[castleofargh]

just for having the will to bother with all that, it upgrades you to the list of people I want to believe. I probably won't because skeptic is my middle name, but I want to

 

[SimonPac]

I agree the test is hardware and software implementation dependent and may even depend on how the computer handles the streaming of different data rates. It's impossible to isolate these variables in practice from the depth/rate variable, though dCS do a nice hardware rate changer for a few grand.....

All round, there is more to get right in h/w and s/w for optimal performance the higher the rate!

 

[bigshot]

Actually, it's very simple. You just create a bunch of different quality sound samples, then you paste them all into a single file at the highest bitrate sample. Then you can skip around in the file as you're playing it and directly compare samples. When you find the sample that sounds the best, you jot down the time code.

 

[pinnahertz]

This one I don't understand. Yes, you cn infer from the files if they are different or not, but isn't the whole point of ABX plugin that when X plays you don't know which file plays? How does the access to the files help you?

It makes the test single blind.

 

[pinnahertz]

I agree the test is hardware and software implementation dependent and may even depend on how the computer handles the streaming of different data rates. It's impossible to isolate these variables in practice from the depth/rate variable, though dCS do a nice hardware rate changer for a few grand.....

All round, there is more to get right in h/w and s/w for optimal performance the higher the rate!

It is absolutely possible to isolate the variables, but it requires an investment in hardware that folks won't want to get into just for a casual test.

 

[pinnahertz]

I can only present results with one DAC and it's a fairly cheap one, a CA DACMagic 100. I'm at the mercy of s/w rate change of course. I intend to get my wife or daughter to randomly choose a file and note it, I've just added -Hi and -Lo at the end of the titles so they can do that.

That will introduce uncontrolled variables and potential bias. Not double blind.

 

[pinnahertz]

Actually, it's very simple. You just create a bunch of different quality sound samples, then you paste them all into a single file at the highest bitrate sample. Then you can skip around in the file as you're playing it and directly compare samples. When you find the sample that sounds the best, you jot down the time code.

But you can't do a direct ABX with a single file, and you'll also be dealing with possible double resampling, which introduces more uncontrolled variables.

 

[pinnahertz]

I can hide the rate lamps easily enough. This will take a while because I'm aware the differences (if any) are fairly subtle. My memory is that they are fairly clear if you quietly listen to a whole album. The threshold for detection with a quick comparison with most gear is probably below 44.1-16 and not above it, agreed.

Detection time is not fixed, nothing requires it be quick. All your gear should be out of sight and earshot it it had any sort of muting relay that might click on a rate change. If your DAC requires time to re-lock to a different rate you have a sighted test.

 

[bigshot]

But you can't do a direct ABX with a single file, and you'll also be dealing with possible double resampling, which introduces more uncontrolled variables.

I don't do ABX in this case. I give the subject a range of quality levels and have them listen carefully and rank them from best to worst sounding. It's not testing for a difference. It's testing perceived quality, which is what they want to test. The end result is a threshold of transparency. He's saying that his level of transparency is 24/192. I think it's much lower. I'd test for that by establishing if he can discern a range of quality levels from MP3 192 on up.

 

[pinnahertz]

I don't do ABX in this case. I give the subject a range of quality levels and have them listen carefully and rank them from best to worst sounding. It's not testing for a difference. It's testing perceived quality, which is what they want to test. The end result is a threshold of transparency.

You need to do the ABX first to determine if there is difference before attempting to quantify the difference then to develop preference. A preference test is a VERY different test structure.

 

[bigshot]

I already know there isn't an audible difference.

 

[SimonPac]

Not doing very well so far! Running the Clapton Album through on a Foobar Repeat Playlist Cycle with both resolutions blind and avoiding listening near the transition I couldn't get a very meaningful correlation. I just went to the rig at various odd times during the day and equated it to the times each version was running later.

On a single HD tracks demo track looping in the same way I managed to lock correctly into the cycle after three passes of the track but that's not very significant.

 

[SimonPac]

I'm going to concede defeat on the Clapton one, I can't seem to reliably tell them apart blind with my fairly ordinary gear. If there's a difference, it's not detectable reliably by me A/B. I down- then up- converted so everything went through at 192.

I tend to get a caffeine overload and a twitch in one eye after a while! I don't find all this very enjoyable to do.

Caveats are that some effects could sink in over a period of time when you're not under pressure, but even if that were true, how would you prove it? Also perhaps your brain 'fills-in' after a while. Usually I seem to think the second one was better, something I've noticed before in myself with small differences. Also the DAC I used tends to smooth things over compared with the one in my CD player even at 44.1.

I've done some searching on blind tests, and yes, you're right. It seems most to all people can't discriminate on blind comparison and neither can I in the situation I have described. Sorry if I've wasted your time.

There are errors of a few uS in one zero crossing point I looked at between the various files when examined in Audacity. In fact re-sampling entirely removes one small cycle through the zero point I looked at. I haven't looked for patterns in time error, but we're looking at a file created presumably with a wideband LPF.

Leaves me wanting to repeat the analogue vs ADA test and wondering if it was for real!

 

[Strangelove424]

Well, your honesty is appreciated. The discomfort you experienced is common. The concentration required to parse such small details is not natural for humans to maintain for long periods. The brain does in fact fill in details which is what makes subjective evaluation so difficult. Your mind will emphasize certain sounds on one listen through, and other sounds on another, even if they were perfect mirrors. It is impossible for the human brain to take in all of the sonic details at once, and because our audio memory is so short, the qualities we think we hear are often an emphasis inside of our own heads.

If you are questioning the DAC, blind test that too. You’ll probably find similar results. The power of the mind to interpret and frame our experiences is quite powerful.

Even if you found evidence of quantization errors, the question is whether it is audible or not. I’m sure there’s variation between playbacks in an analogue chain too, probably worse, but the question is whether or not they are audible.

You certainly did not waste anyone’s time, in fact you are now a participant in the methodology of science. It’s been valuable, and I wish more people had the courage to do ABXing. I’ll bet you in the long run your music enjoyment will increase because of this.

 

[bigshot]

Try encoding it at AAC 256 VBR. I bet you can't discern any difference in that either. Bigger numbers don't mean better quality after the point of transparency. Beyond that it's just inflated file sizes for no reason. The idea of testing is to determine the line where you don't need to worry about it any more. Then you can focus on the quality of the mix and mastering and be secure that the format itself is audibly perfect.

Knowing stuff like that removes a lot of the nagging doubt that makes audiophiles spend money foolishly. This test will save you money in the long run.