Quote from: Arnold B. Krueger on 2010-03-09 14:40:37

The right way to do an experiment like this is to take that fine high quality 24 bit source, and dither it down to 16 bits. Then you know for sure that the basic source material was the same.  Been there, done that, no joy - for reasons described in a post I just made.

I understand, I'm just not sure what the original poster was going for. If the question is "Can I ABX 24-bit vs 16-bit of the same material?" the answer is most likely no in normal situations.

However if the question is "Can I hear a difference between off the shelf DVD-A and CD?" the answer is most certainly yes.

I highly recommend anyone who is in to good much check out DVD-As (if their chosen styles of music are on DVD-A) because of the better production that is done. Only downside is that is can make the brickwall limited CDs of today even harder to tolerate .

Quote from: enry2k on 2010-03-11 09:22:04

thank you for the latest posts!
great discussion!
Don't you think that it would be interesting to use the following special sample signals to esperience the full dynamic range of a digital system?

a 2.5 khz low level square wave obtained by swiching on and off only the least significant bit during half period, in other words the lowest possible square wawe.
a 2.5 khz high level square wave obtained by switching all the wodcode bits on and off during half period. in oher words the highest possible square wave.

These samples can be created manipulating a blank linear pcm wav sample in a hexadecimal editor.

I have chosen 2.5 khz frequency because it is positioned within the highest sesitivity region of the human ear.

You could experiment the low level square wave in different volume and noise coditions compared to the loudest signal level.


What do you think about?

You should use dither, and choose the lowest level signal that's still audible against the noise floor. It'll most likely be lower than your "lowest possible square wave".

a 2.5 khz low level square wave obtained by swiching on and off only the least significant bit during half period, in other words the lowest possible square wawe.
a 2.5 khz high level square wave obtained by switching all the wodcode bits on and off during half period. in oher words the highest possible square wave.

These samples can be created manipulating a blank linear pcm wav sample in a hexadecimal editor.

The widest dynamic range source material I've ever found for regular commerical sale had about 85 dB dynamic range. Doing that without fudging is no mean trick, either. And, it is still 15 dB shy of the century mark.  A typical recording made by fairly natural means has maybe 70 dB dynamic range.

And that room with 10 dB SPL noise still forces you to listen to 110 dB peaks if you want to hear the benefits of that 100 dB dynamic range. The ear is most sensitive at more like 85 dB for most people. So, you still won't be listening at your peak.

I'm trying to decide if there's any benefit in releasing music in 24bit format, compared to the traditional 16bit.

Audacity has a Generate tool that can create signals & waveshapes.  Probably any audio editor can do it!

How did you get at these numbers? Is there a easy method to calculate the dynamic range of a recording?
...
note: a microphone can record ~120db in a studio environment. why aren't there any (commercial) recordings that are beyond 70db?

Is there a easy method to calculate the dynamic range of a recording?

Why does the noise floor matter if you want to experience the benefits of high dynamic range? Sure, that's kind of the definiton of dynamic range, but i mean this: the benefits of high dynamic range are also expressed in terms of the accuracy of the recording (ie: a smoother transition between amplitude, or does interpolation make this irrelevant)?

I'm trying to decide if there's any benefit in releasing music in 24bit format, compared to the traditional 16bit.

I'm more concerned about recordings that were generated using software synthesizers, that can utilize the full dynamic range of the format (i'm assuming this is the case).

note: a microphone can record ~120db in a studio environment. why aren't there any (commercial) recordings that are beyond 70db?

I have read in the specifications of a Creative amplifier speaker a 75 db signal to noese ratio, so I guess it more than enough for 16 bit quatization?

The designer of Harbeth speaker stated recently in the Harbeth user group that Harbeth speakers quantization is 11 bits. Harbeth is an audiophile grade speaker maker, renowned for the clarity of its sound. I doubt other brand of speakers could be any better.

Quote from: Yee on 2010-03-15 04:41:52

The designer of Harbeth speaker stated recently in the Harbeth user group that Harbeth speakers quantization is 11 bits. Harbeth is an audiophile grade speaker maker, renowned for the clarity of its sound. I doubt other brand of speakers could be any better.

Trying to figure out what you mean by that.

Jack hammer at 1m     100 dB SPL
[...]
Very calm room    20..30 dB SPL

70dB + 20dB  = 90dB. So in that recording, there's almost a hammer captured at 1meter, plus the noise of a very calm room.

What do you miss outside of this range?


The problem with dynamic range is that most of the time, we cannot listen to it. If you play a record at 60dBSPL (a level acceptable at home), your dynamic range is around 40dB's. If the record has a bigger range, you have difficulties or simply cannot hear it.

Quote from: Yee on 2010-03-15 04:41:52

The designer of Harbeth speaker stated recently in the Harbeth user group that Harbeth speakers quantization is 11 bits. Harbeth is an audiophile grade speaker maker, renowned for the clarity of its sound. I doubt other brand of speakers could be any better.

Trying to figure out what you mean by that.

A rare case, where high dynamic range would be useful, is if you have a recording done at low volume. Say a field recording in nature, where you want to hear every sound wave. Under these conditions, a low gain might be necessary to prevent clipping of a sudden loud noise. For example: You want to record the faint sounds of insects that are miles away, and also an occasional loud sound (which would be removed later on). Of course if you were to release it, you'd amplify it to regular levels and then might aswell reduce it to 16bit.

For the record, I found a record that has -89 dB. I analyzed the first 15 seconds of Autechres album Oversteps. This is a 24bit recording released on the web.  The first raise in amplitude starts at ~1216 Hz, well within the range of 16bit.

Quote from: Sycraft on 2010-03-15 09:20:07

Quote from: Yee on 2010-03-15 04:41:52

The designer of Harbeth speaker stated recently in the Harbeth user group that Harbeth speakers quantization is 11 bits. Harbeth is an audiophile grade speaker maker, renowned for the clarity of its sound. I doubt other brand of speakers could be any better.

Trying to figure out what you mean by that.

Speakers are quiet when you're not sending any program to them. But, when driven, speakers make all kinds of noise in addition to the signal they're trying to reproduce. You've got harmonic distortion (which I assume he is excluding from his 11 bit figure) but you also have noise unrelated to harmonics from moving parts and moving air.

This was a very rough approximation based an some estimations. First, let's have a think about what we mean by 'resolution'. We infer by the use of the word something to do with our ability to detect and measure external stimulii. We could be referring to the resolution of photographic film or the resolution of our eyes and it's ability to discriminate very small shifts in brightness and/or colour. We could be talking about the resolution of a digitised audio stream (music) or we could be talking about the resolution of our sense of touch, smell, hearing or any other means of grading the external world.

    Thinking about our place in the universe, it's clear to me that as human life is only viable over a very narrow temperature range of about -10 degrees to +50 degrees (range roughly 60 degrees C). We can compare those to the temperatures found across the universe, these range from -273 degs. to + 200,000 degs (range roughly 200,273 degs.). So, the temperature range found 'out there' is at least 3000 times wider than that which can support human life. That means, in effect, that we humans here on cosy warm little earth are sensing only one three thousandths of the temperature range across the universe*. Or put another way, our resolution of temperature is very poor, or more accurately, our sense of the dynamic range of temperature is extremely limited.

    This extremely limited capability applies equally to all our senses. And industry has been very successful at creating products which are perfectly acceptable to our poorly-resolving senses. Look at photographic film under the microscope and we see that the image is actually made up of clumps of colour dyes (pix to follow). Providing that we don't use a magnifying glass on the film to see the limitation of the resolution we can be perfectly contented with the results. And the same applies to TV. Providing that we don't sit too close to the TV we don't see how limited the resolution is. And with sound - providing that we don't turn up the volume when playing an LP we may not be troubled by the hiss, clicks and crackles. But if we do 'zoom-in' on the photo, the TV or the analogue sound we can clearly detect the limits of resolution that define those technologies.

    In the photo, if the light from the object that hits the film has finer detail (hence is smaller) than the dye molecules, the molecules will clump together and the fine image will turn into a smear. And as nothing can be reproduced on an LP that is below the hiss and crackle the same applies - the stylus just doesn't reveal what's buried under the random noise. And of course, it's exactly the same with digital audio. When the CD was invented Sony/Philips knew that whatever technical standard they settled on for resolution would make or lose them billions of dollars. Too low a technical resolution and the format wouldn't be marketable as high fidelity - too high, and the discs wouldn't hold an hour of music. So, they settled on a resolution of 16 bit, 44kHz. It's back to our analogy about temperature again - we know that the temperature range 'out there' in space is about 3000 greater than we can live in, but so what? We'll never be expected to resolve that range. And when Sony/Philips settled on the spec. for CD they settled on a resolution that was/is better - far, far better - than any human could detect. So, job done.

    Where does that leave us with speakers then? First, speakers use heavy mechanical parts to create sound (the cone, coil etc.). Those parts have inertia, and that inertia disinclines the speaker cone from moving, and hence from producing sound. So there is a certain minimum amount of energy that you have to put into the speaker (let alone the peculiarities of the human hearing's ability to resolve bass at low levels - see ISO226). And so, just as with the film and the LP record, the speaker has a certain resolution limitation - it just cannot yield the super-micro-details in the sound. Even a good speaker is certainly nowhere near as good as the CD's 16 bit resolution, but it's better than 8 bit (which sounds robotic) so it's probably somewhere about 11/12 bit, which is, incidentally, just a little under the 14 bit NICAM digitisation system used to distribute analogue TV and FM sound around the country. Again, the spec for that wasn't just plucked from the air - listening tests concluded that it was good enough for the purpose.

    My point is this - the film, the TV, the LP and the speaker - and of course, our entire sensory system - is only capable of very limited resolution. What matters is the quality of performance and minimisation of coloration within the limited resolution. So we here strive to make a clean, life like sound from the reliable cone technology we have. Obviously, the polypropylene cone has less resolution than our RADIAL™.

    If you're interested in the resolution of (analogue) film v. the digital camera this is interesting here. Clearly, no matter how much money you spend on your lens, the film just cannot resolve the fine details. No matter how much money you spend on your turntable or cartridge, beautiful though they may be, the vinyl record just cannot resolve the fine details. And no matter how much you lash out on 24/96, your ears are not capable of resolving such fine detail. Conceptually, if the sound could be applied to the brain electrically (it will be possible one day) then we could really improve our 'hearing' resolution.

    Alan A. Shaw
    Designer, owner
    Harbeth Audio UK

I highly recommend anyone who is in to good much check out DVD-As (if their chosen styles of music are on DVD-A) because of the better production that is done. Only downside is that is can make the brickwall limited CDs of today even harder to tolerate ><.