INTERACT FORUM

Please login or register.

Login with username, password and session length
Advanced search  
Pages: [1]   Go Down

Author Topic: Instead of a subjective listening test  (Read 5882 times)

sorepinky

  • Guest
Instead of a subjective listening test
« on: February 02, 2017, 03:13:29 pm »

Instead of a subjective listening test, why not do something really interesting like taking the left channel of any high-res track like a 192/24 FLAC and replace the right channel with the left channel (i.e. make it "mono"), down-convert only the right channel to a lower res format like 48/16 and invert that. Then sum left and right channels to find any difference then replace say the right channel with the difference, discard the left channel and look for the right channel on the DSP Studio Analyzer to see if it is empty?  Could even change the scale of the Analyzer to make it about 1dB full scale, or amplify the difference to something that will fit the standard scale.

Assuming it's possible, perhaps similar could be done to compare SoX resampling.

Addendum:  Maybe I'm being naive in supposing that this is even possible (running different sample rates and bit depths at once).  Perhapes a way around such a limitation would be to up-convert the down-converted channel back to the original format before inverting one of them and summing them.  At least then, there would be no argument that the file was not at some point of lower resolution.
Logged

Alobar

  • Junior Woodchuck
  • **
  • Posts: 65
Re: Instead of a subjective listening test
« Reply #1 on: February 04, 2017, 01:03:08 pm »

With storage becoming so cheap I am personally not using any new mp3 at all but still use mostly 16/44.1 ape. Some of my music I have downloaded from HDtracks at 24/96 but I am not sure if that is wasted money over a CD purchased from Amazon or somewhere else, often for much less cost. Very interested in V.2 of MC's listening test that can compare 16bit/44.1kbps to 24bit 96kbps or 192 for that matter. My suspicions are that the grand old CD is pretty close to the top of the bell curve and it would take extraordinary ears and a great system to gain much SQ simply from higher resolutions, but a test would be quite telling. One thing about HDtracks V a CD of the same music is I can never really know if it is from the same master or not so that is where MC's listening test would tell all about resolution, by insuring the comparison IS from the same master. Then of course the questions still arise if the HDtracks hi rez downloads are from better masters or not and whether to buy that over the CD. Can't imagine a test for that!
Keep up the good work JRiver!
Logged

blgentry

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 8009
Re: Instead of a subjective listening test
« Reply #2 on: February 04, 2017, 05:04:47 pm »

In my very, very limited testing with friends, all of them have been able to hear differences between 24/192, 24/96 and 16/44 .  My testing was done with a tiny number of people, in a non-scientific way.  It was all done with the Pono player and it's "Pono Revealer", which allows you to switch, on the fly, between resolutions as you choose.  JRiver was involved in this effort (Pono Revealer).

Brian.
Logged

sorepinky

  • Guest
Re: Instead of a subjective listening test
« Reply #3 on: February 06, 2017, 01:35:57 am »

I've sat in hi-fi rooms with small groups of people and as soon as one of them says that a certain piece of optical cable (or whatever other more expensive thing was just swapped in) provides more "air", soon the rest tend to agree.  I just do this: ::)

Better than such "power of suggestion" sittings and even better than double blind tests IMO is the idea of inverting and summing to find any real difference and play the difference.

Here is someone I know who has built a circuit to invert and sum to find a difference between cheap and expensive interconnects.  Play his test tracks through to see what I mean: http://www.halfgaar.net/interlink-comparison

I played them with my PC speakers turned up to 100% and hear nothing for the end portion of the ones excluding the "sabotaged" cable.  Haven't tried it with decent headphones yet.  But it suggests quite strongly (even with the limitations set out on his page) that these wild differences that people claim to hear may in reality be nothing much.
Logged

tyler69

  • Citizen of the Universe
  • *****
  • Posts: 946
Re: Instead of a subjective listening test
« Reply #4 on: February 06, 2017, 02:53:10 am »

I've sat in hi-fi rooms with small groups of people and as soon as one of them says that a certain piece of optical cable (or whatever other more expensive thing was just swapped in) provides more "air", soon the rest tend to agree.  I just do this: ::)

Well I think that it's not impossible to hear a difference when a component is swapped.
But on topic: would it be possible to add functionality to this feature (or implement it as a new one with a different Name) in Order to compare two manual selected tracks from the library? If so, Media Center should try to allign those tracks.
Logged

sorepinky

  • Guest
Re: Instead of a subjective listening test
« Reply #5 on: February 07, 2017, 03:26:02 am »

OK I realise this is a diversion, so Jim please feel free to shift these posts elsewhere, but as an experiment I have done this in DSP Studio:



The result is inaudibility of any sound whatsoever over the 35 or so dB background noise in my study from both of my speakers when turned up to 100% and playing any stereo track and a blank screen in the "Analyzer"! This is using my basic PC speakers, but it establishes a baseline.

So as an alternative to the subjective "listening test" (where the same track with different formats is compared at different times) why not make another routine called say "Play the difference" (or "What's over Red Book?" or something else) in which the user could select any high-res track and have a routine do this:

(it's intended to compare just the left channel of a high res track with itself after being down-converted then up-converted to where it started, but to play back the difference in both channels)

Make a temp file copy of the hi-res track - save that as Temp File A
Downsample Temp File A to 44.1/16 - save that as temp File B
Upsample Temp file B back to the original format and save that as Temp File C
Delete Temp File B
Use Temp files A and C to make a track with the left channel of Temp file A as the left channel and the left channel of Temp File C as the right channel - save that as Temp File D
Delete Temp Files A and C
Play Temp File D doing the following:
Reverse polarity of right channel
Add left to right
Copy right to left (so that any difference will be heard through both speakers)

Imagine the blank looks on people's faces when listening to nothing and seeing nothing in Analyzer, or the smiles on their faces when they can actually hear and see a difference!
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10661
Re: Instead of a subjective listening test
« Reply #6 on: February 07, 2017, 03:45:23 am »

We already know that there are mathematical differences between various formats, thats not the question and easily proven, and listening to such differences would be pointless, since its nothing but noise if you take out the common signal.
If it would be so easy to debunk or prove high-res audio or whatever, people would've done it years ago. :)

The real question is if there are actual audible differences people can make out between various formats under real listening conditions, or if those miniscule differences between these formats are imperceptible?
Logged
~ nevcairiel
~ Author of LAV Filters

sorepinky

  • Guest
Re: Instead of a subjective listening test
« Reply #7 on: February 07, 2017, 04:38:53 am »

My proposal will play the actual audible difference if one exists.  Simple as that.  If a null audible result is confronting to some people, that would be a mere side effect.  If I'm sitting in front of speakers that are apparently playing nothing, the listening condiions (which are real) are exactly the same as if something is playing as usual.

Apart from removing the placebo effect and all the other problems associated with playing tracks at different times to enable the making of subjective assessments, my proposal automatically removes any imagined differences.  It also removes the need for double-blind test conditions.

The listening test as it stands does neither of these.

By the way, your comment that the difference is just noise may well hold true for the listening test as it stands anyway.
Logged

JimH

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 71212
  • where the buffalo roam
Re: Instead of a subjective listening test
« Reply #8 on: February 07, 2017, 05:39:20 am »

I moved these posts from the Listening Test thread since they were off topic.

Maybe you could try the Listening Test and post your results in that thread.  As Hendrik said ...

The real question is if there are actual audible differences people can make out between various formats under real listening conditions, or if those miniscule differences between these formats are imperceptible?
Logged

sorepinky

  • Guest
Re: Instead of a subjective listening test
« Reply #9 on: February 07, 2017, 03:40:25 pm »

I already did, and whilst Hendrik may well be referring to a "real" question, so am I.  I'm just taking the subjectiveness out of it.

At the moment I am trying to find a way to do the following step myself:

"Use Temp files A and C to make a track with the left channel of Temp file A as the left channel and the left channel of Temp File C as the right channel - save that as Temp File D"

Can someone help me or guide me to free software that will enable that step to be done, or can someone offer to do it for me if I send them files A and C as 176.4/24 WAVs.  I can create File A using dbPoweramp or Audiogate from DSD (yes I still have an old version which allows that).  I can create File C (down-converting and up converting FIle A) using dbPoweramp.

The only problem that I can see in my approach is that if the down-conversion and up-conversion introduced any kind of phase shift (which would be inaudible on playback of file C just as if using the "listening test"), they could affect the summing step to produce an audible difference that could not be attributed to the format difference per-se.
Logged

JimH

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 71212
  • where the buffalo roam
Re: Instead of a subjective listening test
« Reply #10 on: February 07, 2017, 04:05:06 pm »

Please post your Listening Test results.
Logged

tyler69

  • Citizen of the Universe
  • *****
  • Posts: 946
Re: Instead of a subjective listening test
« Reply #11 on: April 06, 2017, 04:38:26 am »

In this https://yabb.jriver.com/interact/index.php/topic,107924.0.html thread please.
This thread should be about other "listening test" functionalities than the ones offered in my opinion. I still hope to see some more functionality sometime..
Logged

Dmaumau

  • Member
  • *
  • Posts: 1
Re: Instead of a subjective listening test
« Reply #12 on: February 11, 2019, 11:52:06 am »

OK... So letīs talk about listening tests...
Logged

JimH

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 71212
  • where the buffalo roam
Re: Instead of a subjective listening test
« Reply #13 on: February 11, 2019, 02:04:21 pm »

Pretty good!
Logged
Pages: [1]   Go Up