# Uncompressed Lossless (WAV) vs Compressed (FLAC / ALAC) - O/T discussion moved from main forum thread

Discussion in 'Sound Science' started by haiku, Apr 30, 2016.

1. Aw man, just do the test yourself.

We don't listen to numbers,statistics, bits or bytes or waves or waveforms. I even strain to listen to logic.

We listen to music. Don't forget the tape measure. Who would have thunk a tape measure was the missing test equipment in the wav/flac debate?.

Next time I hear music with a definite height element (and if I remember), I may do it myself.

2. That's exactly what we listen to.

3. Just because you got zeroes doesn't mean there is no difference.  What if you went to 7! decimal places......HAH!?

reginalb likes this.

4. A recent joke I ran across- NASA calculates navigation using pi out to 15 decimal places.

First engineer - Did you calculate with pi out 15 decimal place?
Second engineer - Yes, of course.
First engineer - Then how can we be off by hundreds of meters?
Second engineer-  meters?

5. Contributor

the space guys have has many true stories as they have jokes on the metric topic.

6. It is true though, not proving a difference is not the same as proving there is no difference.

7. I've done a null test and the wav and wav converted from flac was not totally the same which doesn't really surprise me knowing how good I'm at conducting tests. After adding 100dB of amplification to the signal the peak was at -19dB. After that I tried to null the exact same wav (dragging the same file from my desktop to Audacity and Diffmaker) and I got similar results.
I think there are three ways of looking at this: the software I used is not good enough (can't precisely invert or add amplification for example), I used the software in a wrong way (I messed up somehow and don't know how) or it is impossible to copy data in a way it's going to be exactly the same (lol). I suspect that I just don't know what I'm doing.

8. Contributor

I've had problems with Diffmaker in the past where two copies of the exact same file do not null correctly (ymmv) - I worked out that the alignment was not perfect .

With Audacity (for me on my system with my files) even two copies of the **same** 16/44.1 file with one inverted do not null exactly when combined -  though they do null to a worst case of -107db and an average of -120db. In fact they do null **very** slightly better than wav0 and wav5 inverted which null to an average of -119db and a worst case of -107db

My guess is that the software is just not good enough when combining 16/44.1 files - sadly I've lost my copy of CEP - sigh...

9. I have problems with Diffmaker at times in a number of ways.  Though when it works it is a nifty indeed.

If you guys are doing nulls on WAV/FLACs digitally then they should null to true and complete files of nothing. Sometimes the conversions throw an extra bit or three at the beginning.  Shifting for those should null everything out. Even with Audacity as your software.

10. Something is wrong here.  I have a hint of an idea.  If you export a file in Audacity, it may depending on your settings, apply dither going from one bit depth to another.  You then will get a residual of the dither which sounds like it might be the right level if you are using shaped or TPDF dither.  So if you are comparing the file to a copy you exported that could happen.  Just copy the file directly and change the name it should null to absolutely nothing.

11. I don't know why you guys are so obsessed with null testing. The research clearly states that even though the signals null, they still sound crazy different. And it's easily repeatable. And if you don't repeat it, then you're doing bad science!

12. The fact I can't null the exact same WAV files bothers me really, it hurts my ego.

13. If you mean time alignment it can't be a problem in this case since I've compared the exact same files so they align perfectly I guess. Turning on or off time alignment has no effect on the results.
Also I've figured out if I use difference track signal boost in Diffmaker no matter how much I boost it it's still silent. However if I amplify the difference in Audacity there's still a noticable noise after amplifying it enough. It seems the software matters more than I expected which really puts me off. Or maybe I still don't use it correctly.

14. Would you detail your procedure comparing two identical files that don't fully null?

Here is a null from two identical files that were waves.  BoB Dylan's Shelter from the Storm in this case.  There is nothing there.  Note the cursor value is like -2 billion db which is what Audacity shows for nothing.  No amount of amplifying turns up anything.  I imported one file, imported the same file though it had a different name and location.  Inverted the lower one.  Selected both, and used 'mix and render' and that is the result.

Again if you import something and then export it you may be applying dither to the exported file. You need to export with "other uncompressed files" selected and under options for that Wav (microsoft) 24 bit PCM should be selected.  Then you shouldn't have any problems.

Also, if you are comparing WAV to FLAC to WAV conversions, some software adds some few bits at the beginning.  Even if you line those up you might show a small residual for the whole file in the FFT.  You can avoid that by chopping off the spare bits at the beginning.

15. Sure. Using Diffmaker with the following settings:

Then loading the same wav to both the Reference track and Compared track and clicking on Extract. Diffmaker makes the difference track which no matter how much I amplify by using Diffmaker's playback boost it remains dead silent. However grabbing the wav which was made by Diffmaker from my folder to Audacity and then going to Effects>Amplify it doesn't give me the same result:
Oh by the way I had to check the gain alignment in Diffmaker otherwise it would crash...