With the number of software products that support ogg growing, I decided to give it a try this week. I've been a long time Lame encoder user at the alt preset standard setting, and I've generally been happy with the results but if ogg could create the same quality(or better quality) audio with a smaller file size, I was ready to consider compressing future music into ogg. Unfortunately, my first attempt was at a quality setting of 4.99. It definately created a much smaller file size, but I was very disappointed in the sound quality in the high frequency ranges. I then tried a quality setting of 7, which from what I read should have been at least comparable to Lames alt preset standard, if not better. At q 7 the file size was even slightly larger than the lame mp3 file and I was amazed to find the high frenquencies were still not up to par compared to the lame mp3 file. Without going through hundreds of posts, am I missing something or does ogg simply not work well at high frequencies? I am using the 1.01 version of ogg by the way.

wellcome bfollett ... before somebody points you to TOS #8 violation, are you sure this is not a "placebo" efect? mean, you internally want to like mp3 better or something? have you done some ABX testings?

If you were using standard preset you have to use not 1.0.1 version but GT3b1 version with 'q' value between 5.0 and 5.5 (Target bitrate for q5 is 180 kbs). In most cases of q5.5 filesize will be smaller than APS encoding. Than ABX it. Actualy I can't ABX LAME 3.90.3 APS and Ogg Vorbis q5. But on some cases (rare CD, classical music) I use q6 value - I hope I'll get better acoustics soon :-) and there may be difference.

I wasn't trying to find fault with the ogg file. I fully expected to happily create a smaller file that sounded better than my original lame Mp3 files. I don't even think I have a keen ear when it comes to hearing subtle differences between sound samples. That is why I was so suprised when I first played the ogg file and thought I heard a noticable weakness in the high frequencies. I then placed both the ogg and mp3 file into winamp and started playing them back and forth. I had dragged and dropped both files into Winamp and they both displayed the same information in the song list, so when I first started playing them, I had no idea which was the mp3 and which was the ogg file, but I easily picked the mp3 file as the better sounding in the high range. Also, I know there's nothing scientific about looking at the spectral analyzer on the Winamp display, but you could clearly see high frequecies being graphically displayed on the mp3 sample, and not on the ogg file.
As for de Mon's response, I need to do a lot more reading about builds and settings and such, becuase your response with all its acronyms made little sense to me.

What is 'weakness' ?

I suggest using GT3b2 starting at Q5 and then use abx. If you can abx then move up.


OT - There is a HF boost issue up to Q5.99 that stops at Q6. I can abx Q5.x but not Q6. Many find Q4-5 transparent though..

To get any serious attention here, you'll need to provide ABX results. It is the only way we can be sure that the placebo effect can be discounted.

A link to some information about what an ABX test is, why it is needed:
http://doc.hydrogenaudio.org/wikis/hydrogenaudio/ABX

Just to clarify...

You didn't transcode, did you? [create the ogg from the mp3]

I hope you used the same original .wav for the creation.

WinAmp's spectral analyzer was never meant to be accurate, just a neat-looking display. Several formats show very little frequency activity in the high range when such is actually present in the audio.

That is an example of an unrelated-to-the-sound external factor that may be messing with your 'results', which is exactly why the ABX test is recommended. It exists to greatly minimize the possibility of drawing conclusions based on observations not related to what is actually being tested. Plus it provides you with a handy "probability that you obtained the listening results by chance guessing" so you can be more (or less, if the probability stays high) sure you're actually hearing something.

I was a long time LAME 3.90.3 APS user too. Now... I am a die hard Ogg believer.

I use GT3b1, and while I did ABX, I didn't keep the results (so don't yell at me).

Basically, the ABX program considered me a complete guesser... I couldn't tell the difference between Ogg at q5 and APS. In fact, I was a big fat guesser for Ogg, and a slightly overweight guesser for APS.

And it's not like I have crappy equipment... not great... but a decent headphone amp and headphones.

But just so I feel warm and fuzzy, like de Mon, I encode at higher rates for well liked CD's.

Please keep in mind Ogg Vorbis was designed as a general purpose audio codec, not as a general purpose religion...

I concur

Oh please guys... let's not be too literal here. It was a figure of speech. This is a great forum, but sometimes, a little too much needless correcting goes on. Gosh... now I'm guilty.

No offence was intended. I am a very satisfied Ogg user, so I know what you mean. "I am a die hard Ogg believer" just sounds funny to my ears.

On the other hand I may be guilty of violating TOS #8 because my claim that Ogg Vorbis is not a good general purpose religion came without any prove

Religion is founded on faith not evidence, so there are no proofs of the validity of any religion only beliefs. The analogy is quite appropriate here as there are many faiths (CODECs) and their followers (Zealots) who believe that their faith (CODEC) is the only true religion (CODEC) and those who believe otherwise are heretics or fools (WMA Users).

Love and charity,
Fairy

could you provide an ABX result for that?

No need to ABX - it's obvious the tonal purity is distorted

Heretics.

Xiph bless the United Ears.

Hallelujah!!!!!!!!

I don't think vorbis lacks any HF at q7.
Just tried an abx using the vorbis encoder in Foobar2K 0.7.7

I have the following equipment, with which I can hear
(or at least perceive) tones up to about 19khz:

Terratec DMX 6Fire
Sennheiser HD580

I'm using the headphone output on my 8 channel
Alesis mixing desk.

Sasha - Drempels ( Vorbis Q7.01 )

Very difficult to ABX. I think there was a very slight HF
*boost* in certain sections. 25% Probability of guessing
after 16 samples.

Sasha - Requiem ( Q 5.01 )
Sasha - Requiem ( Q 4.03 )

Could easily ABX 100% by listening for the high frequency
boost in the background hiss of the recording.

I didn't detect any other obvous flaws - stereo image
was intact and no artifacts with this sample. I expect if
the HF issue is solved Vorbis will be transparent with this
sample at Q4.

There's not really any point posting the results, is there?
The HF problem is well-documented after all.