PlayClassics test files to compare file formats

[sandyk]

PlayClassics test files to compare file formats I would like to offer some test files so you can compare the sound of the different file formats and report your conclusions. See if we can find out how much the file format really affects the final experience.

 

In case some of you haven't noticed, Mario kindly provided these files for people to listen to, and report back their listening experiences. This will enable Mario to decide which formats he should supply his files in.

I doubt that his original intention was for people to play around with pseudo scientific tomfoolery in an effort to try and prove that the 24/96 version can't possibly sound better than the standard RB CD version.

Mario ALREADY knows that 24/96 sounds better, or he wouldn't be recording in this format, as his files don't need further processing.

[goldsdad]

deleted

[Mario Martinez]

In case some of you haven't noticed, Mario kindly provided these files for people to listen to, and report back their listening experiences. This will enable Mario to decide which formats he should supply his files in.

I doubt that his original intention was for people to play around with pseudo scientific tomfoolery in an effort to try and prove that the 24/96 version can't possibly sound better than the standard RB CD version.

Mario ALREADY knows that 24/96 sounds better, or he wouldn't be recording in this format, as his files don't need further processing.

 

 

Hi Alex,

 

I am offering these files unconditionally.

 

You can do whatever you want with them. The purpose of this thread is to offer you a tool to figure out how much the format issue means to each one of you. Some people might just want to listen and some people might want to analyse them. I am ok with whatever you need to do to get to your own conclusions.

 

The reason why I am offering different file formats on my web page has nothing to do with the outcome of this thread. By offering these file formats I am not trying to make a statement on the SQ of the different formats. I am just offering these different file formats for convenience. With your donation you can download any format you want. You could download just one format or you could download all 5 formats. It does not make any difference to me. I do not have a commercial interest on this file format issue.

 

Now, I know others do have a commercial interest, and I know it is hard to get good material for comparison. As an end user you can never be sure weather the different formats got different recording (mastering) chains or not, so I am offering these ones that I guarantee have the exact same chain except for the conversion itself.

 

So please, do use them for whatever purpose you see fit.

 

Enjoy !!

[PopPop]

My first listening session:

 

I started with a comparison of the two extremes; the MP3 and the 24/96 samples. I was quite surprised that I actually had to listen closely to detect any difference and it wasn't apparent on the first go around. I gradually perceived a smoother, less harsh tone in the upper frequencies but nothing I could put my finger on.

 

Then I had my wife select either of those to files at random and was able to identify which file was playing 16 out of 20 times.

 

I was never able to distinguish any difference between the 24/96 file and any of the other high res versions. I will give that a go tomorrow.

 

I then went on-line and took a number of hearing frequency tests from different sources. I set my headphone amp at -25db which in my system is a normal listening range. I selected test tones at the same level. The results were pretty consistent - hearing fell off at 12,500khz.

[Sal1950]

My first listening session:

 

I started with a comparison of the two extremes; the MP3 and the 24/96 samples. I was quite surprised that I actually had to listen closely to detect any difference and it wasn't apparent on the first go around. I gradually perceived a smoother, less harsh tone in the upper frequencies but nothing I could put my finger on.

 

Then I had my wife select either of those to files at random and was able to identify which file was playing 16 out of 20 times.

 

I was never able to distinguish any difference between the 24/96 file and any of the other high res versions. I will give that a go tomorrow.

 

That's been both my personal experience and those of the majority of tests I've read on many sites.

Everything being equal, the differences between 16/44 and 24/96 are extremely subtle and require the top notch equipment and hearing to pick out.

[kumakuma]

That's been both my personal experience and those of the majority of tests I've read on many sites.

Everything being equal, the differences between 16/44 and 24/96 are extremely subtle and require the top notch equipment and hearing to pick out.

 

This has been my experience as well. If there is a night and day difference between the versions then something else is different such as the mastering.

[esldude]

My first listening session:

 

I started with a comparison of the two extremes; the MP3 and the 24/96 samples. I was quite surprised that I actually had to listen closely to detect any difference and it wasn't apparent on the first go around. I gradually perceived a smoother, less harsh tone in the upper frequencies but nothing I could put my finger on.

 

Then I had my wife select either of those to files at random and was able to identify which file was playing 16 out of 20 times.

 

I was never able to distinguish any difference between the 24/96 file and any of the other high res versions. I will give that a go tomorrow.

 

I then went on-line and took a number of hearing frequency tests from different sources. I set my headphone amp at -25db which in my system is a normal listening range. I selected test tones at the same level. The results were pretty consistent - hearing fell off at 12,500khz.

 

Good work.

 

Something people who don't agree with blind testing don't experience is something similar to what you describe.

 

You will often be able to correctly detect a difference blind even when it has gotten small enough you don't think you are hearing a difference or certainly when you are feeling the difference is iffy or fleeting. Blind testing of some things can be much more discerning than is believed by those who think it is worthless.

[PopPop]

So, here is a question. How would those results have differed if (if at all) if my hearing range was closer to, say, 15khz? Is there any correlation?

[sandyk]

So, here is a question. How would those results have differed if (if at all) if my hearing range was closer to, say, 15khz? Is there any correlation?

 

I can hear clear differences with format comparison tests such as those from SoundKeeper Records and those posted by Jon P a while back, although my hearing is WAY down, but my DIY gear is WELL above average.

[kumakuma]

So, here is a question. How would those results have differed if (if at all) if my hearing range was closer to, say, 15khz? Is there any correlation?

 

This is a very good question. I want to say "no" because 15kHz is still below the 22.1kHz level supported by 16/44.2 but this answer seems too simple.

[kumakuma]

I can hear clear differences with format comparison tests such as those from SoundKeeper Records and those posted by Jon P a while back, although my hearing is WAY down, but my DIY gear is WELL above average.

 

Did you hear the differences of the same magnitude with Mario's files?

[esldude]

So, here is a question. How would those results have differed if (if at all) if my hearing range was closer to, say, 15khz? Is there any correlation?

 

I couldn't answer that definitively. My guess is the differences would be more obvious. Such things can be tricky however. As your hearing acuity goes with age, the filters in your cochlea will get less sharp which can change the masking profile. Some types of issues which might include those from MP3 processing may become a bit more obvious as a result. Your ability to hear into noise would be lessened, but your odds of being bothered by added noise of some types might be heightened.

[PopPop]

I can hear clear differences with format comparison tests such as those from SoundKeeper Records and those posted by Jon P a while back, although my hearing is WAY down, but my DIY gear is WELL above average.

 

That is something I have to keep in mind. I think I have quality gear but not real on the far end of the scale by any means. I did expect to hear more of a difference between the MP3 and the hi-res though. I am happy with the gear I have and this simple test answered some important questions for me.

 

This is a very good question. I want to "no" because 15kHz is still below the 22.1kHz level supported by 16/44.2 but this answer seems too simple.

 

OK, thanks. I could extend the frequency range by about about 2khz mby increasing the volume in the test tone but that would have been too loud for music listening.

[sandyk]

Did you hear the differences of the same magnitude with Mario's files?

 

I haven't tried with Mario's test files as the music isn't the type I would normally use for this purpose.

However, I did listen to "Iberia" in 24/96 and it was a superb recording.

[kumakuma]

I haven't tried with Mario's test files as the music isn't the type I would normally use for this purpose.

However, I did listen to "Iberia" in 24/96 and it was a superb recording.

 

That's too bad. It would have been interesting to compare your results to others in this thread.

[徐中銳]

I did expect to hear more of a difference between the MP3 and the hi-res [flac] though.

 

Thus, Protools has been unfairly maligned by many because almost identical performance here regardless of which of these files from Mario.

 

So, Mario, what settings and processes within Protools did you use ?

[Mario Martinez]

Thus' date=' [b']Protools[/b] has been unfairly maligned by many because almost identical performance here regardless of which of these files from Mario.

 

So, Mario, what settings and processes within Protools did you use ?

 

 

Yes, that is what it looks like

 

This music was originally recorded at 24/96, so the 24/96 version has no format conversion.

 

The 24/48, 24/44, 16/96, 16/48 , 16/44 are the ones that had to go through the conversion. The better the software the less of a difference you will hear from the different file formats. If you cannot distinguish them from the original is because protools is doing a good job.

 

The MP3 was encoded using LAME. Again, If it sounds so similar to the original (given that it is mp3) is because LAME is also doing a good job.

 

Protools offers four different setting for conversion quality. The highest one being "Tweak Head". That is what I used.

[Jud]

Yes, that is what it looks like

 

This music was originally recorded at 24/96, so the 24/96 version has no format conversion.

 

Hi Mario, I want to very gently disagree with you here and say that I believe the 24/96 version would have no additional format conversion once it is output from the ADC. But ADCs are like DACs - they nearly always do format conversion internally. Unless you were using one of the (rare to begin with and increasingly so these days) Pacific Microsonics ADCs or something like that, the ADC you recorded with would first take the analog and run it through delta-sigma modulation, producing a DSD-like bitstream; from there this bitstream would be decimated internally to the desired output rate, often 24/96 as you have done.

 

Recordings intended for CD will be further decimated to 16/44.1. A very few studios will take the bitstream from the delta-sigma modulator without internal conversion to PCM, then convert a portion or the entire file externally to PCM as needed for editing; and a very few others will decimate only to 352.8KHz PCM ("DXD") or 192KHz.

 

The 24/48, 24/44, 16/96, 16/48 , 16/44 are the ones that had to go through the conversion. The better the software the less of a difference you will hear from the different file formats. If you cannot distinguish them from the original is because protools is doing a good job.

 

Protools offers four different setting for conversion quality. The highest one being "Tweak Head". That is what I used.

 

From SRC Comparisons, it looks like Protools "Tweak Head" filtering (at least with the settings tested) is linear phase, doesn't introduce much ringing, but has some mild aliasing.

 

Edit: Sorry, I forgot above to include the very few studios like Blue Coast, which edit in analog rather than converting to PCM.

[Jamesroy]

I would like to offer some test files so you can compare the sound of the different file formats and report your conclusions. See if we can find out how much the file format really affects the final experience.

 

For the purpose of this thread, I have made one track containing about 16 minutes of music. Piano solo, soprano and piano, tenor and piano and flamenco voice and guitar.

 

All this music has been recorded exactly the same way. Our physical setup (hall-stage-mics) is a fixed setup, and our recording chain is also a fixed setup from microphones to master. So it would be like listening to one long song with different instruments in it.

 

We do not do any mixing or mastering. Our masters are simply the raw sound of the take. Left mic is left speaker and right mic is right speaker. There is no dynamic range compression. What you are hearing is the full dynamic range of the performance. And we record everything with the same gain, so once you find a realistic sound level on your equipment you should be able to listen through all the excerpts without the need to adjust the volume. This way you will be getting a sense of the real levels of the performance.

 

I have uploaded 1 zip file to the server containing this one track in 7 different formats:

flac 24bit 96kHz

flac 24bit 48kHz

flac 24bit 44kHz

flac 16bit 96kHz

flac 16bit 48kHz

flac 16bit 44kHz

lame mp3 320kbps 48kHz

 

These test files have been directly processed from the 24bit 96kHz master on the protools session. The volume is exactly the same for all of them. If you want to download just say so and I will send you a PM with the code and the link to the files.

 

P.S. I would rather not give an opinion on this. I would not want to influence anybody's perception in any way. I am just happy to be able to provide the test material and I even more happy to find out what you all think.

Yes I would like to try the test. [email protected]

[Sal1950]

Jamesroy, I would edit your email out of that response. Ebots comb forum for emails to add to spam lists. Mario will PM you the file location and password.

[Mario Martinez]

Hi Mario, I want to very gently disagree with you here and say that I believe the 24/96 version would have no additional format conversion once it is output from the ADC.

 

You are right, maybe I should have specified.

 

The ADC conversions occur at the beginning of the chain before the signal is recorded by the protools session. These ADC conversions affect the 24/96 Master which is the starting point to create our Test Files. I was referring to the conversions needed to get the different Test Files out of the 24/96 Master.

 

The 24/96 Test File does not have any format conversion coming out of the protools. The other Test Files were converted using the protools SRC on Tweak Head mode.

[Jud]

You are right, maybe I should have specified.

 

 

It's OK if you don't, I am always around to do it.

 

My apologies for being a little pedantic.

[Jamesroy]

Is AuI-ConverteR-48x44 the same as dBpower amp?

[Jud]

Is AuI-ConverteR-48x44 the same as dBpower amp?

 

Nope, completely different.

[Jamesroy]

Thanks another confirmation of being a newB.