No! 20Hz----20kHz is NOT the range of human hearing!

[blue2blue]

The difference between 20 bit and 24 bit AD is, for most of our gear, inconsequential to pretty much non-existent.

 

20 bit AD affords 120 dB S/N. Few of us here have an input chain anywhere near quiet enough to 'take full advantage' of that.

 

That said, using deeper bit depth and/or floating point for any internal processing makes good sense.

[Beck]

The difference between 20 bit and 24 bit AD is, for most of our gear, inconsequential to pretty much non-existent. 20 bit AD affords 120 dB S/N. Few of us here have an input chain anywhere near quiet enough to 'take full advantage' of that. That said, using deeper bit depth and/or floating point for any internal processing makes good sense.

 

 

Right, and I find there is more difference between converter designs @ 20-bit and above than whether it is 20-bit or 24-bit.

 

IME, 20-bit is a big improvement over 16-bit though. However, S/N ratio isn't my primary concern when looking at bit depth. "CD quality" is already as quiet as having no noise. Hell, even my old (and still working like new) Tascam 246 cassette portastudio, with its 95 dB S/N (A-WTD /dbx on) is pin drop quiet. At some point less noise just means less than the no perceptible noise you already had (didn

[MikeRivers]

 

As I recall Rupert claimed that he and/or Geoff Emerick heard a circuit oscillating at 50 KHz. But what speakers can output 50 KHz? None! At least none available at that time. So it's clear what was heard was more likely IM distortion that aliased down into the audible range.

 

Yes, and I think he acknowledged that he wasn't hearing a 50 kHz whistle, he was hearing distortion that he couldn't account for when the high frequency square wave was going into the console. That's frequencies that weren't in the pure sine wave (or first and only humanly audible harmonic in the square wave), but not harmonically related.

 

If you switch a typical function generator from a sine wave to a square wave, the peak voltage stays the same but the RMS level at the fundamental frequency changes. So of course the sound will change, though it's due only to the level change and not the presence or absence of ultrasonic content.

 

That was the obvious reason, and one that anyone can detect, but you're right - they need to be matched so there won't be a level change.

[blue2blue]

 

Right, and I find there is more difference between converter designs @ 20-bit and above than whether it is 20-bit or 24-bit.   IME, 20-bit is a big improvement over 16-bit though. However, S/N ratio isn't my primary concern when looking at bit depth. "CD quality" is already as quiet as having no noise. Hell, even my old (and still working like new) Tascam 246 cassette portastudio, with its 95 dB S/N (A-WTD /dbx on) is pin drop quiet. At some point less noise just means less than the no perceptible noise you already had (didn

[MikeRivers]

This is basically what I was trying to say for like 30 hours when we were discussing the methods used by Meyer and Moran in their Audibility of a CD-Standard A/DA/A Loop... study. :poke:

There's a difference between measurements and gathering of statistics. In a paper presented by one of Ethan's panelists in a different session, she stated that with the unusually small sample size (number of listeners) used in most of the published listening tests, conventional statistical analysis methods don't apply and that special techniques are used to draw conclusions from the data. Statisticians seem to think these are reasonably valid, but don't have really high confidence levels. It's a sophisticated and systematized version of "fudging."

 

Why are sample sizes so small? Because it's too expensive to find a sufficient number of qualified listeners who are capable of learning what to listen for and providing valid results based on their hearing, not on their biases, moods, or understanding (or not) of the experiment.

[Rudolf von Hagenwil]

according to Ethan's statements

he works with a 14-bit converter

[Beck]

Well, we're not necessarily talking about random or semi-random noise here (although that's part of it and that's most of a properly functioning analog chain's noise floor)... patterned noise from alias error is generally considered to be a more noticeable problem for a given 'amount' of noise. The reason, I suspect, you feel like you have a much more clear sense of improvement when comparing 16 bit vs 20 bit as opposed to a 20 bit v 24 comparison (as far as AD goes), is because you're almost certainly not making 'full' use of the dynamic bandwidth of the 20 bit in the AD process, let alone the 24. Remember, adding more bit depth to a given signal adds that extra bandwidth to 'the bottom' of the signal space... extending the "footroom" (if you'll pardon that expression sure to generate rolling eyeballs around the room).

 

 

Right, analog tape falls into tape hiss in refference to S/N, while digital falls into distortion as the bit depth drops with signal level. So theoretically with a 16-bit converter you only have 16-bits at 0 dBfs. Thus I've always thought we should have LED meters on digital equipment with yellow and red LEDs at both ends, with green in the middle. That way you know both what not to exceed and what not to fall below for best sampling results.

 

Many that complain about tape hiss have no idea that bit depth drops as signal drops, thus so does quality of the sample, which can be audible at very low levels.

[Rudolf von Hagenwil]

mama mia, kiddies,

no wonder some of you believe that analog tape recorder sound better then the best digital converter

[Rudolf von Hagenwil]

Many that complain about tape hiss have no idea that bit depth drops as signal drops, thus so does quality of the sample, which can be audible at very low levels.

 

 

 

that's how it is with PCM, but not with DSD

 

and it is not audible with PCM, that is false thinking

[blue2blue]

 

There's a difference between measurements and gathering of statistics. In a paper presented by one of Ethan's panelists in a different session, she stated that with the unusually small sample size (number of listeners) used in most of the published listening tests, conventional statistical analysis methods don't apply and that special techniques are used to draw conclusions from the data. Statisticians seem to think these are reasonably valid, but don't have really high confidence levels. It's a sophisticated and systematized version of "fudging." Why are sample sizes so small? Because it's too expensive to find a sufficient number of qualified listeners who are capable of learning what to listen for and providing valid results based on their hearing, not on their biases, moods, or understanding (or not) of the experiment.

 

Training can help reduce the effects of cognitive biases but you can't really learn your way free from them. They are part of how we're built.

 

But at the point where you're carefully picking your subject base, you are imposing other biases on any potential test results that will greatly limit their extrapolative value. The ideal, in order to be able to derive potentially valid extrapolation, is a random sample within the target population you want to be able to derive conclusions about.

[Rudolf von Hagenwil]

just checked Ethan's make all the world happy sonic quality...

it is actually 13?-bit and sampling rate 32 kHz

[blue2blue]

 

Right, analog tape falls into tape hiss in refference to S/N, while digital falls into distortion as the bit depth drops with signal level. So theoretically with a 16-bit converter you only have 16-bits at 0 dBfs. Thus I've always thought we should have LED meters on digital equipment with yellow and red LEDs at both ends, with green in the middle. That way you know both what not to exceed and what not to fall below for best sampling results. Many that complain about tape hiss have no idea that bit depth drops as signal drops, thus so does quality of the sample, which can be audible at very low levels.

The nice folks at SSL make -- and give away -- an intersample aware level meter that also shows fluctuating bit depth 'used' of your signal: http://www.solid-state-logic.com/music/x-ism/index.asp

 

(It's a little CPU-hungry, so I usually only insert it when I'm 'mastering' my finished work.)

[blue2blue]

just checked Ethan's make all the world happy sonic quality... it is actually 13?-bit and sampling rate 32 kHz

OK... what are you on about, Angelo?

 

I can never tell anymore when you're engaging in your piquant and sometimes surreal humor and when you're actually making a point. I'm thinking the inclusion of a 7/8 fraction there may signal the former rather than the latter, but, you know... I'm a little at sea trying to make sense of some of your comments I've been trying to track out of the corner of my eye.

[Rudolf von Hagenwil]

well,

Ethan stated somewhere that a frequency range of 16 kHz in the delivery is sufficient - that's a sampling rate of circa 32 kHz, which equal circa a quasi 13?-bit bit depth of a MP3 (MPEG-1 Audio Layer 3) 128/kbps lossy compression algorithm IISO/IEC 11172-3, ISO/IEC 13818-3, respectively a frame header value 3, subband 16 bis 31, intensity-stereo: on

[Beck]

There's a difference between measurements and gathering of statistics. In a paper presented by one of Ethan's panelists in a different session, she stated that with the unusually small sample size (number of listeners) used in most of the published listening tests, conventional statistical analysis methods don't apply and that special techniques are used to draw conclusions from the data. Statisticians seem to think these are reasonably valid, but don't have really high confidence levels. It's a sophisticated and systematized version of "fudging." Why are sample sizes so small? Because it's too expensive to find a sufficient number of qualified listeners who are capable of learning what to listen for and providing valid results based on their hearing, not on their biases, moods, or understanding (or not) of the experiment.

 

 

My problem with those sorts of surveys anyway is that it

[Beck]

well, Ethan stated somewhere that a frequency range of 16 kHz in the delivery is sufficient - that's a sampling rate of circa 32 kHz, which equal circa a quasi 13?-bit bit depth of a MP3 (MPEG-1 Audio Layer 3) 128/kbps lossy compression algorithm IISO/IEC 11172-3, ISO/IEC 13818-3, respectively a frame header value 3, subband 16 bis 31, intensity-stereo: on

 

 

 

I don't know about Ethan, but I stated that 16 kHz is usually sufficient for all practical purposes when it comes to music. But I didn't say that means you should sample at 32 kHz. I advocate higher sampling rates. The impulse to reduce sample rate to match a reduced top frequency is wrong thinking. That's not what Nyquist is about.

[LeftCoast]

But if I can't hear above 20Khz (actually for me, at 45 years of age, it's about 16Khz), what do I do with my expensive speaker cables that are, directional, impedance matched and designed to minimize skin effect and strand jumping at audio frequencies?

 

I think it is important that audiophile rats, mice, bats, dogs and cats experience my music to its fullest.

 

Actually - I bought one of those utra-sonic rat repellers to keep the rats out of my attic. I wonder if interferes with their perception of music?

[Rudolf von Hagenwil]

There's nothing wrong with my hearing or all the other people's hearing that could hear it on old CD's as the music faded. I know you've been around at least as long as I have. I don't know about Ethan, but I stated that 16 kHz is usually sufficient for all practical purposes when it comes to music. But I didn't say that means you should sample at 32 kHz. I advocate higher sampling rates. The impulse to reduce sample rate to match a reduced top frequency is wrong thinking. That's not what Nyquist is about.

 

 

I wonder what you hear when the bit meter drops to less then the nominal bit rate when the dynamics are low.

 

I am not so much into "practical purposes when it comes to music", we record in the best quality, the compressions to all sorts of delivery format we make from those masters, this in audio as well video productions

[Rudolf von Hagenwil]

I think it is important that audiophile rats, mice, bats, dogs and cats experience my music to its fullest.

 

 

a real pro !!!

 

i always have a dog in the studio when I record with 28-bit converters and sample rates in the megahertz range

[Beck]

mama mia, kiddies, no wonder some of you believe that analog tape recorder sound better then the best digital converter

 

 

I only believe what I hear when listening to any format. That's all that matters. We can shoot the breeze about why we think something is so, but why we think something is so does not change whether it is so, even if we are mistaken about why it is so. It's all in the hearing.

 

Music is made for people, not test equipment. I wonder sometimes how many people have forgotten that.

[blue2blue]

well, Ethan stated somewhere that a frequency range of 16 kHz in the delivery is sufficient - that's a sampling rate of circa 32 kHz, which equal circa a quasi 13?-bit bit depth of a MP3 (MPEG-1 Audio Layer 3) 128/kbps lossy compression algorithm IISO/IEC 11172-3, ISO/IEC 13818-3, respectively a frame header value 3, subband 16 bis 31, intensity-stereo: on

 

Ah... got it!

 

Thanks. I figured there was a there there. Somewhere.

 

[Beck]

Actually - I bought one of those utra-sonic rat repellers to keep the rats out of my attic. I wonder if it interferes with their perception of music?

 

 

Probably not, because it drives them out of the attic into the den where they can sit right in the sweet spot between the speakers.

[Rudolf von Hagenwil]

I only believe what I hear when listening to any format. That's all that matters. We can shoot the breeze about why we think something is so, but why we think something is so does not change whether it is so, even if we are mistaken about why it is so. It's all in the hearing. Music is made for people, not test equipment. I wonder sometimes how many people have forgotten that.

 

 

 

do you think the universe cares about us humans?

[LeftCoast]

 

a real pro !!! i always have a dog in the studio when I record with 28-bit converters and sample rates in the megahertz range

 

 

 

Does you dog like Rush?

 

My 8 year old son has great hearing but he likes Lady Gaga, so I can't really trust his ears.

[blue2blue]

 

I only believe what I hear when listening to any format. That's all that matters. We can shoot the breeze about why we think something is so, but why we think something is so does not change whether it is so, even if we are mistaken about why it is so. It's all in the hearing. Music is made for people, not test equipment. I wonder sometimes how many people have forgotten that.

 

 

Well, listening is the final destination (at least if we sort of containerize everything that happens after sound wave meets ear), the receipt of product, if you will, but just as many a craftsman extends his own visual and spatial perception with various measuring tools in order to help assure that his eye hasn't led him astray, as well as to refine his knowledge and understanding of the physical elements before him, so, too, does a smart recordist not willfully throw aside the potentially more precise and almost certainly more consistent and reliable measurements provided by good measurement and test gear.