Yes Tom you just showed the point I was tying to make exactly. 14 bit fullscale audio for a 1bit window is only... 1 bit. The audio you're listening to doesn't automagically take advantage of all the bits, only the human ear does that in the real world with real dynamic audio, or an equivalent bit depth which is 43 bits.

We can hear in the whole band at the same time, but the strongest signal is the only one that's going to be noticed. In a delicate ensemble, only the effective dynamic range is important. Most people are deaf to such nuance because the dynamic range of typical modern music is so incredibly high it can't take advantage of the nuance of human hearing. You have to have eclectic tastes in music to notice.

Fidelity is all about the precision of the reproduction at EVERY range that is acoustically noticeable at a given moment. For the audio range of human perception 24 bits is perfectly reasonable.

I was making the point that we may have 130dB full scale, we may be only able to hear 16 bits of that at any time (or whatever you want - maybe 24 bits.) But we certainly cannot hear 43 bits. I think you misunderstood the full scale range, just like the human eye (which can see 16 million distinct colors, or about 7-8 bits - but, the gamut is much wider than say sRGB and of course we also have rod cells which pick up luminance), the human ear cannot understand all 43 bits of full scale range. Additionally, the human ear has a logarithmic response, in that a really quiet sound can be overwhelmed by a louder sound.

My el cheapo multimeter has approximately 12 bits resolution (4,000 counts) and measures from 100µV (1 count on 400.0mV range) to 1000V (1000 counts on 1000V range), so would you say it has 80 dB range and therefore ~26 bits of resolution? Of course not.

Now, I'm sure there are additional advantages to working in 24 bits over 16 bits. The audio is more preserved if changes are made to it, and it's less sensitive to additional error (e.g. noise, or error in the last few bits.)

 

Say that the amp can place +/- 80 volts on the speaker wires. 16 bits is really 15 bits signed, so you have positive and negative voltages following the DAC output.

80 volts /32768 = 2.44 mv, which is the voltage response to the LSB with the volume maxed. 2.44 mv/8 ohms = 305μamps. Not at all sure that is audible.

20 bits, by the same figuring, has an LSB voltage weight of 15.3μV. That produces a whopping 1.9μA of voice coil current.

24 bit response is going to be hard to detect by any means.

 

Tom, you never did link the study you were talking about, I wanted to read it.

 

It's fun to consider 24 bit sound going from A to D. We have to be signed again, so there are only 8,388,608 states. For a 0 dB signal (kinda rubbery as to what that is, but we will accept 1.25 volts), the LSB is all of 100nV.

 

Tom, you never did link the study you were talking about, I wanted to read it.

Sorry, I can't find it. I did find one study which states that it is impossible to determine over 118dB (approx. 19.6 bits) of audio information, but it was produced by Ampex, which have an interest in audio storage, so it could be biased.

Either way, 24 bit audio is pointless. 20 bits is the limit of current systems, because current technology simply cannot reach the <100nV (as beenthere suggests) noise floor.

 

Really tom? Blueray does 16 20 and 24 bits, Intels HD audio standard goes up to 32 although I've never seen anything that actually uses it.

Where are you coming up with 118db and 19.6 bits? Those numbers don't make any sense out of context of the study you're referring too which you didn't reference.

 

ADC's are far more complicated than these simplistic calculations.
You never get full performance out of an ADC, noise and errors in the conversion make you lose several bits. If you hear LSB's you hear artifacts.

If your original signal has high SNR then you need some minimum resolution such that the ADC quantization error doesn't dwarf the original noise.

If high resolution systems weren't required then they would not be often used. It's a real challenge to actually get high resolution performance.

 

Slightly off-topic of off-topic, fairly often these numbers are forced up for no real reason apart from marketing.
Putting a 24 bit DAC in something won't cost much more than 20 bit, but if it persuades someone to spend a premium on it, then it's worth doing, regardless of if it is better.
The same thing happens in the MHz race in processors, and in the megapixel race in cameras.
Often it's possible to make a processor have higher MIPS with a lower MHz.
Cameraphone lenses are incapable of resolving anywhere near the detail required for a 10 Megapixel sensor. They would almost all be better with a 3 Megapixel sensor because they would capture the same detail and work better in low light.
Unfortunately consumers see Megapixels as quality.

Rant over.

 

Ghar, I wasn't talking about an ADC. I was talking about the human ear.

Want to know the difference between an audiophile and someone that only looks at consumer audio equipment numbers and simple scientific studies? It's pretty easy, simply take a person to a full orchestral production and make sure they're in 'good seats' for an audio performance at that theater, and then play the same performance back on a the highest quality multichannel audio equipment that can be currently produced by man.

NO living human being that doesn't have severe hearing damage would say there was no difference, some of the better quality surround systems come close, at least enough to provide a solid facsimile... if you hold your head still enough. The difference between an audipphile and the masses is audiophiles KNOW what they're missing from an acoustical reproduction.

By the way, please NEVER even for a split second link me up with those fruity nuts that are buy ungodly expensive audio equipment that doesn't actually produce better audio. I know what my ears tell me.

I wish I could provide references as it's been quiet a while but a simple Google search will provide many proofs to what I'm referring to hence.

I have seen some websites that hint at some of the details the truly nail down the difference between the human ear and an acoustical reproduction and it's because of the shape of the ear itself, it's weird looking and shaped such for a reason, each outer ear is actually a wave shaping device designed to focus some frequencies and cause phase shifts (facing forward for humans cats and may other animals can rotate their ears) this allows the perception of location in a limited fashion, the primary way we can tell is the bulk phase change, but the individual channels of the outer ear cause phase change at different frequencies, the most useful of which are in the human voice hearing range and otherwise common 'alert' ranges of audio that have been evolved over time.

You'd have to Google it but I've seen experimenters that have taken recordings from a facsimile of a human ear with microphones where the auditory sensors might be and played back, as long as a person could keep their head still the position audio placement cause people that heard the playback to do double takes, optimally the speaker actually had to be in the ear canal (earbuds) and that's only a sliver of the complexity of human hearing.

The best of the best of the best of audio reproduction technology today can't hold a candle to hearing the real thing in person.

 

Ghar, I wasn't talking about an ADC. I was talking about the human ear.

I know, my comments weren't really to you.

 

As I said before, I can't find the reference for the 14 bit figure. I found it on the internet a long time ago, and understandably, I have forgotten.

However, here is the Ampex reference: http://www.aes.org/e-lib/browse.cfm?elib=11981

I don't mean to associate you with audiophiles.

Of course, what works, works, use whatever you like. I couldn't care; I think you're wasting your money on something that really isn't worth it for whatever improvement (if at all) it brings.

 

I personally think the best audio advancement in my memory is the advent of in ear silicon sealed earbuds. You have to have a really nice home theater to enjoy high quality audio, but a good set of earbuds will get you 99% the way there. The bass performance is actually too good, it tends to need a bit of EQ'ing in order to get a realalistic playback. I was playing around one day and found out that I could FEEL my eardrums increasing and decreasing in pressure like you would in a car when the headphones were driven even down to 10hz.

I read that link you provided at least the cover of it (I'm not paying 20 bucks for it). And you either misquoted it horribly or interpreted it dead wrong. It states that 118db's worth of signal is required for subjective noise free reproduction, that means it's the MINIMUM required not to notice the noise floor, it has nothing to do with the total possible usable dynamic range of human hearing, just perception of white noise, not signal.

 

I read that link you provided at least the cover of it (I'm not paying 20 bucks for it). And you either misquoted it horribly or interpreted it dead wrong. It states that 118db's worth of signal is required for subjective noise free reproduction, that means it's the MINIMUM required not to notice the noise floor, it has nothing to do with the total possible usable dynamic range of human hearing, just perception of white noise, not signal.

I found the link on Wikipedia as a reference.

I do not know enough about audio engineering to argue the point further, so I shall stop now.

I wouldn't and didn't pay for the article either.

 

By the way Tom, again I'm sorry, I'm not trying to argue, I know I come across as harsh online, but I also know intuitively what you're saying doesn't make sense to me, that's why I'm trying to find out where you're getting your information from so I can either correct myself, it or at least understand a little better.

Since neither of us know how to progress from this point, that pretty much says it for the conversation =) Unless anyone else has anything to add. I love being wrong because that means I've learned something, and this requires discussion, I don't like to argue, I like to discuss, I try to avoid (and fail) at looking stupid as well =>

 

How about putting it back on track to audiophiles?

OK maybe 24 bits is useful (I don't know), but what about those ultra tweeters? Or those cable lifters? There truly is a sucker born every minute.

 

I've seen worse. 6 foot RCA cables being sold for a few thousand dollars, don't have the link anymore but it was scary. If they sold even ONE set it would justify the entire enterprise. The scariest thing for me to think is that they might have actually sold a pair.

 

I generally don't talk to true audiophiles, it is exactly like talking religion with a true believer who thinks you are ignorant because you don't believe. I prefer the science of it.

I had a discussion on Arron's forum about speakers and clipping. I questioned why a speaker would blow because the signal was clipping (and not from the speaker). You would think I had invoked heresy. If the phenomenon exists I'd like to know why, instead of "just because". I left the forum, it wasn't for me, and found this one several years later.

 

My goodness. We used to take an Emerson, Lake, and Palmer album with us to all the stereo stores in San Diego. One of the tracks had some bass on it that would clip any amplifier. We did it to annoy salespeople, but we never managed to pop a speaker.

If you are applying a signal level to an amp that will cause clipping, I think the amp will usually shut down or blow the outputs before the speaker voice coil opens (assuming the speaker is rated for the amp). I also don't think I want to experiment to find out.

 

There are some obscure harmonic compunctions that can manifest in a real world wide band loud speaker that would play just fine in a multi speaker system with the proper crossover, even if the total power output of the multi speaker system was 1/4 of the single speaker system.