Hi,

If I record an audio file from CD using 2-pass WMA 9 encoding at 128 Kbps will this provide a significant quality boost when compared with a single pass file? How does this increase in quality compare when the single pass file is at a higher bit rate? I've been told that a 2-pass 128 Kbps file is equivalent to a single pass file at 256 Kbps, is this correct?

Best Regards, Simon.

Please note that my feedback is not specific to WMA but my understanding of 2-pass encoding in general. Also note that most of it is theoretical.

2-pass is useless for CBR, where every second of audio takes up the same amount of bandwidth (Constant Bit Rate). 2-pass is more appropriate for ABR (Average Bit Rate). In this process, the first pass scans the entire file and uses the information during the second pass. The end-file should have a similar size to CBR, but the bits are distributed more intelligently across the file.

2-pass ABR should be a higher quality file than a CBR file of the same average bitrate. There's no easy formula (AFAIK) to say that 2-pass 128 ABR is equivalent to a CBR file of XXX bitrate. The best way to measure that is to listen for yourself. In my humble opinion, I would say the "128 2-pass = 256 CBR" claim is pretty outrageous, but I have no test data to back that up.

From the sounds of your question, you have a quality goal in mind. If so, it may be worth exploring VBR settings as well if quality is more important that a consistent, expected filesize.

Your questions are honestly difficult to answer accurately as there haven't been any listening tests for WMA 9.2 Std.

However, it would be reasonable to expect that 2 pass should give a quality boost vs. 1 pass BUT not as much as doubling the 1 pass bitrate.

I would not go so far as to say that 2-pass for CBR is useless. If it was, then why would it be presented as an option?

Without commenting on quality, it should be easy enough to see if a CBR file encoded using a single pass is different than one encoded using two passes.

I agree with LANjackal: I seriously doubt you're going to get 256kbit single-pass performance by using 2-pass at 128kbit!

Have a look here:


This should explain the reason for providing 2-pass CBR encoding, though it's unlikely that the little variation of the bitrate justifies the time-consuming usage of a 2nd pass.

Besides, is there any serious reason for the OP to use CBR encoding? If it was for streaming purposes peak bit rate-based ABR (which is called "peak bit rate-based VBR" in Microsoft's encoder) might be the better option. It provides better quality than CBR mode and resolves the issue of bitrates possibly exceeding certain bandwidth-related limitations.

Well, actually, 2-pass CBR allows much better buffer control.

Any assumption of "little variation" needs to be demonstrated, too.

I don't think so. Since the variation is negligible enough for Microsoft to still call this mode "constant bit rate" I guess it's safe to say that there's no serious need to back this assumption up.

Edit: Hm, maybe it is... now which company is the one claiming that 64 kbps WMA 9 were close to CD quality?

DING!

On the other hand, the link quas posted differentiates two-pass CBR from the three VBR formats: quality-based, bitrate-based, peak bitrate-based. Little (or at least "less") variation would also be my first guess at a distinction.

From here:

I just don't get how it can be called CBR if a second pass provides added benefit. Can the same number of bits be used against a single frame differently (more efficiently perhaps)?

If every frame is the same size (my definition of CBR) then the only variable left is the use of the bit reservoir. Assuming that the bit reservoir only extends over a few frames then all that is needed for "2-pass CBR" is to look ahead this number of frames to determine how to allocate the bit reservoir in the current frame, thus giving "2-pass CBR" in real time.

Could be something like the bit reservoir in MP3 and a using a second pass can improve how bits are allocated there if that is what is meant by "buffering".

Really, now.

And exactly how did you document the instantaneous (block by block) bit rate, pray tell?




So then you conclude that there is absolutely no MP3 that is CBR?

No AAC bitstream?

NOTHING is CBR that's in common use, except perhaps oddities like OG that I know nothing about, by your definition.

I don't feel like providing any proof for guesses based on the Microsoft statements quoted by quas earlier.


If that's not sufficient in your eyes, so be it.

Well, perhaps one might want to mitigate starvation situations mostly. This is, actually, just a guess.




So, in other words, the Microsoft statement does not agree with your own interpretation, but you will still stick to your own interpretation when regarding their software.

I understand very, very clearly what you're saying, indeed.

I've never claimed the opposite to Microsoft's official statements, it's "fairly constant" vs. "little variation". Both are analogue to each other, a newspaper article that's only fairly correct is in fact a little erronous. Buffering as well as the bitrate reservoir mentioned by Sebastian Mares weren't taken into account in my statement, because actual information about these two points isn't available in the material linked to by quas. With this information lacking I can only conclude that the usage of 2-pass encoding is rather questionnaible, especially since ABXing two samples using foobar's ABX tool didn't provide any useful results. I couldn't tell the difference between 1-pass and 2-pass CBR in both of these samples. Maybe someone else could - if yes, this would justify the two passes, no matter what theoretical advantages they actually provided compared to single pass encoding.


My "own interpretation" about "their software" is of no concern to you, because we're discussing about an individual case here, not their software in general. Don't even try to generalize anything based on individual cases. It's pointless.