44.1 v 96 kHz

Ricardo · Post by **Ricardo** » Thu Oct 09, 2003 7:40 am

Considering a previous thread about 32 bit recording, I'm curious about the difference between 44.1 and 96kHz. I haven't investigated 96Hz properly yet, so has anyone any views?

voidar · Post by **voidar** » Thu Oct 09, 2003 11:51 am

The sampling rates we use are based on the Nyström theory which means that the highest frequency we are able to record equals half the samplingrate. 44.1 KHz thus spans the range of 1 - 22.050 Hz, while 96KHz spans the range og 1 - 48.000 Hz.

A perfect human ear responds to 20 - 20.000 Hz, but most people have problems hearing anything above 19.000 Hz or even 16.000 Hz, and your ears may most likely vary.

So why record frequencies we do not hear? Well, sound sources (even our own human-made instruments) emmit frequencies well above our range of hearing and these upper harmonics are very important in defining an instruments sound. This is because these upper harmonics shape and interact with the lower, audiable frequencies, and thus i.e. a violin can sound very dull on a recording in contrast to in real life. This is due to that the upper harmonic content is removed when recorded. Upper harmonics are very important for a sound not to appear "muddy".

This is the theory behind the use of 96KHz and even 192KHz..

[ This Message was edited by: voidar on 2003-10-09 12:53 ]

Shayne White · Post by **Shayne White** » Thu Oct 09, 2003 11:19 pm

But *my* question is this: what's the difference between our ears filtering out the upper harmonics, and a digital recorder filtering out the upper harmonics? Doesn't it end up being the same?

Shayne

voidar · Post by **voidar** » Thu Oct 09, 2003 11:43 pm

Ok, so you are being difficult...

.

The difference is that our ears do not filter the sound, they only respond to it whether it comes from a recording or some other source.

What is important is the _resolution_. A recording only captures some of the sound spectrum of the original source, but the most significant part, while realtime, original sounds come in their full spectrum.

Obviously, recording in a higher resolution format will give a more precise result. Higher samplingrates also smooth out the sound. Analog gear for instance is totally smooth sounding due to the fact that the concept of samplig rates is a no-issue - there is no sampling. Though that is not to say that a 100% analog recording is 100% true to the original recorded source/sources. Here "tape"-dimentions and "tape"-speed are important in determing the highest possible recorded frequency.

[ This Message was edited by: voidar on 2003-10-10 00:53 ]

bassdude · Post by **bassdude** » Fri Oct 10, 2003 12:45 am

After doing a little more research into this I'm tending to remain nuetral about the whole thing for now. I'm not ready to blindly follow just yet!

For instance my desk runs up to 48kHz. But the convertors actually do 64X oversampling on the input so it's already sampling way way way above the nyquist frequency. The critical bit is the quality of conversion algorythyms and filtering components to rebuild the picture of that oversampled waveform accurately at 44/48.
This explains why say a high quality 44kHz convertor may sound just as good or better than a 96kHz converter. And buying a cheap 96kHz doesn't automatically mean that you have the equivalent to an expensive high quality 44kHz converter!

[ This Message was edited by: bassdude on 2003-10-10 01:51 ]

voidar · Post by **voidar** » Fri Oct 10, 2003 12:53 am

Yes, you do raise some valid points. I am not really promoting the use of higher sampling rates, just explaining the use. I would rather use more space on a higher bit-rate instead.

Immanuel · Post by **Immanuel** » Fri Oct 10, 2003 5:52 am

The theory about the need for frequencies above over hearing ability is in my opinion speculative. Somebody heard a difference and invented an explanation. Please prove me wrong, if you can, but this is, how I see it.

I have another (also speculative) theory, witch I like more. Playing back a 11.025KHz note with 44.1KHz give 4 steps - this is nowhere near a sine wave. Also if you play another high frequency note, you get stuff, that never looked anywhere near a sine wave - it will vary a bit though from cycle to cycle. Still it is not good. Now, if you go to 96KHz sampling frequency, you actually get almost 9 steps of that 11.025KHz note. That is quite a difference. And saying that 44.1KHz can reproduce 22.05KHz sounds apears to be a commonly accepted LIE. 44.1KHz can reproduce 22.05KHz SQUARE waves - NOTHING ELSE.

Also have a million times oversampling does not change the fact, that your signal is stored in 44.1KHz (or whatever format you use). It just lessens the chance, that you get errors, in that it converts a million times and chooses the value it came up with the most times.

spiderman · Post by **spiderman** » Fri Oct 10, 2003 8:16 am

this why there is low pass filter on DA .
if you look at the wave form before DA . right it's a square at 22kz but the wave after the DA ( the wave form that come out the speaker ) is a pure sinus . if not your DA is a piece of shit

. you are thinking about waveform but DA don't think .they reproduce spectral component of sound . the nyquist theorem is a mathematical theorm so it's an exact thing not at all a lie or a empirical thing. it could be demontrate with equations . now theorem and practice are two thing . And there is technological issue to do a perfect DA with perfect filter etc .. so there is good one and bad one .

Post by **hubird** » Fri Oct 10, 2003 8:28 am

On 2003-10-10 00:19, Shayne White wrote:
But *my* question is this: what's the difference between our ears filtering out the upper harmonics, and a digital recorder filtering out the upper harmonics? Doesn't it end up being the same?

Shayne

sounds still smart to me

if the hi's do change the lows, a recorder will record it...?!

Post by **garyb** » Fri Oct 10, 2003 11:53 am

you hear with more than your ears.

resampling is not cool,it makes artifacts.use a higher bit depth if you can,but don't use a higher sampling rate than you will end up in unless the destination is an even division of the rate you are recording at(for 44.1k use 88.2k).that is the only way to avoid signal degradation.

Shayne White · Post by **Shayne White** » Fri Oct 10, 2003 5:29 pm

So a 22khz frequency at 44.1khz sampling rate is a square wave, true. But the DAC is going to reproduce that as a triangle wave, filling in the difference in the two samples, and your ear can't hear above that anyway (if your ears can possibly go that high), so you'll only hear a sine wave.

Shayne

astroman · Post by **astroman** » Fri Oct 10, 2003 6:36 pm

On 2003-10-09 12:51, voidar wrote:
... This is because these upper harmonics shape and interact with the lower, audiable frequencies, and thus i.e. a violin can sound very dull on a recording in contrast to in real life. This is due to that the upper harmonic content is removed when recorded. Upper harmonics are very important for a sound not to appear "muddy".

This is the theory behind the use of 96KHz and even 192KHz..

sorry, but imho you're completely wrong with this assumption.
I recently bought a totally analog high quality recording of a violin concerto of which I have a digital counterpart.
You guess what violin sounds more realistic

In fact, with all respect to classical recordings and their historic value -compared to the latest digital achievements with their precision and vivid reproduction of the performances, those black disks simply can't keep up.

The reason why we perceive the violin differently in the concert hall is that our ears pick up a lot of other information than just frequencies.
There is nothing available to reproduce the spatial resolution of the ear, not even surround can do it, although the impression gets closer to realism.
Simply said an acoustic hologram would do the job, but I don't know if that is even theoretically possible.

Anyway, while regular folks have difficulties to distinguish 44 khz from 96 khz recordings, all agree that a surround version sounds 'best' - regardless of samplerate and bitdeepth.

Back to 2 dimensional stereo:
the best we can do is to keep the phase information of the signal as good as possible. Spiderman already mention the filter circuitry, which has a heavy influence on phase (any filter afaik).
I once read about a high-end CD player using just 14 bit converters, but having a high quality LC filter that made it sound superior.
When a signal is digitized there are always artifacts (I forgot about the math details) related to the sampling frequency.
The higher the sampling rate, the easier it is technically to filter those artifacts out.
Read: the cheaper the circuit.

On the other hand I'm not at all against high quality gear. I've heard it on some Sony classical recordings myself what that stuff (I won't be able to afford this in a lifetime) can do, but the number alone doesn't make it. As you already wrote : better a good 48k than a bad 96k.

my 2 cents, Tom

voidar · Post by **voidar** » Sat Oct 11, 2003 2:05 am

/*sorry, but imho you're completely wrong with this assumption.
I recently bought a totally analog high quality recording of a violin concerto of which I have a digital counterpart.
You guess what violin sounds more realistic
In fact, with all respect to classical recordings and their historic value -compared to the latest digital achievements with their precision and vivid reproduction of the performances, those black disks simply can't keep up.*/

It is expected for the digital version to appear clearer and perhaps more cliniqal. There is so much you can do to enhance older analog recordings. Newer recording techniques are in no doubt more faithfull to the original source than their predecessors. I never claimed otherwise. Old analog gear has a less clear hi-end. But the point regarding instrument harmonics is still true.

As an exsample I can refer to old folk recordings, preferably usiong violins. They might sound quite horrible in regards to what they would have and probably did do live, even to the point of putting people off on the very instrument itself.

Music Manic · Post by **Music Manic** » Sat Oct 11, 2003 7:58 am

Analogue and digital are both dead formats.Wait for AEL,it's supposed to be a combination of both which is cd based.DVD audio is a con.

voidar · Post by **voidar** » Sat Oct 11, 2003 9:35 am

Could you refere to some information on AEL?
I doubt we will see any Creamware AEL card soon though

.

Music Manic · Post by **Music Manic** » Sun Oct 12, 2003 6:39 am

It's analogue enhancement,and it is supposed to supersede SACD and be way better than DVD audio,which never was better than cd in the first place.It unites digital and analogue.Will let you know more about it.

astroman · Post by **astroman** » Sun Oct 12, 2003 9:44 am

whatever it is - it will not be able to record and transmit a faithful spatial image of an acoustic performance.
That technology simply doesn't exist, or did I miss something ?

It's one of the reasons that people still visit concert halls and opera houses - ok, there might be some other reasons for that, too

Any improved recording technique is (can only be) targeted at the audiophile classic market. A fairly small segment btw.

For synthetic sounds that discussio is beyond any sense because we use digital sources and in some cases even love digital artifacts.
They count as a valid soundsource on their own, let's say a SID chip or an EMU 12bit drumbox.

Vocal recordings are anything but faithful in current productions. You certainly wouldn't want to hear the original voice of 95% of top 20 performers

It's a kind of sound painting, fidelity isn't required at all.
But I admit I like this useless kind discussion about an improvement every producer (in the widest possible context) needs and no consumer finally can hear

cheers, Tom

Stige · Post by **Stige** » Sun Oct 12, 2003 10:57 am

even 5KHz sampled at 44.1KHZ looks ugly and definitely not sine wave. Our D/A converters upsamples it x times and then comes anti-alias filtering. The result what we hear is a clean sinewave.
Still all the audio inside our computer are those ugly shapes of original sound. Think about mixing and processing these waveforms together.. every process changes the sound more far away from original analog counterpart.
It makes sense to do all mixing and processing at higher samplerates because then the waveforms are more close to the original signal when they are inside our computers. Not all the work has left for A/D converters to make it sound 'real'.
It's not bulls**t why UAD-1 Pultec eq upsamples all incoming singnal before processing.

Post by **garyb** » Sun Oct 12, 2003 11:22 am

no it's not b****t,but when you downsample,it must be to an even division of the higher samplerate or all the gain of the higher rate will be lost in the process.to go to 44.1 from 96 involves inventing samples that never existed,hence artifacts.going from 88.2 to 44.1 just involves losing half the samples,no harm done.but it really doesn't matter because 44.1 sounds fine.hell,some listen to all their music on mp3s these days!

really,i DO like using the best sounding system i can afford,for the best results in any playback situation,but let's not get fanatical about it.....

nice post astroman.

Shayne White · Post by **Shayne White** » Sun Oct 12, 2003 1:04 pm

Yeah, I don't even listen to CDs anymore...I rip everything to my Ogg Vorbis library on my HD!

Shayne