Accepting Pluto's Challenge.

[STHLS5]

If you can get these two SACDs sent to me I am willing to wager that I can convert them to 16/44 and you will find such conversions indistinguishable from the originals - and that includes the conversion from DSD to a PCM-based format!

…If I can convert your DSD SACD to 16/44 and you cannot distinguish the two, my point is proven..

You are suggesting that no matter what the Hi Rez is it will sound identical to the 16/44 down sampled copy of it. Therefore, I should be hearing the same sound from a true SACD player and a CD player.

Let’s take one step at a time. You can download the free version of either the 16/44 (Copy A) and/or 24/96* (Copy B) from BlueCoast (quick registration required). I would suggest Keith Greeninger & Dayan Kai’s Looking for a Home from the CAS 2010 downloads under the free download header. Let’s analyze this two formats to see if you can find any manipulation to make one to sound better than the other.

Thereafter, let’s make one copy of 24/96 in 16/44 format (Copy C). According to you, Copy B and Copy C should sound identical. However, my player is limited to 16/44 (no upsampling or other fancy things) and therefore for a valid comparison I suggest that you redo Copy A to another copy of 16/44 using the same method that you used to create 16/44 (Copy C) version of 24/96. Let’s call it Copy D.

Once again:

Original downloaded 16/44 is Copy A
Original downloaded 24/96 is Copy B
Original 24/96 to 16/44 downsample made on your PC is Copy C
Original 16/44 to 16/44 downsample made on your PC is Copy D

According to you: Copy C and Copy D should sound identical. Let’s see if we (most likely it is going to be myself) can tell any difference. Though I prefer a much higher resolution but let's stick to what we have now.

ST

{* This is a huge download: 199MB}

[A.S.]

This sounds like an interesting challenge.

I'd like to make one important point though before we get stuck into a full analysis. This forum is that of the Harbeth speaker company, and whilst we here are highly motivated to find real, pragmatic and affordable ways to improve the fidelity of our systems, we must take care to remain objective about third-party products and services. So, if we are going to proceed to analyse in fine technical detail the offerings of another company, can we take care to present our findings in a way that leaves final interpretation up to our readers. I'd suggest that, in the interests of fairness, we individually download and analyse, and present our findings after a few days of deliberation.

The first step, following your suggestion, is to register and download your recommended track. Bearing in mind the recent fiasco surrounding hi-def CD recordings which were apparently nothing more than standard recordings with a different printed label, may I suggest that a sensible starting point is to evaluate the high frequency bandwidth extension. 44k sampling sharply cuts-off at 22kHz (half sampling frequency) so if a recording truly is 96k sampling, assuming that there is energy in the performer's instruments beyond human hearing, there should be no evidence of cut-off at 22kHz - the limit should be half of 96k = 48kHz. That's twice the audio band width of standard CD.

If that cannot be proven, then I suggest no more effort is expended.

Here is a free audio editor/analyser that can be used to make a frequency analysis of the audio. Be sure that the Project Sampling Rate (bottom left corner) matches the claimed sample rate of the wav file before analysing. It supports 96k sampling.

[STHLS5]

After starting this thread, I realize that my player is limited to 16bit 44.Khz. Logically, I shouldn't be hearing any difference with the 24bit/96 KHz recording!

OK, now about the Audacity and Spectro I find the readings to be confusing. Since Audacity cannot read WMA files I have converted all the original files to WAV using Switch Sound File Converter. In Audacity graph the frequencies end somewhere near 25kHz but in Spectro it is far beyond that.

Edit [Attachments removed.]

ST

[Pluto]

Original downloaded 16/44 is Copy A
Original downloaded 24/96 is Copy B
Original 24/96 to 16/44 downsample made on your PC is Copy C
Original 16/44 to 16/44 downsample made on your PC is Copy D

According to you: Copy C and Copy D should sound identical. Let’s see if we (most likely it is going to be myself) can tell any difference.

Right - I've downloaded A and B, and I've made C. I don't understand what you want me to do for D - what do you want me to change? I can confirm that B has a surprisingly large amount of content above 20kHz but I'm confident that my C will be indistinguishable from B by listening alone. Spectral examination will, of course, reveal the truth. The provider's A has a clearly visible drastic anti-aliasing low pass filter; my C does not exhibit such.

Where do we go from here?

Of course, all this has nothing to do with DSD but, as I've said before, I'm confident that DSD per se is indistinguishable from PCM at sensible sampling rates.

[STHLS5]

Right - I've downloaded A and B, and I've made C. I don't understand what you want me to do for D - what do you want me to change? I can confirm that B has a surprisingly large amount of content above 20kHz but I'm confident that my C will be indistinguishable from B by listening alone. Spectral examination will, of course, reveal the truth. The provider's A has a clearly visible drastic anti-aliasing low pass filter; my C does not exhibit such.

Where do we go from here?

Thanks. I want A to go through the same process of conversion that B has undergone. I want to see if the orginal 16/44 file and your 16/44 file file sounds the same. BTW, how are we going to get the file uploaded? I have MYBOOK LIVE where you can upload the file but I need some time to set it up. Ideally, it should be uploaded here but I think the file is too big. Not sure if we can upload it to Window Live.

ST

p.s. [I don't really understand how these files work!. I just downloaded to my laptop (W7 64bit) and after extracting the file I get WAVE files but earlier with my old desktop XP I got WMA files!?]

[A.S.]

... p.s. I don't really understand how these files work! I just downloaded to my laptop (W7 64bit) and after extracting the file I get WAVE files but earlier with my old desktop XP I got WMA files!?

WMA is a compressed file format. That will introduce another unwelcome variable.

Incidentally, I had some difficulty opening the 96k file in Audition. Either it or I was trying to be too clever and it couldn't or wouldn't accept the file structure. After three attampts at recognising it, it automatically opened correctly.

I'm tempted to plump for the Pluto Super Prize and suggest that neither the 44k version nor the 96k version can be distinguished by human ears from a 256k MP3 file made from the 96k clip (or the 44k clip if you prefer).

Pluto commented ....

I can confirm that B has a surprisingly large amount of content above 20kHz

I agree. But as I mentioned a few posts back, if your hearing cuts off well below the CD top-cut at 22kHz (I doubt that there is a human anywhere on earth aged 50 who can hear over about 16kHz at normal levels) can you illuminate the advantage of a frequency response that extends to 25k, 30k or even 50k when the ear cannot detect those frequencies even if the speaker could generate them (which it can't and shouldn't).

[Pluto]

I suggest that all files transmitted in the course of this experiment be sent in 7zip format. This is a free, multi-platform compressor that provides a useful amount of data compression but, more important, it wraps the file so that any corruption on the journey will become readily apparent. In order to prevent any other software from interfering with the contents (as the 7z format is open-source), I suggest that all packages be assembled with the password Harbeth - upper-case H, lower-case arbeth.

The files are likely to be too large for normal private e-mail, so we need to find somewhere to host these files short term (the duration of the experiment). It needs to be somewhere discreet because the material is not our property so therefore must be treated with respect.

[STHLS5]

I'm tempted to plump for the Pluto Super Prize and suggest that neither the 44k version nor the 96k version can be distinguished by human ears from a 256k MP3 file made from the 96k clip (or the 44k clip if you prefer).

).

What a coincidence! I burned the 16/44 and 24/96 and another wav and a 256K MP3 of a Steinway recording downloaded from here to make an audio CD using WMP. Played them random and tried to guess which track was playing. Out of 25 or so guesses the result was just about 50% correct guesses.

I must be doing something wrong.

ST

[A.S.]

... Out of 25 or so guesses the result was just about 50% correct guesses. .... I must be doing something wrong. ST

I'm just quietly wondering to myself if you have learned more about human hearing acuity and psychology than about grading audio recordings.

Far from doing something wrong, wearisome as it is to repeat this mantra yet again, have you considered that maybe you're doing something right?! The ear is an unbelivevably poor scientific tool because it is hard wired to a heap of water and emotions. Instruments should not be part of us: they should be outwith our own bodies, cold and objective.

[Pluto]

The ear is an unbelievably poor scientific tool because it is hard wired to a heap of water and emotions.

You know, that is nearly as depressing a point of view as the possibility, unlikely though it may seem, that humanity is alone in this vast universe.

[EricW]

You know, that is nearly as depressing a point of view as the possibility, unlikely though it may seem, that humanity is alone in this vast universe.

And why just pick on the ear? Presumably one could add the eyes, the tongue, the nose, the skin ... they have their abilities, but are fatally limited by the sad heap of water and emotion they're attached to. Maybe there's nothing we can do that a scientific measuring instrument can't do better.

[STHLS5]

I'm just quietly wondering to myself if you have learned more about human hearing acuity and psychology than about grading audio recordings.

Far from doing something wrong, wearisome as it is to repeat this mantra yet again, have you considered that maybe you're doing something right?! The ear is an unbelivevably poor scientific tool because it is hard wired to a heap of water and emotions. Instruments should not be part of us: they should be outwith our own bodies, cold and objective.

After decades of chasing high fidelity, it is difficult for one to be turned over with one or two uncontrolled self tests. I need to be convinced.

Back to the topic. Today I ripped the audio CD tracks form the CD-R to identify the tracks.

Something did go wrong. It is now obvious that when I burned the the files to the CD-R using WMP it did not write the full resolution. Please see the images below:-


You can see HF content above 20kHz in the original 2496 files.

Attachment 1906



Now, I presume track 1 is the original 1644 files that was burned to the CD using WMP in audio cd format which was ripped again using WMP. You will notice that the HF contents missing significantly.



Attachment 1907


This is track2 extracted form the CD and I believe it is the original 2496 file. (I didn't mark the files when I burned them to avoid being influenced with HiRez label attached to it.)

Attachment 1908


I think we need a better way making the audio-CD to preserve the original files integrity.


To Pluto,

We cant use Skydrive of Window live because it limits the size to 100MB per file. Zip file is okay with me since the original files came zipped but we are looking at 300 to 400MB storage.

The second issue is how and what software to use so that the actual files get transferred to the CD-R without any modification. Obviously, WMP is altering the files. Any IT experts here can help us?


ST

[Pluto]

For ST and anyone else interested, my 16/44 conversion of the 96kHz track we discussed is available to

download here

If I've done it right :)

[STHLS5]

I can download and use the password to open the folder but cannot extract. The message I am getting is

! C:\Documents and Settings\xxxx\Desktop\File C - Pluto conversion - 44.1kHz.7z: Unknown method in Keith & Dayan - File C - 44.1kHz.wav
! C:\Documents and Settings\xxxx\Desktop\File C - Pluto conversion - 44.1kHz.7z: operation failed

ST

[Pluto]

Try the download again - it works for me!

[STHLS5]

I can download and use the password to open the folder but cannot extract. The message I am getting is

! C:\Documents and Settings\xxxx\Desktop\File C - Pluto conversion - 44.1kHz.7z: Unknown method in Keith & Dayan - File C - 44.1kHz.wav
! C:\Documents and Settings\xxxx\Desktop\File C - Pluto conversion - 44.1kHz.7z: operation failed

ST

Thanks Pluto. It can be extracted using 7Zip but not Winrar. Ok will update you tonite but still need to find the best way to burn it to the CD-R.

ST

[Pluto]

still need to find the best way to burn it to the CD-R.

Why? Surely the best thing to do is play it using exactly the same method as you would use to play the 96kHz download - otherwise you are simply not comparing like with like which renders the whole exercise pointless.

For burning WAVs to CD, use the free ImgBurn

[STHLS5]

*I used the ITunes to play Pluto's and the original 24/96 using my simple Dell Inspiron Laptop connected to NDS-1 which was connected to my old 16/44 Theta Pro GenIII DAC. After spending about half an hour I accepted defeat. I couldn't tell any difference though I feel Copy C appears to be "faster" somewhere near 3' 30 but couldn't find a way to just listen to that segment instantaneously to do comparisons with my ITunes.

Then I used the Laptop's soundcard analogue output and connected to the PreAmp's analogue input. The soundcard was set to 24/196. I listened again and I picked up the slight difference in the first 30 seconds. Proceeded to create a playlist in ITunes consisting of the original 24/96 and Pluto's Copy C. I used shuffle to pick up the tracks randomly.*

The results:-
27Mar2012, 9 12pm .

1) Yes - correct answer.
2) No- correct answer.
3) No- correct answer.
4) Yes- correct answer.
5) Yes- correct answer.
6) No- correct answers.
7) No- correct answer.
8) No- correct answer.
9) Yes- correct answer.*
10) Yes- wrong answer.

Note: Yes means 24/96 file.*

I stopped after getting No 10 wrong and took a break. Unfortunately, I lost concentration and couldn't get it right after that. The easiest way to tell the difference is to stop listening for differences and let yourself enjoy the music and you will 'feel' the difference. Will try again and see if I can get it as close as the first results.

This is going to be controversial. ;)*

ST

[Pluto]

...Then I used the Laptop's soundcard analogue output and connected to the PreAmp's analogue input. The soundcard was set to 24/196...

Do you mean 96kHz or 192kHz?

Either way, you were asking the laptop's sound card to sample rate convert my file. Having created that file using SRC resources worth $$ four figures, do you really think that having a PC sound card re-convert amounts to a realistic test?

The only way of conducting this test fairly correctly is by playing the digital data from either file directly into a DAC equally capable of handling either sampling rate*. All you have actually proved is that the sample rate conversion in your laptop's sound card isn't transparent. I hardly think the analogue output of a laptop is a useful source when discussing cutting-edge sound quality.

* Actually, in the above scenario, to attempt to eliminate the possibility that the DAC performs differently at 44kHz and 96kHz, having completed the down-conversion from 96 to 44, I should then have re-converted the 44 back up to 96 so the playback DAC was always operating at 96k, but the file size would have been twice what it is already.

[EricW]

*I used the ITunes to play Pluto's and the original 24/96 using my simple Dell Inspiron Laptop connected to NDS-1 which was connected to my old 16/44 Theta Pro GenIII DAC. After spending about half an hour I accepted defeat. I couldn't tell any difference ...
ST

I'm confused. If your DAC is limited to 16/44.1 how could you possibly hear a difference between the 16/44.1 signal and the 24/96 signal? Everything's going to play back at 16/44.1, isn't it?