It looks like you're new here. If you want to get involved, click one of these buttons!
Subscribe to our Patreon, and get image uploads with no ads on the site!
Base theme by DesignModo & ported to Powered by Vanilla by Chris Ireland, modified by the "theFB" team.
Comments
This will read like complete b*llocks but it was more the sense of space or I suppose openess rather than the music itself that was more pleasing to listen to.
My dac comes in 3 boxes one being a totally disconnected battery psu. Totally blind on headphones i could pick within 3 flicks of switch if it was in battery or ac (even though a hifi reviewer reckoned the battery psu made very little difference. it did abd was certainly noticeable and easy to pick).
I reckon AB with ability to choose when it flicks is best test as you cab flick as often or as little as you like between the two. Now pick best and repeat test a few times with A and B allocated randomly and see if pick same one over and over.
it is also easy to hear it doesn't all sound he same by how people build systems. Give two people same budget and tell em to buy a hifi. One may come back with all one make and nice sounding system... The other comes back with auditioned separates which sounds night and day better than other system. Well if we can't hear it and difference doesn't exist how come people can put together separates that sound much better than a random selection? Cos you can hear it. Yes i am sure one day to next and mood, or moving head, or expectations can come in to it but differences also exist and can be heard too.
Randomly allocated AB would seperste wheat from chaf. ABX is flawed as its 3 options and your brain can hear one to next but can't then compare 1st to 3rd so easily. And there is no need for the X when it really offers no advantag over a blind random AB with repeatable results .
i remember i went to get an amp once and auditioned a pile of s/h hi-fi amps with high end atc monitor speakers. Everything sounded good. But the underdog we just chucked into mix as was in budget cleanly blew rest away. My mate was with me and as soon as track we'd been using started we both looked at each other at same time and went holy shit. It was a Sony ES amp which destroyed Marantz, quad a few others and even sounded better than a small krell integrated. In my younger days when I could still heat bats I would happily have taken a blind AB test on hi-fi components and repeat and give same results. Not sure if I'd be up for 192 v 48 these days though. May try it at some point on good headphones and see if i can identify out a random a/b without being bothered which wins.
again it could be your particular dac hardware just handles 192 better than 48 rather than the file itself being the issue
If anyone is wanting to do this, the media player foobar2000 comes with an abx tool built in.
The only point of doing this is if you want to know the truth about what you can hear for yourself, it's pointless to prove anything to anyone else. Some people will believe things regardless of scientific proof as we all know.
I'll just reiterate that doing this made a big positive impact on my productivity and enjoyment of mixing music and think it's worth doing for any mix decision you'd consider subtle.
Any rendering would need to interpolate lots of data points and I can't see how it could do that as accurately as just sampling at 44.1khz (or a multiple of it) in the first place.
Unless I'm missing something?
https://www.npr.org/sections/therecord/2015/06/02/411473508/how-well-can-you-hear-audio-quality
if final product is to be 48 then use 96, 192 etc
if i can pick same item as best time after time in a random hidden AB then it also proves there is a difference.
I don't think it could be possible to compare ab blind with the same 2 pieces of audio a load of times and pick the same one every time but then fail an abx.
If it's actually known in science that this is possible please do link any articles that explain it.
Chance.
What could happen, though, is that someone could pass an ABX test and hear a subtle difference when comparing 3 files where 2 of them are different to the other but then in a blind AB test of preference, not necessarily prefer the same one each time if they have no reference and are just listening the the audio as a whole. That's probably why ABX is used as the standard - a quick Google didn't really suggest AB is considered as an alternative to ABX.
The process itself would be much different for me - listening through the AB and choosing which one I preferred would be quite abstract and only at the end, the results may prove there was a definite difference causing me to prefer the same one each time. With the ABX, I'd know each time if I was confidently hearing which was X or if I was just guessing. Of course, the results would be all that mattered as the confidence could have been wrongly felt but if I knew I was guessing I'd know at the time.
If there is software you know of that does the blind AB test I'd appreciate a link. Would be interesting to see if I could pass an ABX on something very subtle like 320kpbs MP3 vs 192kbps MP3 but then fail the AB version.
Might be a fantasy thinking I'll actually have time to do this to be honest but I'd appreciate the link none the less.
It's been years since I looked into it but I've been doing a big of scouting around, trying to find old forums and articles.
My understanding is that you'd be right, if converting to half the sample rate just involved dropping every other sample point. Then you can be happy, because the samples you're left with are all real samples that were actually measured by the converter as you recorded them.
But it's not that simple. If you did that, it would be effectively like you recorded at the sample rate you're converting to, but had your nyquist filter set twice as high as it should be; have I explained that well enough? You've let signals into the digital domain that you're now not going to be able to reproduce because you don't have enough sample points any more - you've thrown half of them away. And in doing so you're not just throwing away the top octave of your bandwidth, as those higher frequency signals will still be present, modulating the samples you have kept hold of and producing aliasing.
So, no matter what ratio of downsample conversion you do, the process isn't that simple because you still need to first brick wall filter the signal before the downsampling actually occurs. Now every sample point has been moved in amplitude. And that filtering process is far less transparent than interpolation of the signal's sample points. In fact, because filtering introduces phase shifts... your audio signal has been moved in time. Now, by the logic of avoiding interpolation, you're in hot water. Because *every* sample is now shifted relative to the audio signal* by different amounts for each frequency, by some tiny amount through the entire audio range! So you're not left with the actual sample values recorded, you're left with an entirely interpolated signal since the wave it's representing isn't the same any more.
But that doesn't matter. because once you're in the box, once you're safely in the world of pure computation and you don't need to worry about the real-world practical side of digital sampling - analogue filters, quality of op-amps (distortion, slew rate), keeping power supply noise out the audio path, clocking errors - sampling theory is *perfect*. It's as certain as 1+1=2.
So, the computer can calculate interpolated sample points when converting to a non-dividable target rate to an insane accuracy. It's got the whole signal perfectly recorded, and it can calculate any point along the theoretical band-limited analogue wave than the sample points would generate up to whatever precision you use to do the calculation. I'm not a maths or a programming guy, but I wonder how many decimal places you have to calculate to before any error is reduced below the ability of a human ear to pick out. And I wonder how many decimal places beyond that modern SRC processes go. I'd be interested to find out.
Apparently in the olden days of digital sample rate conversion was a different process so simpler conversions used to be better, but these days it just doesn't matter. Personally, I'm happy that any error introduced will be so utterly insignificant that it can as well be discounted.
It's also worth pointing out that as a mixer I'm mangling the audio in all kinds of ways - plugins do their own up and downsampling using different methods, the mix engine of my DAW moves audio about however it's been programmed to, dither may or may not be applied when it should, bit depths are constantly changing... there's so much going on.
Now, on the subject of higher sample rates, this guy has made some fantastic plugs that I love;
https://www.gearslutz.com/board/mastering-forum/968641-some-thoughts-quot-high-resolution-quot-audio-processing.html
And he makes the point that by recorded a wider band than you want to end up with, you can end up introducing more distortion in the final product than if you'd ignored those higher frequencies from the off.
Frankly, I think the whole world of digital sampling is fascinating and I'd *love* to understand it better.
*or rather, the audio signal has been shifted relative to the sample points, which obviously don't move. But it's all relative!
Bandcamp
Spotify, Apple et al
So the tests would have to be done enough times to rule out chance as realistic. I think a surprisingly low number of tests would be needed before the probability of luck can be seen as very unlikely.