[00:06:43] Dan Burnett leaves the room [00:08:08] Erik Norvell leaves the room [00:20:50] Kevin P. Fleming joins the room [00:22:49] Kevin P. Fleming has set the subject to: Codec WG Meeting, IETF 77 | Audio Stream @ http://videolab.uoregon.edu/events/ietf/ietf773.m3u === [00:23:06] Kevin P. Fleming has set the subject to: Codec WG Meeting, IETF 77 | Audio Stream @ http://videolab.uoregon.edu/events/ietf/ietf773.m3u === http://tools.ietf.org/wg/codec/agenda?item=agenda77.html [00:25:35] Kevin P. Fleming has set the subject to: Codec WG Meeting, IETF 77 | Audio Stream @ http://videolab.uoregon.edu/events/ietf/ietf773.m3u === http://tools.ietf.org/wg/codec/agenda?item=agenda77.html === http://www.ietf.org/meeting/77/webex.html [00:31:28] stpeter joins the room [00:32:31] so are we preferring the webex to jabber+audio for this meeting, or is the webex just to present slides? [00:33:04] wait... there isn't a webex session listed on that page [00:33:09] not for this meeting anyway [00:38:44] jmspeex leaves the room [00:39:09] JonathanLennox joins the room [00:39:18] sayer joins the room [00:39:28] Dan Burnett joins the room [00:42:03] jmspeex joins the room [00:45:05] Florob joins the room [00:45:30] running a bit late? [00:46:13] jon-ietf joins the room [00:46:16] Erik Norvell joins the room [00:46:45] Kevin P. Fleming leaves the room [00:48:19] linuxwolf joins the room [00:48:44] Kevin P. Fleming joins the room [00:48:46] markus.isomaki joins the room [00:49:39] Joe Hildebrand joins the room [00:49:40] magnus joins the room [00:52:10] YJS joins the room [00:52:10] meeting started yet? [00:52:46] Yes, are there no audio? [00:52:58] just nothing here, not on the audio stream yet [00:53:03] Ok [00:53:18] is the webex live? there's no link on the webex page [00:53:37] Alan Johnston joins the room [00:54:05] ahh... i see the emails now [00:54:18] bemasc joins the room [00:54:22] No, I think they skipped it. Try http://videolab.uoregon.edu/events/ietf/ietf773.m3u [00:57:23] suzukisn joins the room [00:58:16] gmaxwell joins the room [01:02:38] brian.bnsmith joins the room [01:04:03] Constant delay is _sometimes_ available. [01:04:31] linuxwolf leaves the room [01:04:39] linuxwolf joins the room [01:05:55] Kevin P. Fleming leaves the room [01:05:56] Can someone point me to the presentation download link? [01:06:29] http://tools.ietf.org/agenda/77/slides/codec-5.pdf [01:09:21] as0-d91k joins the room [01:09:22] lminiero joins the room [01:09:24] The "echo" is due to feedback from speakers into microphones on a bidirectional link [01:09:34] An example application is teleconferencing. Two rooms with audio equipment, and a low latency (e.g. single provider, or a metro service) network. You can run without AEC or only minimal AEC. [01:09:52] Dale Worley joins the room [01:11:54] It's completely reasonable to achieve very low latency (<5ms network delay) within a geographic area, even on the public internet— so long as some considerations are made (e.g. using a single provider) [01:14:01] Dale Worley leaves the room [01:15:27] Dale Worley joins the room [01:19:13] that partial redundancy would only work properly if packetisation and codec work together [01:19:55] I am not disagreeing that is not possible. However, I am trying to see the chain the audio will take. To get an echo it seems that you are talking about microphone -> buffering -> encoding -> packetation -> transmission -> jitter-buffering -> decoding ->playback at far end -> microphone -> buffering -> encoding -> packetation -> transmission -> jitter-buffering -> decoding ->playback (locally) before it becomes an echo. [01:20:41] magnus: yes, and apparently 25 ms is seen as a rough upper bound on that total round-trip latency before echo cancellation is required [01:21:12] and even over 25 ms, smaller latencies allow the use echo cancelling algorithms that induce fewer artifacts. [01:21:57] Dale Worley leaves the room [01:22:12] Dale Worley joins the room [01:22:31] so for very high quality links (e.g. Internet2) and relatively short distances (a few hundred kilometers) these latencies are achievable, but only if the algorithmic delay is short. [01:23:09] Yes, which argues for a sample based codec where you can packetize as little as you want. [01:24:03] magnus: Even on tremendously high-bandwidth links, pushing hundreds of KB/s or even 1 MB/s is not possible with low latency [01:24:15] There is a middle ground. [01:25:40] gregory joins the room [01:25:54] gmaxwell___ joins the room [01:26:30] One thing I'd like to point out is that in the room we are, the reverberation time is on the order of 1 second. [01:26:43] oh, I can hear it now [01:27:03] So you have acoustic echo of up to 1 second and you still don't notice it that much. Well, you notice the reverberation, but it's not as annoying as echo [01:27:03] I think the true time constant might be shorter, but yeah, there's lots of reverb [01:27:36] So if you had -20 dB echo with ~200 ms delay, it would actually blend in the reverberation [01:27:46] the mic is hugely clipping, btw. [01:27:50] There is definitely tradeoffs here. The question is how important is low delay compared to compression efficency. [01:28:04] Or you could have -10 dB echo with 100 ms delay and not notice either [01:28:06] magnus: I think we should provide a knob. [01:29:36] YJS leaves the room [01:30:56] a knob on the codec implementation, like an environmental variable, or a knob on the end-points UI? [01:31:17] You definitely don't want knobs on the UI [01:31:17] I mean it in an abstract sense, a knob on the spec. [01:31:55] linuxwolf leaves the room [01:32:03] linuxwolf joins the room [01:33:02] right... I was trying to inductively get you to think/explain a bit about how such a "knob" would be practically useful. Should have been more direct. [01:33:34] stephanwenger joins the room [01:34:23] dogs can hear much higher frequencies, which MP3 tends to code minimally or discard [01:34:27] wrt fish, depends if they are in/out of H2O [01:34:40] The ear geometry difference would make perceptual codecs not work so well for animals, sorry. (You can even measure masking difference between people. :) ) [01:34:47] gmaxwell leaves the room [01:35:41] I think fish out of H2O have bigger problems than how MP3 sound [01:36:09] depends [01:36:22] if it is their death march, they may care greatly [01:36:27] but we digress... [01:36:32] jmspeex: the characterization of the impact of early vs late reverb has been fairly well studied in the area of room acoustics. I can send some citations to the working group list if people are interested. [01:36:37] Dale Worley leaves the room [01:38:40] Dale Worley joins the room [01:39:39] RjS joins the room [01:44:03] gmaxwell joins the room [01:44:06] jack joins the room [01:44:09] jack leaves the room [01:45:11] ford5636 joins the room [01:45:13] ford5636 leaves the room [01:46:17] how about MOS degradation in BV codecs for non-random packet loss? [01:47:58] Jack Moffitt joins the room [01:48:28] johnson24320 joins the room [01:48:58] stephanwenger leaves the room [01:49:30] who is speaking? [01:51:27] johnson24320 leaves the room [01:51:29] mjgraves joins the room [01:52:40] what is the name of the speaker? [01:52:48] harrison30308 joins the room [01:53:16] harrison30308 leaves the room [01:53:47] gmaxwell___ leaves the room [01:54:02] where are the disclosures? [01:55:38] IETF IPR disclosures are at http://www.ietf.org/ipr/ [01:55:55] You need to put in the I-D filename and you will see any listed [01:57:11] I believe the presenter of the SILK codec was Koen Vos [01:58:19] Thank you. [02:00:24] jr112 joins the room [02:02:42] jgunn joins the room [02:04:05] vanburen4036 joins the room [02:04:11] It's worth pointing out that on JM's test graphs the 7khz / 3.5khz graphs are lossless wide/narrowband, so they reflect the best performance that wide/narrowband codecs can deliver. [02:08:00] markus.isomaki leaves the room: Replaced by new connection [02:08:03] markus.isomaki joins the room [02:08:45] Dale Worley leaves the room [02:08:49] cleveland60038 joins the room [02:09:08] gmaxwell: Actually, Koen convinced me that it's not the case. You can actually do is keep the whole 8 kHz [02:09:18] suzukisn leaves the room [02:10:39] jmspeex: I suppose thats fair. [02:11:40] Joe Hildebrand leaves the room: Disconnected. [02:14:32] patents are easy to work around as long as you know about them [02:15:40] We shouldn't be tolerating this fud. [02:15:54] but overbroad patents are unenforceable, and so not terribly dangerous [02:15:57] The speaker should be asked to disclose the patents which he is apparently aware of. [02:16:20] subversion!? git! [02:16:47] cleveland60038 leaves the room [02:17:45] linuxwolf leaves the room [02:17:54] linuxwolf joins the room [02:19:04] linuxwolf leaves the room [02:19:04] would multiple channels impact that? [02:19:44] think ambisonics, >4 channels [02:19:51] is tandeming now out of scope? [02:19:58] For that syncing comment: Codecs need to have knowable (or constant) algorithmic delay. This should probably be mentioned as a requirement (though it's the case for all the codecs under discussion!) [02:20:32] gmaxwell: hmm... variable packetization? [02:21:41] ford58324 joins the room [02:22:15] Dale Worley joins the room [02:23:58] bemasc: For example, I could imagine a latency-minimizing vorbis implementation which actually had variable algorithmic delay. But thats not true for most things, and not true for SILK/CELT/BV.. etc. (unless you count the change in serialization delay from VBR operation) [02:24:44] gmaxwell: if you pack variable numbers of frames into an RTP packet, then you can easily get variable/unpredictable delay [02:25:35] Delay would be a simple function of packing, yes? So variable, but I don't see unpredictable. [02:25:37] It's not unpredictable: You're choosing it by how you packetize. I also think thats out of scope in the same way that considering the ip/udp/rtp overhead is out of scope. [02:28:59] ok [02:29:49] btw, I don't think it's fair to say that a "big switch" represents two codecs. If all implementations are interoperable, then it's clearly one codec. [02:30:08] ietf will still be here, but we might be on IPv14 [02:30:32] jr112 leaves the room [02:30:48] 2^1024 addresses, due to SmartGrid and Internet of things [02:30:48] hehe [02:31:00] Dan Burnett leaves the room [02:33:43] Dale Worley leaves the room [02:33:44] Dale Worley joins the room [02:34:53] gregory leaves the room [02:35:06] gregory joins the room [02:35:59] jr112 joins the room [02:37:26] JonathanLennox leaves the room: Computer went to sleep [02:37:48] Jack Moffitt leaves the room: Disconnected: connection closed [02:37:48] as0-d91k leaves the room [02:37:52] RjS leaves the room [02:37:59] Erik Norvell leaves the room [02:38:02] jr112 leaves the room [02:38:04] Alan Johnston leaves the room [02:38:07] gregory leaves the room [02:38:10] ford58324 leaves the room [02:38:11] bemasc leaves the room [02:38:19] jgunn leaves the room [02:38:23] sayer leaves the room [02:38:57] Dale Worley leaves the room [02:39:11] lminiero leaves the room [02:42:39] gmaxwell leaves the room [02:42:55] Florob leaves the room: Disconnected: connection closed [02:43:04] quincyadams9325 joins the room [02:43:16] magnus leaves the room [02:44:17] mjgraves leaves the room [02:45:57] stpeter leaves the room: Computer went to sleep [02:48:48] jmspeex leaves the room [02:49:01] François Audet joins the room [02:50:18] jon-ietf leaves the room [02:52:52] François Audet leaves the room [02:53:03] quincyadams9325 leaves the room [03:01:00] markus.isomaki leaves the room [03:13:41] vanburen4036 leaves the room [03:14:08] gregory joins the room [03:16:44] jmspeex joins the room [03:17:48] jmspeex leaves the room [04:47:25] Jack Moffitt joins the room [04:51:08] Jack Moffitt leaves the room: Disconnected: connection closed [04:51:35] Jack Moffitt joins the room [04:56:54] Jack Moffitt leaves the room [07:08:30] gregory leaves the room [07:09:43] gregory joins the room [07:10:08] jmspeex joins the room [07:12:35] gregory leaves the room [07:13:50] gregory joins the room [07:13:51] jmspeex leaves the room [07:14:26] jmspeex joins the room [07:19:36] gregory leaves the room [07:20:36] gregory joins the room [07:21:21] jmspeex leaves the room [07:21:35] jmspeex joins the room [07:24:44] gregory leaves the room [07:25:51] jmspeex leaves the room [07:25:54] jmspeex joins the room [07:26:05] gregory joins the room [07:32:12] gregory leaves the room [07:48:52] jmspeex leaves the room [09:41:31] jmspeex joins the room [13:23:05] weihongbo joins the room [14:07:53] weihongbo leaves the room: I'm happy Miranda IM user. Get it at http://miranda-im.org/. [14:38:50] stpeter joins the room [14:38:54] stpeter leaves the room [14:52:52] reagan28408 joins the room [14:52:53] reagan28408 leaves the room [15:27:56] gregory joins the room [15:31:04] gregory leaves the room [15:47:10] jmspeex leaves the room [16:00:15] Dale Worley joins the room [16:00:26] Dale Worley leaves the room [16:00:28] Dale Worley joins the room [16:00:30] Dale Worley leaves the room [16:16:47] markus.isomaki joins the room [16:17:07] markus.isomaki leaves the room [17:33:00] jmspeex joins the room [18:34:38] jmspeex leaves the room [19:14:59] brian.bnsmith leaves the room