it's not a waste to use any cnc for manual work if you don't have other work for it
we have a big damn horizontal, an OKK
letting the machines be idle is waste
i want to use that.
i like the control
and i'm more comfortable with either a full manual or a full blown cnc, not this damn in the middle thing
part of it was me frustrated and not paying attention
struggling with the conversational control
all i wanted was "g1 x10"
and it would not do it
damn thing needed all sorts of crazy shit
I see a iocontrol.0.user-enable-out is there a iocontrol.0.user-enable-in ?
jmkasunich: just got back
wanna talk about analog stuff?
pwm to +/-10 to be precise
jmkasunich: I dunno -- I'm really not that serious about building a board to do it, since I don't have any of the motors
if you do decide to build a board, the circuit you posted is probably overkill
jmkasunich: so if it's a choice between trying to teach me about op-amps, or spending the time on 5i20 stuff instead, I'd encourage you not to waste time on me
you would do fine with one or two op-amps per channel, not 4
jmkasunich: how's your index?
one IC package hardly feels like overkill
A to Z and everything in between
jepler: two or four channels with one package is nicer ;-)
cradek: index works fine for 2.2
haven't started looking at 2.1 yet
nor have I checked to see if my encoder is the reason for the lost counts
I forgot about that
I doubt mesa is as badly broken as that
you're probably right
I'll be bummed if its this encoder though
sounded like it's an antique
1991 I think
jmkasunich: so you think a 1-op-amp "difference and gain" would be adequate? one thing I don't grasp at all is just how good the circuit needs to be. H&H propose all kinds of complicated schemes to decrease offsets and the like
(horowitz & hill that is)
depends on your requirements
sources of offset = common mode input * resistor mismatch and op amp offset voltages
your pot tweaks resistor mismatch, but not op-amp offset
and your buffer opamps actually add more offset
I have a bias against pots because of my day job (adjusting them sucks in a commercial environment), but for this kind of thing they're probably not bad
I'd still try to get well matched resistors instead
the main piece I'm still missing is why one adjustment can't cover offset of whatever type
what makes op-amp offset different from resistor mismatch
cause resistor mismatch isn't really offset, its common mode rejection
assume perfect resistor matching, and assume that IC3B has 3mV of offset (assume the buffers and the output amp are perfect)
pins 1 and 14 are at exactly zero volts, so pin 5 is at zero
due to offset, the opamp wants pin 6 at 3mV
so pin 7 will be at 6mV
you can adjust R8 all you want, pin 5 will still be at zero, and pin 7 at 6mV
my approach for a diff amp (thats what you are building) would be to match R2 and R3 as well as possible, likewise R1 and R4
ground the bottom of R4 (no pot)
to trim offset, run the pot from + supply to - supply, and run a meg or so from pot wiper to pin 5
I think that was what archivist was suggesting too, but I was having trouble understanding him
that can apply a small offset to pin 5 (small = supply * (1Meg/(4.7K/2))
to keep things matched, you also want to put a meg in parallel with R1
you can get 4.75K 0.1% 0805 resistors from digikey for $0.20 each, minimum quantity 10
so that takes care of matching pretty well
and I bet a sub-2mV opamp costs less than a pot, so I'd be tempted to forget about offset comp as well, unless you need really low offset
that's a question of the servo drives, I suppose -- and I know jack about them
it would be handy to be able to null out offset in the amp as well as in the circuit
and I guess the other question is gain adjustment -- whether to have one, or whether to set it by the installed components
I'd probably do it with installed parts
gain accuracy rarely needs to be better than a few percent
1% resistors are a dime a dozen
its possible to do everything with one opamp, but that requires that you change the value of the precision resistors to change gain
I'd keep the output stage in your case, so you can change 1 or 2 non-precision resistors to change the gain
so IC3A/IC3D get removed and replaced with a wire .. R1..R4 are spec'd as .1% .. and R5 is replaced with 2 1% resistors which set the gain
if you remove the buffers, then R6 and R7 also need to be precision matched
and they affect the gain
* jmkasunich waffles about the buffers
assuming you have only one value of 0.1% resistors, you could make R6 and R2 the same value, with C6 connected to their junction
but if R1 is also the same value, the net gain of the diff-amp stage is 0.5
you'd rather have more gain, not less
maybe 2.2K 0.1% for R2,3,6,7, and 10K 0.1% or so for R1,4
cap values not critical?
also, you can put caps across R4 and R1 for another pole of filtering
the first pole is (R6 in parallel with R1) * C6
matching of C6 and C5 affects AC common mode rejection, but with up/down PWM, you'll never have an AC common mode signal
one is switching, the other is zero
the acex datasheet showed Vol might be as high as 400mV (though that was for a TTL load)
acex is the FPGA?
for X3-3, the load is always referenced to ground
I bet when the output is low, its withing a few mv of ground
for X3.1 its not so simple
hmm, those buffers are looking nicer and nicer
assuming you find a good and cheap opamp
good = low offset
or you use an offset trim pot
I haven't tried to pick an op-amp yet
digikey's selection guide sucks
they let you select on slew rate, not offset or supply voltage range
who cares about slew rate?
high swing, fast signals do
ds2: I know
my point is, in many applications, offset is at least as important, if not more, and in just about all applications, supply range is important
mouser lets you choose various things including supply voltage but not offset
Or start with LM324
ISTR the 324 is rather mediocre
they are cheap and around
horrible specs.... probally a 30+ year old part but...
bandwidth 0.0055MHz ?
Unity bandwidth is 1MHz
no, it's one of the choices on the mouser "bandwidth" selection
[01:47:04] <ds2> http://www.national.com/pf/LM/LM324.html
does 5.5kHz have some significance for video or sound?
I have no clue
sometimes these databases have non-sense specs in them
hope now...that's in the middle of the audible range
i prefer to go through the manufacture sites instead
slew rate at unity gain: 2.5V/ms
gawd thats slow
it's very low power though "<1uA per channel"
that splains it a little
I'd just like to know the intended application
you could do worse and consider a 741
324 can run single ended
jepler: application for the slow one might be measuring temperature or something else that doesn't change quickly, and runs for a year on a battery
its a shame you can't get large volume pricing on the OP4177 - thats a nice opamp
is it happy with single ended operation?
I believe so
but jepler's application would have dual supplies anyway
the hitch is its $6 for a quad
$1.50 per amp is steep
if you need the analog specs...
the 324 is about 10 times cheaper, but has 100 times the offset
324 is a horrible but common device that you might have in a junkbox kit
is offset a big deal for his application?
what is his application?
don't want the motor wandering off while you are commanding zero speed
ds2: PWM to +-10V servo amp
motor wandering is a feature, not a bug
although the postion loop corrects for it in theory
toastydeath: how so?
i don't know, nobody's ever challenged that statement before!
jepler: aren't you using that mostly as a comparator?
(because i just made it up)
toastydeath: so that the pid loop doesn't get bored?
jepler: it keeps the machine at temperature
minimizes thermal effects!
toastydeath: ah yes good thinking
(i'm not sure if i would call it 'thinking')
my father said he's got a pretty expensive oscilloscope sitting about, $6k
that i can have
doesn't one of you have a billion dollar scope though
i feel as though most of you probably have that kind of thing just sitting around
so it was more of a "hey guys i'm in the club now too!" statement
SWPadnos has a fanch scope
like the retarded kid joining the lunch table
I don't have any cool stuff
i am hoping it has the crap i need
it probably will
it will probably be able to calculate pi to a bajillion decimal places
and i just won't know what i'm looking at
jmkasunich: appears OP4177 is replaced by OP4277 which is $7.32 qty1 on digikey
10uV offset voltage, 1nA bias current
digikey has the 4177 for a bit less ($6ish)
be nice if you could split the difference
1mV at $2 would be nice
if I was building 3 boards for myself I wouldn't blink at a $7 chip
if I was looking to have a product line obviously I would
the alternative is the $0.50 chip and a couple pots
as long as the $0.50 chip doesn't drift over time
pots are probably $0.50 - $1.50 each (I saw some smt ones for 0.20 each, but they would probably be a royal pain)
* jmkasunich stops to get some perspective
10mV of offset on a +/-10V signal is 0.1%
of course, you are starting with a 3.3V signal, not 10B
my op amp selections are biased the other way, I've recently been working on circuits that take +/- 10V and compress it into 0-5V for a 12 bit DAC
with the goal of less than 1 LSB of error
jepler: maybe LM324 plus offset pot is the way to go
its not like its gonna see 50C temperature swings - adjust the offset once and forget it
at $0.50 for a quad, you can afford to buffer the inputs, that will reduce error associated with the Vol of the fpga, and decouple the filter resistors from the diff-amp (precisely matched) stage
ugh -- "*** glibc detected ** /usr/bin/python: free(): invalid pointer: 0xb7d480e0 ***"
ugh here too
swapping the encoder signals didn't change anything - encoder 01 still counts correctly in the + direction, and loses steps going -
so I wired encoder 00 in parallel - same signals go to both
they don't count the same
is your card bogus somehow?
I sure hope not
ok, rotated the encoder till the index lights light up (on the 7i31 led board)
reset both encoder channels
then wiggled the shaft back and forth 20 times or so
then caretully rotated it back to the point where the index light turned on
channel 00 shows 394 counts, channel 01 shows 1474
do the signals look sane?
I haven't scoped them
at low speeds the LEDs sure look ok
it just can't be that broken :-/
I'm thinking about using stepgen in quad mode to provide signals to count
then I can compare command and count and trigger the scope when/if they diverge
are you using that input board with the differential encoder inputs?
no, I'm just directly driving the cable
but I don't have any noise sources here
no motors or anything
also, channel 01 counts EXACTLY right as long as its counting up
I gotta wonder if there is a timing constraint that isn't quite being met
reversing the wires and turning the encoder the same way does seem to rule out an encoder problem
is this your new superfast machine, or a normal one?
I gotta wonder if there is a timing constraint that isn't quite being met
I'm more and more glad that I used the pluto for my lathe - it seems to do everything right, with no screwing around
that should be a real compliment to jepler, since it's the newest and least tested device of the kind
(not that this helps you with the task at hand)
( well pluto does have a nice quad counter...)
i have my pluto here on the desk, im looking for female idc connects (pcb mount). it would be nice to "plug" the pluto down onto three idc connecters on the interface board
curiouser and curiouser
I sent sim-encoder signals out thru 5i20 generic output pins, and looped them back in to the encoder inputs
I also rounted them to a software encoder block
the two hardware counters and the software counter all agree
count up, count down, reset on index - they match
maybe it IS my encoder
either that, or something that only happens when the singals are changing faster than the sim-encoder can make them change
although I doubt I've exceeded 10-15 KHz turning the shaft of this encoder
scope scope scope
lazy lazy lazy
Linux emc 188.8.131.52-magma #1 SMP Fri Jun 22 22:04:24 CDT 2007 i686 GNU/Linux
(maybe this'll fix my keyboard problem)
well, the scope says this encoder makes nice quadrature
of course I can't look at every single pulse it makes
OR CAN YOU
hmm... at 20nS/div I see some ground bounce and crosstalk
but the 5i20 encoder counter has a multi-clock filter
clocked at 33MHz = 30nS
with a DSO and sufficient memory, you sure can look at every pulse ;)
the bounce I see is less than 20nS long
I have a dso, but not sufficient memory
can dso's really do that?
its mid-90's technology
no dudes this is a serious question
I mean my scope is 90s
i've got a thing i need to scope, and the event will happen over say, five seconds
there are scope that can acquire 100,000 or more samples
i'd like to look at a low resolution chart of the whole five seconds
the problem is examining all that data
well to be fair i'm just looking at the overall doodad
not very high resolution at all, i just want to look at noise
over five seconds
if your scope can acquire 100,000 samples, then over 5 seconds the sample rate will be 20KHz
if the things you are looking for are significantly slower than that, you're are golden
that's way more than i need
like, i could eal with 360 samples
well, most scopes don't do 100K samples anyway
mine does 1000 I think
so over 5 seconds I'd get 200 samples/second
if all you need is digital data, a logic analyzer would do better
i need analog data
but it usually won't catch odd analog stuff
well i'm going to get a chart recorder too
my scope will turn down to 5 seconds/div or 50 seconds per screenfull
how much was your scope, jmkasunich
if a chart record will do, a DAC on a parallel port might be sufficient
it won't, i've looked
plus it's bulky to use a pc
in this application
uh... bulky? see UMPC?
toastydeath: I paid $200 at a work auction of surplus equipment
what was the cost of it new, i mean
I got a good deal
if you know
I dunno, its pretty old
that is a great deal
i've got a free o-scope coming
analog scopes used go for that
i believe it's a dso
what is this "new" thing you're talking about?
jmkasunich: is this the one that runs windows?
ds2 - you gotta be kidding
they made ones that ran windows
2440 is several generations older
ah.. no up on things I can't afford like Tek scopes ;)
i've got a scope coming to me but i don't know what features it has
but it is free so i don't care
the going rate on ebay for a 2440 is $400-500
i was trying to compare going rates
i probably should have asked what model it was
that would have made my job easy
does anybody ever use a function generator?
can function generators (in general) do really slow things
they make ones that run at AF
i'm talking slower than that
most will go down to a few Hz
some will go down to 0.001Hz, but thats less common
sounds like a job for a DDS, DAC on Parallel port or like
like i said, pc based stuff is not going to work
if you still find that clumsiy....
it's not that the thing itself is clumsy
it doesn't matter how big or small it is
Tek FG501 goes down to 10^-3 Hz
the fact it's a pc and not native hardware with knobs within easy reach
it what makes it clumsy
looks like the 2440 was introduced in 1988
anyway not buying it!
i may get a DAC as like, a long term logging thing
but not as the first set of test equipment
[04:11:21] <jmkasunich> http://cgi.ebay.com/Tektronix-FG501-Function-Generator-Plugin-TM500_W0QQitemZ330136946499QQihZ014QQcategoryZ50966QQssPageNameZWDVWQQrdZ1QQcmdZViewItem
10^-3 would be great!
signal generator that will do very low frequencies
basically i need a reference signal
to do differential and subtrative things
get a eval board for an ARM proc
on the oscope
think ADI has one that you can get a sample program to do function generation controlled by a POT
I'm quite fond of tektronix TM500 series instruments
do these things trigger
they make functions
sine, triangle, square, etc
but i mean, can i time the start of the signal
you can trigger on the signal they make
with another signal
jmkasunich: you are seriously my new hero
you just solved so many problems for me
you can trigger the scope on the signal from the generator
I'm not sure you can trigger the generator from something else
depends on the particular generator
the scope and the generator have to trigger on an external event
that for-sure cannot be triggered
and must be the event
you might be able to use the modulation input
like, what i'm trying to do here is subtract a saw wave from a rising signal
to produce a difference on the scope
a sync input should be sufficient?
i am not sure what that is, sir
thats kind of esoteric - a triggerable sawtooth isn't something your average function generator will do
mabye just a rising sawtooth, if it's slow enough
what are you trying to accomplish?
a sync input lets you time the start of a signal with an event
yes, that would work
but these are all high end features
i'm trying to look at fine differences in acceleration
what analog signal to you expect to be scoping?
tekronix math function, one trace on the sawtooth, other on signal, math will subtract sawtooth from signal and display it ( ymmv ) you'd still be faced with syncing the sawtooth tho
position isn't analog, its from an encoder
linear and rotary measurement devices
analog linear and rotary measurement devices
with analog outputs? thats extremely uncommon
it's very common in high precision stuff
analog usually isn't very accurate
the amps they use are good, and they sure charge you enough
very short range
also an interferometer
if you are handy, you can build one with the MAX038 chip (if you can find it)
an analog interferometer, because they're "accurate enough" and also "not a million dollars"
ds2: build which
the function generator
i don't have a problem purchasing a used one for a couple hundred bucks
if it will do what i need
if all you need is an adjustable slow ramp, thats an integrator
with an analog switch to reset/sync it
a function integrator?
an integrator - apply a constant input, get a constant ramp
problem is finding one with a sync input
op-amp with a capacitor from output to inverting input
and an analog switch across the cap
i'll need other stuff
sine waves and triangle waves
but sawtooth most importantly.
actually, you might not need a sync input
might be able to do this with an async single.. play games with X/Y mode
i am going to profess my ignroance again though
what does an async input do?
lassojou (sp) patterns
nothing. it just is not sync'ed with your signal
but that just introduces a phase shift
I think hes gonna want sync
i mean the only reason i need a function generator is to provide a reference to subtract
or add or whatever
on the scope
especially if the rates are slow to start with - if its async, only one in many traces will be usefull I bet
so like, the reference needs to start at the same time the lvdt or interferometer starts to ramp up
so we can check stuff like leadscrews
so like, i know it's going to raise X volts and at Y rate, and set the generator for that
set the storage up on the scope
and see what happens
can check pitch error AND vibration/etc
on that axis
all at once
so you need a very linear, very low noise ramp from the function generator
either high rez dds or constant current charged caps then
what kind of range are we talking about?
probably up to about 20 volts
way, way tops
does the lvdt or interferometer go from -10V to +10V over some specificed distance?
the analog amp determines that
it's a set voltage per unit distance, but you can change the range
thus, if you set it up weird, it can go pretty high
so something like 1mV per micron, which would give you 1 volt per mm, and +/-10 mm of travel before it hits the supply rails
something like that
(analog stuff rarely goes much beyond +/-10V)
i don't know what the interferometer is going to spit out
just what the lvdt's and cap probes spit out
I bet they're nonlinear
at least over large distances
from what i understand they're real linear
because they don't know absolute distance
maybe... with a storage tektronics you can 'sync' a loaded signal and subtract it ( move the sawtooth till it 'aligns', then subtract it ) i know i could overlay stored signals and compare, but never subtracted one
might be 1mV per micron for the first 500microns (1/2mm), then nonlinear
or I could be full of crap
i think it cycles up and down
but i'm not sure
because that's what the real input signal is doing
but it's all speculation until i have one in my hands
and try to use it
who knows, it might only be analog for the unit itself, and the output may be digital
[04:33:52] <tomp2> http://www.macrosensors.com/ms-lvdt_faq-tutorial.html
tomp2: not lvdt's
the lvdts are not an issue
but that is an awesome link anyway
you guys know a lot about the oscope/generator stuff
thank you for your answers
well, I counted 7 million counts at 30KHz, in software and in hardware (2 channels), and all three agree
whatever my problem was, it must be the encoder
(some of us have been using oscopes for 20+ years)
hopefully i will get my noob hands on this stuff and learn me some tricks
seems to be a dying art
tiem for bed I think
im not quite sure where all those pictures came from either
I stole em
jmkasunich, have you opened the encoder and checked for dust?
jmkasunich, one of my motors had a small piece of dust that was pulled into the encoder when rotating clockwise but ejected from the encoder when running counter clockwise
a-l-p-h-a2 is now known as a-l-p-h-a
jepler, have you finished the pwm->voltage design?
$ apt-get build-dep python2.5
The following packages will be REMOVED:
onboard python-xml serpentine ubuntu-desktop
wow it has to uninstall ubuntu-desktop to build python? that's f**ed up right there
well, it is ubuntu ;)
* robin_sz wonders whether it does that in Debian
Build-Conflicts: tcl8.3-dev, tk8.3-dev, python2.5-xml, python-xml
that is slightly crazy, after all .. python is an interpreted language
it might depend on it, but its not linked to it
python2.4 (2.4.3-7) unstable; urgency=medium
* Reupload, depend on python-central (>= 0.4.15).
* Add build-conflict on python-xml.
-- Matthias Klose <email@example.com> Wed, 14 Jun 2006 18:56:57 +0200
debian people decided to add this, not ubuntu people
I am curious about why it is required
can't I blame ubuntu anyway?
you can do whatever you like, but in this case the smell is coming from debian, or maybe the python developers
presumably there is a "just do it and ignore that stuff" option
the things it wants to uninstall are not that important
i never really understood what a "desktop" was supposed to do anyway
it pulls in all the packages that should be part of the desktop
it exists solely to require other packages for installation
ah crap -- now I've filled up the 3GB drive I allocated for feisty
back to step 0
how many linux versions supports emc?
umm, depends ...
in theory, you can run it on ANY version, with a suitable patches to the kernel etc
and a bit of work
any "sufficiently modern" linux should be able to run emc2 in simulator mode; any "sufficiently modern" x86 linux with a realtime kernel should be able to run emc2 to control hardware
for minimum pain, the Ubuntu route is the way to go
the emc developers build packages and realtime kernels for only two flavors of ubuntu
what he said.
the two flavours of Ubuntu being something like "farting rabbit" and "neurotic guerilla" or something
before I head out for breakfast, I just wanted to say:
robin_sz - don't be an ass :)
nah, stick to what you are good at. thats what I say
oh, in that case, continue.
what are the Ubuntu flavours named after?
but it would have to be something like "Rutting Rabbit" and "Neurotic Newt"
ubuntu "autistic ass" linux
I know debian is toy story ...
they're more-or-less random adjective-animal pairs, with the same first letter, first letters going in alphabetical order
that's after the first two or three, where they hadn't adopted that method
so the next will be gargling gorrilla then?
gotta run. see you later
* robin_sz waves