EDIT: I just want to be very clear on this point: Digital copies of recordings are potentially forever. Not vinyl, not CD, not anything physical that we actually make. I do not want anybody's record to be a crap quality MP3 100 years from now. There is every reason to save the best quality copies we have. Now, on to what I wrote:
So, you say you can hear the flaws in 128kbps CBR MP3s? Hear a difference between FLAC and WAV? 320k CBR and 320 VBR HQ? Here is any easy way for anyone that has good (or good enough) sound sourced off a Windows computer to really find out what they can actually distinguish. Whether or not you feel like sharing your results, please give it a try for your own benefit.
If you do enough trials, you can use this method to definitively prove that you can tell the difference between two files! You cannot, of course, prove the opposite.*
I have tried to make this excessively detailed, but please let me know if anyone gets stuck on a step and I'll revise the instructions. Note that most of these steps are one-time efforts.
1. Install the foobar2000 audio player, and examine the settings to your satisfaction: http://www.foobar2000.org/
1.5 Also, install the ABX comparator http://www.foobar2000.org/components/view/foo_abx. I think the ways to do this vary depending on your version of windows, not sure, but there are many. you can right-click the file and choose "install", you can go into the foobar menus and do it there, you can do it by running the foobar installation and picking it then, etc.
2. Install Audacity: http://audacity.sourceforge.net/
3. Install Lame encoder: http://audacity.sourceforge.net/help/faq_i18n?s=install&i=lame-mp3
4. Start up Audacity, and import (you can just drag it to the window) a lossless track that you want to use.
5. Go to file>export and select a format (MP3/etc) that you want to test yourself with. Click options in the lower-right to set specific parameters for the format (bitrate, etc.). Save the file somewhere convenient, and repeat this step for any other formats/options you want to try.
6. Start up foobar, and drag the files you have just made into the default playlist (or make a special one, doesn't matter). Select (highlight) two files to test yourself with by holding down ctrl and clicking each of them on the playlist. If you need to be able to see file names or bitrates or etc. on the playlist in order to choose your two files, right-click the column headings to add more information to the playlist. The playlist isn't used during the test, so don't worry about what you can see.
7. Now right-click either of the highlighted files, and go to utilities>ABX Two Tracks. Leave the replaygain and DSP options unchecked unless you already know you need them.
8. Now the test is on. A and B are your two tracks. X and Y are also your two tracks, but you don't know which corresponds to A and which is B. Most of the controls on this screen are just options for you to use comparing the tracks, which you can do however you like for as long as you like.
Nothing you do counts toward the test until you click a choice AND click Next Trial. Whether or not you check the box to hide your results during the test is sure to affect your behavior during the test, but it's up to you. You can get valid results either way. You need to do at least 10 trials, but you can keep going as long as you want. Remember to relax. It's your test, and you can always start it over.
When/while you show the results, you need to get a 5% or lower probability of guessing to demonstrate that you can tell the difference between A and B. Yes, really. No, really, I mean it. 5%.
*Interpreting Your Results:
1. If you passed a test with flying colors, this does mean you could hear a difference. This doesn't necessarily mean you heard a difference due to what you wanted to test. It could be something happening with the downsampling, changing DAC performance, etc., etc. If you want to really track down what you heard, you can play scientist and start figuring out how to eliminate more and more variables. It is passing your initial test that makes this worth investigating in the first place, though.
2. Relatedly, if you failed a test, it's always possible something prevented you from being able to tell the difference. It's also possible you could pass the exact same test a different day. People tend to lose interest in investigating after failing an ABX, but if you don't, nothing is stopping you from trying more things. If you've got a microphone, try using your own recordings! Keep using blind ABX as a deciding factor. It is the correct type of test for "difference vs. no difference" issues. If you don't trust this software, it's not too hard to figure out how to set up a blind ABX in most pro audio software. Just resist the temptation to skip statistical significance when you don't have a tool keeping track for you.
Have fun!
So, you say you can hear the flaws in 128kbps CBR MP3s? Hear a difference between FLAC and WAV? 320k CBR and 320 VBR HQ? Here is any easy way for anyone that has good (or good enough) sound sourced off a Windows computer to really find out what they can actually distinguish. Whether or not you feel like sharing your results, please give it a try for your own benefit.
If you do enough trials, you can use this method to definitively prove that you can tell the difference between two files! You cannot, of course, prove the opposite.*
I have tried to make this excessively detailed, but please let me know if anyone gets stuck on a step and I'll revise the instructions. Note that most of these steps are one-time efforts.
1. Install the foobar2000 audio player, and examine the settings to your satisfaction: http://www.foobar2000.org/
1.5 Also, install the ABX comparator http://www.foobar2000.org/components/view/foo_abx. I think the ways to do this vary depending on your version of windows, not sure, but there are many. you can right-click the file and choose "install", you can go into the foobar menus and do it there, you can do it by running the foobar installation and picking it then, etc.
2. Install Audacity: http://audacity.sourceforge.net/
3. Install Lame encoder: http://audacity.sourceforge.net/help/faq_i18n?s=install&i=lame-mp3
4. Start up Audacity, and import (you can just drag it to the window) a lossless track that you want to use.
5. Go to file>export and select a format (MP3/etc) that you want to test yourself with. Click options in the lower-right to set specific parameters for the format (bitrate, etc.). Save the file somewhere convenient, and repeat this step for any other formats/options you want to try.
6. Start up foobar, and drag the files you have just made into the default playlist (or make a special one, doesn't matter). Select (highlight) two files to test yourself with by holding down ctrl and clicking each of them on the playlist. If you need to be able to see file names or bitrates or etc. on the playlist in order to choose your two files, right-click the column headings to add more information to the playlist. The playlist isn't used during the test, so don't worry about what you can see.
7. Now right-click either of the highlighted files, and go to utilities>ABX Two Tracks. Leave the replaygain and DSP options unchecked unless you already know you need them.
8. Now the test is on. A and B are your two tracks. X and Y are also your two tracks, but you don't know which corresponds to A and which is B. Most of the controls on this screen are just options for you to use comparing the tracks, which you can do however you like for as long as you like.
Nothing you do counts toward the test until you click a choice AND click Next Trial. Whether or not you check the box to hide your results during the test is sure to affect your behavior during the test, but it's up to you. You can get valid results either way. You need to do at least 10 trials, but you can keep going as long as you want. Remember to relax. It's your test, and you can always start it over.
When/while you show the results, you need to get a 5% or lower probability of guessing to demonstrate that you can tell the difference between A and B. Yes, really. No, really, I mean it. 5%.
*Interpreting Your Results:
1. If you passed a test with flying colors, this does mean you could hear a difference. This doesn't necessarily mean you heard a difference due to what you wanted to test. It could be something happening with the downsampling, changing DAC performance, etc., etc. If you want to really track down what you heard, you can play scientist and start figuring out how to eliminate more and more variables. It is passing your initial test that makes this worth investigating in the first place, though.
2. Relatedly, if you failed a test, it's always possible something prevented you from being able to tell the difference. It's also possible you could pass the exact same test a different day. People tend to lose interest in investigating after failing an ABX, but if you don't, nothing is stopping you from trying more things. If you've got a microphone, try using your own recordings! Keep using blind ABX as a deciding factor. It is the correct type of test for "difference vs. no difference" issues. If you don't trust this software, it's not too hard to figure out how to set up a blind ABX in most pro audio software. Just resist the temptation to skip statistical significance when you don't have a tool keeping track for you.
Have fun!
Attachments
Last edited:
). Fixed now. Happy new year's eve, probably about time to get partyin'.
