fenn_ is now known as fenn
BJT's system is misconfigured in a way that's asking for occasional trouble if it can't handle full acceleration up to full velocity
cradek: someday I'd like to look into fixing that initial overshoot on synced moves
I think basiclly what it needs is an offset added to the target position that covers the accel delay
right now, pos = spindle_pos * pitch + starting_pos
if it was spindle_pos * pitch + starting_pos - accel_dist, I think it would avoid the overshoot
similarly one of these days I want to work on configurable "position lag tweaking" for rigid tapping
I should play with that at workshop
on this thread (M12x1.75), there is about 2 turns or more of unusable Z before it steadies out
jmkasunich: my high accel machine hides that very nicely
does it get better if you go slower?
I almost ruined a part by starting too close
I'm not gonna try - these parts take close to 20 mins at 330 RPM
I'd like to be going faster, but the facing operation is the limit
gotta get that VFD so I can do CSS
yeah you really need that
new motor too right?
yeah, its a big project
I think ballscrews come first
CSS is very cool.
the sound of threading with that overshoot is very distinctive
time to change parts
you know as much about spindle sync as anyone now - please do play with it
#15 is running, I think that will be the last one today
I hate turning the corners off of square stock
yeah that never works very well.
my "roughing" cuts for the facing operation are only 0.030 deep and 1.5 ipm
bout 0.0045 per rev
must be a big diameter if you're at 330 rpm
2" square, so 2.8 at the corners
241 SFPM if I did the math right
its 12L14 steel
yeah something like that
woo that's pretty fast then - using carbide?
so I'm definitely pushing it (HSS tool)
soon to be a HSS ex-tool
carbide probably wouldn't like the interrupted cut
meh, 6 corners on a $2 insert...
and how many times to I have to restart the program after a corner chips off
fenn_ is now known as fenn
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emcdevel/2008-03-28.txt
ok - I tried to wing it as far as I could. the card loads but I am not getting any pins.
[13:26:04] <skunkworks_> http://pastebin.ca/960765
not sure how to set the address and dir
where did you get 0800c0h? I see e400 in your lspci output
heh - each port has an address - c0h is one of the plugs on the board.
0800 was from the lspci
dir=0 seems like it should give you all in or all out
fails with e400 will not load. 0800 will
does it say anything helpful in dmesg when it fails to load?
* cradek <- newbie
now I get a bunch of errors in axis - 'error registering pci8255.0...
with the 0xe400 address. That seems promising.. :)
rtapi_print_msg(RTAPI_MSG_ERR, "registering %s ... %x %x\n"
looks like I left some debugging crap in there
I bet this shouldn't be ERR...
or... not :)
there are a lot of ERR in there that shouldn't be
I am getting pins now.. and a realtime error :)
if you want to recompile and get rid of them, simply do a global search and replace of RTAPI_MSG_ERR with RTAPI_MSG_DBG
I'll check it in, I've already done that
cradek: that would be great
I think I should play with this on a dapper machine.
EMC: 03cradek 07TRUNK * 10emc2/src/hal/drivers/pci_8255.c: oops, these are debug info, not errors
cradek: it needs to be changed in 2.2 as well
jepler: how does the dir work? is each 0 for 4 bits?
EMC: 03cradek 07v2_2_branch * 10emc2/src/hal/drivers/pci_8255.c: oops, these are debug info, not errors
skunkworks_: The 'a' and 'b' ports are switched 8 bits at a time; the 'c' port is switched 4 bits at a time. reading the source, I am not sure which bit is for 'a' and so on...
but it's a bitmask, right?
yes, it's a bitmask
4 bits for each of the 3 connectors
so if you specify it in hex, each hex digit will control a,b,c1,c2
of those 4 bits, one of them switches the direction of "A", one switches the direction of "B", and two switch the direction of "C", 4 bits each
put 0x at the beginning just like you did for io=
should be easy for skunkworks_ to figure out the rest :-)
try 0x111,222,444,888 and see what bits you get
then, write a manpage...
that'd be great
[14:07:09] <jepler> http://emergent.unpy.net/index.cgi-files/sandbox/direction.out
whoops, forgot dir=0; fixed
what a strange ordering
1=c, 2=b, 4=c4, 8=a
ah - so the debug outputs where probably what was throwing the realtime delay error?
it's related to the way the 8255 direction register is specified
skunkworks_: no, it shouldn't be related
WRITE((dir&3) | ((dir & 0xc) << 1) | 0x80, ioaddr, 3);
oh - well - I think I will try it in a dapper machine. Then I can easially get trunk and play.
(the low 3 bits to the 8255 direction register are the same as the low 3 bits of the dir specification)
I see that (low 2)
C-low and B are sort-of related, and C-high and B are sort-of related. There are various modes in which you can use them a bit like a data bus, with C-bits acting like read/write/strobe or something
I forget the details -- the hal driver only uses them as dumb I/Os
will it also control the relay?
* skunkworks_ just needs to play with it.
skunkworks_: it's supposed to, yes. I never tested it very heavily.
hal_pin_bit_newf(HAL_IN, &(inst[i].relay), comp_id, "pci8255.%d.relay", i);
sweet. I should have seen that when it finally exported the pins.
I don't think my hardy crashes are related to load; I had the system running all night rebuilding the kernel package
cradek: does this look accurate to you? http://emergent.unpy.net/files/sandbox/homing-types.png
I think so
EMC: 03jepler 07TRUNK * 10emc2/docs/src/config/ini_homing.lyx: a table to clarify the different types of homing (e.g., how to get index-only homing
EMC: 03jepler 07v2_2_branch * 10emc2/docs/src/config/ini_homing.lyx: from trunk: a table to clarify different types of homing
EMC: 03jepler 07v2_2_branch * 10emc2/debian/changelog: note documentation improvements
one last time.. What is the command to run halcmd from trunk? .mumble
skunkworks_: . scripts/emc-environment, then just type halrun or halcmd
thanks - exactly what I needed
I have had hardy lock up on me once.. Screen when black. I don't know if it was static though.
huh, apparently automake already works for nonrecursive make (long as you write all your stuff in one .am file?) http://www.mega-nerd.com/erikd/Blog/CodeHacking/nonrecursive_automake.html
jepler: do you remember what the 12v terminal block is for?
jepler: dapper system says http://pastebin.ca/960918
skunkworks_: I think that is power for the relay
disabled is bad?
* skunkworks_ is just trying to get the relay to work :)
I cannot get the relay to fire.. http://pastebin.ca/961064
heh - the relay is showing as an 'in' bit
the digital I/o seems to work :) cool
I should say O
so the card is working
* skunkworks_ cannibalized a floppy cable
* skunkworks_ found an un-keyed one.
wow - it outputs 0v for logic 0 and a hefty 5v for logic 1
* skunkworks_ seems to be talking to himself
don't feel bad; we all do it sometimes
so the question is - what am I doing wrong with the relay.
I am running 12v into the card from the computer power supply.
skunkworks_: if I remember looking at the board, the logic-level output from the PCI bridge chip operates an opto that can allow the 12V to flow through the relay coil. No 12V, no relay operation.
double-check the polarity?
I thought maybe I needed the write relay function - but that didn't help
I will check again. went by the polarity written on the board.
yes - ground is towards the motherboard
With my jerry rigged setup - I sure can get it lost easy. It seems to go into open collector mode. Then only a reboot will fix it.
It acts like it does until emc is loaded.
what goes into open collector mode
right now I only have a dvm. When the compter is booting - the dvm voltage jumps around. Then when emc starts - it goes to 0 volts.
when I zap it - or what ever i to - the voltmeter jumps around like (not physically) like it does when the compter is booting.
I have not gotten it to do it for a while.
I have flood coolant turning on bit 0 of the first header. I am excited :)
restarting emc did not reset the card. Reboot did
mist and flood.. I am the man.
* alex_joni was wondering what skunkworks_ is on
high on friday I guess
* alex_joni heads to bed
good night all
jepler: when you get a chance - could you try your pci8255 card? I put it in a different computer and still don't get the relay to trigger. The port changed on this computer 0x1000. The normal header I/O's seems to work great.
skunkworks: what's the relay for?
fenn_ is now known as fenn
(I mean - I don't know what it was ment to do.)
I should be able to activate it thru emc but can't. Either my card doesn't work - or I am doing something wrong or the driver isn't quite right :)
I mean jeeze - it is 72 i/o plus the relay - I want it all. ;)
is there a modern (meaning works with Gnome) version of "write" - so one computer can send a message to a user at another?
having the CNC pop up a dialog saying "drill a hole, dummy" doesn't work when the dummy is staring at a different screen
jmkasunich: maybe using pidgin as an IM client
I could just send an IRC message to myself
I think there's a cmd line IRC client that can just connect, send, disconnect
is the other machine a linux machine too?
there is a thing called WinPopup
supposedly it can work on linux too
smbclient has support for sending messages over the protocol
cat mymessage.txt | smbclient -M MACHINE
on the remote machine you can set up a smb server which does something when the message is received
e.g. opens a gedit with the message loaded
* skunkworks thought alex was going to bed
yeah, well.. that's how it goes