there is the minor matter of splitting it
crap & marsh-mellows...
yeah it smells bad...red oak wet does
yes...how big are the chunks..
2 man to start with or handleable??
way big. Some is split, but the other is a bucked up 4 foot diameter tree
I left a 16 foot saw log
but it is so heavy we may not be able to get it out
i hate when i do that
It's too big for the wood mizer mill
use the roughing chainsaw on it first
mill about a mile away could cut it
or go there
I am trying to get my cousin to rough it out with a big chainsaw
it's gettin it on the truck that is the problem
i used to be THE cousin...
take the saw to the tree
it's 4 feet at one end...about 3 at the other
the woodmizer is on a trailer
but the log is too big
slabs for mounting fish for the next 10 yrs...LOL
all the limbs firewood of course
take the saw start at the big end and nibble her down..
tons and tons and tons
36" max on the woodmiser
has a 24 hp honda engine
make 3-6 " slabs or as long as 1 pass will take... then roll 90 degs
1 pass on your chansaw blade
it's things like "roll 90" that are issues
hope we caN DO IT
chain block... or come along... block up the other end so it wont move but can roll around
My tractor FEL can only do 220
do a wrap over and pull it over..
but cousin has a real big one
2200 not 220
We'll tackle it as soon as it warms up and I can get away from the lab work a little
but as I said...Im under a bit of pressure
breath... 3 deep breaths...its what we wanted...
right.. its gonna be GREAT
no try... just DO
I think this one is a winner.
They don't all work...but this is looking real good.
thats the process variabilities...
mine looks REALLY good too.. thats why in installing it at my fathers cottage first.
well I have to invent it but also invent how to make it
all I can say is...
there will be lots of computer automation
just like the mirror heaters.
id like mine to be a pc and monitorable over the web... i can adjust a system in timbuckfudge
well...time for dinner and music room.
Don't spin your hang glider!
not so bad in the HG ...BAAD in the pg...
for what it is, not a bad price... damn sears whats $40 for a all plastic one http://www.harborfreight.com/cpi/ctaf/Displayitem.taf?itemnumber=3888
didn't turn on the sound system
I had a nasty with the high power ultrasonics today anyway. A bad reflection. I didn't really here anything. but just felt a funny dizziness when I turned the resonator knob.
I'm gonna melt my brain or something...
hear not here
les_w Ear Plugs, use em, love em
no I think this went through my head
I hasd on ear protectors
muffs + plugs maybe?
don't know...not sure ears were involved at all. didn't hear anything.
turn the knob....fell a surge of kind of dizziness
maybe time for noise suppersion material on the walls to absorb accidental exposure?
it is sound loud enough to burn things.
I can see it now.... "Local Engineer burns himself to death with soundwaves"
just be careful with that shit, alright?
it's about 200 db
beats the music room
* Jymmm shakes his head
normally ultrasonics has no danger to hearing
but this is nonlinear....so can have subharmonics
also the nonlinearity makes multiply tuned resonators complicated.
they can suddenly "lock in"
i.e. quantum effects
the physics is much more like detonation shock waves than linear sound propagation
perhaps I did melt my brain.
well, just so it isn't the pleasure center.
and it'll NEVER be that.... Murphy's Law
what are these, endmills? http://www.harborfreight.com/cpi/ctaf/Displayitem.taf?itemnumber=34640
it doesn't say
they are either pcb router bits or drilss ro both
I have a bunch
for $6.50 I want them to be endmills =)
bad catalog...mystery tool.
I guess I can call them and ask
I think they are drills...from the number size description
bummer, I could have really used micro endmills
for that cheap, I can play around
plenty at drill bit city.
they raised their prices
kinda ouch too.
at the first of the year.
the things are about $5 new.
screw it, it's only $6.50
These are resharpend ones
um, not from DPC
well, I rarely use them. they break too easily.
My cam won't let me do inlay even with 1/16" bit as it says it's too big
oops skype call
your Columbian Drug Lord?
ah dropped out
they want me to develop some part for a cell phone still
I could sell you my phone and you just reverse engineer it =)
i'm booked up
when it rains it pours
why dont you just open up a engineering service, then outsource things to others
you proof it, then take your 60%
+ royalties of course
are these any good? http://www.harborfreight.com/cpi/ctaf/displayitem.taf?Itemnumber=34233
whoa the skype turned into a lot of politics.
no off skype
still on here
oh, ok =)
just found this too... http://www.harborfreight.com/cpi/ctaf/displayitem.taf?Itemnumber=93301
I think I am going to have an exciting spring and summer.
exciting or swamped?
don't you want the $3000 itw one?
lol, um, sure... why not, I have money to burn
I remember the demonstration.
It did a whole frame in about 3 seconds
I know it's gonna be a POS, but maybe I'll learn something.
yeah, grab it.
did you see the gun?
the EXACT same one is going for $89
I see some problems with that
I think the press type will be better
the tab gun or the v nail?
suspect the v nail will be better
but not sure
i'll vivit amp when I'm up there
the vnailer is for the frame, the tab gun is to hold the glass in the frame
like glacier tabs, but different
I am used to seeing staples
but I am not in that business
let me get a better pic for you...
[02:16:31] <Jymmm> http://cgi.ebay.com/TAB-TABS-REFILL-PICTURE-FRAME-FRAMING-POINT-DRIVER-GUN_W0QQitemZ8260886378QQcategoryZ37575QQrdZ1QQcmdZViewItem
that was the tabs themselves, here's the gun
[02:17:22] <Jymmm> http://cgi.ebay.com/Picture-Framing-Tab-Gun-includes-300-Flexible-Tabs_W0QQitemZ8259498295QQcategoryZ11786QQrdZ1QQcmdZViewItem
ok got it
that auction wants $50 for what HF seels for $15
It was a fluke that I tried HF looking for that tab gun
These tabs are what all the framing places use.
I had some drawings I did and some paintings my grandma did framed but I just took it to my uncle
then I hung em on the wall
A friends of mine paid $60 for a photo, then $80 to have the photo framed.
they vharged $100 or something per
frames are just four sticks, some glass, and a mat, good profit margin.
I figure that I could make some create frames
not alot of lumber involved, design once, reproduce many
easy to store
We have talked about doing a run of frame material with walnut and old gray barn wood
since I have autofeed large shaper machines...no problem
$20 a foot retail
but I AM SO BUSY
damn sticking shift key
well no shop space here, so I have to be reserved about storage
lots of old barn wood here too
cut it up into little squares...and inly it right into the walnut
lol, there ya go
I could make a lot of it
and the machine is not being used
I might have one of my guys take that on
let him figure it out.
damn.... I should move out there and kick your ass to the electric shop and make the wood shop hummmmmmmm
we have talked about something wood coming out
that I don't need to be involved in
frame material would be ideal
Yeah, just have some creative designs and your off
something to keep the spindles turning
they are not turning much, because I have to do other things
and none of the guys you employe have a clue?
one i am thinking of could do this
he is a retired owner of a sign shop
H likes golf so I have been havibg him work on my little golf course we are making
just figure out how to automate the mat cutting and you're set!
so he could do this
yeah I think so
but would it beat a mexican with an exacto knofe?
not so sure
we were talking the other day about doing a run of those butcher block workbenches
but the frame stock might be even better
running a shaper isn't hard
and glue up the barnwood inlay
one guy could make a hundred...no hundreds of feet a day
There ya go
no cnc carving...just shaper work
the machine is just sitting there..
really? how are you going todo the profile?
no big deal
the shaper can run 4" cutterheads
power feed of course
...of course =)
just stuff the stick in
run it slow...like 12 feet per minute...then little or no sanding
it can run faster
Well, sounds like you have a lot of potential to keep the shop busy w/o too much intervention from your part.
but the sanding labor catches up with you
yeah this sounds like something the guy could handle.
materials cost about zero
and shot shoot it as it comes out of the machine
and just shoot it as it comes out of the machine with paint/clearcoat
possibly light sand with a formable wheel
commercial operations often just use a contoured wick to coat it
looks almost like carpet material
well us po folk just use spray cans
the wick just rubs on the finish
the formable wheels are way good
ever seen them?
never heard of em
let me check
found em yet?
it is a flap wheel that takes the shape of a contour just by feeding a metal template into it...saw a demo at iwf
used in machines like this:
[02:48:39] <les_w> http://www.dellevedove.com/
wow.... that's kinda cool
so I ask this...why is retail $20/ft for picture frame material?
because they can
obviously inexpensive machines like that can spit the stuff out fast
how much is a machine like that go for?
who is getting the money?
not more that a car I would guess
* Jymmm was hoping like 500-700 =)
now chinese can run those just fine
again, who is getting the money
I don't see frame shop owners drivin ferraris
becasue it's cheaper to buy from china
and most keep their photos on cd, instead of a frame
i'll have to check my uncle's margin on those frames
most of the frames he sells are cast urethane foam
I would say 90%
It's just a facet that I have to consider, and if I can get the (cheap) tools for under $50, it's worth a shot.
I guess they make about $50k or something in the frame shop part of their business
something like that
plus the actually framing/matting too
they cut mats
glass too I guess
Add on top of that custom designed frames (carved), sky is the limit
the ritchies go for the solid wood barn wood stuff
I never expected my friend to pay more for framing than the photo, so it made me think.
seems like a good thing to check out
This is the SF area, lot of ppl are into that arts stuff
then there's Monterey and Carmel too
you know, when my guy is working on the course he is not turning a profit
and the fairly expensive machine is idle.
well, not expensive.
a few thousand.
an idle tool turns no profit
a screwdriver only costs $0.99, but inthe right hands can make $1000's
well let's try this
i'll talk to my guy
we were trying to find something good for him to do
let him figure out how to dice up old barn wood into inly tiles
well I'm off
03cradek * 10emc2/docs/man/man1/.cvsignore: ignore generate file
giacus is now known as giacus_afk
giacus_afk is now known as giacus
bla bla bla
don't break the silence
giacus is now known as giacus_afk
renlore: now you just need to join #madscientists
oh nevermind looks like he was already there
this looks interesting. Wire an 8255 to the parallel port to get 24 I/Os. http://www.dontronics.com/optus/vicuni/prn8255.html
but one 8255 I/O will require several parallel-port I/Os, so the speed might make it unsuitable for high speed uses
jepler: that sounds nice, btw there is a HAL driver for 8255 based stuff
it should be fine for non-RT outputs
we have some large servos off of a cnc mill. Is there a good source for servo drives? (would love a DIY amp)
[15:31:35] <jepler> http://www.doc.ic.ac.uk/~ih/doc/par/pportexp/pportexp.asc
nice - going to have to save that
skunkworks: how big motors?
12 ft-lbs of torque 10 amps 30 amps peak (I think - give me a few and I will get the exact numbers)
sounds like .5kW minimum
stall - 14.5A peak 59
hold on -
stall - 14v 16.2amps 17 ft-lbs
a dual H-bridge I used for 10A, 12V motors
peak 59v 82amps 80ft-lbs
[16:00:48] <alex_joni> http://dsplabs.cs.utt.ro/~juve/yanor/downloads/bridge.zip
sorry - they are bib
mine weren't that small either, and the bridge should handle about 20A
or more.. but I never tested it
but if I were you I'd get some G340's
did you see - peak 59v 82amps 80ft-lbs?
the 82amps might be a problem ;)
that's 5kW peak.. no way near homebrewed stuff
if you are good with the 1.6kW the gecko can supply, then you're cheap to go
if not, then the only option is commercial servo drive (EXPENSIVE)
hmm - have to look
is the gecko a servo amp step and direction in?
yes, and DC motor+encoder on the other side
the G340 also has a 1:1, 1:2, 1:5, 1:10 multiplier
so 10 steps from the PC, 100 encoder ticks movement (on 1:10)
how do you think those would work - (not really having emc2 doing the loop)
I tried one, and it works like a charm
it does PID internally
saves you the cost of the Motenc or similar
and it provides a signal for ferror
ideally you would have a DRO board too, and feed the encoders to emc too
have to do some more research
thanks for the circuit - what was the zip file?
orcad schematic & layout
pcb layout I mean
the circuit you sent me - that would work with how jmk was playing with servos out of the parrallel port?
pwm signal for forward and reverse?
pwm signal for speed
and bit for direction
in my version
[16:17:36] <alex_joni> http://dsplabs.cs.utt.ro/~juve/yanor/images.php
<- that's what I ran with it
that is another reason why I think you are a proffessor
are you in the picture?
the guy in black
some reason I thought you where older ;)
how much older? :D
that was a few years ago .. but not that many..
I had a picture in my mind of a 40+ year old.
wow, even I'm not that old
so was that a final project for college?
those sharp corners look like they would destroy everything nearby including ankles
cradek - some reason I thougt you where in your 30's - I think we had talked about this before
skunkworks: going on 26 right now
cradek: I wrecked a wall actually ;)
cradek: with those sharp corners
skunkworks: in 9x I was working on the Apache project and one of the guys in the core team one day gave away the fact that he was 14
it was a big surprise to everyone
so what did you say? too old, we need a 12yr old ;)
someone who's not intelectually biased
no, some jokes were made and we all moved on
If he was good it shouldn't matter.
cradek: nowadays you find lots of young people with better skills than veterans
that's one reason I like the online world
otoh, I wouldn't trust a 14 yr old designing a 5kW servo amp
I am 32
alex_joni: or a web server?
- I built a 200 watt per channel audio amp in highschool - using darlington transisters and op amps.
had issues with thermal run-away ;)
always wanted to use fet's - but they where so expensive at the time.
orcad is still around - I remember playing with a demo when I was in highschool
renlore is now known as renlore-sleeping
renlore-sleeping is now known as renlore-away
cradek: software maybe, that's not that experience-related
skunkworks: nowadays they are pretty cheap
but that was real rms power - not he peak to peak to peak power - how ever they calculate that ;)
they are getting better though.
How you doing with your machine?
good - I added stuff to the oword page.
Are you using stepper_inch?
was looking for servo amps. alex gave me some good info
I saw some overshoot the other day. Have you been seeing any at high speeds.
[16:47:38] <giacus_afk> http://www.artwooddesigns.com/supplies.html
nice tool :)
no - not recently - did quite a while ago. Are you adding headroom in the ini file?
"Can Be Worn With Full Beards"
(one of the respirators on that website)
Yes for both accel and max vel. They didn't seem to help at 160 IPM.
hmm - my ini is set to a max of 200 ipm right now. I don't notice any overshoot.
let me look at my ini
Ok. I'll keep testing. Thanks
I have quit a lot of headroom mostly for accelleration - I have 12 for the ini 20 for the stepgen
I sure wish someone would try to fix the tp
3.33 for velocity and 3.9 for stepgen
cradek: ditto, but I think simple_tp should be enhanced ;)
that would be so much easier
but I think we need to expand the interface to motion
to let TP know about joint limits
that's pretty hard to do, because it needs to run kins to check that
I would test :)
I'll try to help if you want to work on it
cradek: I need to pull myself together for that ;)
right now I'm not in shape to do it.. too tired
too tired is easier to fix than not smart enough, which is my problem
it's not that complicated I think
you're overappreciating it
rayh: what are your settings for max and stepgen max? for accel and velocity?
are you running the latest emc2?
I have enough trouble understanding/planning a one-axis trapezoidal velocity profile because of the quantized time
^ as my wife says "you are talking in code again" ;)
cradek: lets sit down sometimes and see if we can pull our heads around it
alex_joni: I've asked jepler for help too, and he says "it's obviously easy" and then doesn't do it for me :-)
I can do the simple calculus to figure out the continuous-time case
we need to do something about that
we'll get him with something one day, then he'll have to do it
I mean jepler
he's like me - has to get pretty interested in something in order to bother
or, be very offended by it (like ugly screens)
lets make a GUI for the TP =))
not offended, more like digusted
well, AXIS turned out great ;)
thanks, we started from scratch, not knowing the insides of any of the existing guis, which I think was a good idea
sometimes studying the existing helps, other times it dooms you to making/keeping the same mistakes.
but yeah, if you can help me with simple_tp, that would be great.
I think once it plans the trapezoids perfectly, blending them won't be too hard.
cradek: haven't you fixed that trivial bug yet??
cradek: what's still wrong?
I can't figure out how to get the decel profile right, and end at the correct position
because of the quantized time
seems like you have to adjust the decel OR the cruise speed if there is one
in order for the last step (dV = -maxaccel) to put you at the right place
so without doing that it will end a tick earlier or later?
what I need is a function that takes in distance, maxaccel, maxvel, period, initial and final velocity: gives number of accel periods, number of cruise periods, number of decel periods
(any of those numbers may be zero)
* cradek waves his hands
alex_joni: it's worse than being a period off - you are likely not able to land at the specified coordinate at exactly a period boundary
that doesn't sound that hard
I know, that's what I thought when I started too
I'd calculate the number to reach max_speed (in periods)
the number of periods to reach final_speed
hrmm.. not that trivial as it would seem
I think my contention was that you should not quantize
yes, you often can't do those in the given distance
jepler: but you need the exact destination on a period
at the end you get a piecewise function x(t) which you can call with the 't' values that correspond to periods
your output is "how far do I move this period" and your answer has to be "same as last period, +- < A"
If the ending velocity is 0, tack on an extra cycle that outputs the final commanded position
if it is not 0, use the slightly wrong final position as the start of the next segment .. it'll be wrong by some small amount, in the vicinity of final_velocity / period
jepler: how do you mean that?
Sorry skunkworks I was on another machine. Let me look.
ahh.. ok, got it
but that easily gets messy when you start to do blending
and if the next move is not of the same type
linear vs. circular
* alex_joni goes looking at cradek's sources
In traj I've got DEFAULT_VELOCITY = 0.0167
MAX_VELOCITY = 3.0
DEFAULT_ACCELERATION = 2.0
MAX_ACCELERATION = 3.0
and in each axis I've got MAX_VELOCITY = 2.5
MAX_ACCELERATION = 3.0
STEPGEN_MAXVEL = 3.2
STEPGEN_MAXACCEL = 7.0
try putting the same max_vel all over
rayh: you are getting position overshoot?
it makes sense to have axis maxvel < traj maxvel
Not if you are using a 4th defined in degrees.
That was the purpose of individual axis max's
rayh: you misread what I said
rayh: "it makes sense"
even with all axis in the same units
it still makes sense because TRAJ is tooltip vel
Traj has to be 40-60 for max vel if a fourth is defined in degrees.
cradek: I see you have some duplicated code in TpRunCycle()
yeah, that makes the issue much worse
// max change in velocity per cycle allowed by accel constraint
max_dv = tc->accel * tc->cycle_time;
// displacement under which we consider the motion done
tiny_dp = max_dv * tc->cycle_time * 0.001;
Then the individual max vel reduced for linear
alex_joni: tiny_dp was stolen from motion.c
yes, but it's twice in there
Is what we are saying is "that we can not use different units?"
but when you have different units on different axes, traj maxvel has no useful meaning
rayh: no, what we are saying is "we can use it, but it always was defined awkwardly, so you might need some strange values for that"
rayh: is there a particular case (geometrically) that gives you the overshoot?
cradek: I see, it needs to get recalculated (if there's a new tc) !?
I get the overshoot on all motions in the same vector
I get them on circles at the quadrant.
when gcode is 1/4 arcs
cradek: although tc->accel & tc->cycle_time don't change..
rayh: are you talking about position, vel or accel overshoot?
position overshoot then return then start the new block
Do you think this is a motmod problem or a stepper problem?
I was thinking I'd try freqmod next
I'm sorry but I still don't understand what you're describing
First case is a rapid to a feed in the same axis.
ok so they're colinear
how do you see position overshoot?
rapid overshoots returns then starts the feed
the axis reverses?
about a 1/4 inch overtravel, then reverse then starts the next move
oh holy crap
so you have g0x1 g1x2, you get a motion to 1.25, then back to 1, then to 2?
the only time i saw that was when all the axises used the max accel from the traj secion. Cradek had fixed that.
I can see it in backplot
AXIS's backplot uses commanded position
Right and the tickle backplot uses actual position.
it would be nice to know if this mess is in commanded
cradek: if it's stepper_inch, then I don't see why it wouldn't be
We could watch actual and commanded in halscope.
that sounds like a good idea
or following error ie the difference between
I think that will tell us if it's tp or stepgen
I thik it's both, tp telling crap, and stepgen rattling along at it's max
rayh: what are your ferror's ?
yeah I'm afraid we're going to find that it's tp.
no change from dist
alex_joni: did you mention crap?
rayh: try increasing the maxes for stepgen, and see if it changes
If it were tp, then the servo stuff would show crap also.
ValarQ: you may lurk again
ValarQ: kidding ;) there is enough crap for everybody
rayh: PID might swallow something like this
indeed there are
rayh: filter it out I mean
I'm crying on the inside
wanna talk about it?
lol @ alex_joni
no, it's lunchtime
Is emc2's tp significantly different from emc or bdi?
Jymmm: not that way ;) I mean seriously..
rayh: no, it's the same crap
ok, maybe crap is a bit harsh
rayh: I fixed the accel problems with different maxaccels in different axes, that's it
but it's not 100% ok
Then I should see the same sorts of thing there.
there is one exception
stepgen won't go faster then it thinks it should
rayh: pid will absorb in emc1 even for steppers
even if TP commands it to
Let me work a bit on this. I can compare with univstep.
what kinda computer is this on and what is your period?
rayh: I would think commanded position will look the same no matter what the driver
athalon 64 with base_thread of something like 10
and what is your base period
he just said
must be nice ;) - wow.
BASE_PERIOD = 0.000010
SERVO_PERIOD = 0.001000
TRAJ_PERIOD = 0.010000
could not quit get there. could you send me your stepper ini - I could play with it here.
skunkworks: sure. dcc or your eaddy?
headed your way
skunkworks: it's best not to put email addys in here.. remember we are beeing logged ;)
it's best not to live in fear of your e-mail address getting out
jepler: if you have a good spam killer, sure
but most don't
alex_joni: who doesn't?
most I know.. :(
jepler it's not a matter of getting out, it's a matter of getting harvested.
jepler: and a matter of beeing sold ;)
I mean the addy on some obscure lists ;)
that is my junk email - use it when I don't care. thanks though
rayh is now known as rayh-testing
ray - so with that ini - if you do a g0x5 it will go to something like 5.25 then back to 5?
I think it only happens when G0X5 G1X6 (e.g. blending 2 moves)
(I changed the base period to .00002 for my computer)
alex_joni has changed the topic to: "Welcome to the Enhanced Machine Control forum -
Support and development of a linux based CNC control. |
Home:www.linuxcnc.org | Regular Developers' meetings every Sunday
14:00-18:00 GMT | wiki up @ http://wiki.linuxcnc.org
| EMC usage map:
[18:02:42] <alex_joni> http://www.frappr.com/emctheenhancedmachinecontroller
alex_joni has changed the topic to: "Welcome to the Enhanced Machine Control forum - Support and development of a linux based CNC control. | Home: linuxcnc.org | Regular Developers' meetings Sundays 14:00-18:00 GMT | wiki up @ http://wiki.linuxcnc.org
| EMC usage map: http://www.frappr.com/emctheenhancedmachinecontroller"
boy - not seeing that at all. still playing though
i've given quite a bit of thought to how to do arbitrary kinematics in realtime..
there is quite a lot more to worry about than appears at first glance
i think emc2 should focus on getting the tp to work for cartesian machines, as it will take a long time before kinematics will work well
the most elegant and simple way to do it would render most of the current code base obsolete
but i'm not sure that it will work - that is, if i can always use an N by M transformation matrix to get from cartesian to joint space
but there are a lot more complications like redundant axes and self-collision and arbitrary work envelopes and optimization - you dont have to worry about any of this with cartesian machines
ray: - could you send be the gcode that is doing it?
i think till started working on his redundant axis plotter to try to break down the multiple new challenges a hexapod architecture presents
into multiple smaller problems
fenn: I agree, TP for cart. for now
[18:25:08] <fenn> http://fenn.dyndns.org/pub/camera/DCP_0602.JPG
<- one month and 3 days later, the stuff that dreams are made of arrives
hey fenn, alex
i shoulda gotten some trim pots
I am not seeing it on backplot - or the corrdinates
skunkworks: I'm using cds.ngc
I see some of the overshoot and odd blending even with univstep.
les have you considered using mems-type techniques for your thingybob?
hey, i got a patent issued today. It's the encoder.
[18:32:02] <les_w> http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=/netahtml/srchnum.htm&r=1&f=G&l=50&s1=6,974,949.WKU.&OS=PN/6,974,949&RS=PN/6,974,949
oh, gratz btw
made sure the client's name was first
I tell themwe will do that
he gets a raise
ah yes i saw a similar design based on RF
I pulled some optical tricks
all made with emc btw
if you have accel and decel periods (in time), how do you calc the distance traveled
it amazes me that anything at all is patentable.. it all seems so obvious once you explain how it works
you have the initial vel
that's probably why there are so many stupid patents out there
they cant tell the difference
double integral of accel + initial velocity
hey ray - have you tried a period of .00004 - do you get the same thing>
cds.ngc is a gcode file shipped with the distro.
les can your encoder do >360 degrees?
les_w: integral (initial_vel + integral (accel))
integral of initial velocity+integral accel
yeah, I can see that, but how do you put that into a c-formula?
crossed in the wires
integration is just adding
les_w: doing discrete math
(initial_vel + accel * time) * time ?
add values multiplied by time increment
yeah that looks right
and a derivative is just a subtraction divided by time increment
* fenn curses the uspto website
so i guess i cant see any pictures
let me see
ray: could you send a screen shot of your back plot?
when I click images I see white cows in a snow storm fenn
perhaps because it just issued
many are basically toolpaths for emc
skunkworks: How much of an image do you want to see. The whole cds thing?
just so I can see what you are seeing. As far as over-shoots
les_w: when i click images i get a "click here to download plugin" image :) either that or it locks up the browser
(you said you could see it on the backplot.
I am running it right now.
with your ini files
les so, the "optical cam" has a different exit angle once you go over the hump.. how do you keep the LED pointing at the sensor and not have an abrupt transition?
Okay. The problem is that the overshoot and the next line cover each other.
if you did a g0x1 > y1 would it over shoot past the 1 on the x and then back?
then go the y direction?
Not as a regular thing. Let me see if I can capture an overshoot. Will be a few minutes.
ok - thanks
les_w: here's an easier question.. how thick is the optical member?
sorry was out a second
the trick I pulled is temperature compensation
leds, photodiodes, and plastic all change their optical properties a bunch with temperature
I devised a differential system where all that cancels
even the plastic properties
All optical cam profiles are machined on emc!
cost for the device is under $3.
We sure don't sell em for that though.
bah it was supposed to be $1
It's a throttle position sensor.
oh i thought this was the battery sensor
no that's another one...
I used the lm2900 series op amp
under hood temp qualified
and a quad cost $.11!!!!
LM2900 is discontinued?
3 cents per op amp
no not at all...there is a fistfull of them in every car
oh i bet it's because of ROHS crap
newer rail to rail ones are easier to design with
but they cost much more
especially if they are -40-125c
I did have a tough time sourcing photodiodes
unlike phototransistors they tend to be expensive.
I insisted in using pin pd because this is analog
PT are all over the place
I note law dogs wrote up phototransistor though
just to cover the bases
i thought op amps worked on voltage, not current
les_w: could you explain why PD vs PT?
ray: I am running this program - seeing if it reverses - nothing yet
Photodiodes have much more predictable output/light intensity specs
phototransistors are better for on/off stuff
more predictable wrt temperature changes?
* fenn is an information sink
fenn <--- infowhore! =)
What speed is your g0?
hey now them's fightin words
fenn: The truth hurts, don't it =)
nah i left a channel with tons of useful info the other day on principle
I thought a whore was someone who had it and was selling it.
no that's a pimp
les is an infopimp
No that's a broker.
I'm a RamWhore... I'll take all the ram I can get.
That's the context I say it in.
ray - it is your ini file - 180
Okay. I got a shot of about .100 overshoot in z. I'll edit and send.
I'd like to see it too
how did you do it and can you reproduce it? (short snippit of code?)
read what I wrote you?
curried vegetables with rice
yes, I think you're on the right track
should we try it?
with sim and some ddts it's easy to get a nice halscope plot of position, velocity, accel
ok, I'll give it a shot in an hour or so (finishing a movie first ;)
I don't see where distance comes into this
I think you're missing a few pieces of the puzzle still :-)
in the time check
skunkworks: image on the way.
ok I'm not quite seeing it
but I believe you
I guess I'm not either ;)
You guys seeing overshoot with sim?
rayh-testing: we were talking about TP (not the current TP)
rayh-testing: our guts tell us the current TP is broken so we're back to working on the new one
I wish you'd been with paul, matt, and I when we spent a day with Fred talking about this issue.
His impression is that there is a blending problem.
there certainly is
that's everybody's impression
you where single blocking this? so it ran to 2.1 - and over shot? went back to 2.1?
So why not work on blending rather than a new tp?
because the whole structure is bad, and I've cleaned it up on a branch
But your stuff only single blocks
and because a redesign usually is better then patching something
There is no way that exact stop will make a decent part finish\
rayh-testing: we're figuring out how to do proper blending too
and i thought jmk said end of month.
rayh-testing: getting exact stop 100% right before adding blending is necessary
might not be emc-2.0.0
but it would be good if we ever figure it out
I guess it's a matter of priority. Fix a bug or ...
machines tend to only get faster lately
rayh-testing: the bug is there for the last 5 years?
No it is not.
We never had this kind of overshoot with the emc
I can say that for a fact.
yes, because PID ate it away
but the output from TP was just as bad
BS It's ;there in univstep
and there is pid
we might want to take this discussion elsewhere.
rayh-testing: the code is the same
You said five years, tp hasn't changed in that long?
I was asking.. not saying
That's crap cause paul changed it as did others.
paul didn't change it
he took parts out of it which he didn't understand
which made it better in some cases, and worse in other cases
That does not constitute a change?
no, that's just fiddling with it.. not fixing it
It's a change whether we want to admit it or not.
fixing presumes understanding it first, and it seems no-one does (I'm not sure about FredP, never got the chance to talk to him about this.. :(
This arg is crap. The TP code has been modified
Like it or not it has and at the same time that these modifications were made it changed behavior.
[20:07:40] <alex_joni> http://cvs.sourceforge.net/viewcvs.py/emc/emc/src/emcmot/tp.c?rev=1.20&view=log
yes, I agree that it changed, and that functionallity has changed
but I don't see how that helps, I don't think it's better than it was before
but this is all talk without proof
you guys dont have to be so melodramatic, btw
rayh-testing: it's possible you could check out old versions of the kinematics directory from cvs, in order to compare the operation of earlier code
rayh-testing: they may need a small bit of tweaking to build
ray - can you reproduce it?
if you single step though again will it do it?
skunkworks: single step won't do it
does a one line program (or mdi) overshoot and reverse?
it's a blending problem
and it might be related to SWP's re-enabling the code paul took out of the TP
but again.. it's only guessing, because I have nfc what's going on in there..
<skunkworks> you where single blocking this? so it ran to 2.1 - and over shot? went back to 2.1?
<rayh-testing> So why not work on blending rather than a new tp?
<cradek> because the whole structure is bad, and I've cleaned it up on a branch
I assumed the yep was for me
skunkworks: what is single blocking?
skunkworks: I got lost in your conversation with ray - did you reproduce this overshoot too?
Paul took out something Will put in in 2000 to correct a problem of sometimes not honoring max velocity and accel
cradek - no - not yet
will's fix seemed to force honoring the numbers...but broke the TP
les_w: it's broken either way
les_w: I've studied it a lot (at every change) and I've never seen a proper parabola come out of it so far
single blocking is running one line at a time.
skunkworks: ok, I understand
about the only thing it does right is follow a path
skunkworks: but single stepping does NOT turn off blending
skunkworks: we'd have to use MDI or g6x to turn it off
les_w: thank goodness that part isn't broken
les_w: will alex and I finally get you on emc2 if we fix it?
so by removing the 2000 fix there is some kind of bad overshoot or undershoot in reversals?
removing the 2000 fix causes accel constraints to be violated by 2x
I was getting overshoot - unrelated to blending before. (pretty much as ray describes it) but have not seen it in a long time
any path errors noted?
skunkworks: I think that was a stepgen bug long since fixed
les_w: don't think so
- If I remeber right it really happend when you had the max accelleration in the traj section higher than any one of the axies max accelleration..
if the accel is violated up to two times the ini number, then divide the ini number by two internally?
we were getting (accel at -A in 0,0,-1 direction) + (accel at A in 0,0,1) = accel at 2A in 0,0,1
Is that what we are talking about? or are there othe rproblems
les_w: nobody understood the code that paul took out, and taking it out changed it from "wrong" to "differently wrong" so it's unclear how to fix it
les_w: another hack fix would be to block blending for things that are reversals or nearly reversals
At the time that Paul did that, he was in some sort of funk with WillS and refused to ask what the code did.
question: other than accel violations what othe bad things have been noted after the #ifdef?
I'm sick and tired of funks
rayh-testing is now known as rayh
rayh: it's not like we don't want to fix it..
I know that.
les_w: I noticed colinear lines were still misblended (I could hear the vel spike)
but after all that time I feel the effort to try to understand & fix it might be bigger than just writing it again
My concern is for the path to fix that we are taking.
I don't know if he really wants me to say this, but Paul claims to have a totally fixed tp.
rayh: cradek & I are trying it out in a branch, is that bad?
You are certainly free to do as you choose.
well so far only cradek ;) I want to jump in and see if we together can get it shiny :)
les_w: too bad he's not cooperating with us anymore.
Well I've heard a lot of stuff go by about new "shiny" trajectory planners.
rayh: agreed, I don't trust them till I see them
Seems like most everyone has a better way
well I for one put my money where my mouth is
I have not seen pauls work
same goes for threading, since you just mentioned "shiny"
ray - did you try upping your period - .00004 would be more than fast enough for your feeds.(just to see if that is part of the problem)
skunkworks: don't think that will do any difference
rayh: how about that plot of commanded vs. actual ?
Didn't know how to do it.
ok, I must say I'm just about on the level with you :D
didn't play with halscope much either
I would like to know too - using halscope?
I believe that is the definition of following error.
well, so far I hear an accel violation only. I can type in half the number I really want in the .ini, or throw in a /2 somewhere...
cradek knows ;)
les i believe the problem is it overshoots the endpoint
in a reversal only?
les_w: It only seems to happen in some situations.
other times it issues the correct stuff.
ok. So the problem has not been characterized fully yet I guess
Is it know to be tp and not some other thing like step generating?
ray - so you don't think you can make a short program that will reproduce it consistantly?
les_w: not fully decided, but I incline TP
les_w: that's why we want someone to plot commanded vs actual position
skunkworks: This cold has me pretty heavy headed but let me try.
rayh: do you think I can run your ini without hardware connected?
Sure you can. It's a stepper_inch slightly modified.
Want me to send the file.
cradek - I am running it here no problem
skunkworks: ray's ini??
are your emc versions different?
mine is the latest
head or testing?
ok, so testing
ok that's the latest testing
which I think is the same as head except for some man page stuff
yes they are the same except for mini, man page stuff, a few makefile fixes
rayh: I don't see anything..
I'm almost positive I saw hard-coded exact stop for any segment angles over 90 derees.
My net is down, Sneakernet takes a while.
you were using tkemc?
No mini. It has a bit easier to use plotter.
got it here now
Normally it should of course never rech a reversal point in blending mode, unless pid is mistuned to be underdamped.
cradek: replicated it?
no, got the ini in my email
I'm not sure I can run it here
This is athalon64. You may need to increase the base_period a bit.
my available machine is 266 MHz
I'm going to try to write a short program that will show the overshoot issues.
ray - did you download the 64 ubuntu? or are you running the 32 bit verions?
are you running ubuntu?
hrmm.. still can't see it..
cradek: can you mail me the ini too?
it was also sent to you
oh wait, ray already did ;)
I sent it again just in case
oops, now you'll have two
ouch, I forgot to change period
.00001 runs on mine - barely
on mine it's just fine
Was thinking...If you clamp accel too late in the tp process...it will not calculate a slower average speed to allow path compliance
so a clamped decel will overshoot
there is another clamp somewhere else
rayh: now I just need some idea of what you were seeing, I can't see anything..
were you using feed_override?
I immediately get joint 2 following error
with ray's ini
probably because of missed RT deadlines
I changed it to 50ns
or because you use the speed he has in there
there's no way I can run it here
you might scale them all
but that's .. error prone
that has never gone well for me
yeah. that's an understatement
I get an ferror on FO=300%
but only after those lines ray mentioned
as far as I know, FO won't command anything F9999 in the program won't
<skunkworks> ray - the 32 bit verions?
I have the 64 but havn't tried a rt kernel.
rayh: can you describe what you were seeing?
The first place that I saw fe when period was to large was on the first z down to feed.
Where the z down at rapid switches to feedrate is where overshoot is very prominent.
It is also prominent at the quadrant changes during the circle mill.
ok, so you basicly were seeing fe?
I'm not sure I know what to look after
It stops, sometimes backs up a bit and then begins to feed.
and you see that how?
big zoom on the backplot?
by watching backplot.
ok, trying 4 now
rayh: it would be great if you could grab that corner in halscope
I captured one overshoot while in step
Let me try.
you could trigger on Z position less than [right above the corner]
I've got a sample program that shows stop in x and a reversal or two in a circle.
I'll copy that over and send.
tp.c #ifdef is in on this right?
les_w: currently paul's change is taken out
still can't see it at zoom=4
because with it out my full circle arcs stopped at the beginning of each quadrant
but it was g coded that way...4 90 degree arcs
with the #ifdef in it was one smooth circle.
so paul's change helped your machine a lot
oh, yeah...I wouldn't dream of taking it out
but I have not found any of this revrsal overshoot
les_w: in emc2 we're much more sensitive to constraint violations now - they immediately trigger following errors
les_w: your machine won't ferror even if you're having twice the requested accel (I guarantee you are sometimes)
rayh: is that surely the ini file? I see only 3 axes, and [traj]MAX_ is the same as [axes]MAX_
I know I have (probably) accels and velocities that violate my .ini....but if the path is followed not much problem really.
cradek: at least related to steppers
les_w: ok I understand; with steppers it's not the case that you can violate constraints and get away with it
It does it okay there though.
right...I can go 10 in/sec and accel 150 in/sec^2....
files on the way
so I can kinda let the tp pick what it wants
Thank you guys.
back in a few
rayh: I ran it, still can't see it
you sure it's not a backplot problem?
I see that I'm running with feedrate override at 300%
I wonder if this is a problem where scale gets multiplied in.
Let me run here with tripple the feedrate and scale at 100
at fo=300% I can't see much.. (zoom=3)
the updates in mini's backplot are about 3cm long on my screen
giacus_afk is now known as giacus
ok, now I SEE it
G0X1, and it turns black at the end, readout reads 1.2275
g0x3, readout 2.7725
no, I think that's step-related
The run at f120 and 100 scale does not show any delays at the blends.
I think at high speed I can see some delays at blends too, but I'm not sure what that is..
it might be display related
I'll try to scope it
You can increase display rate if your box will handle it.
I'll try after I run this test with halscope running
[21:35:27] <giacus> http://www.getdemocracy.com/
ok, I see no ferror whatsover
[display] CYCLE_TIME = 0.050 shows it really well here.
does this error show up in AXIS with the same gcode btw?
Axis displays commanded not actual position.
rayh: I just made a recording with halscope
there wasn't any difference between commanded & actual
good thought, fenn
no difference I can measure on halscope (probably something like .00001" or less)
alex_joni: do you see reversal?
"actual" is the output of stepgen?
fenn: feedback from stepgen
actual is what is passed back to emc as current position.
I don't know what stepgen does to produce that.
ok, now I see it
I'm of the opinion that what I'm seeing is a feedrate override problem.
in the backplot
it slices it into little bits and makes sure you dont exceed some velocity/accel constraints
If I run 300 percent of feed it is really obvious.
let me see if it's in halscope aswell
rayh: running 300% now
for others thats 300% of 40 ipm
I see no blending issues at 120 ipm and 100% override.
he was here recently
i got the quote for the FMS i WAS considering... 1.35 Million US donaros..
15:20:56 <les_w> back in a few
(that's 27 minutes ago)
rayh: I got it, will investigate it, and let you know when I know more.. ok?
I didn't say the time units...
i can set up alot of individual machines for that $$$
reading back to see what happened...
The effect of override is very small at low feedrates in the circle.
you have to RE-MEMBER yourself...from one life to the other though
But it is really exagerated on the inline switch from rapid to feed.
very peculiar??? hmm
override is a plc function?? correct.. and the TP may be farting into the wall... where G01 and G00 transition maybe a pos'n window overlap or someting like that
try an exact stop prior to the change over??
at 120 ipm and 20 or so in/sec^2 massive misblends were the rule before the #ifdef
however it followed the path
so you might not see it on a backplot
but would a CMM plot show it??
in the part
well I could see it in the part
it's very very violent
on a big machine
i just finished a part i KNOW is gonna be BLUE dots all over..
so this is a feedrate overide issue again?
is that what I am reading?
I could be...but we found no evidence of that before
at 100% is all sweet??
not sure from what ray said
and at + 10% it starts...
@ 50% its wild??
my feed overide was always fine
bcz it is possible to over run the drive systems
bbl - late
dmessier: at 300% it's a bit wild
i have GOOD lookin' part... with a row of green (good) dots ad a red dot -.02" in the middla of them all
dmessier: you need to look very hard to see this
and push the machine hard, you won't probably see anything
at 300% overide YOU programmed the wrong chipload in the first place...
our NDT inspectors find EVERYTHING
etch & mag on all high strength steels
54-58 RC... machined post h/t
one part in house this week 62 Rc .. m/c all over post h/t and ndt and plate... i was given 5 days
i told them on a wish and a prayer it was 3 weeks minimum
Alpha.... we gotta talk
I was about to go to my car... and get my laptop.
you wanna make some $$$$
sure I do.
who do I have to kill... or impregnate?
im building a team to man a plant.. part ime..
what's the job criteria?
keep typig, I'll brb.
oh you werent the the one that got to my bitch... as i forgot her out last nite..
one of our divisions.
burnese mountain dog is warm to cuddle
I say our...should be their...
I am a separate company
but have an ITW 401k....
things getting a little confusing
use them to flux your magnets... ; )
'ight, so like what'd be the job requirements?
ndt coil shots are GREAT...
be there.. sort out the problems on the machines....keep her going...
dmessier, that's it?
req'm... initial investment.... and commitment of time...
like, make sure the motors are working... on the job training?
get' her done attitude...
safety equip paid for?
do I get to wear a red spiffy shirt?
goggles, steel toe boots?
oh i have a red shirt for you.. QRT .... Quick response Team... SHUT UP or ill bury ya...LOL
hey got a patent issued today robin
I'm wearing a red sweater right now.
les_w: coo, on the ferret mating spray?
but I look good in it. :)
dmessier, 'ight so what are the hours like?
you interested..??? the team a bunc a brainiacs
dmessier, give me the details... I'm interested.
no, it was on the female obedience brin beam
sun is in my eyes
les_w: or on your ultrasonic finger-burner?
no project before
[22:18:14] <les_w> http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=/netahtml/srchnum.htm&r=1&f=G&l=50&s1=6,974,949.WKU.&OS=PN/6,974,949&RS=PN/6,974,949
finger burner is PAF though
got the name order right too
coo. a sort of analogue pot, but without a carbon track ...
first name is the guy that sends the checks
it's mostly for a sneaky temperature comensation scheme
you know, this works. I just flat tell the client "i'll get you a patent"!
All he does is send the po through
he gets his name on it and a good review...
sorta a varying track and a fixed reference
I get $100k.
les_w: do LEDs get dimmer as they age?
im having similar patent fun on some s/w at the moment ...
and all this time I've been burning my fingers the old fashioned way.
personally, I cant see how its patentable ...
les_w: of course phosphor-based ones do
but, the boss has paid some guy with an office on Park Avenue to tell him that it is .. so, I guess it is.
reference is mothing new...but puttinf a sample of the optical cam material in such a way that temperature induced changes in refractive index and specular reflection is
there is a special prism of the material in the reference optics...
that piece is moulded on the same sprue in the injection mold
as the cam
and you can make a cam thats linear, or at least linear enough?
so basically it is temperature stable from -40 to +125...but costs 3 bucks.
I can do 1% custom tapers
we have sold a bunch already
and when a spec of dirt gets in? it goes full throttle?
haha just dont tell them that
ok, when a loose bit of crud left in during manufacture falls across the sensor?
I get sued...
I also designed the production testers
its not like it's a fly by wire system.. (is it?)
just put a bit of double sided tape on the inside of the lid :)
that's sort of a neat idea.
but then I like anything involving LED's
so, can I have your mouse then?
no, I like it. (it involves led's)
and all your servo encoders
oh wait ..
I read thats as i DONT like ...
on my mouse at home I replaced the blue wheel-light with a color cycling led.
the double sided tape is a cute trick used in laser servicing ...
after a mjor blow up, you replace the mirrors with blanks of double sided tape, vac it down and run hte trubines overnite
so all the cruft sticks to the tape?
how's life at the stables?
full of small boys :)
and lots and lots of work
took on a new welder this week too
thats two of them now
one of the production jobs would do great with a bot
hehe.. they always do :)
I can imagine how a bot welds nicely up am internal corner
arc voltage sensing and weaving
what about external corners ... does it need cpacative height sensors onthe faces for that?
usually laser scanning for that
but that's 1..1.5 the price of the bot ;)
realistically, a second-user basic bot sells for?
I wouldn't buy one that's cheaper
long call about voltage coeficients vs ESR on PZT materials
told the ceramics blenders to maximize V^2/ESR
in the 31 direction
alex_joni: like this maybe?? http://www.goindustry.com/offerfiles/839000/839416_1.jpg?TimeStamp=0x000000000F805BE4
10k tops for that
with welder included
did ray vanish? what of the tests?
alex_joni: well, it has no welder ...
alex_joni: current bid is sub 1000 :)
like err, 300 eur
Hi Les. Was making supper.
We have figured out how to cause it consistently.
thawed corn beef and cabbage frome the big cook last sunday for me.
scale or feedrate override is the offender.
yes, but TP is the problem, once again
But it seems to interact differentially with different types of blends.
well feed overide is an integral part of the tp
I.E scaling in tp/tc
I still suspect it is clamping early in the pt process, then checking and reclamping later
that would be consistent with the #ifdef ANd feed overide inputs
the check might not see the overide clamp earlier
tp not pt
tp is a program...pt is a car
and a damn ugly one too.
well, both are.
just so I understand....
and with that in mind, I go to bed
paul #ifdef in.....overshoot depending on feed overide
#ifdef out, no overshoot but horrible misblend?
the tp still not right?
not in emc2.
It's better in emc 1
It does violate max accel by 2 to one to get the right path.
that is a non issue.
hey, divide by 2.
les_w: same in emc2
alex_joni: seriously for a monent ...
2:1, but that's an issue with stepgen (which is very strict)
well apples and oranges, but mostly the same dna huh?
a bot like that could be made to weld?
no, TP is the same one
robin_sz: probably so, but it would be non-trivial
so there are weld bit and "other" bots?
there are bots.. and there are a 'few' weld bots
is it just a software thing?
most of the manufacturers make them for all app.fields (like fanuc, motoman, yaskawa, etc)
this a "univeral" bot :)
very few (I know of one: cloos;) makes them ONLY for welding
I may be looking at bots to do resonator trim...looking at impedance curves in a fourth order system and moving poles...with lasers perhaps
but I think you might be able to handle one that has other tasks in the SW aswell
theres another with a Fanuc R-J2 control
No doubt I'll hire one of you guys to write the code...
* alex_joni is off to sleep
robin_sz: don't know R-J2.. sorry
any wavelength blast bright aluminum robin?
unless I black anodize it