somebody else might have an opinion anyway...
I'm looking for a cheap (<$50) microcontroller board (already built) with a Free compiler, etc for linux
need to count a couple encoders, read some analog signals, generate a bit of pwm or maybe just some dig outs, etc
so far this one is looking good: http://www.futurlec.com/ET-AVR_Stamp.shtml
but I wondered if anybody here has favorites?
jmkasunich: I recently played with mbmgecvt/mbmgecvt.c: ASCII Pascal program text
[01:24:19] <jepler> http://arduino.cc/
atmega128 is a much bigger chip but no faster than the one on arduino boards
gnu gcc for avr is pretty good
I was looking at the arduino
seems to use some strangish language
you can program it in "C" if you like
but yeah part of the arduino gimmick is this little IDE that is easy to install on windows
how fast do you need to read quadrature -- motors, or mpgs?
the prices I saw for arduino were in the $33 range
wheels on a floor
I haven't done all the math yet, but I think its 10's of KHz max
maybe single digit khz
engineering week project
mousetrap powered vehicle with a twist
cool - I would not want to be going against you :)
it has to do stuff, not just go
pop two balloons, pick up three washers from the floor, and stop inside a 24" square 50' from the start line
the balloons and washers are NOT on a straight line from the start area to the target square
all forward motion must be powered by the trap
jmkasunich: sounds fun
but you can steer, etc, with other stuff
so the trap doesn't have to power the uP - thats good ;)
although it will probably result in late nights and frustration as time runs out
skunkworks: yeah, that would suck
I'm thinking two encoders made of mouse guts, on two unpowered wheels
trap powered front wheel (tricycle), with steering, maybe an RC servo
dead reckoning using the encoders, plus some line following - the course is outlined by black electrical tape, and the balloons are on the border - the washers are in the middle though, so that has to be dead reckoning
the washers are known locations I assume?
the course is 8 feet wide
both balloons are along the right edge
the washers are at 3, 4, and 5 feet from that edge, and 25, 30, and 35 feet from the start
the ballons are at IIRC 15 and 45 feet from the start, or something like that
it doesn't have to be kinetic does it? you can power the wheels with some sort of gear increaser (string to a gear box of some kind) ?
so you go right, pop a balloon, then head for the middle for washers, then right again for the other balloon, and back to the middle for the end target
speed is worth points, but not lots of points
jmkasunich: the analog isn't particularly fast, but you can start a conversion and continue running code, polling it or trigger an interrupt conversion complete
I'm thinking of using the analog for line sensing
"up to 76.9ksps" at reduced resolution. there's also a penalty for switching channels
maybe that's still plenty fast
yeah, if I can sample each of 3 or 4 sensors at 1KHz that will be plenty
I suspect that even if we go for "blazing speed" the thing will be doing 3-4 feet per second
so thats only 1/20th of an inch per millisecond
(that's atmega168, the chip on arduino; atmega128 may be different but I suspect it's pretty similar)
I hope this gets videotaped..
it probably will, but probably won't be posted on a public website
if its posted internally and not huge, I'll try to snag a copy
the atmega128 is same conversion rate. looks like it has differential inputs with 10x and 100x gain, though
does the 128 have those, or is that the 1280?
this document says atmega128
the gain is no big deal - I can do op-amps if needed
I'm more interested in your thoughts on tools (compiler, etc) and ease of use
hmmm - I've never noticed that in the '128 manual
gcc for AVR
in the past I've used a laptop running linux+rtapi+hal, but I think that would be too heavy, and also I/O limited
nice dev environment tho - halscope and everything
jmkasunich: sometimes the flash programming can be fiddly
you have an opportunity to brick each chip when you program the fuses to set clock source
is that an "every time you change everything" kind of thing?
the avr tools package (which I can't remember the name of at the moment) uses the $36 Atmel programmer, which should work reasonably well
change anything I mean
no, only once
if you have a board with a crystal on it, then it's a lot easier to not-brick chips
you can re-flash without setting the fuses again
(and even "bricked" ones can be re-flashed if you apply a clock to the clock pins (except for the AT90S1200A I think)
- there is a programmer at the bottom of the poage
it has a crystal
there's no clock connection on the standard ISP header though
there's an SPI clock, but not an oscillator connection (that goes to XTAL1 and XTAL2)
[01:45:25] <SWPadnos> http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=ATAVRISP2-ND
given that the board has a crystal, and I'm not interested in slow clock power saving modes, I'm not too worried
is that CD linux friendly?
but there's a package if you turn on universe
search for AVR
I saw lots of AVR stuff in synaptic
I'm liking the $5 programmer that goes with the futurlec stamp
avrispv2 is supported by avrdude, the common avr programming software
right - avrdude, that's the one :)
those prices seem too good to be true - 4.90 for the programmer board, including the cable?
so is "pony prog stk200"
jmkasunich: the programmer is just 2 digital I/O from PC, 1 to PC
jmkasunich, I wouldn't use the futurlec one - it depends on parallel port timings, and as such may (will) change characteristics when you change PCs
jmkasunich: the board is signal conditioning
that's been by experience with dumb programmers anyway
still, a board, a ttl chip, a few resistors, leds, and caps, and a cable, for $5 is cheap
besides the arduino bootloader I've only used dumb programmers .. they're a bit finicky, maybe it's why i brick all my AVRs
the AVRISP has a microcontroller (two, actually), which takes caer of all timing
jepler: hello. i'm hooking up a 2nd 5i20 and more motors to simulate a 5 axis machine. just like the one in your utube video. could i borrow the gcode that runs in that simulation?
the second micro is the "bootloader" for the one that does programming - it takes over and programs the main chip if you start it up in update mode
Roguish: I believe it's "cone.ngc" from the w_tool_length branch on CVS
jmkasunich, how big (small) does the board need to be?
no real size limits, other than if its as big and heavy as a laptop, I'll use the laptop
smaller is good of course, but not critical
the AVR butterfly is pretty cool - has an LCD even
I think that's the one anyway
when do you need this?
I saw that in passing, but it seems to have much more limited I/O - things are dedicated to the sensors on the butterfly instead of availalbe
I probably want to decide what I'm getting in time to order this week, so I have it next week
that's partly true, but the pads around the board can have headers put in there, and they give access to something like 30 or 40 I/Os
the contest is feb 22, but it takes time to come up to speed, and to build the rest of the hardware
you'll have a lot of I/O on the et-avr, less on the arduino
I can bring you a butterfly if you want it
thank you. found it.
SWPadnos: offer appreciated, but it will be too late - our wicheta weekend is the last weekend before the contest
ah, ok :)
a $20 or $30 board isn't gonna bust my budget anyway
the trouble with AVRs is no quadrature input
yeah, I figure I'll have to count in software, like hal-encoder
can these things handle a 10-20KHz interrupt routine?
yep, but you'll probably want interrupts for that, with tight filtering / counting code
yes, but it really depends on the interrupt code
its supposed to be 16ish mips, so that would be 800 instructions per 20KHz interrupt
assume 30% for the ISR and the rest for the main program
thats 240 instructions
that's 800 to 1600 cycles from interrupt to interrupt .. not an absurdly large budget
yeah, you may need to do that in assembly, and tell the compiler to leave some registers alone
a lut based encoder counter can be done in a few dozen clocks
the interrupt prologue that gcc writes is pretty big, because it stores and restores the whole register file
the AVR is really good for multibyte simple math, unless you have to fetch/store to SRAM
I just remembered that GCC assembly language is this really screwy strange dialect
you mean inline asm? you can also write it in separate ".S" files .. then it's just the operand order that is surprising to DOS refugees
ah, thats good to know
(and I think that with modern gas you can specify intel order if you prefer it)
it'for register operands, it's one cycle per byte of precision for addition and subtraction (in some cases, only one cycle for two bytes), but when you have to load/store, you add two cycles per byte per access, so 8 extra cycles for a 2-byte load then store
its been long enough ago since I did x86 assy that operand order won't matter much
50-50 shot there ;)
encoder counting is "get current bits, mask in old bits, use as index to lookup a code, and based on that code, either inc, dec, or do nothing with the count
if the code is +1, 0, or -1, the inc/dec/nothing is just a (wide) add
you can be more efficient by using bits to decide whether to increment or decrement
there is a single-cycle add/subtract to /from word instruction
no branch penalties eh?
there are also skip instructions though
I've been doing CISC too long I see - I tend to avoid branches
so skip if ... / addw / skip if not blah / sub takes 4 clocks
are bit tests fast, or do you have to mask and such?
I should look up those adiw / sbiw commands - they may be two clocks
bit tests and moves are pretty fast, shifts aren't
err - sbrs
one bit shifter, not barrel
definintely want to use assembly, C would make you do ugly things
* SWPadnos whistles in ignorance
if using assy, no shifts would be needed at all
you can do 4-bit "shifts" with the swap instruction, which swaps nibbles
there's also the "t" status bit, for moving bits around
the address for the lut (16 bytes) would be <oldA><oldB><newA><newB>
and the value in the table at that addr would be <newA><newB><inc><dec>
so I just store that looked up value
I bet once you get the table done, and you've pushed/popped pointer registers, and loaded base pointers, and then done the logic, you'd be just as well off with a mask / eor approach
next time, I replace <inc> and <dec> with <newerA> and <newerB>
the Z register (the only one usable to point into flash, which you need for a table lookup) will require two pushes and two pops, plus two immediate loads
that's 4+2+4 cycles for the push/load/pop
is pointing to ram easier/faster?
if you're really cycle counting, you'll want to move the LUT into RAM during startup .. a bit faster to access than program memory
its got 4K, and I need 16 bytes
you can optimize that by aligning the table on a 256-byte boundary (note that that's a 128-word boundary), and load the low byte with the table offset
no, program rom and SRAM have the same 2-cycle hit
maybe ROM is 3
but there are more registers that can point to sram, right?
ok, it is 3
SWPadnos: what's the difference between RAM and SRAM?
yes, there are 3 that can point to RAM, but there are no two-pointer addressing modes
jmkasunich: yes I think you can index with any of X Y or Z to (S)RAM, while only Z can index for program memory
you can do one pointer + a constant (compile-time constant) offset
fenn, in this case, I'm using them interchangeably
the RAM in an AVR is SRAM
but you can't do Z+Y
yeah .. but just page align the t
are X, Y, and Z addressable as bytes?
so that the computed part is the low byte and the constant part is a high byte
so the pointer math is identical, but you do save the cycle because load from ROM is 3 instead of 2 for load from RAM
yes, that's what I said before :)
jmkasunich: yes, X is shorthand for r27:r26
jmkasunich, yes, Z is R30 and R31
* fenn doubts any of this matters for a mouse bot
there's also a W (R24+R25), which can be used for some word functions, but isn't a pointer
I'd put the LUT on a 16 byte boundary, load the high byte of the pointer with a constant, load the low byte with <prev_cycle_lut_value>, stuff in the low two bits with I/O pin values, do the lookup, and store the new value away
such as add immediate to word
since the algorithm only uses the low 4 bits of the LUT value, the high 4 bits can be bits 4-7 of the address
but again, once you do all that, then actually act on the two important bits (do I add, do I subtract), then add the overhead of a possibly extra set of register saves, the advantage of a LUT may be lost
you'd want to zero those, and use a 256-byte boundary
(using 256 boundary seems unneccessary)
if there are bit writes, I don't need masking
you have two bits (old), which hopefully would be stored with other bits as zero
you or in the two new bits
this is the low byte
with a 16-byte boundary, you also have to or in the high nibble
not with a clever table
that's probably a lot more clever than necessary, since there's no penalty to using a 256-byte boundary :)
(yes, table values have the high nibble of the table address ...)
perhaps, but clever doesn't cost cycles
it may cost readability. consider the C / ASM code to put that nibble there ;)
and with only 4K of ram, I'd rather not force any large chunks to be wasted
you don't waste anything
unless you need 3.75k+1 byte, and there's nowhere to put the table
it still uses only 16 bytes, it's just got to go somewhere on a boundary
as long as the compiler is smart enough I guess thats true
but I shouldn't say that, as I don't know the linker controls with gcc
I'm thinking back to masm days, when you issue an align directive, it skips ahead to the next address that meets the alignment restrictions
actually, what I do is specifically start the data segment somewhere
and put some things where I want them using .org statemenrs
but I've even seen the (IAR) linker put short code blocks inside the interrupt table, when I didn't use all the vectors
ok - clever tools replace clever programmer
an 8-byte function would get stuck between interrupts 5 numbers apart (so 10 words)
I don't know if there is an align directive - there should be
jmkasunich: i hear the futurlec programmer doesn't work with standard pinouts - you need a patch or something (it might be included in avrdude now though)
fenn: it sounds like I might want to get a better programmer
you'd have to build a crossover cable to get that board to work with the "standard" programming pinout though
that board being the ET-AVR
oh, a proprietary header, even better
well, the header matches the programmer, no doubt
that could be standard, but with the parallel port pin conections changed
you are talking about the programming header on the ET-AVR, right?
the cable from there to parport will probably be handwired anyway, so that doesn't really matter
I'm more concerned about timing sensitivity
if fenn is correct that people have to change things to get the futurlec programmer to work, then that means that there are two places where the screwup can be
* jmkasunich fires up synaptic and installs everything with AVR in the description
one is the parport to programmer connection, the other is the programmer to target connection
I've had tons of bad luck with PIC and AVR programmers that used the parport, and almost none with serial and USB programmers
so it may not be luck ;)
maybe I want a better programmer - but I kinda like the futurlec stamp
did you notice their development board for that?
[02:23:32] <SWPadnos> http://www.futurlec.com/ET-AVR_Stamp_Board.shtml
I saw it - I've never been fond of solderless breadboards, so I kind of skipped over it
heh - it's got other stuff, like lights and buttons, that could be useful
I can do lights and buttons
one thing I don't like is that I haven't been able to find the datasheet for the ET-AVR, so it's not possible to check pinouts
there's also a solder-able set of holes under the solderless breadboard
I figure I'll be doing a perfboard anyway, with signal conditioning and drive electronics
oh btw futurlec takes months to deliver sometimes
[02:26:11] <jepler> http://www.ett.co.th/download/03AVR/03A12-00279/ET-AVR%20START%20KIT%20V1.0%20%20EXP_Schemetic.pdf
of course - no docs from futurlec, but from the manufacturer
deja vu again
fenn: well, that would be the nail in the coffin
futurlec has great prices but that part is certainly a project killer
they have to ship from thailand to australia to us
jepler: the schematic you posted is the breadboard, not the stamp itself
jmkasunich: oh oops
I didn't look too close at what I found :(
yeah - I'm looking for the stamp now :)
but I only saw the jtag connector, not the normal SPI programming connector
if you want a "big" AVR, sparkfun has http://www.sparkfun.com/commerce/product_info.php?products_id=35
jmkasunich, I missed your actual I/O needs - what are they again?
hah you didnt expect specifications did you
2 encoders, a few analogs (photocells or such for detecting black tape on the floor), and a few digital outs (either RC servo waveform, or PWM, for a steering servo or motor), plus a couple more digital outs for misc, like maybe a solenoid to make the thing stop
well, no, not really ;)
I have a couple of boards I've designed that could do that, but I'm not sure I put chips with A/D on them
I see that futurlec is based in AU, and has no phone number so I can't call and say "I want this only if its in stock in the USA"
so they suddenly are very un-interesting
they've got things like transistor outputs (for solenoids), terminal strips for wiring, switching supplies for wide-range DC supply etc
SWPadnos: I'd rather have something cheap and off-the-shelf replaceable
especially if it turns out to be a fun toy, I might want more later
well, you can't beat the price, if I send you one or two fro free ;)
get you hooked ;)
oh, and a serial port on one of them :)
what is AT90CAN128?
hmmm. that's an interesting point - do you need comms with a PC so you can tell it how far to go or whatever?
it's got a CAN controller in it?
I need to be able to do some kind of debug and development, so a PC would be handy
yep - it's more or less a 128 + CAN controller
[02:37:05] <SWPadnos> http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=ATSTK500-ND
it's $83, but it has all I/Os available, buttons and lights, connectors for all ports, and the programmer is built in (as well as a DC supply)
I think it comes with 2 processors as well, maybe megas, maybe not
damn. those were $300 when I bought my first one
sockets for dips...
I'd rather have something stampish
buttons are a waste of space - I will need to wire MY stuff, not their stuff, to the I/Os
all I/Os are connected via 10-pin headers that have 8 bits + power and ground
you need to cable to the buttons and lights, else they're just on headers
so I need cables to go to my perfboard(s) instead of just plugging it onto my perfboard
yes, or you can plug the perfboard into the expansion headers - those have all the I/Os on them
but the perfboard would need to be bigger
that sparkfun board jepler found looks nice
then you can fiddle with RTNet too
$37.95, 1.85" square, all I/Os accessible, crystal oscillator onboard
ethernut looks interesting - haven't found prices yet
it looks like they use the 6-pin programming header
I wonder how long it would take to get an Arduino
hmmm. the arduino may not have enough I/Os
looks like a 28 pin dip instead of the plcc or whetever that other package is
yes - 23 I/Os max
including analog comparator and A/D channels, and USB/serial
bummer. I used the ATMEGA162 on both of these boards, so no A/D
just a single comparator
the sparkfun board is nice and simple - they have schematics and such - it uses the Atmel in circuit programmer pinout
I think there are 20 usable I/O on the arduino
jepler: where did you get your arduino?
sparkfun has them listed at 34.95, but out of stock at the moment
to be honest, for $3 more the mega128 board seems nicer
there's also the "stamp" version of it, a few bucks more, smaller, no USB .. http://www.sparkfun.com/commerce/product_info.php?products_id=8164#
no USB or anything though
still low I/O count
oh, it says it uses the mega168, IIRC swp said that has no A/D
no that was 162 .. there are too many models of avr
168 has 6 A/D
right -0 sorry for the extra part number in there
[02:58:43] <SWPadnos> http://www.sparkfun.com/commerce/product_info.php?products_id=31
just add chip ;)
[02:59:34] <jmkasunich> http://www.sparkfun.com/commerce/product_info.php?products_id=35
chip included and twice as fast
twice as much money too....
oh - I have dozens of those (or something almost identical to that)
12.288MHz crystal though, I think
atmega32 is the same speed as 128, just less ram and program space
(and it comes in DIP)
wow - bummer. those would be much much less expensive for that product
i think almost all avr's are 20 mips now
megas and newer tinys are, I think
the actual chips are pretty cheap - $4.93 for a mega168
hmmm. nope - there are a bunch, but nowhere near all of them
well, the new ones
hmmm. nothing with more than 64KBytes of flash can go to 20 MHz
if a 28 pinner would be big enough, this is hard to beat: http://www.sparkfun.com/commerce/product_info.php?products_id=29
plus a $5 chip
that's the same as the one I linked to, except for a smaller chip
same price even
the chip for that is $8.80
newer chips have more flexibility with interrupts too
I wonder why those boards have 8MHz oscillators? the chips can do 16
there is a clock multiplier
all chips can use the 8 MHz osc, even the low voltage ones ??
uh - I don't think so. there is a divider in some of the chips
all chips are low voltage, just the ones marked low voltage have been tested to function correctly at low voltage
and the LV ones aren't specced to run at the higher frequencies
when operating at low voltage
uh, i might be wrong about the multiplier
well, the specs for a given chip are generally either low voltage and low freq, or high voltage and high freq
but they're probably the same chips, so you could likely run the LV chips on HV at HF :)
all that matters is that I'd find myself looking for a source of 16MHz crystals if I bought one of those boards
jmkasunich: you dont need the crystal actually
you do for 16 MHz
the internal osc is either 1 MHz or 8 MHz
woah did the datasheets always have hyperlinks in them?
heh - yes, for internal links. not as much for web links
(but I started with the AVR slightly before it hit the general market, so they may have been there "forever")
gonna want a few of these: http://www.sparkfun.com/commerce/product_info.php?products_id=8181
I think whatever I get I'll get it from sparkfun so I only need to place one order
I still think the butterfly for $21.28 is a good deal. there's already a phototransistor on it, and all the I/O is available if you want it
I need at least 3 phototransistors, and I need to mount them on the vehicle, not on the board
but I should take a closer look at the butterfly anyway
there are 3 10-pin headers, which I'm sure are 3 I/O ports
one is probably the one with A/Ds on it
(they're not labeled on the baord, and I don't have the pdf open at the moment)
why is it the RC oscillator can do 8MHz but not 16MHz
it may also be a low voltage thing
damn - he doesn't have the "I tinker, therefore I am" tee-shirts anymore
does have this one tho: http://www.docsmachine.com/tees/teegraphics/tee16.jpg
I like the way a friend of mine put it:
"I'm thinking about it, therefore I might be"
EMC: 03cradek 07TRUNK * 10emc2/src/emc/usr_intf/axis/scripts/axis.py: fix the extents sometimes not updating right away when the origin changes
EMC: 03cradek 07v2_2_branch * 10emc2/src/emc/usr_intf/axis/scripts/axis.py: fix the extents sometimes not updating right away when the origin changes
sparkfun order placed - got the 128 board, some sensors, and a parport programming cable
EMC: 03seb 07TRUNK * 10emc2/src/hal/drivers/ (m7i43_hm2.comp m7i43_hm2.h):
EMC: This adds very basic support for steppers.
EMC: Also: improved docs and misc bug squishes
steppers are weird
heh. what leads you to that conclusion?
i'd never really used steppers before this week... It's wierd to have a P controller in the device driver.
also it's wierd how they get all hot just sitting there not moving
hmmm. what kind of driver?
motor drive that is, not the softwae/FPGA with P in it
It's a Lin Engineering stepper with integrated driver
I think I've seen some like that advertised
[06:31:09] <seb_kuzminsky> http://www.linengineering.com//site/products/silverpakD.html
do you have current reduction turned on?
it's only 33% (reduction, though they could mean down to 33%)
ehm, i think so but i'm not sure...
i think the driver uses reduced holding current by default, unless you special order a motor without it
I'm not sure how these motors are configured...
do velocity-mode steppers generally get hooked up to an external PID controller?
it says "Selectable Current Reduction of 33%"
well, that's an interesting question :)
that's how Jon Elson's USC works
the problem is that steppers have the wrong torque response (compared to servos)
if a servo lags, you tell it to try harder
steppers have high torque at low speed, and torque falls off as rpm increases, right?
if a stepper lags, you need to slow it down (and you have to detect the lag before a stall occurs or it's over anyway)
while servos have constant torque up to their nominal speed?
yes, and even more for brief periods
how do you detect stepper lagging? with an encoder?
usually the peak torque is 4-6x the continuous
with a special driver and an encoder
Mariss (of Geckodrive) is working on just such a beast
how do you get a servo to give up that extra 4-6x of torque? say the motor bogs down, speed falls, the controller detects the lag and increases pwm, which increases current to the motor and it gives more torque?
yep, that's how it would work
it's a little different if you're using step-to-servo drives such as geckos though
then the pid happens on the gecko
you need to set the current limits correctly for servos, they'll destroy themselves if you ask them to and the drives aren't limited
yes, though they have a fixed I term
you can only adjust P and D (and I'm not sure they're quite P and D - they're called proportional and damping)
brushed dc motors have resistance in the windings, doesnt that limit the current for you?
if the ps and the motor are matched, shouldnt it "just work"
most DC motors have 1 ohm or so of resistance, less for smaller motors
if all the power went into winding resistance (and out as heat), there would be none left to output fromthe shaft as mechanical energy
that's a "stall", right?
err - no, for a DC motor, a stall is when the amount of current (torque) on the motor can't overcome the load, so it stops
one of the little servos i'm testing with has a "stall current" of 2.5 A (19V, 7.5 Ohm winding resistance)
oh, maybe that is it then
if you're givign it full current and its not moving, the only place the energy can go is into resistive heat i think
7.5 ohms is very high. is that a really small and really high speed motor?
yes, that's true
yeah, 7800 rpm, about 1" dia & 2" long
ok, so it has no torque :)
it's similar to the servos cradek has on his sherline lathe
7800 no load?
right, no load
1.6 oz*in cont torque, 7.some peak :-)
needs like a 100:1 belt drive :-)
19/7.5 ~= 2.5, but if all the power is going into heat, then the rotor shouldn't spin
also, the back emf gets subtracted from the supply voltage before calculating resistive losses
so that makes no sense to me
(but then again, I'm not a motor expert :) )
you only get back emf if the motor's spinning i think
the back emf rating of a brushed dc motor is measured in V/rpm
but at no load, 7800RPM certainly would qualify as spinning, no?
sure, or more commonly V/kRPM
at no load, the motor is spinning at 7800 rpm, big back emf, low current
now fight the motor, rpm drops, back emf drops, current goes up, torque goes up
ok, but you said 2.5A, 7.5ohm, 19V stall
2.5*7.5 is 18.5V in resistive losses, plus 1/2V for back EMF, which isn't computing in my brain
though it's late, so it could be me
yes, 2.5A at stall, which is 0 rpm, so no back emf
ok, so those were separate specs ;)
7.5 Ohm is the winding resistance, no matter the shaft speed
19.1 V is the motor's rated voltage
7800 rpm is the no-load speed
2.5 A is the stall current
the 7800 rpm and 2.5 A are mutually exclusive, but the voltage and resistance dont change
at no load, there's very nearly 19 V of back emf, and the motor is just sipping current
right - that's why I said they're spearate specs
I guess I'd need to (a) look it up or (b) go to sleep to know what the definition of stall is for a DC motor :)
ok i think we mean the same thing now :-)
oh yeah, this conversation started when you said you need current limiting on servos
my little toy servo doesnt need it, because 2.5 A is well within the range of my power supply
but big servos...
have much lower resistance, around 1 Ohm you said, which would pop my ps and/or my amp...
sure, and you can't overcurrent yours (unless stall is peak, and you're only supposed to use ~0.6A continuous)
i think that's the case
well, that is if you use a 19V supply
24V, but what's 25% overvoltage between friends?
hmmm. about 50% more heat? :)
ok off to bed for me
night night. cool stuff with the 7i43
skunkworks605 is now known as skunkworks_
I'm going to e-mail the developers list as well, but is anybody still actively working on an outstanding bug for 2.2.3? if not, I might try to get a release out this weekend.
I got all of mine done
(except O-repeat not working)
SWPadnos_ is now known as SWPadnos
interesting. I have two (probably unrelated) problems trying to boot the liveCD (version 2.2.2-1 I believe)
first, it doesn't boot - some PCI resource allocation problem. I
I can fiddle with BIOS / boot options to fix that
second is that just before the boot menu, there's an error message saying (something like) "error: unrecognized keyword in config file"
bad download? bad cd?
this is only visible for a half second or so, so it's entirely possible that I'm seeing it only because this is an LCD display and mobile graphics card, so there may be no re-sync going on (like with a CRT when you change graphics modes)
no, I don't think so. I've burned from this image and booted successfully several times
I'm running a CD check right now
well, no I'm not, because it needs to boot to do that
I did verify with k3b when I burned it though
I have not seen that.. I have booted the live cd on quite a bit of hardware. A lot of the newer hardware I have to set up the video manually
some systems have broken PCI devices when CONFIG_PNPACPI is turned off -- for example, on one of my laptops this kills the network interfaces
all ACPI is turned off in our kernel, because some ACPI is power-management-related or latency killer..
I've tried booting in safe graphics mode, and separately with noacpi and nolapic boot params, neither worked
If you remove 'quiet' from the boot line do you get any additional information?
oh, maybe - one sec
I was thinking of the change to nosplash :)
I'm not 100% sure what option I mean
I think in grub it's to remove 'quiet'
I think it would be the same for the live CD though
yes, there are two options, I had confused the one you asked about with the one I had actually changed
oh, there's a suggestion to ty the option "pci=biosirq" in the extra messages
what finally fails -- can't mount the cdrom?
btw, have you booted regular dapper livecd on there?
the system fails to boot, at a text screen
I'm not sure I have booted normal dapper, I have certainly booted 7.10 (that's what's installed on the HD)
without quiet, it happens to stop after the list of registered schedulers, but I think those aren't the problem
not that it helps, but here are my dmesg lines after that: http://pastebin.ca/872629
(previous line was 'io scheduler cfq registered")
(that's not a magma kernel though, it's ubuntu generic)
dunno - I'd assume that the liveCD wouldn't write to the hard disk, and I can't get to a prompt since the liveCD boot never finishes
I'll try "single" to see if it's a graphics thing
the live cd will use swap it finds
trying the stock CD now
damn - that's one unbalanced CD
sounds like a milling machine
but it boots
oh well, I guess you won't be running emc on that machine anytime soon
heh - it's my new laptop - I just wanted to demo stuff
you don't need realtime for that
it seemed like the easiest way, since I don't have any EMC stuff on this machine yet
time to gtfooo
hmmm. this is the laptop I plan to bring to Wichita - any suggestions on the best way to get it "emc2-able"?
considering that it's a 7.10 SMP 64-bit install at the moment ;)
sim builds on 7.10 .. debian/configure sim then look at debian/control and see what packages are listed there
doesn't sim just work?
ignore the doc ones (e.g., latex, lyx) and install the rest
sim probably does work - I admit I haven't ttried it yet
if it doesn't work, I want to hear about it
I generally prefer to work on non-sim for developpment though, because I usually work on the lower-level stuff :)
bah whatever you say
you can't plug a mesa card into it
USB 7i43 :P
actually you can
but it's expensive as hell
it has expresscard though, no pcmcia
ah, so only pci-x then
pci-x <-> pci bridge?
seems that's a grossly wrong machine to use then... :-/
you can develop non-driver comps, though...
that's my point
anyway .. leaving now
yes, I know it's not great for RT or driver development
hmm, working tomorrow.. for a change
btw seb_k e-mailed me and says he wants to put the 7i43 driver into 2.2.3 so that may lengthen the timeline a bit
though I should be able to do modbus
SWPadnos: maybe installing it, and then pulling in the kernel will make it work
heh - We have a note pad with http://www.emcins.com/
there's still a chance the live-cd works differently than an installed system
well, it certainly doesn't boot
oh, ok then :)
* alex_joni hates 6MB changelogs
how are you supposed to read them?
look for stuff you're interested in?
err.. what if I don't know what I'm looking for?
then you have to read the whole thing!
time for dinner - bbl