First, good morning to all
. I hope we keep in mind that we are discussing a hobby and not any matters of life and death. So let's keep emotions out of these posts and just talk about the technical topic.
As to your point, your interpretation of what I post is not correct. Both theory and reason I post what I post is not understood.
A lossy audio compressor is constantly analyzing segments of audio (called frames) looking to see how much it can compress it while staying within its bit budget (320 kbps/sec in this instant). As such, the level of distortion varies constantly in sub-second intervals.
Let's look at an extreme but easier to understand example of 5 seconds of complete and total silence (digital "zero"). The entropy coder (back end of the codec) can compress this down to nothing. It wouldn't even need 64 kbps let alone 320 kbps. Ditto for a pure tone. Simple redundancy of data allows it to crunch such content with ease.
At the other extreme but difficult concept to understand are high frequency transients. There is no redundancy per se in a signal that all of a sudden sharply shoots up for a few milliseconds and then goes back down. The lossy encoder's frame now expands before and after the transients most likely. As it attempts to requantize (roughly speaking truncating frequency bands), it will spread that distortion to before and after the transient. What is after the transient will be masked usually. But not nearly as much for what comes before the transient. We call this type of distortion "pre-echo," i.e. echo that happens before the signal itself. As you can imagine, this type of distortion can be quite annoying depending on the amount of it.
What does this mean in this context? There is no such thing as hitting play and instantly being able to recognize the compressed version. Despite my training, there distortions that are inaudible to me as much as it is to you.
What I have shown in the test results is the full process from me starting the test to finding a critical segment that sounded different per above description. I am being transparent here showing you the failures until I found the difference. I could have of course restarted the test when I found the difference and show you perfect scores. But I thought there is education in learning how this type of testing works.
As to why occasionally I fail to tell the difference like that sample in the middle, again, remember this is not life and death discussions
. I am listening to these files on my day to day laptop I am using for these posts. I am sitting in our living room while the TV is on and someone else watching it. We also have two dogs that have had a lot of practice barking
. So it is not the ideal situation for critical listening.
Also, the way I do these tests is I play A and then B until I can classify in my mind which one sounded which way. As I run the tests, occasionally I confuse A for B because I don't play them again. I play X and Y and based on what I hear, I vote. The smaller the differences, the more I can make mistakes remembering what the difference was through so many trials.
Remember that the standard in such tests is 0.5% probability of chance. That is, we accept some amount of mistakes such as above. Outcomes with 0.5% error are considered proper and "scientific." So there is no requirement for perfect scores.
My results were: Total: 22/26 (0.0%). As you see, I achieved 0% probability of chance. Per above this exceeds the standard. What this means is that there should be no doubt that I found differences and could reliably tell the files apart. To call my results "rediculous" makes absolutely no sense. These are superlative results for double blind tests.
As I have said before, we are not at all used to seeing positive results in double blind test. As such our ability to analyze them is quite low as this discussions shows. Let's put aside emotions and allow the information to add to our audio knowledge and advance our learning. It takes a lot of work to keep answering these challenges (above was Krabapple's claim that I could not tell 320 kbps apart). This is rare kind of data/insight that we should celebrate having rather than showing such angst against it.