I made a new rtai package with the modified rtai_smi in it
but I'd like at least some testing before distributing it
the package is here: http://dsplabs.cs.utt.ro/~juve/tempdebs/
but the rtai_smi.ko included is a bit different (there is an insmod param called: smiDisableMask which needs to hold the bitmask for the desired disables)
for disable ALL it should be:
insmod rtai_smi smiDisableMask=1
0000:00:1f.0 ISA bridge: Intel Corporation 82801DBM (ICH4-M) LPC Interface Bridge (rev 01)
any idea if this is a machine where the module would function?
sounds like PCI_DEVICE_ID_INTEL_82801DB_12
can you check lspci -n ?
it should be in linux/pci_ids.h
yep I'm looking at that file now
hmm DB_6 is not in the smi-module.c
only DB_0 and DB_12
I'm not sure which out of all these devices I should check
what does lspci -n say?
the number is in there
about which device?
[01:10:47] <jepler> http://pastebin.ca/290470
0000:00:1f.0 0601: 8086:24cc (rev 01)
yeah, that one
now grep pci_ids.h for 24cc
#define PCI_DEVICE_ID_INTEL_82801DB_12 0x24cc
ok, that's the one in smi-module.c too
jepler: what is this?
so rtai_smi should work on that
skunkworks: that's my laptop
do you have latency issues with it?
he's using sim :D
alex_joni: back when I had breezy my computer would pass the latency test for hours, but fail a few times a day
without any smi stuff
yeah, with this it might be better
yeah right now I don't have realtime because the ipw2200 wouldn't work without ACPI_PCI_mumble
and i like to talk to you guys
"might" as this is highly advanced magic
or should I say wizardry
I have been running 3d chps over and over. No issues so far
skunkworks: yeah, you're a lucky sob
I have never had the unexpected delay if I have the module loaded.
does it go away when you close emc?
have not looked
the rtai_smi module?
it probably should
jepler: custom kernel? :D
but I guess it's too much trouble
alex_joni: I did that on breezy but sim is so much better
yeah, I agree
although I recently saw evidence that some people think RT kernels with ACPI are sometimes better
ACPI on, APM & power management off
I think there are at least some parts of ACPI that are good to turn on -- the one that I need, for instance, just influences how the kernel finds which interrupt corresponds to which hardware...
but it's a PITA to do it properly
and I would like battery monitoring on my laptop
so I guess I'll take the non-RT approach too
yeah I turned on stuff without regard to whether it affected the reliability of realtime
CPU frequency scaling for instance
well.. off to bed now
good night all
this "PDM" waveform looks like a big blur on my 'scope and it makes my speakers buzz
what are you doing?
did you follow the discussion yesterday about "interleaved PWM"?
I implemented a PDM/"interleaved PWM" mode for pluto, although I'm not sure whether it's actually useful
no I really didn't
normally, the pwm output is: out = a > b where a is the set level and b is the counter
"interleaved pwm" uses bit_reversed_b instead, giving pdm-like characteristics
the Mesa implementation apparently reverses only the high-order bits, while I reversed all of them
ok I saw something about that - I didn't study it enough to understand it
it's too fast to see on your scope? isn't it 100MHz?
I did turn on "full bw out" and dial to very fast scan rate -- but the waveform is still ugly. it's sure not a square wave
but I'm not sure this probe I bought is 100MHz
it probably won't look very square...
you mean the "full bw out" button is out right?
don't you have one good original probe?
actually the probe says 200MHz
maybe it's not adjusted right
I changed it to not reverse the low 4 bits and I get a square wave but with a LOT of ringing
the ringing appears whether or not I use PDM mode so it's not exclusively due to the high frequency of changes on that output
probe switch is on 10x?
I just moved the GND clip and the signal is good now
do you get more levels then?
no it looks like a square wave
OK, and now the "pwm+direction" mode is working too
ve7it is now known as LawrenceG
jepler: what were you expecting PDM to look like?
on a scope it will be very ugly
but if you are trying to filter it to an analog signal, it will take less filtering than a regular PWM signal
do you have the ability to run one channel of PWM and one of PDM (I assume all pluto outputs use the same master counter)
try putting the PWM on one scope channel, and the PDM on the other, trigger on the PWM
adjust the scope so one or two PWM periods are stable on the screen
jmkasunich: unfortunately I have to switch all channels to PDM together
then the PDM should be stable too, but it has a much more high frequency content
jmkasunich: but the problem was that I didn't have GND
oh, that helps
after that the waveform was much squarer, not so much ringing
btw, how did you go about learning VHDL? online info, bought a book, trial and error?
* jmkasunich wanders off...
jmkasunich: I worked from a few examples and some introductory articles. the pluto firmware is written in verilog hdl, not vhdl, by the way
(they're two very different languages)
Lerneaen_Hydra: PDM is "pulse density modulation". It produces shorter pulses than PWM does, which can be better in some cases
(e.g., to produce an analog output by putting an RC filter on the digital output pin)
oh, so lots of pulses could be 1, no pulses 0, and stuff inbetween would be, well, stuff inbetween 1 and 0
so to get a 2.5V analog output, you output 5V 50% of the time and 0V 50% of the time
so does that give you twice the data throughput?
what you do is change the digital output at a very high frequency (say, 40MHz) to create an average analog level signal at a much lower frequency (say, 20kHz)
I can't seem to get why it's better than pwm
in PWM, each 50uS cycle has all its "1"s at the start and all its "0"s at the end. In PDM, the 1s and 0s are scattered through the whole time
if you are switching an inductive load, PWM is better because it has fewer switch-ons and switch-offs
if you're doing something else, like generating an analog voltage, PDM will stay closer to the desired output voltage than PWM
oh, for a certain RC
in the case of pluto, the PDM mode would be useful for someone building an interface to a +-10V analog servo amplifier -- with an RC network you easily get a good 0-3.3V analog signal, which you can convert to the 0-10V range with some op-amps
er, -10V to 10V range
+-10 is the "standard" voltage range?
strange voltage level
I've never had such a servo drive but people often refer to this voltage range
you'd think the manufacturers would go for +-5,12,24 or so
I bet that with a nominal +-12V supply and non-rail-to-rail op-amps, +-10V was a good range choice
hmm, that's true
then again, +-15 and +-12-13 is more common than that :p
never heard of +/-10V signal ranges?
+/-15V power supplies thru generic op-amps gives "about" +/-12V swings, if you want some margin, and a nice round number, you use +/- 10V
somewhat more modern op amps can reliably give +/-10V swings from +/-12V supplies
my guess wasn't far wrong, then
what I found amusing is that you had to guess
(some things I tend to take for granted - like "everybody knows that +/-10V is the standard for analog signal"
I didn't take up electronics (as opposed to computers) until 2003 or so
not long in geological time
and to be honest, I should say "+/-10V _was_ the standard"
lots of stuff is going to 0-5V or worse
is there some advantage to an analog-input servo amp, vs a PWM input one?
mostly is a partitioning thing
analog input means the PWM frequency and other details of the power switching are decided by the amp designer
the analog signal is also better for sending some distance
I had thought of the first (leaves more up to the amp designer)
most good amps have overcurrent protection and such built in, thats harder to do (but not impossible) with PWM input
"back in the day" all of the circuitry in the amp was analog, so the analog input was the simplest way to do things
now I bet many amps send the analog input to an ADC....
is it christmas there?
it must be nearing midnight
Sun Dec 24 23:20:44 EET 2006
where's the appropriate place for an example .hal file? I have a script I use to "test" the pluto and I'd like it to be in CVS.
[21:41:28] <jepler> http://emergent.unpy.net/index.cgi-files/sandbox/pluto.hal
I use 'halrun -I pluto.hal' and it makes the LED glow
there's another "halrun"nable thing in src/ already: ./src/hal/classicladder/projects_examples/example.hal
jepler: how about a config dir for the pluto?
and having the hal there?
there is nothing for pluto in configs/ yet
perhaps when chris finishes he'll put his lathe configuration there
jepler: now it's after midnight
hooray -- merry christmas
yeah.. same to you
although I got my gift a couple of days early :D
is it comfortable when full?
yeah, just scary big :)
I was somehow imagining it smaller
I have no problems with it when going on a trip, but it's not good for everyday use
what's your gift alex?
a backpack for my photogear and laptop
[22:38:10] <alex_joni> http://www.lowepro.com/Products/Backpacks/notebook_camera/CompuTrekker_Plus_AW.aspx
13W x 5.3D x 18.9H in.
hey folks. got a couple of questions on group buy administrivia
merry christmas alex (and anyone for whom 12/25 has arrived
same for you guys (I'll be safe asleep later :)
first, I'm in favor of insurance on the inbound package, especially because it belongs to multiple people
* alex_joni goes along with the rest
agreed unless the cost is unreasonable
that actually nearly doubles the shipping cost though - $40-$50 total (roughly $4 each)
$4 each probably isn't too unreasonable
not at all
I was figuring it would be $1 or $2 each
total value is $4K (approx)?
outbound is in the $9-$11 range for everyone
I don't care about +-$10, whatever you guys want is fine
so $20 or so for insurance (the 40-50 was total shipping + insurance)
I can live with that
ok - I think it's a very good thing - nobody would want to be the one who doesn't get their presents due to a forklift error ;)
on outbound, you can have the choice, but since it's $1.50 (minimum) or a max of $2.50, I figure everyone will probably want it anyway
($2.50 since nobody but me has >$500 worth of stuff)
oh, and skunkworks ;)
I'm close to that I think
next question: Shall I just round to hte next whole dollar amount?
yes, but still under for insurance purposes
SWPadnos: fine by me
jmk-st: still around?
jmkasunich: still around?