Hmmmm... This probably depends on the quality of the DAC (especially its filter). I'd say it's fairly easy to tell between 12-bit and 16-bit, but 16-bit and 24-bit (equivalent quality), you need a really low-noise system after the DAC. I stand to be corrected, but as far as I know, this is the case.
Often a 24-bit DAC isn't as good quality as another 16-bit DAC, then the 16-bit one sounds far better, and sometimes 24-bit DACs sound much better at 16-bit.
Your ability to discern a voice has very little to do with the "quality" of that voice. There are a host of clues your brain uses indicate who is speaking.
Even a toy-like 8-bit audio system can capture enough info to let you recognize who is speaking.
Where do you get that conclusion from? In the second part of the article, the 24-bit and 16-bit were reported as indistinguishable. At best, people INCORRECTLY identified by 52 to 47% the 24-bit sample. And again, a coin flip would be right 50% of the time.
In other words, a massive difference in quality, easily detected by simple audio test equipment, could NOT be discerned by human ears.
The difference is opamps is MUCH MUCH MUCH smaller than the difference in 16 versus 24-bit.
Quite so but equally quite irrelevant as the issue was your claim about long term memory sucking.
No, there were not. People (experienced audiophiles) failed overwhelmingly in the article trying to discern the difference between 16- and 24-bit.
That difference is readily detected by cheap test equipment. And differences in audio quality opamps CANNOT be detected by even expensive test equipment.
The conclusion is plain as day...
Pick any confidence level you wish. Nobody in the article comparing 24- and 16-bit achieved anything remotely better than a coin flip.
The article refutes this. 140 people that fancied themselves as you do failed to detect a dfiference on their very expensive systems.
Then please cite an ABX study were human ears are able to outperform cheap test equipment? Because absent that, this is just your assertions.
You can bet that the 140 avid audio folks in the article with >$1000 systems believed the same too
Every test can have issues with methodology. But the fact is, after decades and decades of all this debate, we've never identified a person that can consistently demonstrate skill in ABX testing given some pretty big flaws introduced into the source material. Sure, they can CLAIM to do it in the their living room, but they all seem to disappear when asked to display this skill in public.
And hey, I can bench press 500 pounds in my private weight room. I just can't do it when others are around.
You are right if you are given free reign for exploring large data sets. But the source material for the ABX didn't include silence. They went to some effort to ensure that you couldn't detect the 24-bit by just looking at file sizes
"Sucking" is relative. I don't need great memory to discern between a long-lost friend's voice and a stranger's voice. I need phenomenal memory to discern between an opamp at -135 dB THD and -115 dB THD (with the former being a good opamp and the latter being a poor opamp) listened to 60 seconds apart.
Yes, its the information coded in what is been said I am referring to.
You can tell who is speaking because the cues are significant ('gross') and because we've evolved to be able to differentiate voices like this. However, in the context we are talking about re high end audio, if I changed the timbre, or added some harmonic which altered the signal by less than 50 ppm and then asked you to do an ABX test, after some time delay of minutes or hours, to pick out which was the original, you would fail.
Have you got a meaning for your introduction of "quality" by any chance ?
Again so and once again irrelevant unless you're listening to sinewaves. Who uses their audio system for sinewaves primarily?
By 'gross' do you mean there's a measurement which can reliably identify a particular speaker's voice characteristics? If so, what is it?
I don't see the relevance to high end audio of either of those modifications you're suggesting to a signal. Care to fill out more details?
There are no humans that can discern the difference between -135 dB and -115 dB distortion as far as I know unless there's been some serious brain/ear evolution taking place the last few decades since the introduction of high fidelity.
No, I mean the differences are significant enough so that you can easily detect who is speaking - I thought that was clear.
To the 2nd point, what I am trying to convey is that if
1. I made minute changes to a recorded voice you would not be able to tell the difference. Ergo, same holds for differences in opamp performance spec of a few dB down at -110 to -130 dB.
2. Further, if there is a time delay of some minutes between the tests, and I asked you to recall which was which, its not likely you would be able to.
So whats wrong with opamps? Nothing. You are not going to be able to better an IC opamp for 99% audio applications. The only two I can think of are mic amps, and if you are fixated with noise, then to go from ~79 dB to 85 dB S/N on an RIAA with a real world MM cartridge you will have to go discrete (same holds for MC as well). But then you are going to have to trade off distortion, PSRR, lots of additional complexity and cost.
Oh, there is one other reason. Because your customer base believes the only way is the discrete way because that's the mantra of the high end. That's perfectly legitimate, but it does not make that approach technically better than IC opamps.
So then, only subjectively the differences are easily detectable? There are no measurements to correlate with those subjective identifications?
'Minute' determined by measurements or subjectively? Its not at all clear which domain you're speaking about here.
I am sure you could measure the differences between different human voices . . . but that's hardly the point I am trying to make here.
Okay, after all this argument back and forth, I decided to go back and double check my hearing. I have an old recording I made of a guy playing acoustic guitar with some fan noise somewhere in the background. The original recording was at 24/44.1. I used Voxengo R8Brain Pro to dither it down to 16/44.1. Can I still hear a difference? It's been awhile since I did a test like this and I'm my HF hearing is worse than it used to be.
Results: I couldn't hear any difference using Sennheiser HD 600 headphones. Switching to the NS-10s, no difference at 3 feet away. However, when I got about 8" from the NA-10 tweeter and carefully listened to the fan noise in the background, I could hear a difference. At 16-bits the HF fan noise and guitar harmonics are more grainy sounding. I had forgotten the exact difference after a few years of not trying. Years ago I could easily hear the difference 3 feet from the speakers, but not now. Even in the past, at first I had to listen very carefully to the HF and not get distracted by paying too much attention to the music itself. After lots of practice, recognition became automatic and that ability lasted for quite a few years.
If anybody wants, I can see if I can post the two files somewhere and you can try for yourself.
Results: I couldn't hear any difference using Sennheiser HD 600 headphones. Switching to the NS-10s, no difference at 3 feet away. However, when I got about 8" from the NA-10 tweeter and carefully listened to the fan noise in the background, I could hear a difference. At 16-bits the HF fan noise and guitar harmonics are more grainy sounding. I had forgotten the exact difference after a few years of not trying. Years ago I could easily hear the difference 3 feet from the speakers, but not now. Even in the past, at first I had to listen very carefully to the HF and not get distracted by paying too much attention to the music itself. After lots of practice, recognition became automatic and that ability lasted for quite a few years.
If anybody wants, I can see if I can post the two files somewhere and you can try for yourself.
This doesn't sound like listening to music for pleasure. As you point out you don't hear it in headphones or listening at a normal distance in an ambient room environment. Any estimation for frequency response/distortion changes by adopting extreme listening positions?
Experiment to support theories...
How about an experiment to support your theory?
Many people cannot pass the Klippel distortion test while for me it was too easy. Yes, it is far from -110dB, but I have good ears here, so think about this: If I could pass the test, I might be able to give some valuable information...
Everyone can theorize how, when and why a voice can be recognized. But if it comes from those who can't hear, what is the value? Get what I mean?
(1) I have a way to recognize voice. Let's see if it will work with small alteration.
(2) Regarding opamps, someone experienced with opamps had mentioned in this thread that it is hard to make opamp circuit that is perfectly stable. This instability may be one of the reason why opamps sound different. LM4562 is known to be "unstable". Can you record clips using LM4562 and any other opamp? I will tell you blind which one is the LM4562 based on its unique sound characteristics and provide an ABX (Foobar) result.
How about an experiment to support your theory?
Many people cannot pass the Klippel distortion test while for me it was too easy. Yes, it is far from -110dB, but I have good ears here, so think about this: If I could pass the test, I might be able to give some valuable information...
Everyone can theorize how, when and why a voice can be recognized. But if it comes from those who can't hear, what is the value? Get what I mean?
(1) I have a way to recognize voice. Let's see if it will work with small alteration.
(2) Regarding opamps, someone experienced with opamps had mentioned in this thread that it is hard to make opamp circuit that is perfectly stable. This instability may be one of the reason why opamps sound different. LM4562 is known to be "unstable". Can you record clips using LM4562 and any other opamp? I will tell you blind which one is the LM4562 based on its unique sound characteristics and provide an ABX (Foobar) result.
- Status
- Not open for further replies.