WEBVTT 00:00.000 --> 00:15.000 We are improvising at the law, so this was the one that was kept until late yesterday, 00:15.000 --> 00:25.000 but we thought, well, we really wanted to let these guys show what they have done in the last year. 00:26.000 --> 00:32.000 What you see here is a photo of the new Muslim video box, and this is what we are going to introduce you today. 00:32.000 --> 00:35.000 I mean, you see it in action pretty much all weekend. 00:35.000 --> 00:41.000 I would like to share some details because I think it's really an awesome piece of technology. 00:41.000 --> 00:44.000 These guys deserve all of this for that. 00:44.000 --> 00:51.000 So you may remember, in almost video boxes, they were fully made out of wood. 00:51.000 --> 00:59.000 They were the first iteration, I think we built them in 2014, and they've served us well for a particular number of years, 00:59.000 --> 01:06.000 but last year we just, well, for a 20.4 edition, we had decided that, well, it was decided to say goodbye to them, 01:06.000 --> 01:09.000 simply because they were falling apart physically. 01:09.000 --> 01:17.000 I mean, the screws were falling out, and the hardware inside was absolutely impossible to replace. 01:18.000 --> 01:21.000 There were components in there that were in the infrastructure anymore. 01:21.000 --> 01:24.000 So any spare parts were super expensive to think. 01:24.000 --> 01:33.000 So last year, within a little bit of an in between solution with something in the same case, 01:33.000 --> 01:36.000 as that you see today, it was all off the shelves apart. 01:36.000 --> 01:43.000 So it was just a USB 4-core ethernet switch and physical interface. 01:43.000 --> 01:49.000 The USB hub, the USB HDMI crever, and then some ports at the outside. 01:49.000 --> 01:51.000 So it was no, you know, very functional. 01:51.000 --> 01:56.000 It needed a laptop to operate, so every room needs to have two laptops. 01:56.000 --> 02:02.000 There were some issues with this cover, as well as running this setup. 02:02.000 --> 02:10.000 So it decided, well, let's improve for next year, all of you were planning to land this, maybe one or two editions. 02:11.000 --> 02:18.000 And I had a long, well, I had long been dreaming of the solution. 02:18.000 --> 02:23.000 That's what we wanted to come up with. 02:23.000 --> 02:28.000 And so during last fall, I think even Sunday evening. 02:32.000 --> 02:36.000 I started talking to these two guys in the lock, 02:37.000 --> 02:41.000 because it was pointed out that, well, maybe these kind of snow, 02:41.000 --> 02:44.000 if thing were to be a button designing hardware and at least, 02:44.000 --> 02:47.000 or at least if we tend to very well think how. 02:47.000 --> 02:51.000 So that's when, you know, I started explaining what I had in mind, 02:51.000 --> 02:56.000 and actually my dad had independently of whatever we were going to be dreaming, 02:56.000 --> 03:01.000 or he started working on something similar with the idea of hopefully in mind. 03:01.000 --> 03:09.000 So speaking of, you know, the synchronized brainwaves or whatever we call it. 03:09.000 --> 03:14.000 Anyway, I'm going to let these guys introduce to you what they've built, 03:14.000 --> 03:18.000 how they did it, and we'll give you some more details. 03:18.000 --> 03:22.000 So this is Dexter, and this is my dad. 03:22.000 --> 03:25.000 I think there's also a really big round of applause. 03:32.000 --> 03:38.000 Okay, well, I am going to go over some of the components here, 03:38.000 --> 03:45.000 but basically what this set of all does is take a video stream 03:45.000 --> 03:50.000 from a capture card, which on this diagram is here, 03:50.000 --> 03:57.000 and send that video stream through a computer, the RageXA export, 03:57.000 --> 04:02.000 and then, you know, my performance. 04:02.000 --> 04:08.000 And then, have that video streamed over network. 04:08.000 --> 04:12.000 In addition to that, we have some integrated, 04:12.000 --> 04:20.000 we have integrated some components that make the life of the video team a lot easier. 04:20.000 --> 04:24.000 One of them is the network switch that allows the loud, 04:25.000 --> 04:29.000 daisy-chaining the boxes, so we don't need to deploy extras, 04:29.000 --> 04:32.000 which isn't the room, as you can see. 04:32.000 --> 04:37.000 We don't have a switch here, we just connect one of the boxes to the other one. 04:37.000 --> 04:45.000 And also, we have the audio mixer integrated inside the box, 04:45.000 --> 04:52.000 so that we don't need an external mixer in order to master the audio. 04:52.000 --> 04:57.000 So, in that case, we don't need that much external hardware, 04:57.000 --> 05:03.000 and the whole setup is quite cheap, even though the box itself is not so cheap. 05:03.000 --> 05:12.000 So, the first component we're going to talk about in a bit of detail is the audio mixer, 05:12.000 --> 05:18.000 that was mainly designed by my friend, so we can say a bit about it. 05:19.000 --> 05:24.000 Yeah, I started the initial design of the board after last year, first time, 05:24.000 --> 05:28.000 because I saw the really nice new boxes with a thing button top, 05:28.000 --> 05:33.000 but besides that, I came back up the room still and locked the audio mixer. 05:33.000 --> 05:38.000 And while the laptop could be managed remotely by the video team, 05:38.000 --> 05:41.000 the audio mixer had to be managed in room, 05:41.000 --> 05:44.000 and that was not ideal for audio monitoring. 05:45.000 --> 05:48.000 So, I found the TNC, the board is in here above, 05:48.000 --> 05:54.000 and that has a pretty neat library that you can really easily make an audio mixer 05:54.000 --> 05:58.000 from this board by just working in various audio components, 05:58.000 --> 06:04.000 and to make it work for first time, I decided to build this board that's around it, 06:04.000 --> 06:09.000 and it has a few X-LAR inputs, so that the wireless microphone sets. 06:09.000 --> 06:12.000 I'm using that now, can be plugged into the box, 06:12.000 --> 06:17.000 and then we mix together and set off over the network to the video operation center. 06:19.000 --> 06:21.000 Let me just enter check. 06:21.000 --> 06:27.000 I think I can show you, yes, I can probably show you the control interface. 06:27.000 --> 06:31.000 No, let me see, like this. 06:31.000 --> 06:37.000 This is the control interface for the video setup that we all see 06:37.000 --> 06:39.000 from the video control room. 06:39.000 --> 06:44.000 So, you can see that we can remotely control all of the levels. 06:44.000 --> 06:47.000 We see a graph of how we've been talking. 06:47.000 --> 06:49.000 We see previews and things like that, 06:49.000 --> 06:51.000 basically those sliders, 06:51.000 --> 06:57.000 they're end up sending commands to the audio mixer board that does the next thing. 06:57.000 --> 07:06.000 Yeah, yeah, this is basically, 07:06.000 --> 07:10.000 initially, based around the example board from teensy, 07:10.000 --> 07:12.000 they have a shield you put on top of a teensy, 07:12.000 --> 07:17.000 and it gives you a single stereo input and stereo output for headphones, 07:17.000 --> 07:22.000 and I basically took that design and devoted to have more channels 07:22.000 --> 07:27.000 and add some interfacing chips in between to give it balanced audio 07:27.000 --> 07:32.000 to interface with all the more professional equipment that's connected to it. 07:32.000 --> 07:37.000 And there's a few more features that are not currently in use at first 07:37.000 --> 07:38.000 them. 07:38.000 --> 07:46.000 Like, you have the connector to have internally an audio jack connected to the HDMI frame grabber, 07:46.000 --> 07:50.000 so we can have the audio from the laptop from the people preventing, 07:50.000 --> 07:54.000 but that is not working this year, hopefully next year. 07:54.000 --> 08:00.000 And all the way here, there's an extra header that can have a decks receiver, 08:00.000 --> 08:03.000 because the wireless microphones are actually 08:03.000 --> 08:06.000 technically wireless telephones, 08:06.000 --> 08:08.000 and the plan must to make a receiver there, 08:08.000 --> 08:13.000 so we don't have separate receivers on the box here to have all the audio, 08:13.000 --> 08:18.000 so less components and less things can go wrong. 08:18.000 --> 08:22.000 I think that's pretty much accomplished the audio board. 08:22.000 --> 08:31.000 Yeah, and we also have this file that is basically doing everything 08:31.000 --> 08:36.000 that wasn't already done in some other place in the box. 08:36.000 --> 08:39.000 It has four ports to charge microphones, 08:39.000 --> 08:42.000 so with these cables connected there, 08:42.000 --> 08:45.000 we just charge the microphone receivers, 08:45.000 --> 08:51.000 so that you don't need to run extra cables in order to power the receivers. 08:51.000 --> 08:54.000 There's also a USB pass through port. 08:54.000 --> 08:57.000 The charging ports are actually isolated, 08:57.000 --> 09:01.000 as to not to close ground loops through the microphone ports, 09:01.000 --> 09:06.000 and we also have a temperature sensor on that board. 09:06.000 --> 09:13.000 We have a fan controller that controls the fans that have been making 09:13.000 --> 09:16.000 a really annoying noise in this room today, 09:16.000 --> 09:19.000 but this should be fixed for next year. 09:19.000 --> 09:22.000 Yeah, basically this is the catch all board, 09:22.000 --> 09:26.000 that's everything that's not done somewhere else. 09:26.000 --> 09:31.000 And we also have the network switch that is extremely interesting, 09:31.000 --> 09:34.000 also designed primarily by more than 10. 09:34.000 --> 09:39.000 Yes, it is actually pretty much the same thing as the network switch 09:39.000 --> 09:43.000 used last year, because that is a 5-board switch 09:43.000 --> 09:47.000 that's a USB network interface in it to make it appear like it's a 4-board 09:47.000 --> 09:50.000 USB network interface, 09:50.000 --> 09:53.000 and that one was not very reliable. 09:53.000 --> 09:58.000 So this is basically a copy of the design of the very common 5-board 09:58.000 --> 10:00.000 switch you might find online, 10:00.000 --> 10:03.000 and it's basically the same design, 10:03.000 --> 10:07.000 but with a few components, beefed up, so it's more reliable, 10:07.000 --> 10:09.000 the other ones are prone to overheating. 10:09.000 --> 10:14.000 And more importantly, one of the 5-metic ports is facing backwards, 10:14.000 --> 10:18.000 so we don't have to have a cable around the outside of the case 10:18.000 --> 10:21.000 to plug in the computer that's inside. 10:21.000 --> 10:24.000 In addition to that, there's this connector 10:24.000 --> 10:26.000 that connects over to the computer inside, 10:26.000 --> 10:30.000 and that allows this switch to actually be a managed switch, 10:30.000 --> 10:35.000 and that is used to show the status of all the switch ports in the front. 10:35.000 --> 10:38.000 So network issues can be developed easier, 10:38.000 --> 10:41.000 and also, again, this is used for remote monitoring 10:41.000 --> 10:44.000 of all the hardware and for them. 10:44.000 --> 10:51.000 Yeah, and also we have this bit that is actually the computer. 10:51.000 --> 10:55.000 This is off the shelf, rather, XAX4 single-board computer, 10:55.000 --> 10:58.000 that is X86 based, 10:58.000 --> 11:03.000 so there's a proper Intel and 100 processor in there, 11:03.000 --> 11:08.000 so that we can make use of the video encoder that's inside, 11:08.000 --> 11:13.000 and they have achieved the exact Raspberry Pi layout, 11:13.000 --> 11:17.000 so the board has the exact phone factor as a Raspberry Pi, 11:17.000 --> 11:20.000 and also the GPIO header, 11:20.000 --> 11:27.000 and the way they implement that is that they have an onboard RP2040 microcontroller, 11:27.000 --> 11:31.000 that actually controls the GPIO, 11:31.000 --> 11:36.000 and is connected to an internal USB to the Intel processor. 11:36.000 --> 11:41.000 And we use that in order to control different things around the box, 11:41.000 --> 11:45.000 like the fans, the display, that's on the front, 11:45.000 --> 11:50.000 and to monitor status. 11:50.000 --> 11:55.000 And the display basically looks like this. 11:55.000 --> 12:00.000 It's an off the shelf, we've shared this play, 12:00.000 --> 12:03.000 that has three components left, middle and right, 12:03.000 --> 12:06.000 and basically we showed the status of the box, 12:06.000 --> 12:09.000 the video screenshot, and things like that, 12:09.000 --> 12:13.000 and also levels from the audio board, 12:13.000 --> 12:18.000 so that audio can easily be debugged. 12:18.000 --> 12:24.000 And yeah, this is also connected to the GPIO header, 12:24.000 --> 12:29.000 so it's images are sent to the display over SPI. 12:29.000 --> 12:34.000 And one of the more challenging aspects of the entire box 12:34.000 --> 12:38.000 was doing the whole mechanical assembly in a way 12:38.000 --> 12:43.000 that allowed us to produce more than 70 of these devices 12:43.000 --> 12:52.000 and just several days of assembly lines done by volunteers, 12:52.000 --> 12:55.000 by both the volunteers, 12:55.000 --> 13:00.000 and the whole thing is completely open source, 13:00.000 --> 13:03.000 hardware mechanically, 13:03.000 --> 13:07.000 and also all of the PCBs inside a open source 13:07.000 --> 13:10.000 apart from the HDMI capture card, 13:10.000 --> 13:14.000 which we're planning to open source next year. 13:14.000 --> 13:19.000 And you can view both the mechanical 13:19.000 --> 13:23.000 and the PCB designs on GitHub, 13:23.000 --> 13:26.000 including the aluminium base plate, 13:26.000 --> 13:30.000 that's on the bottom of each box that was designed by us, 13:30.000 --> 13:35.000 and then manufacture by laser cutting. 13:35.000 --> 13:41.000 Also we had some interesting things we have to do, 13:41.000 --> 13:44.000 while manufacturing, for example, 13:45.000 --> 13:51.000 we had to do sort of assembly lines for assembling the box, 13:51.000 --> 13:53.000 and for preparing different components. 13:53.000 --> 13:58.000 We had to disassemble the off-the-shelf capture cards 13:58.000 --> 14:01.000 in order to get the PCBs out of them, 14:01.000 --> 14:07.000 and also the RADXA computers had the wrong whole diameter, 14:07.000 --> 14:10.000 just by 0.1 millimetre, 14:10.000 --> 14:15.000 so we had to drill through all 420 holes 14:15.000 --> 14:20.000 on all the RADXA computers we have. 14:20.000 --> 14:24.000 So I will now show you a video from the volunteers assembly line 14:24.000 --> 14:28.000 we did while drilling the holes. 14:40.000 --> 14:52.000 Yeah, and to do the whole assembly process, 14:52.000 --> 14:54.000 this had to be planned in advance, 14:54.000 --> 14:59.000 and we had made this flow chart that lets us decide 14:59.000 --> 15:02.000 what to do before the next task, 15:02.000 --> 15:06.000 and basically in the end we have an assembled box. 15:07.000 --> 15:13.000 Yes, I think that said, get heavy. 15:13.000 --> 15:18.000 Well, so now you have a little bit of an idea of what is in the box, 15:18.000 --> 15:22.000 but I really want to emphasize these guys they've done 15:22.000 --> 15:24.000 the most wonderful job. 15:24.000 --> 15:27.000 I had a lot of ideas and suggestions that, 15:27.000 --> 15:29.000 hey, maybe we should add this or that, 15:29.000 --> 15:32.000 or even in the middle of the design station, 15:32.000 --> 15:35.000 maybe we should add a post through USB 3 port, 15:35.000 --> 15:37.000 because there's another conference, 15:37.000 --> 15:40.000 which uses a slightly different set-up from Fauston, 15:40.000 --> 15:42.000 and I thought by adding that port, 15:42.000 --> 15:44.000 we can also support their use case. 15:44.000 --> 15:48.000 So these guys have been amazing. 15:48.000 --> 15:51.000 I mean, they always say they don't know what they're doing, 15:51.000 --> 15:53.000 but I suspect they're alive, 15:53.000 --> 15:57.000 because they came up with this all by themselves, 15:57.000 --> 15:59.000 so it's seriously, thank you guys, thank you so much. 15:59.000 --> 16:09.000 And so one of the things that we've been, 16:09.000 --> 16:13.000 I mean, one of the ideas behind this box is that, 16:13.000 --> 16:15.000 I mean, it's as open sources, 16:15.000 --> 16:17.000 we could get it at this stage, 16:17.000 --> 16:20.000 there's still the rough set, 16:20.000 --> 16:24.000 which is central part, 16:24.000 --> 16:28.000 but then there's the HDMI capture card, 16:28.000 --> 16:31.000 as Dexter mentioned. 16:31.000 --> 16:34.000 We, I mean, my friend has a reference, 16:34.000 --> 16:38.000 as already assembled some kind of design to replace that, 16:38.000 --> 16:40.000 because that's currently still an off-the-shelf part 16:40.000 --> 16:42.000 for which we don't really have all the designs, 16:42.000 --> 16:45.000 and we've not really reverse engineered to firmware yet, 16:45.000 --> 16:47.000 but we're going to get there next year, 16:47.000 --> 16:50.000 and so next year there's going to be even more, 16:50.000 --> 16:53.000 so I'll start with one of the ideas 16:53.000 --> 16:56.000 that we have is that we want to encourage 16:56.000 --> 16:58.000 all our conferences to use our hardware, 16:58.000 --> 17:00.000 so we've built extra boxes, 17:00.000 --> 17:04.000 and we'll be lending them out of the given them away, 17:04.000 --> 17:06.000 in some cases, because it makes more sense, 17:06.000 --> 17:08.000 and just shipping it back and forth all the time, 17:08.000 --> 17:11.000 to all our conferences who can then hack 17:11.000 --> 17:13.000 and improve all our setup, 17:13.000 --> 17:16.000 so that's one of the design that we have here. 17:16.000 --> 17:21.000 I think there's maybe some time for questions, 17:21.000 --> 17:23.000 if you have questions. 17:23.000 --> 17:24.000 Sure? 17:24.000 --> 17:27.000 One question, would you have the design? 17:27.000 --> 17:32.000 Is it possible to display the rules, 17:32.000 --> 17:36.000 to project some of the ways that comes from 17:36.000 --> 17:39.000 a video on your room, right? 17:39.000 --> 17:40.000 If we go into an architect, 17:40.000 --> 17:42.000 if we represent a computer, 17:42.000 --> 17:45.000 it's just as deep there after the computer, 17:45.000 --> 17:49.000 or we can output from the controller to the computer. 17:49.000 --> 17:53.000 Currently the video is going straight 17:53.000 --> 17:56.000 into the capture card, and it has a loop output, 17:56.000 --> 17:58.000 and that is used for the project, 17:58.000 --> 18:01.000 or so we cannot really do anything with a signal. 18:01.000 --> 18:04.000 But what we have designed now for it, 18:04.000 --> 18:06.000 that can maybe replace it, 18:06.000 --> 18:10.000 has an HDMI switch included after the loop out, 18:10.000 --> 18:14.000 so we can decide whether we show the feed from the laptop, 18:14.000 --> 18:17.000 or show the feed from the internal computer, 18:17.000 --> 18:20.000 so we can at least have something on the screen 18:20.000 --> 18:23.000 when there's no percent of the laptop floating. 18:23.000 --> 18:26.000 I don't actually have an overlay of stuff over there, 18:26.000 --> 18:27.000 that will be really hard, 18:27.000 --> 18:30.000 because that would require some feed of mixing hardware, 18:30.000 --> 18:32.000 and that gets quite expensive. 18:32.000 --> 18:35.000 Would you have sorry to get the question? 18:35.000 --> 18:37.000 Oh yes, the question was, 18:37.000 --> 18:43.000 can the current hardware do overlay all the projected image? 18:43.000 --> 18:46.000 So we can read, well, we could, 18:46.000 --> 18:48.000 with the next board replace, 18:48.000 --> 18:51.000 the projected image, but I'm overlaying something, 18:51.000 --> 18:52.000 so it's not in the, 18:52.000 --> 18:54.000 do we have a use case for it? 18:54.000 --> 18:56.000 I mean, it's something that you would point. 18:56.000 --> 18:59.000 We want the attendance, 18:59.000 --> 19:00.000 what? 19:00.000 --> 19:02.000 All right. 19:02.000 --> 19:04.000 Okay. 19:04.000 --> 19:07.000 So far, these guys have never said that something isn't possible, 19:07.000 --> 19:13.000 so we did test doing, 19:13.000 --> 19:17.000 actually we managed to do an overlay by accident, 19:17.000 --> 19:20.000 during the, 19:20.000 --> 19:23.000 because we were testing, 19:23.000 --> 19:28.000 if we can get the video go through the Rode XA 19:28.000 --> 19:30.000 through an elephant peg pipeline, 19:30.000 --> 19:34.000 and then out of its port with minimal latency, 19:34.000 --> 19:39.000 and we could get it down to 500 milliseconds of latency, 19:39.000 --> 19:45.000 and at some point we accidentally blended with elephant peg, 19:46.000 --> 19:49.000 an image onto the video. 19:49.000 --> 19:52.000 So it should be possible with this CPU, 19:52.000 --> 19:58.000 but I don't think latency will be good enough for a generic use case. 19:58.000 --> 20:00.000 Maybe for remote talks, 20:00.000 --> 20:01.000 it will be fine, 20:01.000 --> 20:05.000 if you manually delete the audio by this amount, 20:05.000 --> 20:06.000 but yeah, 20:06.000 --> 20:08.000 it would be Jankey. 20:08.000 --> 20:10.000 Oh yeah? 20:10.000 --> 20:11.000 We'll have to repeat it, 20:11.000 --> 20:14.000 but you might want to look at the, 20:14.000 --> 20:17.000 the Rode 5B plus, 20:17.000 --> 20:20.000 maybe eventually it's a bit more expensive, 20:20.000 --> 20:24.000 but it has its own HDMI capture. 20:24.000 --> 20:25.000 Okay. 20:25.000 --> 20:28.000 That I got about 2.0. 20:28.000 --> 20:30.000 It does have, 20:30.000 --> 20:31.000 it's my capture, 20:31.000 --> 20:33.000 but it's an arm platform, 20:33.000 --> 20:36.000 so video encoding on it is pretty hidden miss, 20:36.000 --> 20:41.000 and it's usually the last thing they had support for on Linux. 20:42.000 --> 20:46.000 So all of the vendor stack on it, 20:46.000 --> 20:48.000 it's open source and media. 20:48.000 --> 20:50.000 There's no blocks in this one. 20:50.000 --> 20:52.000 So it's a good one. 20:52.000 --> 20:55.000 And there's many people that are at the conference, 20:55.000 --> 20:58.000 that are actually working this year on the encoder, 20:58.000 --> 21:01.000 so we can have mainline Linux support for them. 21:01.000 --> 21:03.000 This is going to be, 21:03.000 --> 21:05.000 I say look at it, 21:05.000 --> 21:08.000 because this is going to be a platform we will support, 21:08.000 --> 21:12.000 and that for anything of related WebR-Dc's remains, 21:12.000 --> 21:14.000 but it's going to be a good platform. 21:14.000 --> 21:17.000 Okay, so what was the exact board you were proposing, 21:17.000 --> 21:20.000 so that I can be to the top 5B plus plus, 21:20.000 --> 21:22.000 it's just a fixed version of this. 21:22.000 --> 21:24.000 Okay, the Rode 5B plus, 21:24.000 --> 21:26.000 so we might have missed that one, 21:26.000 --> 21:29.000 but basically for the last five years, 21:29.000 --> 21:33.000 every year we've been doing this research bit, 21:33.000 --> 21:35.000 that's basically, 21:35.000 --> 21:37.000 is on there yet, 21:37.000 --> 21:39.000 in terms of encoding, 21:39.000 --> 21:42.000 and usability for encoding, 21:42.000 --> 21:44.000 and the... 21:44.000 --> 21:46.000 Sorry? 21:46.000 --> 21:50.000 So the question was which code that we use, 21:50.000 --> 21:53.000 we're currently doing H264, 21:53.000 --> 21:57.000 but we might want to move to something more efficient 21:57.000 --> 21:59.000 and more open in the future. 21:59.000 --> 22:03.000 But, yes, so far the verdict has been 22:04.000 --> 22:08.000 that our support is still janky enough for it to cause it, 22:08.000 --> 22:11.000 to cause issues during something like pause them, 22:11.000 --> 22:14.000 and we're not ready to commit to that yet, 22:14.000 --> 22:17.000 but maybe in a couple of years, 22:17.000 --> 22:22.000 if the guys who support this drivers 22:22.000 --> 22:26.000 stuck get it toward reliably enough. 22:26.000 --> 22:29.000 Well, look at that, I think it's a bit... 22:29.000 --> 22:30.000 Yes, I think. 22:31.000 --> 22:35.000 This is the worksheet RK3588, right? 22:35.000 --> 22:36.000 Correct. 22:36.000 --> 22:39.000 Okay, yes, the year ago when we looked at the support 22:39.000 --> 22:41.000 was not there yet, at least. 22:41.000 --> 22:45.000 Yeah, having mainline, 22:45.000 --> 22:49.000 then the support is our requirement that we've been posed 22:49.000 --> 22:50.000 on this project, 22:50.000 --> 22:53.000 and that's why we're using X86 at the moment. 22:53.000 --> 22:56.000 Yes, there's also a problem with that. 22:57.000 --> 23:00.000 I'm curious what is the total post of part of this process? 23:00.000 --> 23:04.000 So the question was what the total cost of the boxes, 23:04.000 --> 23:06.000 this one's for Harry? 23:08.000 --> 23:11.000 So it's not easy to give you an exact number, 23:11.000 --> 23:13.000 because so many things happened in different phases, 23:13.000 --> 23:17.000 and we've recycled hardware and incurred extra cost 23:17.000 --> 23:23.000 because this was a rush production phase. 23:23.000 --> 23:27.000 I think some things could have been made a little bit cheaper. 23:27.000 --> 23:32.000 I think one box will sit around maybe 450 euros, 23:32.000 --> 23:34.000 450 euros, something like that. 23:34.000 --> 23:37.000 Considering that it includes everything 23:37.000 --> 23:40.000 and the kitchen sink, it's not too bad. 23:42.000 --> 23:45.000 But the problem is we need to scale it to 30 rooms, 23:45.000 --> 23:48.000 so every year we spend this at least times 30. 23:48.000 --> 23:50.000 Thank you so much. 23:52.000 --> 23:55.000 Any feedback on the power plane? 23:55.000 --> 23:57.000 You didn't use the earlier name? 23:57.000 --> 23:58.000 Well, nine. 23:58.000 --> 24:01.000 Okay, so the question is about the power lanes? 24:01.000 --> 24:03.000 I will give that to the extra. 24:03.000 --> 24:08.000 Okay, so you're asking what voltages 24:08.000 --> 24:11.000 and what power supplies we have. 24:11.000 --> 24:22.000 Well, I would say they weren't like huge roadblocks, 24:22.000 --> 24:27.000 but they were like lots of small issues to go through. 24:27.000 --> 24:33.000 For example, one of the start, 24:33.000 --> 24:39.000 we were trying to see what kind of power breaks to use, 24:39.000 --> 24:44.000 and we were using another version of the RadexA computer 24:44.000 --> 24:47.000 that was only powered by USB-C, 24:47.000 --> 24:51.000 but by a fake version of USB-C power delivery 24:51.000 --> 24:55.000 that actually expects 12 volts, 24:55.000 --> 24:57.000 asks for 12 volts, 24:57.000 --> 25:00.000 but doesn't check if it has got 12 volts. 25:00.000 --> 25:05.000 And then works best if you supply it with nine volts instead. 25:06.000 --> 25:12.000 But apart from these issues with the current version, 25:12.000 --> 25:17.000 we basically are using a version of the computer 25:17.000 --> 25:22.000 that can also work from just 12 volts on the header 25:22.000 --> 25:24.000 and that's fine. 25:24.000 --> 25:31.000 We also had some issues related to filtering the power to the fans, 25:32.000 --> 25:36.000 but that's just normal. 25:36.000 --> 25:40.000 We have lots of regulators all over the place 25:40.000 --> 25:44.000 because every single component needs different voltages. 25:44.000 --> 25:50.000 I think the most exotic one is the 1.1 volts for the switch chip. 25:51.000 --> 25:57.000 And we also have done multiple experiments 25:57.000 --> 26:04.000 with filtering and eliminating ground loops with the audio board 26:04.000 --> 26:07.000 while keeping the cost down, 26:07.000 --> 26:11.000 and at the end with that, I think. 26:11.000 --> 26:15.000 In the end, we just decided to power it from USB 26:15.000 --> 26:17.000 and just filter it well, 26:17.000 --> 26:22.000 and that had the best, like, price-performance characteristics 26:22.000 --> 26:24.000 in terms of power supply. 26:24.000 --> 26:26.000 So yeah, it was okay. 26:26.000 --> 26:30.000 There weren't major old blocks there. 26:42.000 --> 26:46.000 I think it actually worked. 26:48.000 --> 26:54.000 I think overall, things went really well. 26:54.000 --> 26:59.000 I haven't heard any major issues with the video setup. 26:59.000 --> 27:02.000 The hardware has been stellar. 27:02.000 --> 27:06.000 I mean, we've identified a couple of areas for improvement, 27:06.000 --> 27:13.000 but basically for what is the very first production run. 27:13.000 --> 27:18.000 This has been excellent. 27:44.000 --> 27:48.000 The other question is, if the network chip on the board 27:48.000 --> 27:52.000 can be used to make a few networking stuff more efficient 27:52.000 --> 27:57.000 like ICMP, multicasting, or IGMP, multicasting, and freelance, 27:57.000 --> 28:00.000 and yes, this chip does support it. 28:00.000 --> 28:03.000 We are not using it for this year, at least. 28:03.000 --> 28:07.000 For now, that's only reading out the status from the board, 28:07.000 --> 28:10.000 but it is possible to make a fully managed switch 28:10.000 --> 28:12.000 out of this switch support for all the features. 28:12.000 --> 28:16.000 I don't think that multicast streaming will be really, 28:16.000 --> 28:20.000 that you will fall on for STEM, because it's basically all the boxes 28:20.000 --> 28:23.000 are doing just a single stream to the operations center, 28:23.000 --> 28:25.000 and there, I think, it's handled. 28:25.000 --> 28:28.000 So we don't have to split any of the signals anywhere. 28:28.000 --> 28:32.000 So multicast doesn't really help in our case, I think. 28:32.000 --> 28:35.000 Let's do this. 28:35.000 --> 28:40.000 QOS, I think, only starts helping if we overload links 28:40.000 --> 28:43.000 or get closer offloading it, and that should not really happen 28:43.000 --> 28:45.000 with a network infrastructure. 28:45.000 --> 28:47.000 All of this is run on a dedicated network, 28:47.000 --> 28:51.000 so we basically have gigabit and to end 28:51.000 --> 28:56.000 from every room to the encoding laptop. 28:56.000 --> 28:59.000 It's not really an issue in our case, 28:59.000 --> 29:03.000 but yeah, it's an option if everything becomes necessary. 29:03.000 --> 29:05.000 We didn't want to add the complication at this point, 29:05.000 --> 29:08.000 because we didn't really see a USB thread, but sure. 29:08.000 --> 29:11.000 I mean, the good thing about this design is that it's completely modular, 29:11.000 --> 29:14.000 so if one day we don't like a certain component, 29:14.000 --> 29:17.000 we just swap it out and don't have to replace the entire box. 29:17.000 --> 29:21.000 So anyone else? 29:21.000 --> 29:23.000 These are some special solutions, 29:23.000 --> 29:26.000 we don't improve objects, we just manage. 29:26.000 --> 29:31.000 So I mean, it's more fine, but sometimes, 29:31.000 --> 29:34.000 I notice it's easier, I don't know, 29:34.000 --> 29:36.000 especially for nice. 29:36.000 --> 29:39.000 So I just need to see what's the source of this problem 29:39.000 --> 29:41.000 and how can we please? 29:41.000 --> 29:45.000 Right, so the question is about the sometimes, 29:45.000 --> 29:48.000 it happens that there is some decentralization 29:48.000 --> 29:50.000 between the audio and the video, 29:50.000 --> 29:52.000 and where that comes from, 29:52.000 --> 29:54.000 and what the source of the problem is. 29:54.000 --> 29:56.000 So I think texture can answer that one. 29:56.000 --> 30:03.000 So the problem with the audio video, sorry. 30:04.000 --> 30:08.000 Synchronization between the two streets, 30:08.000 --> 30:12.000 the speaker slides, yeah. 30:12.000 --> 30:16.000 Oh, between the speaker and the speaker. 30:16.000 --> 30:19.000 This should not be a problem. 30:19.000 --> 30:23.000 So the image capture by the camera, 30:23.000 --> 30:26.000 or the camera, for example, today, 30:27.000 --> 30:31.000 is presentation about, 30:31.000 --> 30:34.000 and people contain them. 30:34.000 --> 30:36.000 It was quite noticeable. 30:36.000 --> 30:38.000 Okay, so this, 30:38.000 --> 30:41.000 and it's a UBC. 30:41.000 --> 30:43.000 Yeah, it is. 30:43.000 --> 30:47.000 Basically, I think that does not implement the timestamp. 30:47.000 --> 30:52.000 From there, so it's unable to remove the delay from the receiver, 30:52.000 --> 30:55.000 and even if you do it, it's not always working. 30:56.000 --> 31:02.000 Okay, so, what we got was that, 31:02.000 --> 31:07.000 FMPEG has an issue with synchronizing UBC streams 31:07.000 --> 31:09.000 from video for Linux, 31:09.000 --> 31:11.000 which is the case, 31:11.000 --> 31:14.000 but we have another issue with the re-streamer. 31:14.000 --> 31:17.000 That's completely unrelated to the boxes 31:17.000 --> 31:20.000 that is basically, 31:21.000 --> 31:25.000 when the network drops between the box and the re-streamer, 31:25.000 --> 31:28.000 the re-streamer catches up in a weird way 31:28.000 --> 31:32.000 that sometimes these synchronizes, 31:32.000 --> 31:35.000 and this is just a software problem. 31:35.000 --> 31:40.000 It's not something that is a problem in general, 31:40.000 --> 31:43.000 it's something that we just haven't got to fixing. 31:43.000 --> 31:46.000 And the other such thing that we haven't fixed yet, 31:46.000 --> 31:51.000 is that the slides and the other part of the talker 31:51.000 --> 31:54.000 are not synchronized at all, 31:54.000 --> 31:57.000 and that's just work to be done in the future. 31:57.000 --> 32:01.000 Those are two separate streams that just need to be synchronized 32:01.000 --> 32:04.000 in the video center. 32:04.000 --> 32:07.000 Yes, all of them are connected. 32:07.000 --> 32:10.000 Yes, all of them are connected. 32:10.000 --> 32:13.000 Yes, all of them are connected. 32:14.000 --> 32:17.000 Yes, all of them are connected. 32:17.000 --> 32:18.000 Yeah. 32:18.000 --> 32:19.000 The pt question? 32:19.000 --> 32:20.000 Oh yeah. 32:20.000 --> 32:22.000 So the question is, 32:22.000 --> 32:24.000 is there software on the box, 32:24.000 --> 32:26.000 and is it available online? 32:26.000 --> 32:30.000 Yes, everything that runs on the box 32:30.000 --> 32:35.000 is available at github.com slash false-dem slash video. 32:35.000 --> 32:39.000 And all of the software that we're written, 32:40.000 --> 32:45.000 so there is firmware for two of the hardware components, 32:45.000 --> 32:48.000 the RP2040 and the teensy. 32:48.000 --> 32:51.000 That's inside the repo public. 32:51.000 --> 33:00.000 And also, we have some software that runs on the RabyxA itself. 33:00.000 --> 33:03.000 It's mostly written in Python, 33:03.000 --> 33:07.000 and it's available, part of it is available in that video repo, 33:07.000 --> 33:11.000 and another part is available in infrastructure repo. 33:11.000 --> 33:15.000 And there is also some software that we've made 33:15.000 --> 33:17.000 to control Vockton Mix, 33:17.000 --> 33:20.000 which is the software we used to mix this streams 33:20.000 --> 33:22.000 in the video center, 33:22.000 --> 33:26.000 and this is also public and the infrastructure repo. 33:30.000 --> 33:31.000 Yes. 33:31.000 --> 33:33.000 There's certainly plans to replace the video camera 33:33.000 --> 33:35.000 without any source. 33:36.000 --> 33:39.000 So the question is, 33:39.000 --> 33:42.000 are there any plans to replace the video camera 33:42.000 --> 33:44.000 with an open source one? 33:44.000 --> 33:47.000 We haven't actually talked of that, 33:47.000 --> 33:50.000 but maybe more than can elaborate on 33:50.000 --> 33:53.000 the status of open source video cameras, 33:53.000 --> 33:54.000 and if they exist. 33:54.000 --> 33:57.000 It is very hard to make an open source video camera. 33:57.000 --> 33:59.000 The hardware is messy, 33:59.000 --> 34:00.000 the software is messy. 34:00.000 --> 34:03.000 A color science is really, really hard. 34:04.000 --> 34:08.000 So I don't think it will be really close 34:08.000 --> 34:12.000 in the future that first-hand can use open source cameras for it, 34:12.000 --> 34:15.000 and especially cheap ones. 34:15.000 --> 34:17.000 It is pretty hard to get hardware for it, 34:17.000 --> 34:19.000 especially if you want something that's comparable 34:19.000 --> 34:21.000 to the current cameras here, 34:21.000 --> 34:23.000 that has a pretty large sensor, 34:23.000 --> 34:27.000 and even getting the parts for that is expensive. 34:27.000 --> 34:29.000 It's more cheap to get a camera. 34:29.000 --> 34:31.000 Can you remove the sensor? 34:31.000 --> 34:34.000 Actually, it's so much cheaper. 34:34.000 --> 34:36.000 The cameras are actually not owned by us, 34:36.000 --> 34:37.000 so we rent these cameras, 34:37.000 --> 34:39.000 most of them we rent, 34:39.000 --> 34:42.000 so that's one part that we don't own, 34:42.000 --> 34:46.000 so we didn't really have the need to look to deeply into it. 34:46.000 --> 34:48.000 Priorities first. 34:48.000 --> 34:50.000 Anybody else? 34:50.000 --> 34:51.000 Yeah, I don't know. 34:51.000 --> 34:52.000 Can I ask you a question? 34:52.000 --> 34:54.000 Can I ask you a question? 34:54.000 --> 34:55.000 Can I ask you a question? 34:55.000 --> 34:57.000 Can I ask you a question? 34:57.000 --> 34:58.000 Can I ask you a question? 34:58.000 --> 35:00.000 Can I ask you a question? 35:00.000 --> 35:01.000 Sure, I mean, I mean, I'm... 35:01.000 --> 35:02.000 I can avoid that, just post it, 35:02.000 --> 35:03.000 but something so big. 35:03.000 --> 35:08.000 Yeah, so the question was other updates of the release papers of this. 35:08.000 --> 35:10.000 And not that much, 35:10.000 --> 35:12.000 but I have all my physical block 35:12.000 --> 35:16.000 of you post yourself about various components of this box, 35:16.000 --> 35:21.000 and I don't have a nice light with the request, 35:21.000 --> 35:25.000 but maybe can be in the metadata for the talk. 35:25.000 --> 35:27.000 It's fine, something else valuable? 35:27.000 --> 35:28.000 Yes, this does that. 35:28.000 --> 35:39.360 just do that. It has a lot of unrelated stuff on it. It's blockedobricid.no. This is my 35:39.360 --> 35:51.800 company website. This is my blog. Let's see if I can scroll and find something for 35:51.800 --> 36:00.200 this is the switch. This is the early iteration. Okay. So this is actually connecting 36:00.200 --> 36:06.200 the prototype for false them up to an arm computer because we had Linux magic. 36:06.200 --> 36:12.000 It is possible to integrate it in such a way that the four external ports here show up 36:12.000 --> 36:18.160 as actual network interfaces in Linux. And those are fully functional network 36:18.160 --> 36:22.160 interface and not like if you have feelings for that and you circle interfaces. 36:22.160 --> 36:28.960 So all the tools like F2 just work on it and you can see the port status and change 36:28.960 --> 36:36.840 like the configuration of if it should be automatically gigabit or not. And if you 36:36.840 --> 36:43.120 then take those ports and put them in the Linux bridge together then the Linux will 36:43.120 --> 36:48.160 configure the switch that those ports will be handled in hardware. So it's hard to 36:48.160 --> 36:53.200 actually write the networking in that way. And that was not needed at all for 36:53.200 --> 36:58.120 false them. But it is possible and I wanted to try it. 36:58.120 --> 37:04.520 Are you relying on the university network or are you using your own custom network? 37:04.520 --> 37:09.200 So the question is whether we're relying on the university network or using our own. 37:09.200 --> 37:13.520 We are mostly relying on the university network. So we have a very close cooperation 37:13.520 --> 37:18.960 with the university network team. And basically they reconfigure half of the network 37:18.960 --> 37:24.000 for us just for the weekend. They have permanent configuration, you know, just partly 37:24.000 --> 37:30.320 disabled, partly just enabled all the time just for this one weekend. So those guys 37:30.320 --> 37:36.000 also really great. Without them we couldn't do this at all. So we use their infrastructure 37:36.080 --> 37:41.920 but we do all of our own. So all of the core is ours. We just back hold everything to our 37:41.920 --> 37:49.280 core and we do our stuff, our own stuff here. Anything else? 37:51.280 --> 37:54.400 Okay. Well, thank you for being here. I'm really glad. 37:55.200 --> 37:56.400 Thank you. 38:02.400 --> 38:05.200 I'm also very glad that Fossil is finally over. 38:05.200 --> 38:09.200 And we're almost almost just a couple of times.