**** BEGIN LOGGING AT Sat Oct 12 02:59:57 2019 Oct 12 03:13:06 . Oct 12 03:16:41 , Oct 12 03:17:08 and all that Oct 12 03:17:10 :-) Oct 12 03:17:21 having fun Oct 12 03:17:35 im trying to decide what microphones to try to attach Oct 12 03:17:38 to the beagele Oct 12 03:31:31 what are the options? I know PDM microphones are ubiquitous, and is basically the only option that doesn't require an ADC (for which you'd probably not want to use the built-in ADC :P ) Oct 12 03:32:11 oh, i2s microphones actually exist too... Oct 12 03:44:11 (but those seem quite rare compared to pdm microphones) Oct 12 03:44:35 https://learn.adafruit.com/adafruit-i2s-mems-microphone-breakout Oct 12 03:45:15 PDM into the PRU Oct 12 03:45:36 mems into the pru :p Oct 12 03:49:27 pdm into pru is one option yeah Oct 12 03:51:55 and a pretty good one if you don't mind a slightly oddball samplerate instead of one of the standard ones (i.e. 48 kHz or 44.1 kHz or some power-of-two scaled version thereof) Oct 12 03:52:12 although you could get standard rates with a bit more effort Oct 12 03:52:14 noi for my project, i dont' mind at all Oct 12 03:52:40 I'm wanting to sample for several minuts and then do FFT on the sample Oct 12 03:53:22 so, especially for a single microphone, if you want fast results then an i2s microphone would definitely be easier Oct 12 03:53:42 fast results is not particularly important Oct 12 03:53:49 that would just require a bit of DT setup and you'd have a linux alsa device to record from Oct 12 03:57:59 if you only want to do an FFT, then assuming you don't actually want to do an FFT across the entire recording (which would be an FFT size in the millions of samples), when using PRU you could also have it fold the data to the FFT size for you, which avoids the need for offloading a large-ish amount of data Oct 12 03:58:50 yes. i dont' need to record the sample Oct 12 03:58:59 i just need the spectral analysis of it Oct 12 04:04:34 right, so perform decimation, apply analysis window, and then record the sample with something equivalent to buf[ i++ % BUFSIZE ] += sample; where buf is zeroed in advance (and uses integers of sufficient width to avoid overflow during the recording time) Oct 12 04:04:50 applying the window seems like the main annoyance Oct 12 04:07:26 (since using a minutes-long precomputed window would completely negate any benefit of doing this pre-processing in PRU, but options to produce a window on the fly are limited...) Oct 12 04:12:14 I guess just recording minutes of audio to a buffer would actually be the easier option after all :P it's not like it's *that* much data... you probably won't even need a ringbuffer or whatever, a static buffer would suffice Oct 12 04:16:29 the built in ADC is uh... tolerable Oct 12 04:17:00 what spectural ersolution do you need? Oct 12 04:17:14 and what bandwidth Oct 12 04:17:40 i suspect i'm not interested in anything > 5kHz Oct 12 04:17:42 depending on how fine, it may be easier to implement a continuous parallel set of filters Oct 12 04:17:43 but i'm not sure yet Oct 12 04:18:00 and ds2.. again not sure. Oct 12 04:18:36 doing FT's in chunks has its own can of worms Oct 12 04:18:56 a "continuous parallel set of filters" sounds like an inefficient DFT implemented in a way that has poor numerical accuracy :P Oct 12 04:19:13 uh no Oct 12 04:19:29 there are algs for computing it w/o a FT of any sort Oct 12 04:19:46 but the bins can be a bit course Oct 12 04:20:25 an N-point DFT can be implemented as N parallel first-order complex (or second-order real) IIR filters Oct 12 04:20:46 yes except you have gibbs doing things to do you Oct 12 04:20:50 it's just computationally expensive and numerically crappy Oct 12 04:21:07 it isn't the computation as much as it is annoying artificats Oct 12 04:21:13 artifacts Oct 12 04:21:30 the size of the sample is also up for grabs Oct 12 04:21:39 just dont' know yet how much will need to be sampled to be useful Oct 12 04:21:51 those are not really dependent on how the DFT is implemented and need to be tamed with windowind either way Oct 12 04:21:56 *windowing Oct 12 04:22:28 windowing isn't needed for a parallel stack of filters Oct 12 04:23:01 I'm curious why you think that, enlighten me :P Oct 12 04:23:59 FIR filters don't need to be windowed Oct 12 04:24:02 nor do IIR Oct 12 04:24:47 but that only works if you don't need a lot of resolution. think the original DTMF decoders or those spectural displays on the "HiFi"'s from the 80's Oct 12 04:25:55 sometimes it makes sense to work in the time domain :D Oct 12 04:27:29 resolution isn't really the problem, it's more that depending on what sort of filter you use I'm not even sure what you're really measuring :P there's some implicit windowing going on in the decay of the filter (the filter corresponding to a DFT has no decay, which is what you'd expect for treating the entire analysis period uniformly), but it's not a great window Oct 12 04:28:05 regardless, it sounds more complicated *and* more computationally expensive than using a windowed FFT Oct 12 04:28:18 so, given that "spectral analysis" was the stated goal, it doesn't seem like a useful option Oct 12 04:29:44 it has its uses, like you said DTMF decoders and such, or spectral displays implemented in hardware Oct 12 04:30:47 (although there are better ways for DTMF decoding nowadays) Oct 12 04:45:49 (actually not really, the algorithm wikipedia mentioned fot DTMF is just the obvious IIR filter) Oct 12 06:23:03 Hello everyone Oct 12 06:26:00 I am using BB X-15 . The first time When I am booting it the serial message is Starting kernel ... Oct 12 06:26:05 and stops Oct 12 06:26:25 wht should I do ? Oct 12 06:27:23 And I am using to supply it 12V 1A adapter Oct 12 06:28:16 self test results are like these Oct 12 06:28:19 U-Boot SPL 2019.07-rc4-00001-g607b5b738b (Jul 06 2019 - 21:01:15 -0500) Oct 12 06:34:59 Starting kernel ... Oct 12 06:35:21 Starting kernel ... Oct 12 06:45:33 any one wants to answer ? Oct 12 08:49:33 wtf this is the dumbest code I've seen gcc generate in quite some time... https://pastebin.com/zt6eaZyk Oct 12 08:50:25 I'm returning an element of a float vector, and it decided it's a good idea to move that element to an integer register and back to (the same!) float register Oct 12 09:07:25 yup, the instrinsic I used isn't really important... it seems that all that matters is that the vector is initialized in a way that's a bit opaque to the optimizer, and it spontaneously forgets how to return a float from a float-vector... https://pastebin.com/4t0JgUqA Oct 12 09:07:45 this is gross, this creates a massive pipeline stall on the cortex-a8 Oct 12 09:13:21 ah it doesn't even have to be asm/instrinsic, any vector-operation seems to suffice Oct 12 13:03:16 m Oct 12 14:29:50 is the x15 discontinued? Can't source it anywhere... :- Oct 12 14:29:52 :-( Oct 12 14:30:37 shouldn't be Oct 12 14:33:49 hmm Oct 12 14:35:49 oh yeah looks like it switched manufacturer? mouser has two versions of the x15: one discontinued, one "new from this manufacturer" .. for the latter they have 494 units on order (expected about a month from now) => https://www2.mouser.com/ProductDetail/?qs=mAH9sUMRCtul7OWq6qLH4g%3D%3D Oct 12 14:52:44 yes. I saw that too. but still 20 weeks till delivery to mouser or digikey Oct 12 14:52:58 I am looking for 4 PRU model, like the x15 Oct 12 14:53:17 thank you zmatt Oct 12 14:54:06 20 weeks? mouser says estimated ship date: 18 Nov 2019 Oct 12 14:54:06 I wonder if the BB-Ai board with the "dual-core" PRU's might meet our needs. The USB-C is great Oct 12 14:54:12 that's 5 weeks from now Oct 12 14:54:29 the beaglebone AI has the same processor as the beagleboard x15 Oct 12 14:54:33 same SoC Oct 12 14:54:39 hence same PRUs Oct 12 14:56:22 hmmm. thank you for the info zmatt. Oct 12 14:57:48 I get the confusion btw: the "20 weeks" is the factory lead time, but mouser already has 494 units on order, and those are expected to arrive 5 weeks from now Oct 12 14:58:06 they don't show that clearly, you have to click on "view delivery dates" Oct 12 16:09:12 Howdy - brand new BeagleBone AI tinkerer here with a boot question. The board boots when the USB-C cable is connected to my laptop, and I'm able to access the Cloud9 interface, but when I attempt to power the board from a phone charger, the PWR LED comes on but nothing else happens. Is that expected? Oct 12 16:10:15 sounds odd Oct 12 16:11:36 unfortunately since it's such a new board, there's not yet much known about things that can go wrong and how to fix them :) Oct 12 16:11:58 That's what I thought as well. Oh well, for now I'll power off the laptop. Trying to get the software updated by connecting to WiFi, we'll see how it goes. Awesome to see so many people here. Thanks zmatt. Oct 12 16:12:51 what's the current rating of the charger? Oct 12 16:13:12 maybe it just can't deliver enough power Oct 12 16:13:38 Argh, yeah, looks like it's time to ditch this one, it's on 1A. Oct 12 16:13:44 That must be it. Oct 12 16:14:09 ah yeah the SRM recommends 3A ... dunno if it needs that much, but it doesn't sound like a bad idea Oct 12 16:17:00 Thanks again, zmatt - looking forward to poking around a bit. Oct 12 18:25:20 Hello Oct 12 18:26:03 I have a query on using swupdate in beaglebone black board Oct 12 18:27:51 I have download and compile SWUpdate natively on BBB (to keep it simple) Oct 12 18:28:49 unable to start the webserver Oct 12 18:29:20 Error: set_ports_option: cannot bind to 8080: Address already in use Oct 12 18:29:29 so pick a different port? Oct 12 18:29:43 either that or disable whatever service is using that port, if you're not using that service Oct 12 18:30:27 which port to be selected for swupdate Oct 12 18:31:02 When i run the command --> "sudo swupdate -v -k /home/debian/swupdate-public.pem -w "-document_root /home/debian/swupdate/www"". I was having the above error Oct 12 18:31:04 why do you expect people here to know this? I had never even heard of this software package before today and had to google it Oct 12 18:31:41 I just thought, some one had experimented on this Oct 12 18:32:11 but it sounds like a simple configuration problem that should be easy to fix Oct 12 18:32:57 tried with different options, but no success Oct 12 18:33:55 I googled "swupdate 8080", first hit was the documentation: https://sbabic.github.io/swupdate/swupdate.html#running-swupdate which indeed shows how you can specify the port you want to use Oct 12 18:35:05 this really is something you should have been able to find yourself easily Oct 12 18:39:19 ok Oct 12 22:36:08 good evening all Oct 12 22:36:19 what is the general feeling about the BBB-AI? Oct 12 22:36:24 are people excited? Oct 12 22:36:34 is it a big step up? Oct 12 22:37:55 I mean... it's more or less a beagleboard-x15 in a beaglebone form-factor Oct 12 22:38:06 and priced in between Oct 12 22:38:19 whether that's exciting or not depends on your needs/application Oct 12 22:39:21 for a lot of applications the beaglebone-black and relatives is sufficient **** ENDING LOGGING AT Sun Oct 13 02:59:57 2019