#linuxcnc-devel | Logs for 2013-09-24

[08:03:42] <skunkworks> The search box is missing on the wiki now...
[08:21:08] <cradek> skunkworks: it's there for me, and it even works
[08:21:41] <cradek> oh was it at the top before? it's at the bottom.
[08:32:36] <skunkworks> cradek, duh - I thought I looked at the botom
[08:37:14] <skunkworks> heh
[08:37:18] <skunkworks> look - easy
[08:37:18] <skunkworks> http://groups.yahoo.com/neo/groups/mach1mach2cnc/conversations/messages/141662
[08:57:56] <skunkworks> micges, any thoughts on why I can only enable encoders and pwm?
[08:58:37] <micges> and you want enable what?
[09:00:14] <skunkworks> stepgens
[09:00:32] <skunkworks> for starters ;)
[09:01:59] <skunkworks> This config
[09:02:00] <skunkworks> config="num_encoders=1 num_pwmgens=1 num_stepgens=1 num_3pwmgens=4"
[09:02:02] <skunkworks> gets me
[09:02:08] <micges> ok
[09:02:16] <skunkworks> http://pastebin.com/jLjFFrgK
[09:02:22] <micges> check with mesaflash what firmware you have onboard
[09:02:39] <skunkworks> oh - I did that earlier - let me see if I can find it
[09:02:46] <micges> ok
[09:03:24] <skunkworks> well - there you go
[09:03:25] <skunkworks> http://pastebin.com/Rc1bdHmx
[09:03:48] <skunkworks> looks like it only has encoders and pwmgen
[09:04:01] <micges> yep
[09:04:12] <skunkworks> should it error though?
[09:04:28] <micges> I have no idea
[09:04:31] <skunkworks> heh
[09:04:40] <micges> probably it should at least warn
[09:05:17] <micges> check 7i80.zip file for some config with stepgen
[09:06:02] <skunkworks> sounds good. Well - I will run this for a while and see if the watchdog bites - but it didn't for the time I played with it yesterday
[09:06:55] <micges> that's good
[09:07:19] <micges> it shouldn't bite
[09:08:32] <micges> rtai version of driver is working but I have no communication :|
[09:08:49] <micges> (at least it doesn't hard hang )
[09:18:00] <skunkworks> micges, cool!
[09:19:57] <skunkworks> and Darn!
[09:56:55] <micges> pcw_home: hi, can you send me latest regmap?
[10:05:08] <pcw_home> sent
[10:05:44] <micges> thanks
[10:59:22] <skunkworks> flashing an led on the 7i80 - Yay!
[10:59:44] <skunkworks> it has been running for the last couple hours
[11:00:42] <skunkworks> It could be that the older nic I tested initally was bad.. It was something I found in the junk box.
[11:01:37] <micges> probably, I have three the same rev 8139 boards and only two works
[11:05:23] <skunkworks> yeck
[11:13:55] <pcw_home> do they work with standard networking with the 7I80?
[11:15:18] <pcw_home> (does mesaflash --device 7i80 --verbose work for example)
[11:16:38] <micges> same with standard eth
[11:16:55] <micges> third one was discovered but not working in both modes
[11:17:24] <skunkworks> hmm - never though of hooking it to the network to see
[11:19:31] <pcw_home> if it does not work in stadard mode I would suspect a bad card
[12:25:48] <skunkworks> still flashing
[12:26:26] <skunkworks> servo max is 697us
[12:26:33] <skunkworks> *397us
[12:26:47] <skunkworks> and that is going to get better?
[12:29:08] <pcw_home> It should be possible to do better
[12:31:25] <pcw_home> mostly by going to a 2 packet mode.
[12:31:27] <pcw_home> Also not sure how optimised the packets are for size (runs of reads,writes to succesive addresses should always use block transfers at the 7I80, not sure if they all do now)
[12:32:33] <skunkworks> neat
[12:34:15] <pcw_home> Probably need to dump some packets or instrument the 7I80 a bit to see where the time is going
[12:34:20] <skunkworks> not going to be able to run faster servo threads. (not that I have needed faster than1khz)
[12:34:59] <pcw_home> I think 2 or 3 should be possible in 2 packet mode
[12:35:06] <skunkworks> again - neat
[12:35:17] <skunkworks> this is pretty awesome work
[12:35:34] <skunkworks> realtime ethernet attached device
[12:36:00] <pcw_home> 2 packet mode means the 7I80 sends the rx data packet just in time for for servo thread (so it normally does not wait)
[12:36:51] <skunkworks> wow - so you have to clock to the servo thread?
[12:36:55] <pcw_home> so packets are txdata with wait for timer, then send RX data
[12:37:18] <pcw_home> yes we phase lock a DPLL to the servo thread
[12:37:24] <skunkworks> wow
[12:38:40] <pcw_home> The DPLL also filters out jitter (read sampling time is de-jittered)
[12:41:53] <skunkworks> again - wow
[12:42:54] <pcw_home> But integrating the 2 packet mode will take some work. It means that you cannot do random reads each cycle but have to read everything you might need (not a lot different that DMA, you dont want to change your scatter-gather list every cycle)
[12:45:46] <micges> driver is close to that besides few hardcoded places
[12:46:05] <skunkworks> I think you may have lost me now - but that is ok... You mean the driver as it is now - keeps asking for what it needs as the driver runs?
[12:47:23] <micges> (hm2_eth under rtai is loading driver and registering properly hostmot2, still have communication timeouts)
[12:48:44] <pcw_home> Yes its a 3 packet mode
[12:48:46] <pcw_home> in the read part of the servo thread the driver sends a request packet for
[12:48:47] <pcw_home> read data and waits for the reply (this is the longest time)
[12:48:49] <pcw_home> in the write portion, it sends a packet with output data only
[12:51:35] <pcw_home> in 2 packet mode there shoud be no long waits for return data (though 400 usec is quite long, it migh be good to instrument the packet sizes)
[12:56:08] <pcw_home> I am about to update the firmware to R14 and I will add debug code to drive the TP0 and TP1 pins with RXPKT and TXPKT active times
[12:56:50] <skunkworks> pcw_home, what sort of times are you seeing in your testing?
[12:57:34] <pcw_home> for ~200 byte packets the 7I80 turnaround is ~50 usec or so
[12:58:42] <pcw_home> (in fact i can ping at >8 KHz with plain old NetBSD)
[13:02:18] <pcw_home> Though ping is quite fast as it does no parsing and in
[13:02:20] <pcw_home> fact just re-writes the header and xmits the RX buffer
[13:03:56] <skunkworks> I rtpinged the card - I remember the info was in us and pretty low - but I don't remember how low
[13:07:38] <skunkworks> <skunkworks> rt_ping is pretty cool - it gives the time as about 60us
[13:08:07] <skunkworks> huh - so the old network card did ping the 7i80...
[13:14:26] <skunkworks> I should ping it now with the new card
[13:38:58] <skunkworks> stupid question - how do I know what pins say the pwmgen uses?
[13:40:16] <jepler> usually there's a way to get the hm2 driver to print it in the log
[13:40:39] <skunkworks> this doesn't seem compleate.
[13:40:40] <skunkworks> the log
[13:40:42] <jepler> the information is in the idrom, pretty sure
[13:40:49] <skunkworks> http://pastebin.com/3nm3Bwue
[13:45:22] <jepler> you could try increasing this in rtapi_global.h: #define MESSAGE_RING_SIZE 32768
[13:46:37] <pcw_home> normally its something like this
[13:46:38] <pcw_home> ...
[13:46:40] <pcw_home> [ 350.979248] hm2/hm2_5i25.0: 34 I/O Pins used:
[13:46:41] <pcw_home> [ 350.979255] hm2/hm2_5i25.0: IO Pin 000 (P3-01): IOPort
[13:46:43] <pcw_home> [ 350.979259] hm2/hm2_5i25.0: IO Pin 001 (P3-14): IOPort
[13:46:44] <pcw_home> ...
[13:49:44] <jepler> (edit src/rtpai/rtapi_global.h and build)
[13:59:01] <pcw_home> If you turn off debugging maybe the pinout will fit
[14:00:10] <skunkworks> oh
[14:00:36] <skunkworks> I could try that. I broke it - how can I go?
[14:00:59] <skunkworks> I put a 2 in front of it and broke it :)
[14:01:14] <jepler> huh no idea
[14:01:27] <skunkworks> but I can put it back to what it was and not set the debug level
[14:02:00] <pcw_home> you _may_ also be able to get the pinout
[14:02:01] <pcw_home> by running /freeby.mesanet.com/readhmid
[14:02:03] <pcw_home> (with normal networking)
[14:05:49] <micges> skunkworks: in xenomai these info is printed to the terminal
[14:12:34] <skunkworks> micges, how?
[14:14:46] <memleak> RTnet with RTAI works with the 7i80 now? I remember last month or so I looked at it a bit but couldn't find a good fix.
[14:15:16] <memleak> micges only had success with xenomai i recall with rtnet
[14:21:49] <jepler> memleak: skunkworks is using xenomai I think
[14:22:38] <skunkworks> yes
[14:22:58] <skunkworks> I don't think the logging isn't very consistant in the new schema
[14:23:19] <skunkworks> * I don't think the logging is very consistant in the new schema
[14:24:18] <skunkworks> I only start getting what looks like useful info with debug=4 - but nothing like hm2/hm2_5i25.0: IO Pin 000 (P3-01): IOPort
[14:26:32] <pcw_home> With the normal driver/RTAI AFAIK you dont need any debugging turned on to get the pinout
[14:26:49] <skunkworks> it goes to dmesg - right?
[14:29:59] <pcw_home> yep
[14:39:32] <skunkworks> that all got sent to the message handler now - linuxcnc.log
[14:40:01] <skunkworks> does it happen when the driver is loaded?
[14:40:23] <pcw_home> Yes
[14:50:02] <CaptHindsight> memleak hasn't touched xenomai
[15:14:25] <alex_joni> anyone seen mhaberler?
[15:46:50] <micges> memleak: 7i80 driver, it's qick and dirty but it works mostly on rtai
[15:47:10] <micges> memleak: with your suggestions what to look for
[15:48:25] <andypugh> pcw_home: Do you have the BiSS regmap yet?
[15:49:09] <memleak> my suggestions?
[15:51:06] <micges> yes, that I didn't search enough :)
[15:52:09] <memleak> i don't recall saying you didn't search enough heg
[15:52:10] <memleak> *heh
[15:52:57] <micges> its ok, I understand it so and it works
[15:53:51] <micges> now besides rtnet working much slower than under xenomai I have this one
[15:54:27] <micges> I must wait in kernel mode in init_module until rt task finishes
[15:55:04] <micges> semaphores doesn't work, probably it needs both task to be realtime
[15:55:10] <micges> memleak: ^^
[15:55:38] <micges> now I'm just delaying 1s
[16:04:28] <memleak> i'm glad you got it going!
[16:04:38] <memleak> i had a feeling just setting a delay value might fix it..
[16:05:02] <memleak> you said the time though from when you need the delay to happen is unpredictable
[16:06:25] <micges> if nothing works then hardcoded timeout will be used
[16:08:30] <micges> yes it's unpredictable but I'll check time on most complex configs and times it by 2 or 3
[16:17:53] <andypugh> pcw_home: And can you explain the "Request Length" register in Fanuc?
[22:00:09] <eliezer> hola a todos quien habla espaņol
[22:00:55] <eliezer> estoy tomando un curso de torno fresados cnc
[22:01:15] <eliezer> y uso linux y quiero descargar un emulador me podrian ayudar