I download songs in APE format and transcode into FLACs to be used on my rockboxed iPod. So today I was curious to see the difference between these two formats so I did an ABX test on foobar. And to my surprise I could hear a difference between the two!! 4/4 times is not just by guessing I guess. However from my understanding, transcoding from a lossless format to another should not cause quality degradation. To my ears the APEs sound cleaner, smoother than the FLACs, a much bigger difference between APE and LAME VBR because I always have a hard time distinguishing between the two.

FYI, I use foobar with FLAC 1.1.2 codec to transcode the songs.

Any idea why this is happening??

Option #1: you didn't hear anything different, you only think you did.

Option #2: you don't know how to do a proper ABX.

Option #3: you screwed up the transcoding.

There isn't really much else to say, the decoded soundfiles are bit-for-bit identical. There are no differences.

Answer to Option #1: I got 4/4 on the ABX test, I don't think I didn't hear no difference...
Answer to Option #2: I believe doing the ABX test on foobar is simple enough for anyone with even a little intelligence.
Answer to Option #3: How could I possibly screwed up the transcoding? Please explain.

Thanks though, I believe there is no difference theoratically, but I do hear a difference in reality, and this has been confirmed in the ABX test on foobar.

I guess you need to consider that statement. You know there's no difference but you hear a difference? 4 samples isn't much of a field for one thing; and it certainly doesn't confirm a difference.

I think you should do a more extensive test than just 4/4 to find out. If you really can hear a difference and can do 20/20 in an ABX test there are only two possibilities

1. There is a bug in the software you are using.
2. You are very very very lucky.

Saying you can hear a difference between two lossless encoders is similar to saying you can see a difference between two pictures, which have been compressed/uncompressed with zip and rar. It's lossless, there should not be any differences, unless there's a bug somewhere.

/Kef

make sure the files are bit-identical, foobar2000 has a bit comparer to do so.
If they are, then you're hearing things (and you need more than four tries for an ABX, try 10 or more), or either the APE decoder or FLAC decoder is screwed up.

If they aren't, there's either a bug, or something being applied to the audio.

Like others have said, it is nonsensical to say you can hear a difference between 2 lossless encodings.

And, 4/4 on foobar is nothing. Just playing around, I've gotten more than that to match just on pure luck. You must realize each time you have a 50% chance of getting it right. So, even for 2 indistinguishable samples, as your number to test cases goes to infinity, your success rate should go to 50%, not 0%. Therefore, it is my opinion that if you really can hear a difference, you should be getting at least 98 of 100 correct.

Any chance that replaygain or something similar is being applied to one format, but not the other, on playback?

-brendan

Hey, no replaygain or anything. I just did a binary comparison and result turned to be negative so could just be my ears (I doubt it.. o.O) or the decoder.

I've never understood how something can sound "cleaner" or "smoother".

Anyway, as has been said, you must be doing something far wrong when transcoding if MP3 is transparent to you while a lossless format is not! How are you performing the conversions?


Do you mean that foobar2000 indicated differences between the two "lossless" files?

Just do an extensive ABX test then.

/Kef

Yes, one could have the RG tag while the other doesn't, though I don't know if the ABX component actually uses RG tags when decoding.

For the binary comparison foobar has to use the ape and flac decoder. So these things are alright.
Just go into another abx session and use 10 trials or more.

derek8555 where are you able to download ape files from?

4/4 (4 correct out of 4 trials) has a p-value of 0.064. There's a 6.4% possibility that your results were due to chance. That doesn't look very high, but let's assume you did one more trial and got 4/5; the p-value jumps to 0.186, making it 18.6% likely that you were just lucky.

http://ff123.net/abx/abx.php

So I agree, some more trials would be better.

You could try decoding both files to wav and then trying the test with the wav files. If you can't hear a difference, then any differences you have 'heard' before are due to either chance or some difference in the playback decoding (e.g. volume).

-Audio Spyder

Easy. "Cleaner" and "smoother" imply a lower noise floor and a more pure audio signal.

Hardware error?

I have seen a computer, that worked normally for six months with normal applications and Ogg Vorbis files. Then the user decided to use lossless audio. APE files played perfect. FLAC files sometimes got a glitch. Turns out the mainboards chipset was overheating.

I am still not sure whether:


. . . infers that foobar2000 indicated differences or not.

Yes I've been wondering exactly the same thing since the original poster said that. Negative simply implies that it was the opposite of either what he expected or what he was trying to demonstrate. So if he was tring to demonstrate that they were identical then "negative" would mean they were different, but if he was trying to demonstrate that they were different then "negative" would indicate that they were identical. So it seems that we need to be able to read dereks mind in oder to make any sense of what he says. Acually that's a little similar to trying to get directions to somewhere from my wife.