#qi-hardware IRC log for Sunday, 2011-09-25

qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-09232011-2040/00:45
kristianpauldamn, i forgot that libc.c dont have float support for printf :-| !!!01:36
wpwrakkristianpaul: you mean libbase ? :)04:48
kristianpaulyes04:52
qi-bot[commit] Werner Almesberger: m1/perf/sched.c: overlook compiler warnings at your own peril, grrr ... (master) http://qi-hw.com/p/wernermisc/a0abd4e09:01
qi-bot[commit] Stefan Schmidt: ieee802154/at86rf230.h: Fix register names for RX_AACK_ON and TX_ARET_ON (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/43adc1110:02
qi-bot[commit] Stefan Schmidt: ieee802154/at86rf230.c: Prepare driver for auto ACK support (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/d8f39e810:02
stefan_schmidtwpwrak: ping19:03
wpwrakpong19:03
stefan_schmidtwpwrak: I have an interesting problem iwth atusb here :)19:03
wpwrakah ? :)19:03
stefan_schmidtwpwrak: I'm implementing auto ack and automatic retransmit right now19:04
stefan_schmidtwpwrak: uto ack works fine already but the sending in TX_ARET mode is failing19:04
stefan_schmidtwpwrak: and it seems its because no slp_tr pin is available on atusb...19:04
stefan_schmidtwpwrak: From what I see it is not needed in normal tx mode but in tx_aret it is :/19:05
wpwrakis it needed ?19:05
wpwrakand you mean in the driver, not the hardware, correct ?19:06
stefan_schmidtas in the driver yes19:06
stefan_schmidtfrom what I found you need to toogle it shortly to 1 and back to 0 to actually start the transmitting of the frame out of the buffer19:06
wpwrakthat's one way to start it, yes19:07
wpwrakbut you should still be able to just send a command, as in basic mode19:07
stefan_schmidthmm, from what I understand I can't do this in extended mode19:07
wpwrakwhat makes you think you can't ?19:08
wpwrakpage 48 has it, much like on page 33 :)19:08
stefan_schmidtthe contiki driver code. And my filing tests :)19:09
stefan_schmidtbut let me check again19:09
stefan_schmidtmaybe contiki uses the slp_tr approach always19:09
wpwraki wouldn't be surprised if contiki did this. in general, it should be more efficient.19:10
stefan_schmidtso one either uses slp_tr _or_ send writes TX_START into TRX_STATE?19:10
wpwrakbut things should still work even without SLP_TR19:10
wpwrakyes19:10
stefan_schmidthmm19:10
wpwraklemme push my unfinished new atusb driver ...19:11
qi-bot[commit] Werner Almesberger: atusb driver for extended mode (untested and unfinished ) (ben-wpan) http://qi-hw.com/p/qi-kernel/815a0e319:12
stefan_schmidtwpwrak: maybe I'm just to tired right now but where is the TX_START done in the at86rf231 driver? I can only see the slp_tr magic in xmit19:13
wpwrakrc = at86rf230_write_subreg(lp, SR_TRX_CMD, STATE_BUSY_TX)19:14
wpwrakright after the slp_tr ;-)19:14
stefan_schmidtwpwrak: but BUSY_TX != TX_START??19:14
stefan_schmidtseems I'm running one-way street here since an hour :)19:15
wpwrakcheck tables 7-3 and 7-4 :) (pages 45 and 46)19:16
wpwrakrunning down one-way streets can still get you closer to your destination. it's the dead-ends you have to avoid ;-)19:18
stefan_schmidtwpwrak: it actually looks like a bug to me that works by accident. Lets see:19:21
stefan_schmidtSTATE_BUSY_TX gets written into SR_TRX_CMD while it is an _state_ and not an command!!19:21
stefan_schmidtand by accident it is also 0x02...19:22
wpwrak;-)) naw, that's intentional19:22
wpwrakstate and command numbers are almost always the same19:22
wpwrakthey did have some bugs in cases where they aren't, though19:22
stefan_schmidtBut TX_ARET_ON is 0x19...19:22
stefan_schmidthmm19:23
stefan_schmidtsame state and command19:23
Action: stefan_schmidt stops for a moment to read it all carefully19:23
stefan_schmidtok, so the problem is I write BUSY_TX_ARET there which is 0x12 and has no corresponding command19:25
stefan_schmidtand page 48 shows TX_START as way to come from TX_ARET_ON to BUSY_TX_ARET19:25
stefan_schmidtso command and state differ for TX_ARET here and with the logic above the state machine fails19:26
wpwrakyes. if you keep the structure of a86rf230.c, you'd need an exception here19:26
stefan_schmidtat leats I'm not completely blind :)19:27
wpwraki don't particularly like the way the driver constantly looks at the transceiver state. state changes are usually announced by an interrupt, too. or you can just wait a little.19:27
stefan_schmidtwell, I never said I like the driver. Just wanted to have another test with aack in my evaluation :)19:27
stefan_schmidtwpwrak: still looking forward to your new driver and new firmware19:28
stefan_schmidtwpwrak: actually after 20th october I should be able to help you with it again. (Deadline for diploma thesis)19:28
stefan_schmidtwpwrak: btw, interested in the evaluation results on the IEEE 802.15.4 layer?19:29
wpwrakyeah, i hope to get around to finishing it by then. got a bit sidetracked with M1 these days, sorry.19:29
wpwraksure :) particularly if they're good - or if you've found some 60 ns difference ;-)19:29
stefan_schmidtwpwrak: no problem. I can do my evaluation this way. And I still have a lot to write so I son't know if I have time to test the new driver for the thesis19:30
stefan_schmidtwpwrak: will prepare something and send it over tomorrow. Don't have it with me right now.19:31
stefan_schmidtwpwrak: short version. _very_ stable and low packet loss even without aack but slow :)19:31
stefan_schmidtwpwrak: interestingly we have some steps in the graphs for different payload lengths19:31
wpwrakstable is good :) slow ... due to the USB delays, the over-the-air data rate, or perhaps protocol delays (which we generally don't implement anyway)19:33
wpwrak(steps) around which kind of numbers ?19:33
stefan_schmidtwpwrak: IIRC around 50 or 60 bytes. Can say for sure and show you the graphs tomorrow19:34
stefan_schmidtwpwrak: I send 400 packet for each payload from 5 to 115 bytes (4 byte have been an internal header for the test)19:35
wpwrak(50-60 bytes) ah, that's a little odd19:36
stefan_schmidtand it is not as linear as I had expected. At least one step, maybe another, smaller, one later on19:36
wpwrakhmm. interesting indeed :)19:36
stefan_schmidtand max values for rtt can be way off while min is near by. That makes sense so. min is limited by hardware and max only by the linux scheduler :D19:37
wpwrakyeah :)19:39
stefan_schmidtmuhahahaha19:41
stefan_schmidtit was indeed only the state change here19:41
stefan_schmidtusing BUSY_TX aka 0x02 made it work for TX_ARET_ON as well :)19:41
wpwrak;-)) congratulations !19:41
stefan_schmidtwpwrak: thanks for bring me back from the one-way street with slp_tr :)19:42
stefan_schmidtwpwrak: thanks19:42
stefan_schmidtwpwrak: btw, if you have your new driver and firmware ready for testing I might be able to sneak this into the thesis. At least if there is something until, say, 14th19:43
stefan_schmidtwpwrak: if not that fine as well will test is after my thesis then19:43
wpwrak(slp_tr) you're welcome :)19:45
wpwrak(new driver) i'll see what i can do19:46
wpwrakstill have a number of other things in the pipeline, though ...19:47
stefan_schmidtwpwrak: don't make any stress about it for me. I'm fine with what I have here and can evaluate it this way. Stable is more important then speed right now.19:50
wpwrakno italian sports cars then ;-)19:51
stefan_schmidtwpwrak: two of the atusb will stay with me after the thesis and finished studies anyway. Motivation enough to test the new driver or even help on it.19:51
stefan_schmidtwpwrak: not for me. Actually since my last car broke down I share the car with my gf :)19:51
stefan_schmidtwpwrak: have to wait until freelancing brings in enough for a Audi A6 now ;)19:52
wpwrak;-) subdued design then19:55
stefan_schmidtwpwrak: you mean the audi or my gf's car?19:56
wpwrakdunno what your gf drives :)19:57
wpwrakwell, or would drive if you hadn't taken possession of the car ;-)19:58
qi-bot[commit] Stefan Schmidt: ieee802154/at86rf230: Add Automatic retransmit form the extended mode (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/54634de19:59
wpwrakbut what i mean is that most of the audis have an understated design. they often have strong engines but don't look very "sporty"19:59
stefan_schmidtwpwrak: VW Polo19:59
stefan_schmidtwpwrak: ah, yes. Friend of mine has an A4 as sport edition. Does more look like a family car but has enough power to have fun with it :)20:00
stefan_schmidtwpwrak: and yes, I'm not to much into cars. And especially not into sport cars. If I had _to_ much money a porsche 911 would be ok thought ;)20:01
wpwrakthe pure sports cars are a bit inconvenient anyway. few places where you can fully enjoy them and hardly enough room for groceries, let alone anything bigger20:02
stefan_schmidtwpwrak: and you feel every small bump from the street20:04
wpwrakthat's a feature, not a bug :)20:05
stefan_schmidtyeah, like taking state instead of commands and screwing with my mind if no corresponding command is there :)20:06
wpwrakthat driver has indeed some peculiarities :)20:09
wpwraki don't particularly like the design. it seems to be overly cautious. better to assume that basic things like communication with the chip normally work, and have catch-all timeouts to recover from situations in which something goes wrong.20:10
stefan_schmidtyeah, once one have identified this chip by reading out some regs I would assume that SPI communication is stable20:14
wpwrakthat may be a tad too optimistic ;-) e.g., in the case of atben, you may very well have some interruptions. but you don't need to check for them on _every_single_ state change :)20:17
--- Mon Sep 26 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!