Why not write to disk whenever it's changed?
jmkasunich: are you here?
Why not write to disk whenever it's changed?
After all, goo tables aren't changed that often.
cause maybe a program wants to make a change that only lasts for the duration of the program?
I have an application where I'd "change" it every run
Is that an alternate way of doing your measure then adjust before final cut?
there are three ways to do it:
1) "touch off" after the measurement: pro - the DRO is accurate, con - all tools are affected
2) correct the cut, not the offset: pro - fairly easy, con - DRO reads funny
3) correct the tool offset: pro: DRO is accurate, only one tool affected, con: don't think it can be done today
4) like 1, except "un touch off" after the cut.
true, dunno how difficult that would be
The offsets *are* available as vars, so you should be able to record the present offset, set a new offset, do the cut, and then change the offset back.
I had the impression that you were talking about a lathe for this. Is that true?
yes, this is lathe work
Can't you do tool diameter compensation on a lathe where "diameter" is actually length?
I don't understand that question
EMC already has X and Z comp for lathe tools
plus nose radius
On a mill, we specify stay to the right side of the path and the tool diameter is .125 (or the tool diameter correction is .001). Can we do that on a lathe?
I'm not sure of all the details for lathe
Isn't that what you want?
Can it be mapped to that?
this isn't about compensating for a known tool length, it is about correcting that last 0.0003 when the length (or radius, or whatever) in the table isn't quite right
whether that is because of spring, temperature change, whatever
so I need to be able to cut, measure, enter measurement, and make a final cut
I understand that the reason is different. But if you specified the "diameter" as .000 for the initial cuts and then changed it to be .0047 for the final cut, would that do the trick?
probably wouldn't be using the diameter term, more likely the X term
and the initial value would not be zero
You don't have a turret, gang, or other automatic way to change tools, do you?
If you had a gang lathe, you could put a probe in one station and do the measurement automatically?
I have a QC toolpost, but I don't have a probe
How about mounting a probe on the back of the cross slide so that you could do it automatically.
nor is the machine accurate enough to measure
it is impossible to measure a diameter with a probe on the machine
because you have to approach from both sides to measure diameter, so lash comes into play
plus, the two measurements are made about 1" apart in X (diameter of part plus diameter of probe ball), so X screw errors come into play as well
You would have to calibrate the probe. Of course, the tolerances you seem to be talking about seem to be pretty tight.
I'm trying to do work to tenths, on a lathe that can only do thousandths "open loop"
I'm basically trying to close the loop by getting very close to final size, measureing with better instruments, and then doing the final cut without changing anything else
Do you have a set of gage blocks to calibrate your micrometer?
not needed in this case tho, since I'm making a part to fit a bore
Is this all for your spindle?
as long as I use the same mic to measure the part and the bore, absolute accuracy not critical
collet back diameter goes into spindle bore
goal is to have something like 0.0002 to 0.0005 clearance
I'd probably be happy to have 0.0001 to 0.0010
Someone on CNCzone had a nice article about tuning ballscrews to fit bearings. Final few tenths done with emery cloth.
yep - I'll probably be doing that - I did it on the first one I made
How many do you plan on making?
but uncorrected, my lathe tends to give me +/- 0.001 or even +/- 0.0015
if I need to leave allowance for sanding, I need to make my target at least 0.002 large
that is a lot of emery
so anything I can do to tighten it up is a win
What is the rough diameter>
the first spindle bore is 0.8805 as well as I can measure, but I think for the rest I'm gonna aim for a 0.8750 +0.0005 -0.0000 bore
since the bores are _very_ to cut and rather hard to measure, I'll probably make them first, and make the collets to fit
so for the bores, I'll just run the program
Are you reaming the bores or boring them?
Or grinding them?
just boring at the moment, might do some grinding or lapping
doesn't help that the spindles are case hardened
the first one I did was 0.8805 in the main part of the bore, but 0.879ish at the front, where the case intersects the bore
I'll probably revise the machining so that front part is 0.885 or so, puts it "in the air"
the design is evolving over time ;-)
On another subject, I'm thinking of using HAL to run the thermostats in my house.
that seems like overkill
I have ten zones of hot water baseboard heat. Zone valves occasionally stick, thermostats go wonky.
After a powerfail, the thermostats lose their settings. (The batteries are always dead.)
after a power fail, PCs tend to reboot
But not notebooks of PCs on good UPS's. And with the right scripts they tend to reboot OK.
yeah, I suppose if you auto-restart the HAL stuff it wouldn't be bad
if you get freezing weather, I'd have a backup t-stat set at about 50F anyway
I'd like to be able to know that a zone valve was stuck open without having to go around to the rooms that are never used and check the temperatures.
you could certainly do that
So, I'd have ten temperature sensors, ten zone valves relays, ten sensors to tell if a valve is open (simple bimetallic sensors at 90 degrees or so.
Then I could change the settings using a slider in pyvcp.
aren't the bimetal sensors redundant?
And monitor things the same way.
hotter than commanded = valve stuck open
Yes they are. True. Or it's a warm day, or the room next door is set higher and air is flowing from one to the other.
"lots hotter than commanded" ;-)
You might be right.
Did you see the discussion about real time serial ports?
didn't make much sense of it, and didn't really care
From a practical standpoint a hal driver for that would be pretty easy. Except that there are a few different chips out there.
the guy never mentioned what he wanted to talk to
That is a bit of a problem. I'm interested because I'll need some sort of data acq system for my thermostats.
getting temperature into HAL you mean?
That has to talk to HAL.
do you have existing sensors of some sort, with A/D converters, and you need to talk to it serially?
Yup. I assume that I can do all of that in user space.
No -- hardware has yet to be sellected.
given your several-second timing requirements, you can probably do everything in user space
I'll probably try to buy some kind of commercial beast. It's not clear yet what to do for temperature sensors.
wiring will be the nasty part
if you have lots of rooms, thats lots of wire
unless you go crazy and do something wireless
I already have regular thermostats in all the rooms. So, I could replace them with something else.
One wire sensors might be neat -- although they aren't real precise, they should be good enough.
regular thermostats with minimum 2 conductor cable back to basement or wherever the furnace is?
Yup to the basement. -- Five foot high ceilings 200 years old. Think dungeon.
New wiring is possible, though.
I don't think HAL has been used to control a dungeon before
how many conductors in the thermostat cables?
Platinum resistance temperature sensors are accurate and cheap. But they require four wires to do it right and I have only two.
I think my tstat has 4 because it can control air conditioning, fan, and heat
That's pretty common.
two wires means you need to send power and signal on the same pair
For the upstairs, I can drop wires from the attic into the walls in most of the rooms.
do you like electronics?
The so called one-wire sensors are actually two wire. They should be fine for that.
yeah, one-wire is marketing speak
I do electronics. But I'd rather buy than make.
they don't count ground - dumbasses
My twin brother is a veterinarian. He does electronics and does most of our electronic layout.
the one wire stuff will involve "make"
they are chips, not "units" that are ready to mount and connect
at a minimum you'd need some bypass caps, terminal blocks, etc
My current long term (other) project is to put together a laser interferometer. I just fired up the laser for the first time, last week.
And a board to put them on. And a driver. And I don't think you can put a dozen sensors in parallel when you have lots of long wires going to them without having rise time issues.
I'm thinking of using a serial pluto board for the laser interferometer electronics. Hey, I could use the same board to interface to my thermostats.
It has lots of I/O. I could use one for each of the one-wire chips. Implement the protocol in hardware.
sounds like a lot of programming and/or hardware
and you still need to build the sensor units
(A job for someone with lots more time than money. Probably also more time than brains.)
I may take a look on the web for one-wire thermostats.
you might also consider 4-20mA instead of digital
My big budget items this year is probably going to be a change over from oil to natural gas for heat.
you can filter the snot out of it on your central board, less sensitive to electrical noise
I hadn't though of 4-20ma, but I bet I can find something that uses platinum resistance sensors that would do the job.
google finds lots of 4-20mA temp sensors, but the prices are ridiculous
Hell, even plain thermistors are pretty good. If I don't mind making little boards, a pic a thermistor, and a little hardware should give me a 4-20ma homebrew sensor.
what is the pic for?
just put the NTC on the end of the two wires
The pic could linearize the sensor.
bias circuit and A/D at the central location
That should work. I've used some 10K 1% thermistors that are pretty cheap.
nice thing about thermistors is they usually have a relatlvely rapid resistance change per degree
compared to RTDs or (shudder) thermocouples
not so great for accuracy over a large range, but that isn't a criteria here
40 to 100 F is more than enough
The ones we are using are good to a fraction of a degree over that range. We use them for temperature correction in our anesthesia machines.
if you get +/- 1 degree around 70F, and +/- 3-4 degrees over the range, that should be good enough
there you go - you already have a source
And at 10K ohms, serial resistance from the wires won't be a problem.
are the ones you use potted or otherwise "ready to mount"? or would they still need a rudimentary board?
I think Digikey has them. Made by GE, I think.
if I was doing something like that project, I'd put one sensor outdoors too
They have an epoxy coating. Solder them to the wires. Heat shrink the connections.
are the existing thermostats passive - just bimetal contacts?
Sure, extras are free.
Some are passive. Others are computerized, multi-setpoints, days of the week, etc. A PITA to program.
mount the NTC inside the thermostat housing (for appearance sake) and wire it across the contacts (if passive)
I may have to add a day of the week component to HAL.
I can even use PID control.
I love it when a plan comes together.
It's past my bedtime. Thanks for your ideas. I'll see you later.
could it be a sign error on the hostmot2 issue?
TP commanding a move in a certain direction, hm2-stepgen moving (or reporting) the other direction?
they eventually match up, so I wouldn't think so
match up? why?
I would expect it to wind up and run away
what if motion commanded 0.502 ?
I wonder why that wouldn't happen for Seb
he commands things by hand from HAL
he said he's not running the driver from emc2
ah - now I get it ;)
the only thing I can think of atm :)
if seb types "setp <blah>position-cmd 1, and the stepgen goes to -1, you'd think he'd notice :)
I don't know if the position-cmd and position-fb are the actual pins or internals, but if they're actual pins, then position-cmd is what motion is asking for
yeah, my scenario only works if they're internals _and_ don't match the pins
sounds like the classic runaway scenario to me :)
hmmm. so here's an interesting question (to me anyway)
what happens if you have a position file and you use something like the hostmot2 stepgens?
I mean, theoretically, position_file tells EMC where the machine is, but then again, in machine off state it updates command position from motor feedback
and what happens to the motor-pos-cmd pins based on position.txt?
SWPadnos: I think that's internal to motion
but even if it changes, it gets changed while machine is not on yet,
it's got to be an internal joint offset I guess
since feedback starts out at 0
I think it is
until we get absolute encoder support
you don't actually use both :P
actually, absolutes may work already, as long as there's some startup code to read them :)
position remembering and absolute encoders :)
the general solution for absolutes is to allow machine zero to be offset from encoder zero, but I don't know if that's common
yeah, homing sure needs a bit of work
you need to startup already homed
heh - see you
was a short l
jepler: did you check that translation update? the diff shows some odd things
(might be my email client, that's why I'm asking..)
-msgstr "Selezione velocit. movimento angolare"
+msgstr "Selezione velocit. movimento angolare"
the first one looked ok, the second one looks like it has a wrong char in it
capital A with ~ ontop for the second, vs. `a for the first
alex_joni: no, I didn't look at it. I don't know Italian, after all.
speed selection for angular motion, of course! :)
the new translations seem nice, but charset seems a bit borked
-"Content-Type: text/plain; charset=ISO-8859-1\n"
+"Content-Type: text/plain; charset=UTF-8\n"
aha, sounds better then
and it probably is my mail reader
the character set changed, so the diff will be nonsensical because it has one encoding in "-" lines, another encoding on "+" lines
yup.. with UTF-8 it looks ok.. sorry to have bothered you for nothing :/
Our sense of "open" is that the authority to make decisions about what gets distributed based on merit and understanding and participation and leadership, not solely on employment or a title or a business plan.
Winifred "Mitchell" Baker, chair of the Mozilla Foundation
[18:01:11] <rayh> http://www.motherjones.com/interview/2008/09/inside-the-firefoxs-den.html
remember the multiple-user-interfaces-several-second-delay thing? it's back (in trunk)
I really thought it was fixed
(or at least worked around somewhat)
UIs are all updated by polling, no?
let me come up with a good how-to-reproduce here
hm, might be AXIS-only
ugh, it's the mode-switch magic
change halui-halvcp to use axis. machine on, see how f3/f5 in axis are responsive.
in halvcp change mode, then try f3/f5 in axis again
it's not just modes
poke feed override + in halvcp, then try f3/f5 in axis again
it's only f3/f5. if you poke f9 in axis, you immediately see the spindle on/off message being sent in debug
the pause is at the first wait_complete() in ensure_mode
ah, so it's timing out waiting for a command it didn't issue?
I thought I fixed all those once before. you shouldn't wait on a command unless you just sent one. it's just a bug to do that.
arg, version 1.1
yes, pretty sure I did this once before, but I must have missed some
it's easy to get wrong
wow, you can get tons (ha!) of good used real machinery, possibly wanting an emc retrofit, for the cost of one of those cnc desktop mills
or about half of a Tormach
the sherline cnc lathe is $2500. If you don't count buying the trailer, my very real lathe + retrofit (if time is free) cost less than that
I know they have different applications, but jeez it's a surprising comparison.
yep, and I'm looking at an HNC also, which would be less than a Sherline (minus the trailer ;) )
it's a very cool machine
does yours say "super precision"?
I don't think so
there's a CHNC-II super precision on eBay I think
I hear, but cannot say for sure, that that name meant more than a difference in the control
I imagine the screws are different as well
here's a nice one, too bad it's 2-3x what I want to pay: http://cgi.ebay.com/Hardinge-CHNC-Super-Precision-Chucker-Model-CHNC4-16C_W0QQitemZ300229030525QQcmdZViewItem?_trksid=p3286.m20.l1116
oh, and 1 ton too heavy :)
much more modern control - with a crt even
heh, 1 ton is next to nothing :)
might even be a microprocessor in there
wonder if it has encoders or resolvers
this is the affordable one: http://cgi.ebay.com/Hardinge-HNC-Chucker-2-5hp_W0QQitemZ190201838136QQcmdZViewItem?_trksid=p3286.m20.l1116
difference between HNC and CHNC is just the control isn't it?
manufacture date and CNC vs. NC
the CHNC are newer in general. I don't know if they made them both at any time
2250 seems high unless it's very close and convenient to get home at no cost...
though the CHNC4 would be a more recent one
(I paid extra for my mill because of its proximity, and it was smart to do it)
not bad at all
a little bad :-)
and interstate the whole way
I went further for the mill
and we grabbeda lathe on the way back through (not for me)
I gotta order shipment for the next robot I'll install in a couple of weeks
that's a bit more of a hassle than 1-2t :P
SWPadnos: what kind of power do you have?
a 240V 40A panel I pulled from the main panel, and a 5HP rotary phase converter
my hnc was stamped as wired 440 but I found it was 240 like the guy said
so you might check before you buy - could be trouble
it's almost universally possible to rewire a transformer to switch between 240/480
unless it's a delta/wye thing
but there are a dozen transformers (really) and only the couple biggest ones had straps
and I bet the coolant motor is not strapped
no, all single phase!
then you can solve it all with a big transformer :)
well I bet you don't care about the control anyway. forget what I said.
if the control works ... :)