Are folks still using the Feb. 2007 build or the latest build and why? Any help would be much appreciated.

I'm using 1.3.3.0. I used to use the Feb 2007 build until that one came out and I find it works better for my needs.

You would be best to download each build and conduct your own blind ABX tests. Personally, I use the latest build as it is able to provide more than transparent results (for me) at -q0.50. The files produced are around 1-2MB (each) smaller than -V 2 Lame. Additionally, the latest build produces Creative Zen and Xbox 360 compatible files (something of which the older builds wouldn't do, you had to go with either Zen or Xbox 360 compatibility, not both).

I use the Feb 2007 build since I had an experience where there was a regression on a sample. Actually it was in my very first experience with 1.3.3.0, which is why I lent it more weight than perhaps I normally would. But like kornchild says, I agree you should try an ABX test to see what your experience is -- this was only mine.
http://www.hydrogenaudio.org/forums/index....st&p=591802

When I was comparing the psychoacoustic "cut-off" between the different versions I noticed that 1.07 causes a pitch shift that's evident in the spectral graphs. Here is an animated gif of 1.07 and 1.13:

If there is one consistency in the current world of lossy compression, it's the development of the Nero AAC codec - eliminating one killer sample after the other. They don't know mercy. I have yet to find a regression.

Forget about your frequency plots. Use the search function to know why.

Your frequency graphs are meaningless without your blind ABX tests results showing that you are hearing what is evident in the graphs. We could analyze frequency plots all day looking for the best encoder and setting but that doesn't mean anything audibly (we don't look at music, we listen to it).

Similar to kornchild2002's reason i use 1.3.3.0 for compat reasons. Since 1.0.7.0 has some issues with certain AAC players which can be fixed by using mp4box though. So far i only found 1 or 2 regression issues with 1.3.3.0 such as Disposable Heroes by Metallica at 0:11 (lost the ABX log), those issues i found are not major but i will post a ABX log later on though.

I would avoid 1.1.34.2 (Aug 07), since it goes to low on analog silence, which causes a annoying metallic hiss.

I must admit, at HA there is one guy who knows even less mercy than the Nero devs. It's /mnt and his uber ears.

The graphs point out out an obvious frequency shift throughout the entire frequency range which should be cause for concern to anyone ripping their entire music library with this particular version. I made no claims about the resulting quality so your canned response is pointless. Besides, whether I can hear an audible diffeence on my equipment at a certain period in time doesn't mean it won't be an issue at sometime in the future on different equipment. In any case, other people who might be interested in this information can decide for themselves whether it's worthy of consideration and they can test it themselves.

This problem exists also in older versions and can be noticed on some samples. It was first time fixed in 1.3.3. It might have been compensated by some other psychoacoustic parameters in 1.0.7 and thus not so noticeable on some cases where it was noticeable in 1.1.34.

It seems i cannot successfully ABX that track that i mentioned last night at the moment, and tried to ABX this other track aswell but no luck atm. It should be a good thing though .



I see, i never noticed it on the older versions; just only on 1.1.34 which had a some horrid metallic noises in the background on silent parts on some analog recordings. Glad its fixed.

The graphs do not matter so long as this behavior is not audible. That is why frequency graphs are never used as the sole judge for audio quality. That is why blind ABX tests are used instead of frequency plots (or along with them). You can search the forums for why things like this aren't used to judge quality. They can be used to enhance someone's blind ABX tests (if the lowpass filter/frequency behavior is the reason for passing blind ABX tests). I can see that my post came off a little rude. I am sorry for that but the use of frequency graphs has been discussed here ad nauseum and why they can't really be used when judging lossy audio quality.

It wasn't fixed, it was improved New handling for such problems was introduced in 1.3.3.

I have to disagree. Like I said, just because it's not audible to me right now using my current equipment doesn't mean it won't in the future using better equipment with a better trained ear. I would be kind of pissed after converting thousands of music files if I started noticing a pitch increase in all of my music, however subtle. It appears to be a genuine flaw, not a psychoacoustic artifact, so I don't think it's fair to compare it to the apparent low-pass issue that's been discussed before. A flaw should be worthy of discussion for the devs as well as potential users.

He means that a visible difference in the graph != hearable difference for human ear.
The whole point of lossy compression is to change ("simplify") the graph in a way that it requires less space to store but sounds very close to the original (at least for humans). If a codec manages to alter the graph with little to no audible difference it is no bug, it's just the intended compression.

Of course. I am simply saying that you can't use frequency plots to try to judge lossy audio quality. As I said, this flaw is enhanced by your frequency plots if you and other people can back it up with blind ABX tests. It doesn't matter if your hearing is going to increase or quality of equipment is going to increase. Using your excuse would mean that no one would be required to conduct blind ABX tests and people could make wild claims and use frequency plots to determine audio quality. No, that is not the case here on hydrogenaudio. Blind ABX testing is used for lossy encoding comparisons, not frequency plots. So again, the "flaw" in your frequency plot is only relevant if it is audible. This can be compared to lowpass filters and various other issues that people try to evaluate with frequency plots. Nothing matters if none of this is inaudible. We are talking about lossy encoding where information that can be heard is key. There is no need to call something a flaw if it can't be heard and backed up by blind testing.

The devs can look at it if they want but there is no need for them to if this performance doesn't result in a audible decrease in quality. You might have discovered a new bug/flaw but we can't use frequency plots alone to determine this.

The changes in the spectrum from the original should be a result of psychoacoustic masking by the lossy compression process and nothing else - any other changes in the spectrum should be considered a flaw. The broad change in pitch I pointed out doesn't appear to be any normal result of psychoacoustic masking (to my knowledge) and it stands out compared to other waveform-based claims such as low-pass thresholds. I have never seen such an accepted compression artifact in another codec, and it doesn't show up in the subsequent versions of the Nero codec, for good reason. Exceptional artifacts that offer no apparent advantage should be pointed out whether one has ABX "proof" of an audible difference or not.

So, is it audible? You might be able to see a change but is it actually audible? Read the TOS if you want more information on this and pay close attention to TOS #8:
All members that put forth a statement concerning subjective sound quality, must -- to the best of their ability -- provide objective support for their claims. Acceptable means of support are double blind listening tests (ABX or ABC/HR) demonstrating that the member can discern a difference perceptually, together with a test sample to allow others to reproduce their findings. Graphs, non-blind listening tests, waveform difference comparisons, and so on, are not acceptable means of providing support.

Do you want to keep arguing that your frequency plots show an error/bug/whatever?

I agree that the images look drastic and bear some cause for concern at first sight. Still you might get a whole new look at graphs vs. actually perceivable differences if you try an ABX test in Foobar. So much has to be thrown out of the signal to achieve 1/6 of lossless bitrates and such extreme filtering is done, it is fascinating that the graphs still look quite equal. Perceivable difference really is all that counts here.

I would also be willing to try this myself if you would upload the samples to HA's upload thread.