Bits more important than kHz?

Posted by: Ebor on 14 September 2013

I was lucky enough to be singing on a professional recording about 18 months ago and, being the good hi-fi geek I am, managed to catch a quick chat with the engineers during a tea break. The 6 microphone signals were being recorded on ProTools on a Mac, and I asked what resolution they were using: the answer was 24-bit and 48kHz. When I asked why they weren't using a higher sampling rate, they said it was pointless due to the Nyquist/upper limit of human hearing argument. He said that it was worth using a higher bitrate than CD on the same basis that, in the olden days of tape, they would record on reel-to-reel with a much higher S/N ratio than the eventual consumer release format. The extra resolution was worth it to be on the safe side, I suppose you might say.

 

As a 16/44.1/CD Luddite, I have no axe to grind on this, just reporting the words of a professional who has been recording for, well long enough to remember the days of analogue tape.

 

Mark

 

 

Apologies if this is too much of an advert, but the recording in question was called Catholic Collection III (not our idea) on Herald AV. I can't make any impartial claims for the quality of the singing, but the Abbey acoustic is wonderful and captured very well.

Posted on: 19 September 2013 by jfritzen

Hi Simon,

 

we seem to agree I think. Perhaps we are both wrong now, but there must be some truth with this temporal resolution theory. It feels sensible. And perhaps 40kHz are the lower limit, as Mr Blackmer suggested, or 100kHz would be better but at least there seems to be a world above 20kHz and a justification for supertweeters.

 

 

KR,

Jochen

 

Posted on: 19 September 2013 by m0omo0
Originally Posted by Simon-in-Suffolk:
[...] This equates to approx 85kHz low pass 4th order Butterworth filter or better. Therefore you can see how 192kHz sampling rate fits in....as a minimum... [...]

Unfortunately... I cannot ! (Except maybe for the fact that, following Nyquist, a 170 kHz sampling rate is necessary to encode properly a 85 kHz frequency, but you've lost me with this filter thing, and with all that follows ! Have you got a link to this paper by any chance ?)

 

Well I have learned something new from the forum.. Good stuff.

And this is not a small feat, considering that it's usually the other way around. Kudos to Jochen !

Posted on: 19 September 2013 by Simon-in-Suffolk

Hi AES paper submissions are not free to access unless you subscribe to the AES library.

Howerever search for 

 

Requirements for loudspeakers and headphones in the "high resolution audio" era 


at http://www.aes.org and search in their library.

 

PS you were spot on your sample rate deduction, and Jochen I now totally agree...

 

Simon

 

Posted on: 19 September 2013 by m0omo0

Thanks Simon.

Posted on: 20 September 2013 by Dozey

Simon raises an interesting point -

"As a follow on in the real world, I wonder how many audio rooms and setups allow for this spatial information to be presented without acoustic stereo imaging smearing from the reflections of the room itself. I suspect not many.

Also surely this information is only truly meaningful from recording a stereo image with at least two seperataed microphones per channel, rather than single microphone channel such as vocal put into the mix and perhaps stereo reverbed. So to take advantage of this surely there would be constraints on how the material was spatially recorded, even if we had the inter channel bandwidth - which of course as you say is debatable."

 

I worked with the team who developed the "Sensaura" system at the EMI Research Labs several years ago. If you are wanting to retrieve all the binaural info recorded by a state of the art microphone system using two loudspeakers you have to do a fair bit of DSP to provide transaural crosstalk cancellation to avoid smearing - so the right ear doesn't hear sound from the left speaker intended only for the left ear. Same applies if you want to locate a panned mono signal accurately in the virtual sound field. It is quite facinating when you get into it. The reflections from the room walls usually arrive sufficiently late that the brain ignores them from the poiint of view of location cues.

Posted on: 20 September 2013 by Aleg
Originally Posted by Dozey:

Simon raises an interesting point -

"As a follow on in the real world, I wonder how many audio rooms and setups allow for this spatial information to be presented without acoustic stereo imaging smearing from the reflections of the room itself. I suspect not many.

Also surely this information is only truly meaningful from recording a stereo image with at least two seperataed microphones per channel, rather than single microphone channel such as vocal put into the mix and perhaps stereo reverbed. So to take advantage of this surely there would be constraints on how the material was spatially recorded, even if we had the inter channel bandwidth - which of course as you say is debatable."

 

I worked with the team who developed the "Sensaura" system at the EMI Research Labs several years ago. If you are wanting to retrieve all the binaural info recorded by a state of the art microphone system using two loudspeakers you have to do a fair bit of DSP to provide transaural crosstalk cancellation to avoid smearing - so the right ear doesn't hear sound from the left speaker intended only for the left ear. Same applies if you want to locate a panned mono signal accurately in the virtual sound field. It is quite facinating when you get into it. The reflections from the room walls usually arrive sufficiently late that the brain ignores them from the poiint of view of location cues.

Just a small remark.

it is my (maybe faulty) understanding that the left / right channel info on loudspeakers is not just intended for left / right ear. Loudspeakes are IMHO not just large headphones ;-).

The mastering, I believe, is done to create a sound image including cross-talk from left to right and v.v.

 

cheers

 

aleg

Posted on: 20 September 2013 by Dozey

Hi Aleg,

 

It depends on how it was recorded. The Sensaura system used a dummy head with microphones in the ear canals. The idea was then to recreate the binaural soundfield you would hear with headphones using two normal speakers. It worked very well.

Posted on: 23 September 2013 by jfritzen

The Tawny IS better, thanks for the hint.

 

Posted on: 23 September 2013 by Michael Chare
Originally Posted by Dungassin:

That agrees with my findings when digitising my LPs.  I finished up ripping to 24/44.1.

 

John

Out of curiosity how did you do the analogue to digital conversion?

Posted on: 23 September 2013 by Felix H
Originally Posted by Dungassin:
Originally Posted by John R.:

I once downloade the very same song in

 

- 16/44,1 WAV

- 24/44,1 WAV

- 24/88,2 WAV

- 24/176,4 WAV

- 24/352,8 WAV

 

and I put all of them on the very same USB stick and did some listening test with my Naim DAC and to my ears there was a great improvement in overall sound quality going from 16 Bits to 24 Bits whereas increasing the sampling frequency from 44,1 to 88,2 only changed the sound very, very minimal ( I am not sure If I would be able to tell them apart in a real blind test to be honest). Going higher than 88,2 did not change anything to my ears.

That agrees with my findings when digitising my LPs.  I finished up ripping to 24/44.1.

 

John

 

Your experiences might also have to do with the quality of DA converters, or the AD signal chain used by yourself or the people who created that download material.

 

I also had difficulty understanding benefits of high-resolution digital when I first experienced it with lower end DACs. But the differences get bigger when the quality of material and reproduction improves: NDX already makes a big difference between CD-quality and higher resolution if the source material is first rate (although I haven't tried various combinations of sample rate and bit depth).

 

Felix

Posted on: 23 September 2013 by Simon-in-Suffolk

Indeed the A to D converter is a critical component and in a 'system' needs to match the D to A converter.

Therefore the qualty and resolution and to therefore to some extent the price needs to match between the two if there is not to be a mis match. If the ADC does not have the precision (note NOT resolution) of the DAC, then there is nothing the DAC can do to recover that information.

A high resolution low precision digital signal will not convey more information, it will simply typically use more data to convey the information than it needs to.

Simon