I updated to 1.4.2_05 (was using 1.3.X prior to this test...sounds would play but switching between sources would sometimes result in the application hanging) and experienced no probs.

I used 1.5.0 beta 2 and it seemed to work OK. I've never seen anyone else mention they were using this version though - is it not considered safe to use? Should I revert to 1.4.2-05?

I've upgraded to 1.4.2_05 and have the same problems, although I have managed to narrow it down somewhat.

If I open an existing session, everything breaks (no sound, trouble rendering the screen) - but if I start the test again, it appears to work normally.

Is there some reason for this, such as an existing session containing info about my sound & graphics drivers perhaps? I've changed both the sound and gfx card in this computer, so if it was trying to draw and play things the same way it would run into difficulty. This is just a stab in the dark though

Joy oh joy

My aunt's laptop (I'm using it now) belongs to her employer and she doesn't have an administrator account. Sure enough, it doesn't come with Java, and I can't install it without admin rights, so I can't decrypt results (not to talk about installing Python for results parsing)

Please don't hate me, I will really find some way

Her desktop, which comes with Win98 (no user rights issues) is on maintenance right now, but I suspect the technician is fooling her ("We need to reinstall the internet, lady", I swear that's what they told her), so she'll bring it back tomorrow and I'll try to fix it myself. I bet a format-reinstall will do it.

Best regards;

Roberto.

I get that same screen rendering error. Often participants send me saved sessions (*.abc) instead of results (*.erf) and I have to convert them manually. Not once the sliders get rendered properly.

And I always forget to report this to Schnofler :B

Are there more results as usual (according to a first impression)? More testers for a less difficult test, or not?

I think you can set up Win98 to not ask for userids/passwords, if that is the issue.

http://support.microsoft.com/?kbid=152104

When I used Win98, there was no problem installing programs as a idless/passwordless user.

ff123

d'oh: I see now that you're talking about two different computers. Never mind

Yes. Shitloads of results. Last thursday I had about 40 results per sample, and countless more arrived this weekend.

(Also, a handful of ranked references, but... oh well )



The desktop, which is supposedly broken, is using Win98. The laptop I'm using right now is on Win2000. I could probably do some evil hacking to get admin access, but then my aunt could get problems if the technician at her work finds out.

Good news!

Maybe a LiveCD, e.g. Knoppix is an option?

Wow, this could turn out to be a really good test with that many results. Sorry about the computer situation, though

Please tell me I'm not one of them. My hearing condition is well and truly a lost cause if that is the case.

Maybe you could ask Frank Klemm to use his PC

I think so...as even myself has done the test (for one sample)...

I can come with my laptop if Roberto pay the travel to Brasil...

I'm on dial-up here. I would be back home before the live CD finished downloading.

Besides, no CD-RW drive here.



I don't think so. I will probably remove the ranked references from the results I'll upload, to make sure mean people won't point at others and laugh


Now, I am finally using the Win98 PC. It´s a slightly crappy Pentium 700 with small screen, but will probably do. I'll download all stuff I need overnight (Java, Python, SFTP and SSH client...), and I plan to finish putting results together tomorrow. I hope my 9-year-old cousin won't force me to go to the movies with her to watch Disney's latest

Talk to you soon.

Best regards;

Roberto.

You leave your home without a Knoppix CD in your bag?!? You're definitely not a nerd.


Just throw them away? Aren't the ranked references somehow significant too? I wonder, if somebody ranked a reference doesn't that mean that for him that encoder performs very well? (Well, as long as he doesn't rank the reference to very low values...)


Cool. Thanks a lot.

Heh. But it would be interesting for the rank owner to know he has mistaken...
At least I'm interested to know it for my result set...

Roberto, you could also wait to be back to your home before processing the results.

You know, the earth will not stop its rotation if you only post results next week...

Can you prove that ?

At least the delay has provided me enough time to complete the test, and hopefully other people have been able to add their results to the list

I look forward to seeing the smallest confidence intervals the world has ever seen

As Gabriel says, take it easy and enjoy your family gettogether.

You have nothing to excuse - on the contrary many people are thankful for the professional way you handle the listening tests.

I'm impressed with your communication skills and the way you manage to obtain near consensus through fact based dialogue, regarding codecs, settings and samples for the listening tests.

The world will be as happy a place to live in and I will still admire you the same, even if I don't get the results of the listening test until later.

Enjoy your life...

Full ack!

If somebody ranked the reference, it might also mean he randomly moved sliders up and down. :/

So, to be on the safe side, I discard all results with ranked references that haven't been ABXd to a confidence of 0.05 (chunky makes it very easy to discard such result files. I can't thank Phong enough for his creation).


Thank-you very much for the words of encouragement, guys.

According to the presentation page, only Vorbis got the special SSRC resampling. If so, was it checked and how and by who, that the other codecs using 32khz sampling didnt have resampling induced quality drop (like vanilla Vorbis) ?

Anyone noticed what you get when inverting end of both words?
Reading quickly, I get this one inverted. The meaning is then completely different.

Nobody showed up to test the other codecs. I am grateful (and the vorbis people should also be) to dev0 for checking it out.

Right. Imo then the correct procedure would have been to use SSRC for all the resamplings, or not use it at all. Now only one codec got the benefit just because someone happened to test it for Vorbis.

As I remember it, you mentioned SSRC resampling needed to be supported in a streaming software to make it's usage meaningful for this test.

dev0 found out it can be done for Vorbis with fb2k. It can't be done for WMA std and QDesign because neither windows media server nor Darwin streaming server use or support SSRC. Shoutcast (probably the most popular MP3 streamer, based on Lame) doesn't use SSRC either. Same thing about Real Broadcaster.

So there.

I said, at least, being a reason in order to justify the use of 3rd party component somehow. Before the test I privately said to you the same as here that all resampling should be done with SSRC or none at all. You said you can't consider any of my suggestions anymore because I'm too biased against Vorbis.
IIRC you also said that streaming server use isn't the point of this test, rather how the codec can do in best circumstances. Non-live streaming is perfectly possible for all codecs with SSRC.
Since the other codecs using resampling didnt get tested or treated equally to Vorbis, we can only hope that there's no resampling issues which would affect the results negatively for the other codecs.

That's exactly what I read first !

If Vorbis performs bad, nobody will care about it. If it comes first, for sure it's an issue !

True. But you gotta consider that I'm only one person that doesn't do listening tests himself and, particularly on this test, was on a limited time frame. In this aspect, Vorbis can be considered lucky that it had an enthusiatic individual that looked into this issue and found a solution. Unfortunately, WMA, Real and QDesign didn't have anyone to look out for them.

It's similar to the issue with the iTunes MP3 encoder in the MP3 listening test. There has been lots of speculation about iTunes maybe being better in CBR at 128kbps, but noone took the initiative to test VBR vs. CBR while the test was being discussed.

http://www.hydrogenaudio.org/forums/index....ndpost&p=181882

Imo it was not similar. In your example it was about internal encoding setting. In resampling issue it was about the use of external 3rd party component which is possible to use for all resampled codecs.

I thought this had been discussed into the ground a couple of pages ago?

Vorbis was the only encoder that was shown to have issues with its internal resampler, so there should be no need to use SSRC with the other codecs in the test.

Eh, how do you know if there was need or not if it was not tested!
It is perfectly possible that this issue is a non-issue, but without testing we simply don't know, thus if not tested, we should treat the codecs equally. This is my opinion.
So, even if this resampling issue turned out to be irrelevant regarding the other codecs, this imo shows somewhat questionable testing practice, especially considering that there is a 3rd party component involved.

Ogg Vorbis is quite new, open source and it's development is in progress.
I understood the SSRC issue always in such a way, that for this test, Ogg Vorbis samples were prepared, that show the potential at 32 kbps in comparison to established codecs.
That's quite fair for a codec in development.

If you ask from Monty, he will tell you that Vorbis is about 10 years old project of his.

Ogg Vorbis is not the only codec that´s in development.

I was away (holiday) the last weeks. And now I would like to make the test, too. But in the first posting you write that the test is scheduled to end at July 11th but the title says "OPEN".

So is the test open or closed already?

Thanx
Big_Berny

I plan to start calculating results this afternoon, so yes, you can pretty safely consider it closed :/

Wll, that's bad news... Sorry, that I couldn't do the test. I hope you have enough results!

Big_Berny

Still no results?

rjamorim you're being very quiet, you OK? or have you gone into shock cos of the results?

Too much wine at the family parties?

erhm... i know 56k is slow, but shouldn't the results be in by now?

anyway looking forward to the results, thanks for the effort.

Hope everything's OK.

me too, the suspense is killing me

Off-topic posts split here.

Hey, what happened to rjamorim? Now, even his website http://www.rjamorim.com/ is down.

EDIT: His site is online again