Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Found a problem-sample (Read 6238 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Found a problem-sample

Hello,

I found a sample RC3 seems to have problems with.

http://metalbanner.virtualave.net/answer.wav  (884 KB)

Artist: Blind Guardian
Album: A Night At The Opera
Song: Wait For An Answer


Just listen for the metallic sound quickly jumping from one side of the stereo-image to the other (seriously bad english, I know...).

At quality levels < 4 the sound seems to be moved to the center and it "feels" much more "rough"/"unclean" (I'm not a "professional" test-listener so my expressions could seem a bit - strange ;-)

Due to my cheap equipment/ears I can't properly detect this artifact in encodings >= -q 4. Nevertheless I'm quite optimistic that more experienced listeners can easily hear it.

BTW, are there any pre-RC4 versions available with audible improvents over RC3?

bye,

Maik Merten

Found a problem-sample

Reply #1 –
Quote
At quality levels < 4 the sound seems to be moved to the center and it "feels" much more "rough"/"unclean" (I'm not a "professional" test-listener so my expressions could seem a bit - strange ;-)


Usually, the best way to go is to perform blind listening tests just to make sure. I really can't tell the difference appearently between lossy coupling method's except for a few samples and you might want someone else's opinion on this. This sample seems as if it has more stationary signals in it and it's tough to detect pre-echo, you can tell when pre-echo is subjectivly noticeable when you have a sample that has a lot of transients in it like castanets, stationary signals are tough though. I really don
't compare using ABX or any physcological testing of that nature. I seldomnly do on very few difficult to encode samples.

Quote
BTW, are there any pre-RC4 versions available with audible improvents over RC3?


yes, there are some binaries available on Benjamin Lebsanft's
site of pre-RC4 releases taken from the CVS repository (xiph.org). It offers an slightly improved window switching algorithm for pre-echo control. Nothing major though. There is also going to be improvments to normalize by band energy on noisy sections of the spectrum where the quantizer would accidently end up giving no bits at all to band's that contained the most audio energy. Inaccurate quantization would result in poor reconstruction of the signal that would seem as though it had poor resolution "muffled" at lower "quality" levels." which means actual with optimizations, latest release is in the left hand column.
budding I.T professional

Found a problem-sample

Reply #2 –
Good sample. Easy to hear the artifact up to 4.99 (abx 20/20) but above that I couldn't hear any difference.

I also tried pre-RC4 from bejamin's site. It was a bit different but not better in my opinion. (abx 16/16)

Found a problem-sample

Reply #4 –
Hello,

did an ABX-test with earphones (yes, the cheap ones that come with a $40 Discman!) at q 4.99: 50/50. At q 5, however, I couldn't distinguish it from the original. So exacactly same result here, ErikS.

bye,

Maik Merten

Found a problem-sample

Reply #5 –
Knockin' On Heaven's Door - Guns N' Roses

ogg -q 6
sennheiser hd 265 headphones
18/24 (p=0.01) (part after 1:55)

prolly something similar (preecho?)
PANIC: CPU 1: Cache Error (unrecoverable - dcache data) Eframe = 0x90000000208cf3b8
NOTICE - cpu 0 didn't dump TLB, may be hung

Found a problem-sample

Reply #6 –
More than likely it may be. What kind of signal is present after 1:55? a stationary signal like a guitar maybe? or is there a transient, i.e cymbal. Pre-echo would more than likely be contributing to the "artifact" your hearing.  Sorry, I don't ABX. I don't constantly perform listening tests all of the time. I only do on a few very diffcult to encode samples though. ABXing might be more statistically accurate. That is my interpertation though. If anyone wants to make any corrections on my part feel free.
budding I.T professional

Found a problem-sample

Reply #7 –
Quote
a stationary signal like a guitar maybe?

castanets type.
PANIC: CPU 1: Cache Error (unrecoverable - dcache data) Eframe = 0x90000000208cf3b8
NOTICE - cpu 0 didn't dump TLB, may be hung

Found a problem-sample

Reply #8 –
The sample as given clips heavily. You need to reduce it's volume and reencode before being able to make a valid comparison.

--
GCP

Found a problem-sample

Reply #9 –
So more than likely the signal that is producing the pre-echo is the result of a transient. Follow Garf's instructutions above and the scale the volume down to an x level so that clipping does not a play a large factor. You should then be able to make your comparison test.
budding I.T professional

Found a problem-sample

Reply #10 –
any1 tryed Knockin' On Heaven's Door?
PANIC: CPU 1: Cache Error (unrecoverable - dcache data) Eframe = 0x90000000208cf3b8
NOTICE - cpu 0 didn't dump TLB, may be hung

Found a problem-sample

Reply #11 –
Hello,

there's now a second version of answer.wav available at

http://metalbanner.virtualave.net/answer2.wav

I reduced the volume by using

# sox -v 0.5 answer.wav answer2.wav

so clipping shouldn't be a problem anymore. I did some ABX-tests on answer2.wav - giving a 20/20 result...

bye,

Maik Merten

Found a problem-sample

Reply #12 –
There's some preecho on the clicks in the center.

Edit: I can ABX this at -q5 as well.

--
GCP