Listen and choose the 8th generation digital copy.

[Blumlein 88]

For a test that determines if people are able to discern differences it is perhaps better to give three fragments. Two identical and 1 different. People have to pick the different fragment.
This is a very reliable method to test if you are able to pick up differences, it is used in food evaluation for instance.
However, comparing three audio fragments might be difficult.
Has this type of test setup been used in audio?

That's the format I used a couple years ago on a similar test. And there were some complaints. That's the reason I did this as a duo trio test.

 

[SIY]

Another format that can work for this is a sorting test, where you have (say) 10 files, each randomly chosen as original or 8th gen (so it may or may not break down to 5 and 5), and the subject has to sort them into two categories. Reference files can also be provided, i.e., a "known" original and a known" 8th gen. The nice thing is that the subject can choose any listening mode he wants, whether A-B, ABX, triangle, whatever. The bad thing is that if the test is unsupervised, it's easy to cheat with file analysis.

 

[Calexico]

Edit
I first had some preferences on the sounds of the tracks and after retry with another headphone what's the more abvious about the tracks is that it's tipically the kind of mastering i don't like.
As i don't like the original sound it was difficult to choose.
Then i say they all sound ugly
I think it's too much compressed music.

 

[Blumlein 88]

Edit
I first had some preferences on the sounds of the tracks and after retry with another headphone what's the more abvious about the tracks is that it's tipically the kind of mastering i don't like.
As i don't like the original sound it was difficult to choose.
Then i say they all sound ugly
I think it's too much compressed music.

Two of the three tracks were recorded with zero compression used at any point. One of the others I don't think used any compression, but I'm not certain of it. Two tracks have quite a bit of processing. I tried to choose a few variations so one of them might sound okay to most people. Some processed little or not at all and some very processed.

 

[Calexico]

Two of the three tracks were recorded with zero compression used at any point. One of the others I don't think used any compression, but I'm not certain of it. Two tracks have quite a bit of processing. I tried to choose a few variations so one of them might sound okay to most people. Some processed little or not at all and some very processed.

Thank you. I agree it's very difficult to see any differences. For my sound taste maybe it's because it's too much clean or too much digital recording i don't know.
To me
Jennifer sounds like plastic
Bob marley too much boomey bass muddy is hard to discern
Cowboy junky sounds quite good this one
Ry cooder: bass is muddy and guitar is too polite like edges are softened and not enough mediums maybe too much digital recording of the instruments to my taste

 

[Calexico]

@Blumlein 88 that's the kind of sound i like

 

[Calexico]

@Blumlein 88 i'm not sure if it's possible to get this kind of sound on modern recordings equipment.

 

[Blumlein 88]

Thank you. I agree it's very difficult to see any differences. For my sound taste maybe it's because it's too much clean or too much digital recording i don't know.
To me
Jennifer sounds like plastic
Bob marley too much boomey bass muddy is hard to discern
Cowboy junky sounds quite good this one
Ry cooder: bass is muddy and guitar is too polite like edges are softened and not enough mediums maybe too much digital recording of the instruments to my taste

On Ry Cooder and VM Bhatt, two types of drums were used.

And one of the stringed instruments was this.

And that is one of the files where it was recorded by one pair of microphones and no processing of any type. You are hearing the direct sound from the microphones.

Reggae music like Bob Marley is traditionally played with an abundance of low end. So its accurate, but I agree muddy and boomy. It was processed in all sorts of way. As was Jennifer Warnes.

The Cowboy Junkies recording was over a single Calrec Soundfield microphone straight into a Sony PCM F1. You know the early Sony ADC everyone knows sounds so bad. I don't think it had any processing done, but I am not certain. Recorded in a church in Canada. Interestingly the voice was sent thru a PA speaker so it would match the loudness of the instruments during the recording.

Phil Woods recorded with a single pair of mics and no compression on it.

 

[Blumlein 88]

@Blumlein 88 i'm not sure if it's possible to get this kind of sound on modern recordings equipment.

I like the same kind of music, but didn't think it was the best recording for hearing differences in multiple generation copies.
I don't know why you couldn't get that sound. The distortion for the most part is in the instruments and amps being used. Tape was used though there are still studios that will use tape if the client wants it that way.

 

[Soniclife]

The Cowboy Junkies recording was over a single Calrec Soundfield microphone straight into a Sony PCM F1. You know the early Sony ADC everyone knows sounds so bad. I don't think it had any processing done, but I am not certain. Recorded in a church in Canada. Interestingly the voice was sent thru a PA speaker so it would match the loudness of the instruments during the recording.

Phil Woods recorded with a single pair of mics and no compression on it.

Full background on this recording at the following link for anyone interested.
https://www.soundonsound.com/people/cowboy-junkies-sweet-jane

 

[3125b]

Interesting thread.
So I`s say:
Warnes: A=reference (pretty confident)

... alright, it ends there. If that is in fact correct (confirmation?), there is an easy way to cheat.

Well, if my assumption is correct that means that I was right with Jennifer Warnes, wrong on Cowboy Junkies, can't tell with Bob Marley (the reference sounds bad already), I guess I was right on Phil Woods(?), Ry Cooder I couldn't really tell though the recording is pretty good.
How and what tonal differences I'm able to pick upon is definately dependent on the very different headphones I use. My DAC/amp don't degrade the signal audibly, so thats out of the picture.

In any case, the conversion doesn't make much of a difference if any, that's much less of an impact than I would have expected - though theoretically it makes sense I guess, 86dB SINAD is not bad, seeing that number I wouldn't really expect much audible difference. Theere is no way most people would pick up on this under normal listening conditions with typical equipment.

 

[Ponyboy]

I can't tell the difference on the Bob Marley.

Maybe the Phil Woods B is the 8th gen.

Cowboy 8th gen is A.

Maybe with some more concentration, listening closely to specific elements, but I don't listen to music this way and these don't sound much different to me.

 

[MediumRare]

Have ever tallied the overall scores to date?

 

[Blumlein 88]

Have ever tallied the overall scores to date?

No, I don't think I did. I guess I need to do that. There really weren't all that many clear responses on it prior to results being revealed.

I've put up a few such listening tests over the years. The response if I make a poll is always disappointing. I've sometimes tracked downloads and files were downloaded hundreds of times and rarely have I gotten even 20 responses of results. People will argue for pages and pages about something they claim to hear, and then timidly go all quiet if they get a chance to demonstrate that. Not so much here, but on another forum the various excuses for not reporting or even listening to such files is highly entertaining and creative.

 

[Soniclife]

@Blumlein 88 Have you ever done one of these where you take a proper hires track, so something with real info above 22KHz, and in one of the ADC steps set it to capture at 44.1 so that is all lost, then gone back to 96 or higher for the subsequent passes, so the final output looks like a high res track but the data has been lost?

 

[Blumlein 88]

@Blumlein 88 Have you ever done one of these where you take a proper hires track, so something with real info above 22KHz, and in one of the ADC steps set it to capture at 44.1 so that is all lost, then gone back to 96 or higher for the subsequent passes, so the final output looks like a high res track but the data has been lost?

No because it would be trivial to see that had happened looking in an audio editor like Audacity.

What I did do once was take a 96 khz file, and filter out everything below 20 khz with a steep filter. And ask people to tell me if they heard anything. I then included slower and slower versions that transposed the ultrasonic sounds down into the audible band letting people say when they could hear it. That was several years ago. When you can slow down this high sample rate material to a quarter normal speed, and all that is left is some barely heard sounds how can any of that possibly matter when in the original it will all be heavily masked by much louder material. Not to mention happening well beyond what you can hear.

 

[Soniclife]

No because it would be trivial to see that had happened looking in an audio editor like Audacity.

How about splicing the top end from one track onto another track, so 20KHz + data is flat out wrong?

 

[Soniclife]

What I did do once was take a 96 khz file, and filter out everything below 20 khz with a steep filter. And ask people to tell me if they heard anything. I then included slower and slower versions that transposed the ultrasonic sounds down into the audible band letting people say when they could hear it. That was several years ago. When you can slow down this high sample rate material to a quarter normal speed, and all that is left is some barely heard sounds how can any of that possibly matter when in the original it will all be heavily masked by much louder material. Not to mention happening well beyond what you can hear.

The only reservation I have with this is if you mute your mains and leave a sub playing on music with limited bass, there is very little that sounds worth while coming out of the sub, but you do miss it when it's gone. Similar but bigger effect to what happens when you just listen to a tweeter without the other drivers, it kind of sounds like you would barely miss it, but the difference is huge if you disconnect the tweeter whilst playing.

 

[Blumlein 88]

How about splicing the top end from one track onto another track, so 20KHz + data is flat out wrong?

I had considered that and a few other things. Adding shaped noise to look like the original ultrasonics in an FFT. Adding ultrasonics from another track. Or even delaying the ultrasonics a few seconds and putting it back. I never got around to doing any of those.

 

[Blumlein 88]

Okay a tally of results.

If someone missed the correct choice of said it sound the same I counted them in the didn't hear a differences side. If they correctly picked the 8th gen copy they go in the heard a difference column.

I come up with a score of 35 heard a difference correctly, and 34 heard no difference or heard it wrong. Not everyone listed all 5 choices. 44 correct choices would meet the p=.05 criteria.