**** BEGIN LOGGING AT Wed Dec 16 02:59:59 2015 Dec 16 03:35:19 help my mini usb plug broke off took pads with it Dec 16 03:36:21 can i solder to other pins on chips to get it back up running Dec 16 03:37:33 depends, the traces may be on interior, deep layers, and that complicates the reconnection. Dec 16 03:37:43 lol nope. Even with a magnifier and hot-air rework, its hard to get a good contact Dec 16 03:38:29 the tracks are likely hair-line thick and very delicate Dec 16 03:39:14 i think just to wires are needed dont know the chip name off hand i looked it up a week ago Dec 16 03:39:37 Yeah, the TI ap note on high speed layout suggests the contacts deep on the other side of the layers for USB. Dec 16 03:39:52 there are likely to be vias Dec 16 03:40:12 Its an interesting point though, that a USB cable could rip it off. Dec 16 03:40:14 vias? Dec 16 03:40:23 vias go 'through' the pcb Dec 16 03:40:35 vias connect up/down through layers Dec 16 03:40:44 zauslande .. I've seen it happen .. albeit with a usb->rs485 breakout board .. similar cost though :/ Dec 16 03:41:28 got it off ebay they had broke it , super glued plug back on ? Dec 16 03:41:37 An external with torquing stress could be a problem for SMT, worth considering in purchasing a board. Dec 16 03:41:44 yea super-glue don't conduct as well as solder ... Dec 16 03:42:18 that's why USB-A and -B plugs are mechanically much stiffer Dec 16 03:43:15 the pins had solder on them , glued plug back on to board Dec 16 03:43:18 and why RS232 has screws :) Dec 16 03:43:39 is how i got it Dec 16 03:45:14 bbb ? solder on 4 pionts , 5-6 wires Dec 16 03:45:30 maybe fabricating a protective case that will take those stresses instead of the chip... but it seems that only works to a point. Dec 16 03:45:38 mark_______ .. I wish you good luck. Come back when you have it working :) Dec 16 03:46:25 zauslande.. I understand they deliberately made microUSB a more rugged connection for phones/etc where there was a heavy duty on them Dec 16 03:47:31 i was going to cut plug off mini usb solder wire right to chip tie wire to mounting hole on the board Dec 16 03:48:22 mark_______ .. if you simply tag wires on, you're likely to violate electrical and timing requirements of usb. Dec 16 03:48:45 two of those wires are a differential pair and need matching impedance and trace length Dec 16 03:48:47 this might be a job for conductive epoxy. Dec 16 03:49:24 GenTooMan .. z-axis tape is aremote possibility if you can get good contact on the pcb .. but if it's been ripped off as I guess, you may have lost pcb trace integrity Dec 16 03:49:25 Its a old truth, for cable torquing connections, use through hole connectors. Maybe they expect double copper weight on that SMT layer for a stronger hold? I remember having to do that back in '83 when keypads were getting 1/10th of their lifetime due to pounding by the people keying in the commands when the lines got long. The through hole buttons were lifting solder trace... I flew up to... Dec 16 03:49:27 ...Lansing to see one in use and they were seriously punding it. Dec 16 03:50:27 i think it was U9 chip on the bbb board Dec 16 03:50:39 well had 8x8 mmm pads shear off a SMT design, the only solution was via stitching on the pads too keep them from shearing off. Dec 16 03:51:31 what is via ? Dec 16 03:51:52 then again we were dropping from 1 meter height. Via is a vertical trace through the PCB. Dec 16 03:52:41 it comes up some where ? Dec 16 03:52:42 mark_______ .. usb mini-B pins go direct to the processor .. if you succeed in patching wires on the underside of the chip, I look forward to seeing your work :) Dec 16 03:53:16 they do also seem to go to U10 too though. Dec 16 03:54:11 though .. I'm sure you can desolder the processor with a hot-air gun .. I've seen worse .. Dec 16 03:54:19 i looked a week or 2 ago i think they go right to u9 chip ? Dec 16 03:54:25 don't bet on anything working afterwards though XD Dec 16 03:54:52 mark_______.. nope that'll be the big USB -A socket host port Dec 16 03:55:06 the one the size of a dime Dec 16 03:55:26 may be u10 Dec 16 03:55:49 coincidentally the chip is the same .. just not the right one :D Dec 16 03:56:40 i know a place that can solder this if i can tell them what chip . pin , Dec 16 03:57:19 u10 is next to the power switch next to the ethernet btw Dec 16 03:57:24 I'm sure that'll be nice and easy :) Dec 16 03:58:00 by S3 Dec 16 03:58:30 do u think it would work ? Dec 16 04:00:22 I'm guessing nope Dec 16 04:00:23 i think they go to a cap right by the , one or two wire , then two wire to u9 or u10? Dec 16 04:00:51 right by the plug Dec 16 04:01:04 well I can see 4 vias there .. that doesn't fill me with confidence Dec 16 04:01:14 a via is a plated-though hole that connects traces on different layers... like a solder straw. On two sided boards they just run through it from top to bottom. In multilayer boards they can exist between internal planes only, unapparent from the surfaces. Dec 16 04:01:26 those tiny specs .. are holes in the board Dec 16 04:03:40 * zauslande is looking at an S-100 board so they're REAL BIG vias on it. :) Dec 16 04:04:41 Ive heard some people building S100 boards (new ones) with FPGA's and massive memory, running their own IO boards etc. Dec 16 04:05:05 i look at schematic bbb looks like i can get to the mini usb point , u9 , cap , plus , grd? Dec 16 04:05:38 * zauslande going to be playing with Arduinos/RPis/Beagles for a couple of months so thought I might use one of them as a S-100 board tester as I restore the old circa '77 to '83 S-100 boards before putting the system back together. Dec 16 04:06:46 GenTooMan: Yep... I'm doing that with a eZ80 and and banks of 32bit wide SRam. Its a SBC... maybe I should call it the SD Systems SBC 400... as that was never made. Dec 16 04:06:47 well thanks for your time Dec 16 04:07:36 veremit Dec 16 04:07:37 GenTooMan: There are a lot of hobbyists puttering around with S-100... not too many professional designers. Dec 16 04:11:42 zauslande well pro's do it for money. Not much money in S-100 systems. However they are adequate for a number of things so nothing wrong with that. Perhaps a BBB S-100 adapter board would be interesting? :D Dec 16 04:11:46 A lot of S-100 people today only want through hole component kits which makes me crazy... take advantage of SMT. Dec 16 04:12:32 zauslande Totally retro it's like buying a Land Rover from WWII because it's a land rover from WWII :D Dec 16 04:13:09 No money in S-100 but a lot of fun... If I'm going to design and built a eZ80 board, why not use a S-100 board outline and offer various means of using it? Dec 16 04:14:29 I'm retro, not vintage... I like the feel and the simplicity of the old systems, I just don't want to spend all my time maintaining its all past its useful lifetime. My #1 goal is to recover all my floppy data to archive. Dec 16 04:15:14 I think a Beagleboard X15 would be a good combination for that task. Should be faster than me designing a special board. Dec 16 04:15:37 And yes, this is a silly waste of time, but I'm old and would be expected to do such things. :) Dec 16 04:15:54 The irony is those old floppies are more stable than the 3.5" units that replaced them. So they are probably still good. Dec 16 04:18:00 I'm in Texas so our envirnoment keeps floppies in great shape. At least visually, but floppies from the late '70s still look pristine. However, I'm designing my floppy ripper to to work fast and not waste any time on the media. Dec 16 04:19:47 zmatt: did you get your assembly dram test code complete yet? Dec 16 04:20:33 I've got about 400 8" floppies and maybe 2,000 5.25" floppies and uncountable 3.5" floppies to work. Some of those are even hard sectored... so I designed a single system to read them all. Only multispeed floppies like Commodore and Apple used some, wouldn't be writable on a standard drive. It should read anything that's left on the media, no PLL or data separator needed. Dec 16 04:20:46 zauslande: the encoding is just MFM data correct? you might be able to decode that in real time the problem then becomes what the disk format is. Dec 16 04:21:42 zauslande: are you usign an adc channel to read the head signal? then demodulating that? Dec 16 04:21:53 I'm doing the decoding in firmware so if can do FM, MFM, MMFM, and GPC technically though I have none of those (yet). Dec 16 04:22:55 yates, Just using READDATA raw. Its not speed dependent. I might later find that I have to boost sensitivity to get everything off the disks... don't know that yet. Dec 16 04:23:32 what is READDATA raw? Dec 16 04:23:42 Technically I could use it an 1RPM or 300 or 360PRM... Dec 16 04:23:50 perhaps i missed the first part of your conversation. just go there. Dec 16 04:24:13 got here Dec 16 04:24:28 READDATE is a signal coming from any floppy drive through the cable that is normally fed into a data separator first before a FDC can read it. Dec 16 04:24:45 READDATA Dec 16 04:25:25 and you're feeding READDATA into an ADC channel on the Sitara? Dec 16 04:25:40 It is the data pulse triggered by an magnetic flux transistion on the media, but the drive extends that pulse to a known period (artificial aspect of the signal). Dec 16 04:25:54 No... doing it all digital. Dec 16 04:26:06 ah. Dec 16 04:26:47 READDATA is a digital signal (open-collector pulse of fixed duration). The leading edge is coincident with the detection of a magnetic flux transistion under the read/write head. Dec 16 04:27:44 It bypasses a FDC (Floppy Disk Controller) so even sectors with failed CRC can be read in and analyzed and in some cases fixed. Dec 16 04:28:01 zauslande: is this just a personal project? Dec 16 04:29:01 zmatt: why isn't the cpsid instruction in the arm architecture reference manual like the rest of the instructions? Dec 16 04:29:23 yeah, Last S-100 board I designed (VFW-III) was the last board SD Systems designed before the market died... so this is kind fun for me... get to do some real-time design and coding which is what I've done most of my career. Dec 16 04:29:53 cool project. Dec 16 04:30:17 i was about to throw out my old floppies. from the 80's, just sitting collecting dust. Dec 16 04:30:57 although I haven't seen an 8in in a long time. we had those on the old Intel MDS 800 systems. Dec 16 04:31:55 Yeah, I started with an eZ80 but while its fast enough to read all the data, its not fast enough to write with precompensation. So I was doing a CPLD to do that (i.e. faster timer) and just thought a Beagleboard X15 would speed the project up. Get my old data archived and then I can do the retro eZ80 on a S-100 for fun. Dec 16 04:32:14 Don't throw any floppies away!!! Dec 16 04:32:18 http://www.classiccmp.org/hp/mds-800/mds-800.htm Dec 16 04:32:54 I used a MDS back in the day. My first real time controller use a 8085 for a SMD disk controller on Multi-Bus. Dec 16 04:33:16 oh crap, "multi-bus" - haven't heard that term in a long time Dec 16 04:33:24 multibus-II? Dec 16 04:33:32 obviously the micro wasn't in the data path. I used a signetics mealy-moore programmable state machine to handle all the data flow. Dec 16 04:33:41 Multi-Bus UNO Dec 16 04:34:15 i used some old ziatech IEEE-488 multibus II cards in an old project Dec 16 04:34:21 In 1982 my design was the only winchester controller that was fast enough to run a Fujitus Eagle. Dec 16 04:34:40 pretty cool. Dec 16 04:35:56 Intel came to us (GTE Government Systems) to present the 8086 circa 1980/81. Dec 16 04:36:01 were you in Silicon Valley? Dec 16 04:37:12 I went to SD Systems do design there winchester+floppy controller, the Versa-Floppy-Winchester III. But they charged $895 each so they never sold that to hobbyists. Their survival strategy was to make IEEE-696 for Industry and Commerce but they gave up after a year and sold off that division. :S I stayed on for another year before hopping into digital radio packet Dec 16 04:37:35 No, did my whole career in Dallas, Tx. Dec 16 04:38:05 ah. I was there contracting for TI about 2 years ago. Dec 16 04:38:14 I should have gone to Silicon Valley but honestly we did cooler stuff in Dallas. No one hears about it though as it wasn't consumer stuff. :S Dec 16 04:38:16 bad drivers! Dec 16 04:39:03 i guess zmatt is makin' zzzzzzzzz's, lol. Dec 16 04:39:34 good night. Dec 16 04:40:01 I was a T.I. draftsman in 1976, but I saw a microprocessor training kit come through the drafting department and switched my plans for ME to EE and left for college right away. (planned to get two years of experience, only did one). Dec 16 04:40:38 ...from mechanical engineering to electrical engineering... Dec 16 04:41:07 later... Dec 16 04:42:53 hmmwhat? Dec 16 04:43:14 oh yates left again Dec 16 04:43:31 you missed all the fun, zmatt Dec 16 04:45:03 I hate it when people don't just leave their client connected Dec 16 04:45:50 .. so inconsiderate .. and yet all these nicks that never speak .. Dec 16 04:56:27 zauslande: I'm too lazy to read the whole conversation, but note that eCAP is perfect for recording when pulses occur (buffers up to 4 timestamps, 10 ns resolution, and supports DMA) Dec 16 04:59:45 zauslande: any reason to believe you'd need an x15 for the task? sounds like a BBB would suffice Dec 16 05:00:17 that's basically what I'm talking about... don't need that type of resolution unless I do ST506/412 era hard drives and the odds of those still running is quite slim. I do have a circa 1988 HD I would like to recover though. Dec 16 05:01:19 The main reason I want an X15 is (1) its newer, and (2) I can stop the OS (likely that is in BBB too) Dec 16 05:01:58 if you need real real-time you should consider using PRUSS anyway Dec 16 05:02:48 if you need more code space and want to do baremetal coding, then I'd actually recommend the BBB even more since the Cortex-A8 is at least still sane to program for Dec 16 05:03:23 Its got plenty (more) I/O and lots of ram storage for buffering and post processing. I'd like to buffer a entire 8" drive (77 tracks, double sided) in Ram walking the drive as fast as possible... 1.1 to 1.-5 revolutions per track and stitched up in post processing. The less time the read write head is on the media, the better the chance of recovery. Dec 16 05:03:36 Exactly PRUs Dec 16 05:04:02 how much RAM do you expect to need? if you go baremetal then you have 512 MB available on a BBB Dec 16 05:04:09 I'm going to a Cortex M meeting Thursday. Dec 16 05:05:24 I did calculations about 6 months ago... I think I need less than 16MB to 32MB depending upon the timer bit width. Dec 16 05:05:25 PRU and Cortex-M are really not remotely similar though... the Cortex-A8 is reasonably similar to the -M if you use a bit of glue to make its oldskool ARM-style exceptions/interrupts more like those of the cortex-M Dec 16 05:06:09 would it be better to save a sampled bitstream, or timestamps of the READDATA pulses? Dec 16 05:06:22 The X15 uses A15, M5s and PRUs. It M5 meeting popped up and I thought I should go to prepare for using ARM and Beagles. Dec 16 05:06:28 M4 Dec 16 05:06:51 but note that only PRU has insanely low-latency I/O capability Dec 16 05:06:52 I time the fluxex on the track. Dec 16 05:06:58 fluxes Dec 16 05:07:53 BBB also has an eCAP instance in PRUSS Dec 16 05:07:53 yeah, I've been reading up on PRUs but The M series (talking about the ST Nucleo mbed) is an opportune meeting. Dec 16 05:08:08 it's good to learn about them in any case Dec 16 05:08:59 if you only need 32 MB then you can probably just reserve that without killing linux Dec 16 05:09:57 then you'd be able to retain linux as a comfy environment for managing/supervising the process Dec 16 05:11:55 if you only need 32 MB of memory that also means you have negligible bandwidth needs, so interconnect traffic from the rest of the system shouldn't really interfere Dec 16 05:12:45 and if you use DMA to transfer timestamps from eCAP to memory (you can set that up from linux userspace) then you have almost no code to write Dec 16 05:12:54 at least for the capturing phase Dec 16 05:14:51 Well I've just considered using a Beagle to speed up the project, for about 5 days now. I'm buying a set of Arduino/RaspberryPi/Beagles for christmas and thinking about something useful they can do, and hit upon the Beagle looking to be the best fit for that project-expedited. Dec 16 05:16:07 Finding something useful for an Arduino to do is painful. 16Mhz?! Considered a DUE (ARM inside) but will do an UNO and a MEGA Dec 16 05:16:11 R3 each Dec 16 05:16:22 the arduino is a weird joke Dec 16 05:16:49 like, the chip they include as usb2serial converter is more powerful than the arduino itself Dec 16 05:16:56 Raspberry's look to be good for cheap bundle smart camera surveillance. Beagle is more my sort of toy. Dec 16 05:17:13 beaglebone is for people who need good I/O capability Dec 16 05:17:19 the rpi is a gpu with an arm core stuck on the side Dec 16 05:17:22 Yeah, I went and saw a ESP8266 project demo last week. Dec 16 05:17:25 (literally) Dec 16 05:17:41 zmatt.. well summarised. Dec 16 05:18:08 (the rpi boots off the gpu, which boots up the ARM as auxiliary processor) Dec 16 05:18:33 broadcom chips which are hard to find anymore. Dec 16 05:19:23 they're not hard to find anymore, broadcom just doesn't do public info (or care about small-quantity customers presumably) Dec 16 05:19:36 A cheap pi with a camera make a nice security bundle. Dec 16 05:19:58 that would be fun to code. Dec 16 05:20:16 it has a csi2 port I think right? Dec 16 05:20:25 I don't know much about the pi Dec 16 05:20:38 zmatt: My read of it too. 6 months after I found a broadcom HDMI chip I wanted to use, it was no longer online. :S Dec 16 05:21:30 note that there's no public datasheet of the broadcom cpus used on the rpi/rpi2 (they're the same apart from the ARM subsystem) Dec 16 05:21:43 I want to play with the DSP chips in the X15 too to see about using them in GPU sort of stuff. Dec 16 05:22:00 the x15 has an insane amount of stuff Dec 16 05:22:36 agreed. Dec 16 05:23:01 (and we don't even get to play with the EVEs!) Dec 16 05:26:36 I noticed last night that T.I. is bundling the X15 with a stand and LCD display/camera. I might do a display with my toy set. Dec 16 05:27:09 yeah, the beagleboard-x15 is essentially the baseboard of the am572x evm Dec 16 05:27:22 not cheap though Dec 16 05:27:28 Guess it would be best to find one that works with Beagle and Pi. Dec 16 05:27:28 (EVMs never are) Dec 16 05:28:07 the beaglebone black has a microHDMI output (or you can grab parallel digital video from its expansion header) Dec 16 05:28:22 yeah, I think T.I. lists the EVM at $700+ so the Beagleboard X-15 at $199-$239 sans display/camera is better. :) Dec 16 05:29:20 zauslande: note that it's much harder to use the I/O of the x15 than that of the BBB Dec 16 05:29:42 Yeah, I looked at the BBB and a few other models a couple of nights ago. I think the X-15 does a full size HDMI connector (wonder if they got the feedback on micro HDMI like the one mentioned up the scroll coming off the board). Dec 16 05:30:21 well the connector sizes are appropriate for the board sizes Dec 16 05:30:28 you can't really compare the BBB with the x15 Dec 16 05:30:30 zmatt: Expected, but I won't know the comparison so I might tolerate it better. The X15 gives me a big toolkit to experiment with over time. Dec 16 05:31:31 zauslande: http://elinux.org/images/2/24/BBX15-BOTTOM.JPG Dec 16 05:31:58 you can't plug jumper wires into those Dec 16 05:32:07 I think the lowly Arduino UNO/MEGA might help test S-100 cards... maybe interface todays kepboard mouse to the old interfaces of S-100. Overkill power for a little interface. Dec 16 05:32:21 they're special connectors suitable for highspeed signals like video Dec 16 05:33:39 Yeah, I wonder if there will be a cape available for breaking that out for I/O related projects... might be faster to do a quick layout? How good is the Beagle cape support? Dec 16 05:34:31 The big shield/cap market seems to be Arduino, then maybe Pi. Dec 16 05:34:44 btw stop saying "beagle" .... there are way too many different boards whose name starts with "beagle" :P Dec 16 05:35:17 well "Beagleboard X-15" is a lot of typing. Dec 16 05:35:22 ok you mean x15 Dec 16 05:35:25 then say that Dec 16 05:35:35 since I initially assumed you meant the BBB Dec 16 05:35:36 Or "Beaglebone Black" or the other varients Dec 16 05:36:02 bbb and x15 are short enough right? Dec 16 05:36:22 but it's rather important to know which of those two you're referring two Dec 16 05:36:23 Well "Beagle" was collectively asking about cape availability (realizing BBB is different than the X15 connectors) Dec 16 05:36:46 x15: right now afaik none except the TI evm thing Dec 16 05:36:52 I expect to buy a X15 Dec 16 05:37:01 I wouldn't Dec 16 05:37:05 for your application Dec 16 05:37:17 the BBB is much friendlier to hardware hacking Dec 16 05:37:28 I'll also buy a Raspberry Pi2, and an Arduinio Uno/Mega Dec 16 05:38:11 if you're going to collect embedded boards, I can list a few more if you want :P Dec 16 05:38:48 well it will be overkill either way. For that project I just need to get the data archived before the data is lost for whatever eventual reason. Dec 16 05:38:54 like this thing is much neater than an Arduino -> http://www.embeddedartists.com/products/lpcxpresso/lpc1549_xpr.php Dec 16 05:40:36 Yeah, Arduino is mostly about being able to people that use them and don't know better. :) Dec 16 05:40:53 the BBB is rare in being a linux board that's friendly to hardware hacking and has cool I/O Dec 16 05:40:56 being able to talk to people that ... Dec 16 05:41:07 it's more microcontroller-like... but with a big enough cpu to run linux Dec 16 05:41:52 and it's now reasonably mature Dec 16 05:41:59 the x15 is still extremely new Dec 16 05:42:31 What I want to avoid is just ripping data and throwing it into a PC for post processing... rather keep in in the microcontroller side. Dec 16 05:42:41 wtb arduino tre :P Dec 16 05:42:52 zauslande: most uCs however also don't have 32 MB of ram Dec 16 05:43:01 Using a PC for post processing is more economical, but not as much fun. Dec 16 05:43:22 true... that's why I'm stepping up to ARMs Dec 16 05:43:45 but if you still want to easily wire external hardware in it, get a BBB, not an x15 Dec 16 05:44:01 That's an important point. Dec 16 05:44:30 Thanks for pointing that out. It could speed development. I'll visit BBB online tonight. Dec 16 05:45:18 note btw that a BBB is still not as resistent to electrical abuse as a typical microcontroller... that's an unavoidable side-effect of being manufactured on a high-tech process Dec 16 05:45:32 fpga cape ftw :D Dec 16 05:46:02 whoah fark .. that's 6am looming and I ain't slept yet .. :/ Dec 16 05:46:17 sleep is for the weak Dec 16 05:46:35 just call it an early holiday Dec 16 05:46:40 yeah I'll hear ya once you done a few 36h days :p Dec 16 05:47:06 not much work happens in december really. better to just let everyone go home. Dec 16 05:47:25 zauslande .. I'll give you my boss's email address .. :D Dec 16 05:47:25 veremit: not extremely uncommon for me Dec 16 05:47:38 zmatt .. how does that not quite surprise me?! Dec 16 05:47:50 where do you work? hereit gets worse in december Dec 16 05:47:54 erm .. double-negative error Dec 16 05:48:02 to much things still to be done Dec 16 05:48:29 I'm a high-latency person... hard to get started, but I also tend to keep going :P Dec 16 05:48:53 zmatt: yup, know how that goes too Dec 16 05:49:21 I used to work long hours on projects at work... in a way, I think it helps creativity with sleep deprevation. :) Bad for mental health according to all the research. Dec 16 05:50:14 yada yada :P Dec 16 05:51:43 veremit: you know UK webshops actually sell caffeine powder in 1/2 kg bags? ;) Dec 16 05:54:16 uhoh Dec 16 05:54:49 (£15.99 + shipping) Dec 16 05:55:02 ouch .. that's half a beagle .. or other arm board Dec 16 05:55:18 zmatt: thanks for the URL: http://www.embeddedartists.com Dec 16 05:55:33 vermit: for the equivalent of about 5000 cups of coffee Dec 16 06:00:38 hmm Dec 16 06:08:42 ok time to retry this 'nap' business .. laters folkz Dec 16 06:08:50 hehe, nite nite Dec 16 06:19:54 A good point for BBB: http://elinux.org/Beagleboard:BeagleBone_Capes Dec 16 06:29:13 it is easy enough to build most of the capes yourself Dec 16 06:43:42 zmatt: you awake? Dec 16 06:43:57 yup, read your email? :) Dec 16 06:44:02 nope Dec 16 06:44:38 ok, thanks Dec 16 06:44:48 sheesh this is a PITA assembly language Dec 16 06:45:32 really? I think ARM asm is relatively clean Dec 16 06:46:13 another question: in init.S, why are you issuing an "mcr" instruction? this is a coprocessor instruction, right? how is it being used to "set vector base" ? Dec 16 06:46:36 the cortex-a8 has no actual coprocessors (or coprocessor interface) Dec 16 06:47:17 well, the neon/vfp unit could be considered one I guess, though tightly integrated Dec 16 06:47:43 that doesn't really answer my question Dec 16 06:47:48 but cp14 and cp15 are part of the cpu.. they're "coprocessors" only for reasons of backwards compatibility Dec 16 06:48:04 most cpu config registers live in cp15 Dec 16 06:48:21 isn't the MMU also a coprocessor? Dec 16 06:48:23 andVFP Dec 16 06:48:27 aha. Dec 16 06:48:56 i'm trying to wrap my head around all these new interpretations. Dec 16 06:48:57 ds2: like I said, neon/vfp could be considered one... but it's not like older RAM cores which had a generic coprocessor interface Dec 16 06:49:16 generic coprocessor interface? Dec 16 06:49:29 imo, "coprocessor" means something like the FPU on the x86 Dec 16 06:49:31 what are you using to distinguish? Dec 16 06:50:06 old arm cores had a specific bus with a well-defined interface on which coprocessors could be attached Dec 16 06:50:18 the old VFP attached to it Dec 16 06:50:45 hmmm Dec 16 06:50:45 cp15 was a management unit also attached to the same bus Dec 16 06:50:55 thought that bus still existed Dec 16 06:51:01 not in the cortex-a8 Dec 16 06:51:03 what is a vmsa versus pmsa implementations? Dec 16 06:51:15 yates: pmsa is Cortex-R Dec 16 06:51:22 ignore it Dec 16 06:51:30 so you are saying no implementation of the a8 can add a custom CP? Dec 16 06:51:35 ds2: correct Dec 16 06:51:58 can't. the correct section of the manual defining p15 depends on it Dec 16 06:52:20 yates: pmsa and vmsa are mutually exclusive Dec 16 06:52:28 pmsa has simpler memory management Dec 16 06:52:58 no virtual memory, only flat mapping. it has an mpu (memory protection unit) instead of an mmu Dec 16 06:53:18 like the cortex-m (when equipped with the optional mpu) Dec 16 06:54:30 i still can't ignore it if i want to find the right page in the manual... Dec 16 06:54:47 perhaps you weren't talkign to me. Dec 16 06:54:51 yates: note that the cortex-a8 TRM also has a listing of its cp15 refs Dec 16 06:55:23 it's typically more convenient Dec 16 06:55:30 noted. Dec 16 06:56:14 since the architecture reference manual includes many registers that are not necessarily present (of course this is noted in the arch ref man, but it doesn't make it easier to browse) Dec 16 06:56:44 i see Dec 16 06:56:56 btw, if you have ref C of the arch ref manual, get ref B instead Dec 16 06:57:14 saves you from having to read around all the blahblah about virtualization which doesn't apply to the cortex-A8 anyway Dec 16 06:57:58 http://www.csc.lsu.edu/~whaley/teach/FHPO_F11/DDI0406B_arm_architecture_reference_manual_errata_markup_9_0.pdf Dec 16 06:59:53 does the am335x have 64kB or 128kB internal sram? I thought you (or someone) said 128k, but the manual says 64k Dec 16 07:00:41 there's 64 KB inside the cortex-A8 subsystem and 64 KB ("ocmc") on the L3 interconnect Dec 16 07:00:54 they're consecutive in address space, so you can present it's one 128 KB memory Dec 16 07:01:04 *pretend Dec 16 07:01:57 the first 1 KB of the cortex-a8 subsystem local sram is reserved for (though unused by) secure world, and is inaccessible Dec 16 07:02:37 i see Dec 16 07:03:05 i am definitely drinking from the firehose here... Dec 16 07:03:08 hehe Dec 16 07:03:44 it takes a bit of time to understand a SoC Dec 16 07:04:10 note btw my linker map does distinguish between the two 64 KB memories Dec 16 07:04:28 this particular SoC is more complex than the ones I'm typically using Dec 16 07:04:41 e.g., the Coldfire V2 Dec 16 07:04:55 ok, noted. Dec 16 07:05:11 that's an MCU, not a SoC Dec 16 07:06:03 besides the fact that someone calls it a MCU instaed of a SoC, what's the difference? Dec 16 07:06:25 a 5000 page reference manual (which still omits many details) Dec 16 07:06:31 :P Dec 16 07:06:34 I thought, in general, a SoC is a MCU plus integrated peripherals like ethernet.. Dec 16 07:07:01 in theory there's no clear boundary I think Dec 16 07:07:05 but in practice there is Dec 16 07:07:45 a microcontroller is something where you can still easily understand what's going on, which are almost always programmed baremetal (or a sometimes tiny rtos), and have no virtual memory Dec 16 07:07:56 a SoC is basically a small PC crammed into a chip Dec 16 07:08:48 semantic games - not fruitful... Dec 16 07:09:54 yates: the distinction seems widely understood though... manufacturers of microcontrollers call them "microcontrollers" Dec 16 07:10:37 the am335x is a bit odd in that it's quite on the small side of a SoC and includes peripherals that are more typical for mcus Dec 16 07:10:47 it's a bit like it stands with one leg on either side Dec 16 07:11:19 if you want to frame things that way, go ahead. Dec 16 07:11:39 I already said, there's no solid definition that distinguishes them Dec 16 07:12:12 but as you noted, this thing is much more complex than you're used to, so you've noticed the difference yourself already :) Dec 16 07:12:32 especially since my little demo prog doesn't initialize the mmu Dec 16 07:12:51 which is absolutely necessary if you want to have non-abysmal performance on a cortex-a8 Dec 16 07:13:38 (compared to what it is capable of, perhaps not compared to a typical mcu) Dec 16 07:14:45 right now every load/store instruction will take around 75 cycles or so, maybe more Dec 16 07:14:56 but for the little demo it really doesn't matter Dec 16 07:18:35 i thought the mmu was for protecting various parts of memory and providing virtual memory? why does it affect load/store performance? Dec 16 07:19:13 i noticed you don't enable data cache, that may be part of it. Dec 16 07:19:30 the mmu also determines other attributes of memory regions Dec 16 07:19:33 like cacheability Dec 16 07:19:59 aha. Dec 16 07:20:17 when the mmu is off, all data accesses are "strongly ordered" Dec 16 07:20:26 which is the most restrictive memory type Dec 16 07:20:43 it means every access is performed synchronously, with the cpu waiting for it to complete before continuing Dec 16 07:20:53 good for dram tests... :) Dec 16 07:20:56 writes are not buffered, nothing is cached Dec 16 07:21:25 terrible for dram tests Dec 16 07:21:34 you want normal uncacheable type for that Dec 16 07:22:16 that's what i mean. Dec 16 07:23:05 or are you saying "normal uncacheable type" != "strongly ordered" ? Dec 16 07:23:10 indeed Dec 16 07:23:39 what's the difference? if you're not caching, isn't ordering guaranteed? Dec 16 07:23:45 you have "normal", "device", and "strongly-ordered" .. the last one is basically "device with synchronous writes" Dec 16 07:23:53 "device" is like the "volatile" keyword in C Dec 16 07:24:23 every access is performed as-is, and in-order Dec 16 07:24:27 no optimization allowed Dec 16 07:24:45 it can't merge two consecutive 32-bit writes into one 64-bit write (which it would normally do) Dec 16 07:25:32 where is this in the arm/trm? Dec 16 07:25:41 somewhere early Dec 16 07:26:03 "Chapter A3 Dec 16 07:26:03 Application Level Memory Model Dec 16 07:26:25 so why is this bad for a dram test? because it's slow? Dec 16 07:26:37 I don't actually recommend reading that chapter btw Dec 16 07:27:04 the formal definition of the memory order model is nearly incomprehensible Dec 16 07:27:13 yes, the dram controller will mostly be doing nothing Dec 16 07:27:43 even in normal uncacheable? Dec 16 07:27:49 no, in strongly-ordered Dec 16 07:27:55 what changes? Dec 16 07:28:33 normal uncacheable will have excellent write speed, poor read (but iirc still not as bad as strongly-ordered) Dec 16 07:29:12 writes will get buffered in a queue, and consecutive writes get merged into 128-bit wide writes Dec 16 07:29:44 strongly-ordered writes will each wait until they are acknowledged as completed by the memory controller Dec 16 07:30:08 so each one takes a full "ping time" to the DRAM Dec 16 07:31:10 i see Dec 16 07:33:24 enough of this tonight - gotta work on a proposal.. Dec 16 07:33:32 thanks as always, zmatt Dec 16 07:33:35 night. Dec 16 07:33:39 I once did some timings on a related SoC... https://docs.google.com/spreadsheets/d/1r6Yx5yHB2jT8heAbp0fEi5iPt_wAPlDvZ2ZwUvPOpqU/pubhtml Dec 16 07:33:44 argh Dec 16 09:18:55 i'm running debin 8.2 (from http://beagleboard.org) on my beaglebone black. but there seems to be an issue with dhcp via usb0. dnsmasq is not serving my pc an ip Dec 16 09:18:58 Nov 12 21:46:55 beaglebone dnsmasq-dhcp[889]: DHCPDISCOVER(usb0) 6e:ac:f8:e5:7d:4d no address available Dec 16 09:19:02 is this a known issue? Dec 16 09:19:23 it used to work fine on older debian images Dec 16 09:39:06 Hi there, I've been trying to submit a post to the forum yesterday but it hasn't been validated yet. How long does it normally take? Dec 16 09:39:36 you mean the google group? Dec 16 10:07:32 @tbr, yes, I meant the google group Dec 16 15:21:34 Blipp. Dec 16 15:21:39 Any news on the X-15? Dec 16 15:26:30 soon(tm) afaiu Dec 16 15:27:24 Last I heard was "delayed one month" > 1 month ago. Last update on homepage is also > 1 month ago. Thus here and asking. I want a couple ordered ;) Dec 16 15:28:39 I havn't seen any news on the mailinglist indiciting imminent availability Dec 16 15:40:36 tbr: okay, thanks. Dec 16 15:59:10 ping rcn-ee Dec 16 15:59:22 morning Dec 16 16:07:33 has anyone used qemu static to modify a beaglebone system image on a x86 system? Dec 16 16:08:45 Sorry, never got around to try that Dec 16 17:24:38 halloo Dec 16 17:37:34 Anyone have any experience reading strings from the device tree? I'm using of_get_property, and end up with "��4-" when I'm expecting "4-bit" Dec 16 17:40:27 doesn't of_get_property return a pointer? Dec 16 17:42:44 it's interesting you seem to be getting the 4- part Dec 16 17:44:07 https://android.googlesource.com/kernel/common/+/android-3.4/drivers/mtd/ofpart.c#68 < has an example of someone getting a string Dec 16 18:06:52 ddrown: I still don't know why that wasn't working, but switching to of_property_get_string fixed my problems. Dec 16 18:07:09 sounds good to me Dec 16 20:25:59 Hi all. Any news on x15? Dec 16 20:26:33 not yet Dec 16 22:29:22 the x15 is a 32bit ARM correct? Dec 16 23:19:21 GenTooMan: yes Dec 16 23:26:54 vagrantc: well thats a bit of a downer but not unexpected I guess. Dec 16 23:28:07 at least it's ARMv7 :) Dec 16 23:51:34 Well I guess their is that. Dec 17 00:01:34 i've downloaded the latest debian 7.9 image. why does /lib/firmware no longer contain all the nice ready to use dtbo files? i can only see some PyBBIO-gpio* overlays Dec 17 00:10:20 hmm, seems like i need to install them myself from https://github.com/beagleboard/bb.org-overlays **** ENDING LOGGING AT Thu Dec 17 02:59:58 2015