[00:47:31] juinhwey joins the room [00:47:40] Hi Raymond! [00:47:49] Hi Jean-Marc! [00:48:01] OK, great this works now [00:48:21] Great. Thank you very much for walking me through it. That's extremely helpful. [00:48:45] As a note, you can see all the logs of this room at http://www.ietf.org/jabber/logs/codec/ [00:49:03] OK, that's a very useful tip. [00:49:37] This room is also used for people who cannot attend the meetings and it's also under the IETF IPR policies (note well) [00:49:47] OK, let's get started :-) [00:50:01] OK [00:50:04] You mentioned you had some quality improvements? [00:50:29] Yes, quite substantially actually. [00:50:30] It's not too late to change the bit-stream and in fact I have done so just today [00:50:35] Nice [00:51:22] Note that I also did many quality improvements recently, most of which were related to transients. Hopefully, your improvements are orthogonal to mine [00:52:03] I only had a rough first-cut and there are still a lot of things that need to be tuned. I have only tried the trumpet and harpsichord signals with this new system. I will try more signals later. [00:52:38] Yes, my improvement is indeed orthogonal to yours, because it mainly improve the steady-state periodic signal regioins. [00:53:09] What did you find? [00:54:14] Hey, JM, a side question: when we chat individually in Pidgin, I could see a note saying you were typing. How come after switching to the IETF jabber room, I don't see such notice anymore? [00:55:06] Not sure, I think the individual IM and the chat work quite differently. I don't think it's possible to see when people are typing when in a chatroom [00:55:16] (which is probably a good thing for crowded rooms) [00:55:56] Ah. OK. In our case with just two people I actually prefer to see such a note. [00:56:28] Yeah, would be useful, but I don't think it's possible. [00:57:00] OK. Guess I will have to live with it, then. [00:57:12] So how did you improve the quality on harmonic signals? [01:04:04] Actually, got the source code to your modified version? [01:06:22] OK, I am sorry about it, but I am currently in discussion with our legal department. This is because I haven't done this kind of IETF work before and I have to be careful in observing my company's policy. Even though the work is for a royalty-free codec and even though I believe Broadcom will donate its related IP royalty-free to IETF, I still need to find out how my company wants me to handle the disclosure of such IP at this point in time. Therefore, I think I have to wait until I get the OK before I can publicly disclose what I did. I apologize for that. I hope to get a clearance within a day or two so I can freely disclose what I did. [01:09:54] OK, let me know when you get the OK from legal [01:14:18] Yeah, I am really sorry about that. I feel quite awkward to have to postpone my disclosure. In fact, originally I was hoping to wait until I get the clearance from legal to go ahead with disclosure and then have a technical discussion with you. However, since you asked me in an email whether I have any quality improvement, I thought it would be impolite to ignore you and not respond to that question, so I responded to that question. [01:17:25] I asked you whether you are open to changing the bit-stream format of CELT at this stage, because if the answer is negative, then there is no point for me to continue my investigation of the CELT improvement approach which requires a change in the bit-stream format. [01:37:18] Don't worry, the bit-stream is still open to changes and I have made many recently [01:39:58] That's good to know. [01:41:01] By the way, I noticed that although your previous CELT 0.9.0 gave significant quality improvement over CELT 0.8.1 for the trumpet signal, it actually made the harpsichord signal worse. Did you notice that? Perhaps it is partially an inevitable trade-off between low frequencies and high frequencies? [01:45:29] I'm not entirely sure, but that would make sense [01:45:54] I've made quite a lot of changes recently, so it's also possible I broke something [01:57:33] The two main things that could have had an impact on the quality is the (old, static) allocation matrix and the new dynamic allocation allocation scheme, what hasn't been tuned much. [01:58:02] May I ask if your changes are mainly new code or modification/tuning of existing code? [01:58:56] Thanks for the info. My changes are mainly new code. [02:01:49] It's past 7 pm my time and I have to go home now to join my family for dinner. Sorry that I have to leave. I will chat with you again another day, hopefully with the ability to discuss technical things freely. [02:02:53] OK, have a good evening [17:00:51] jmvalin leaves the room [17:07:06] jmvalin joins the room [20:44:30] jmspeex leaves the room