so can someone just explain the whole difference between 96 kHZ and others and explain how to make 96kHZ recordings NOT be slow?

THANKS!

http://homerecording.com/bbs/showthr...ghlight=96+khz

[NashBackslash]

I just finished reading that thread. o_O What an ass-hole...

Anyway, I want to test what my highest frequency hearing range. How would I go about doing that?

[LRosario]

I think the link you where given was completely inaccurate. If 96k was a recording gimmick, then no one would go through the trouble of manufacturing units that can achieve 96k and beyond.

The simple fact is, although you may not hear the frequency, it becomes a resolution issue. Simply put, the higher the sample rate, the higher the resolution. Also, you can achieve greater dynamic range with higher samples.

That's why digidesign even bothers to put out a box that does 192k. Problem is, you have to have serious processing power and massive storage, and that becomes expensive.

That's the reason it takes longer to save, cause it's a higher resolution. Usually it's common to record 24/48 and you can get by just fine in the recording world.

But for those whom work with digital and are die hard about capturing the sound at the highest resolution, then anything up to 192k becomes a nessessity.

I myself am guilty of this. I heard the difference once and I never went back

[dgatwood]

Originally Posted by LRosario

I think the link you where given was completely inaccurate. If 96k was a recording gimmick, then no one would go through the trouble of manufacturing units that can achieve 96k and beyond.

The simple fact is, although you may not hear the frequency, it becomes a resolution issue. Simply put, the higher the sample rate, the higher the resolution. Also, you can achieve greater dynamic range with higher samples.

You had it... and then you lost it. Greater dynamic range would be dependent on bit depth (16-bit vs. 24-bit), not sample rate.

Resolution is mostly irrelevant in audio, but it does make a difference in believability if you're doing any processing that involves frequency scaling (e.g. AutoTune). That said, a good interpolation algorithm will do almost as good a job with a whole lot less CPU power....

For the most part, the advantage to 96kHz is because most cheap converters don't do oversampling. The frequency stored is the frequency sampled. Doesn't sound all that problematic, but it actually can be.

The problem with this is that to sample at 44.1/48 kHz, you have to have a brick wall filter from 22.05/24kHz up to prevent aliasing. This filter is an analog filter, which can only do so much as far as generating a steep filter without adding nasty artifacts. And the steeper the filter, the more it costs. Guess where a lot of cheap converters cut corners....

The net result is that for many cheap converters, that brick wall filter may start noticeably rolling off sound as low as 16kHz, which is quite audible. By sampling at four times the desired maximum audible frequency, you generally end up using a different analog filter that starts rolling off at... say 32 kHz---well above the audible range of humans---thus avoiding the loss of high frequencies that would otherwise occur.

By the way, this same filtering occurs on output, so by recording at 44.1 kHz and playing back at 44.1 kHz, you're paying this filter penalty twice.

[LRosario]

Originally Posted by dgatwood

You had it... and then you lost it. Greater dynamic range would be dependent on bit depth (16-bit vs. 24-bit), not sample rate.

ahh thats right, thank you for correcting me on that

[Joel Hamilton]

If you can make it sound good with 44.1 it doesnt matter that much. You will make it sound good with higher sample rates as well.

Remember that 2" tape (even my studer A827) rolls off LONG before 32k, more like 16k...

It does it a litte more gracefully than some cheap converter box, but whatever. I use PTHD and at 88.2 things are not drastically better or worse than tape OR 44.1.... Really. You have to be capable of recoding a source that REVEALS these limitations in the first place.

Basically: Record a lot. That is worth more than any mic pre or microphone or DAW or tape machine. The best way to get better results is to record. A LOT.

Nobody expects a new baseball glove to make them Derek Jeter... So why does this always happen in the audio world?

It wouold be like saying " I really cant play guitar, shold I get a Les Paul? then will i be amazing?"

Save your money, and dont let advertizers convince you of ANYTHING...

[Robert D]

Originally Posted by LRosario

If 96k was a recording gimmick, then no one would go through the trouble of manufacturing units that can achieve 96k and beyond.

I'm guessing you don't work in marketing.

[EleKtriKaz]

Originally Posted by Robert D

I'm guessing you don't work in marketing.

Good point.

[Massive Master]

I'm kind of a stickler for decent sound and I *still* record at the target rate (44.1KS/s for general audio or 48KS/s for video).

Read up on Dan Lavry's stuff... Little or nothing is normally gained by recording at higher sample rates. On many occasions, it's actually worse.

Another point being that a lot of 192K converters are junk. I'd rather record a GOOD 44.1 signal that an "iffy" 192 signal any day.

[bleyrad]

For the most part, the advantage to 96kHz is because most cheap converters don't do oversampling. The frequency stored is the frequency sampled. Doesn't sound all that problematic, but it actually can be.

Please reference a single recording A/D converter currently on the market that doesn't have at least 32x oversampling at 44.1.
Many have 128.
I haven't seen any in like 7 years without any oversampling.

[mshilarious]

Originally Posted by bleyrad

Please reference a single recording A/D converter currently on the market that doesn't have at least 32x oversampling at 44.1.
Many have 128.
I haven't seen any in like 7 years without any oversampling.

My humble 44.1/48 Alesis AI3s are 128x

As applied on the ADC side, the converters are running at a much higher rate with few bits, and the data rate is constructed from that stream. One of Lavry's most salient points is given a higher data rate, the data must necessarily be LESS accurate as there are fewer observations on which to base each data point.

Here's an interesting link:

http://www.alesis.com/downloads/manuals/EC2_Manual.pdf

That's the 96kHz upgrade for the Alesis HD24. Towards the back, they give a good summary of the history of digital audio, and explain why oversampling has taken away the main arguments against 44.1kHz, and essentially dare you to try to hear a difference, which, according to them, might involve buying new microphones, recapping the rest of your gear, and changing your monitors to ribbon tweeters

There are many reports that these converters ARE better, but that could be because of a superior clock or analog signal path. Who really knows? Alesis themselves only measured THD over the standard audible range. What if those ultrasonic frequencies are present, but at 40% THD? How would you know?

I have adequately demonstrated to myself that I can hear up to 17.5kHz, and can sense up to about 18.5kHz as pressure on my ears if sufficiently loud, but it doesn't create a sound in my brain. Beyond that I cannot detect at all.

[ahuimanu]

Originally Posted by bleyrad

Please reference a single recording A/D converter currently on the market that doesn't have at least 32x oversampling at 44.1.
Many have 128.
I haven't seen any in like 7 years without any oversampling.

Hey bleyrad,

What happened to Appleton?

[gullfo]

i think when you consider a number of factors: 96khz has a valid application for its use - oversampling on the ADC is one thing, capturing the stream of numbers off the ADC is another (one of DanL's concerns about higher bit rates is the finite time to convert ADC versus reliably capturing the numbers...). keep in mind the "theory" is many sine waves = complex signals. in practice, i doubt that is really perfect. having worked for a number of years designing and building high speed test electronics for satellites, subs, missiles, etc... that use a lot of complex signal testing, i don't ever recall the design specs considering lower sampling rates for accuracy in complex signal production or capture. then again this equipment had to operate from DC through 20mhz :-) my thoughts on this is that 96khz sample rates allows for more sampling points across the 24 bit word and therefore, complex signal nuisances are more likely captured accurately rather than being extrapolated by the ADC designer. it seems highly unlikely that a 96/24 data set is less detailed or only as detailed as a 44.1/24 data set. not negating Nyquist but we've come a long way since his theory was first laid down and the practical matters of silicon and perception have been examined by many people in many ways...

i threw together some samples on my soundclick site for 96/24 vs 44.1/24 and 16 vs 24 bit. you can take a listen. http://www.soundclick.com/roughcanvas - in the music section - top 4 items. ignore the rest. i deliberately left a lot of room noise in because i think you can more easily hear the overall difference in the noise in addition to the guitar bits... the 16 vs 24 (IMHO) is obvious. the 96 vs 44.1 is subtle but even in MP3, the source difference is there...

[bleyrad]

Originally Posted by ahuimanu

Hey bleyrad,

What happened to Appleton?

Our bassist is in another province!
We are still writing though. Thanks for the curiosity!

[Bob's Mods]

There appears to be a subtle difference between the lower sample rate and higher sample rate. 44.1kHz seems to be high enough to satisfy the many as accurately capturing the source. I know I'm a believer.

But what about improvments in mixing at 96kHz? There has been this ongoing debate regarding mixing in the box (at 44.1k) and analog mixing. My few experiments in the 96kHz range showed improved mixing capability. Any opinions for improved mixing benefits at 96k? Yeah, ney?

[ahuimanu]

Originally Posted by bleyrad

Our bassist is in another province!
We are still writing though. Thanks for the curiosity!

Sorry for the hijack...

Bleyrad... website? I await more material, I really love the two songs I have: "If you want me" and "scream it out"

No B.S. I really enjoyed the band and found out about it from your posts here.

<< back to your regularly scheduled sample-rate discussion here >>

[bleyrad]

ahuimanu... can't seem to PM you.
PM or email me and i will email you some more recent demos if you'd like.

[dgatwood]

Originally Posted by bleyrad

Please reference a single recording A/D converter currently on the market that doesn't have at least 32x oversampling at 44.1.
Many have 128.
I haven't seen any in like 7 years without any oversampling.

None of M-Audio's cards (stand-alone DACs notwithstanding) advertise that they do any oversampling.

[mshilarious]

Originally Posted by dgatwood

None of M-Audio's cards (stand-alone DACs notwithstanding) advertise that they do any oversampling.

Doesn't mean they don't. From a (bad) image I found, it appears the 2496 uses Cirrus converter chips. Here is the .pdf for Cirrus' crappiest currently available chip:

http://www.cirrus.com/en/pubs/proDat...CS5340_PP3.pdf

You'll note a p.1 the reference to decimation, but you have to go to p.17 to actually find the oversampling rate. And of course, the oversampling rate is dependent on the data output rate, because the master clock speed does not change. I imagine that's why manufacturers do not advertise oversampling anymore.

The performance specs don't note a difference between the various sample rates, but of course they only test with a 1 kHz sample rate. The manufacturers could settle this discussion very easily if they liked by testing full audible range data with their vastly superior test gear Something tells me it is not in their interest to do so.

I did note that their filter in 192 kHz mode is designed to cutoff at 50 kHz vs. 47 kHz in 96. Bats worldwide will be severely disappointed

[mshilarious]

Originally Posted by gullfo

i threw together some samples on my soundclick site for 96/24 vs 44.1/24 and 16 vs 24 bit. you can take a listen. http://www.soundclick.com/roughcanvas - in the music section - top 4 items. ignore the rest. i deliberately left a lot of room noise in because i think you can more easily hear the overall difference in the noise in addition to the guitar bits... the 16 vs 24 (IMHO) is obvious. the 96 vs 44.1 is subtle but even in MP3, the source difference is there...

16 v 24 is not at issue, and I don't think too many would argue for 16 at this point--the specs in noise floor are dramatically different.

So I just listened to the 44.1 vs. 96 samples. These performances are too different, both samples are very quiet, and the .mp3 files are low rate. I don't think this is a very valid test--try using something repeatable, like a CD.

Incidentially this is the FFT of the two files, blue is 96 and yellow is 44. The 44 sample is 3.4 dB louder than the 96, both peak and RMS, which I didn't normalize on this graph, because it makes them easier to read.

In theory, we should expect to see a slight attenuation in the 44 file above 18 kHz. It's not there, but again the level up there is -90dB or so, so there may have been practically no signal to be attenuated.

[BigRay]

I record at 24/96 normally (though my recorder handles 192) ..As for me...I was hard pressed to hear really big differences between say 48 and 96 khz...BUT heard HUGE differences between the various Bit Depths...the jump from 16 to 24 was ASTOUNDING!...I still record at 96 khz because in theory it is supposed to sound better, more "true" but the real difference for me lied in the bit-depth, not sample rate.

trb

Originally Posted by LRosario

I think the link you where given was completely inaccurate. If 96k was a recording gimmick, then no one would go through the trouble of manufacturing units that can achieve 96k and beyond.

The simple fact is, although you may not hear the frequency, it becomes a resolution issue. Simply put, the higher the sample rate, the higher the resolution. Also, you can achieve greater dynamic range with higher samples.

That's why digidesign even bothers to put out a box that does 192k. Problem is, you have to have serious processing power and massive storage, and that becomes expensive.

That's the reason it takes longer to save, cause it's a higher resolution. Usually it's common to record 24/48 and you can get by just fine in the recording world.

But for those whom work with digital and are die hard about capturing the sound at the highest resolution, then anything up to 192k becomes a nessessity.

I myself am guilty of this. I heard the difference once and I never went back

[gullfo]

Originally Posted by mshilarious

So I just listened to the 44.1 vs. 96 samples. These performances are too different, both samples are very quiet, and the .mp3 files are low rate. I don't think this is a very valid test--try using something repeatable, like a CD.

I don't think you'll see any attenuation before 19.5K when the low pass filter on the MP3 processing kicks in. both MP3 files are 128K because the original question a few weeks ago was related to "does higher quality in the recording process matter in MP3 files". I realize the performances vary a bit but the real key is to listen to the audio quality difference - you can also hear it in the room noise which is constant between the 2 samples. a CD would not likely be a better test though because it is already 44.1/16 and the recording sources can vary greatly - in this test, it's just raw audio sourced at either 96 or 44.1, no additional level raising, etc...

the 16 vs 24 bit files are examples of how the noise levels are raised and the associated distortion in the 16 bit when increasing the level and having lost the lower bits (in addition to more noise).

[mshilarious]

Originally Posted by gullfo

a CD would not likely be a better test though because it is already 44.1/16 and the recording sources can vary greatly - in this test, it's just raw audio sourced at either 96 or 44.1, no additional level raising, etc...

A CD isn't going to vary 3.4 dB between samples. That's an enormous difference which invalidates any audible change in the noise profile, because it means something significant happened upstream of the converter.

A comparison test is an experiment, which should be done to test a hypothesis, which should be based on some theoretical principle. The principle in question with sample rate is high-frequency response. Most ADCs do not vary between sample rates with respect to noise. Therefore, we are objectively looking for high-frequency attenuation at lower sample rates. It would thus help to have a sample with a fair amount of high-frequency content.


If using a CD troubles you, use a SACD or DVD-A. Or get two of the converters in question and split the preamp output, or use L-R channels. Then post a 10 second 24-bit .wav file.

Or rig up a golf ball to slide down a chute onto a cymbal. That should be easily repeatable, as gravity is a constant

[Robert D]

If a golf ball slides down a chute and hits a cymbal, and there's no one there to hear it, does it really make a sound?