What's up on the west coast?
Thought I would give a little more info on the 7I43
like, when it might be available for purchase? :)
Not much, Our network was down this morning - janitor unplugged a Enet HUB
That must be why Lily didn't get back to me in 10 minutes, like she usually does
OK 7I 43 should be available in about 6 weeks, We have a working proto but need to fix some USB powered startup problems
FPGA is Spartan 3 (Spartan 2 is too expensive)
hmmm - like - hard to start up on USB power when the FPGA is unprogrammed?
No Windows driver cyling the power a few times on enumeration - dont like the I/O bits doing things at that point...
so it's closer to the 5i22 than the 5i20, since it's a Spartan 3
Were just going to put a delay on the FPGA power up from enumeration.
Yes, and it has the same 5V tolerance via a couple of bus switches
in the photo, what's the connector at the top, opposite the USB connector? is that for the parallel port connection?
great. is the pin configuration basically the same as a motherboard parport pin header?
ie, can you use a 26-pin ribbon cable and avoid the DB-25?
Also helpful for debugging the USB connection (shares the same FPGA pins)
ok - that was the next question - can you run it off USB power (and maybe program it via USB), but use the parport for communication
I guess that answer would be no
Yep, thats what we used to test it, a DB25M --> 26 HDR-F flat cable
side question: do you sell 50-pin ribbon cables with twisted pairs and/or shielding?
Yes, possibly, Might take a tweek in the little CPLD on board, so the USB chip is never issued a read
oh, that would be great
getting power for parport devices is always an issue
No, we normally just use flat cables. Havent looked at that stuff for a while, Does Spectra-Strip still exist?
apparently it does (accorfing to google)
I found another company that sells relatively inexpensive twisted pair, and they have a standard distance between flats of 8" or 8.5"
ribbon twisted pair, that is
Also if you get the Parallel port only model, the USB chip isnt there so USB power would always work
[00:59:09] <SWPadnos> http://www.connectworld.net
hmmm. doesn't USB only provide 50 or 100 mA unless the device asks for more?
We have the interface chip ask for 450 mA (the max). We have on card switching supply so 1A of 1.2V core power is OK
oops - that doesn't look like the right website. sorry
here's the twisted-pair cable: http://www.connectworld.net/cgi-bin/iec/CAB050-RI-TW
do you have a delivery timeframe for the 7i37T?
hmmm. that does seem high, doesn't it?
Oh I see 115 Ohm
7I37T's are available but first lot had a problem with 3.5MM headers (for pluggable screw terminals) They are OK but cards warped due to using ThermoPlastic Headers at RoHS soldering temperature
ok. I'll need a couple in 2 weeks or so - I can get by with the breakouts I have for now
We're trying to decide whether to scrap the lot or resolder the connectors or maybe sell them at a reduced price as blems - they work fine just a .1" warp on PCB where terminals are
oh - well I'd take some discounted scratch-n-dent ones :)
my customer can pay full price though
(for unblemished product ;) )
We are re-doing the lot now so it should be pretty soon for perfect ones
excellent. this customer is getting antsy - we've got ~6 weeks to get an analog board designed and built, write custom FPGA code, write a HAL driver for it, and integrate the result into the machine. at least the digital I/O is easy.
Well with analog desgn you can always make it work just by changing component vaukse, Right :-)
yeah - that's the ticket
We just did an interesting custom job, an absolute encoder interface using SSI
I haven't really done much analog design (certainly not in the last few years) - it'll be fun making a 16-bit A/D+D/A board
cool - what encoders?
pcw: good evening .. the new parallel/usb board looks neat to me, as you may know I've been writing hal drivers & fpga firmwares for another parallel-port card
does it track position while de-powered (or re-acquire on power-up)?
<Jepler> Yes since the parallel port part only costs us about 22 cents extra I thought we'd make it more useful for EMC
22 cents! the nerve
pcw: when programming over the parport, is it done a byte at a time?
<SWPadnos> Its an optical absolute encoder 8192 counts per rev/ 4096 revs = 25 bits - no memory all mech/opto
I was just afraid to drop it - kilobucks
hmmm - so it inherently "knows" where it is, even if moved with no power applied - even across turns?
yeah - I'm trying to find an alternative to Yaskawa motors/encoders/drives for a robot
<Jepler> yep just write the config file to it
it gets expensive at $3-4k per axis
(dependong on controler options)
<jepler> same as USB just send bytestream
pcw, is the SSI interface available to others, or is it "owned" by your customer?
<SWPadnos> yep not cheap. But no homing!
Public (I charge more for private devel)
yeah - that's the idea. we want to be able to tell exactly what's going on when we recover from ESTOP or power failure, without dropping $meganucks worth of product
* jepler wonders what "meganucks" means
is it 1,000,000 canadians?
err - yeah, that's it!
(that reminds me of my favorite joke, the one about the three brazillian soldiers...)
nyuck nyuck nyuck
SSI is really simple though RS-422 clock out RS422 data back - first clock parallel loads (sample)
cool. the less I have to "reinvent", the less I have to convince my partner to go for it :)
I think we can make an embedded EMC machine with an FPGA card that will significantly ourperform anything Parker, Yaskawa, Delta Tau, or anyone else can sell us at the same price
it's just a matter of getting all the pieces to fit together correctly
but that "just" is a relatively big risk factor, so it's not a trivial decision
of course, we won't do it on 600 mA at 24V, but that's life
Dont know if its relevent but we have 3 Phase 320V Driver coming out for SoftDMC but it should work with EMC
Well it uses IGBT modules so its kinda hot...
I assume the stepper driver (and this one) needs softDMC or a very fast PCI update rate to work (or slow step rates)?
(I assume this because my understanding is that the PWM is varied to provide microstepping and phase current control, which requires a very fast update rate)
Our little stepper driver needs 30KHz + update rate (Sine and cosine PWM from table) But 3 Phase has brains so can run something like SoftDMC or PVT or just send phase angle and it echos currents
ah - nice
all kinds of cool stuff
Tiny brains (2 DSP PICs)
Were happy with them...
heh. I know a lot of cool things have been done with PICs (I've done some of them), but given the choice, I'd just about never choose them for a microcontroller these days
Yes the're pretty ugly, but for ~$2 16X16 multiply 40 bit accumulator 3 phase PWM+deadzone 1M sample second A-D, hard to beat
yeah - they do have some good motor control-oriented controllers, and I think quadrature is more or less unheard of at that price level (dunno if you've heard of Luminary Micro though)
Heard a little, ARM?
yep - ARM + various motor related peripherals
like quadrature and multiphase PWM
if they last, they'll be pretty good, but as far as I know, the company is pretty new
I like their development kit though - microcontroller and demo board, plus a 2HP 3-phase motor :)
it's basically a VFD development kit
hmmm - maybe it isn't 2 HP - I thought I had seen that somewhere, but I don't see it now
I must say though - I hate it when companies make you register to get their stupid datasheets
Ours is basically for BLDC, which are somewhat simpler (For highest sample rate we do Clark but avoid Mr Park by staying in polar coordinates)
I try to avoid those companies unless there's no choice
they had good demos at ESC, and their dev kits are available from Mouser / DigiKey, but they are annoying
We should be able to drive ~2HP (30A peak module - 11A RMS cont@~300VAC)
Course its still early and it may asplode...
"this IRC chat has forward-looking statements. These are not intended to accurately predict the future" :)
pcw: you mentioned that the 7i43 has "the same 5V tolerance via a couple of bus switches" -- is the pin direction switchable on a per-pin basis, or is it in blocks of pins?
<jepler> the bus switches are bi-directional (just a NMOS FET series switch) so no switching is needed. They work by turning off as the input gets ~1v from the NMOS FET gate drive, therfore protecting the FPGA
pcw: I see, thanks.
(protected from positive inputs, lower than diode drop below ground is stil fryola time)
what interface Mode an Parallel port do you prefer? Looks like EPP is most reasonable (ECP give me a headache). Our little CPLD can support enough of EPP at startup so theres at least a configuration target address/status port
(Meant for <jepler> )
I'd say that EPP is the best we can use with EMC, due to the low latencies needed. ECP seems more useful for bulk data transfer
(but jepler and JonE may know better)
Just for simplicity I'd prefer EPP
sure - more simple and better (or at least no worse) for latency. we're not trying to move 1 MB/second of data
pcw: I have used EPP exclusively.
I followed Jon E's lead in that
I didn't try ECP, but it certainly looked harder
doesn't it more or less add DMA capability to EPP?
Yeah was the headache part
on the systems I've used, it doesn't seem like you get a whole lot more than 1 byte per uS with epp
I always lose my benchmark numbers right after I note them, though
Maybe just 8 bit ISA bus access (or simulated ISA bus access) times
I wonder if chipset-integrated parallel ports can be set to use NOWS, according to this page: http://www.lvr.com/jansfaq.htm
* jepler looks
that's supposed to get rid of 3 wait states, which should speed things up by 2x, or so they say
sounds like it's not something you enable or disable, but something you might or might not have on a particular motherboard (?
or a particular ISA card. that's why I wonder if it's somewhere in a chipset register
not that there would be a BIOS setting for that (Turbo Parallel Port :) )
Q: Which mode (ECP or EPP) would work best with parallel port devices such as the Snappy and the QuickCam Color?
A: As a general rule, because of its FIFOs and DMA support, ECP is good at transferrring big blocks of data quickly (scanners, printers). EPP is good for links that switch directions frequently (drives). Specifically, it depends on the driver for the device, so you might want to experiment if both options are available.
if it takes "a long time" to initiate an ECP transfer (particularly a read transfer) that would be bad for emc
since you always want to read - compute - write
that was my thought
also interesting: he's got a port type detection algorithm on that page
I saw that
my algorithm is "hope for the best" :-P
I wonder if we can use that to automagically implement the parport mode thing (I forgot what it's called)
or something else?
err - no, the module you have to load if you have certain motherboards that exhibit problems
I think it just disables ECP or something :)
probe_parport has to do with configuring ports that are "ISAPNP"
and not configured by the BIOS or something
in hal_parport, this method might be able to determine that there's no port at the requested address
that reminds me of something...
also in hal_ppmc, or pluto
if nothing's connected or the I/O address is not used, after only about 30000 outb()s the pluto hal driver detects that it can't communicate and the module fails to load
if those I/O addresses were for some other piece of hardware, who knows what's happened by then
yeah - not much can be done about that unless you put identification code into the pluto
before you load the FPGA config ;)
[02:44:22] <SWPadnos> http://www.fapo.com/cenmode.htm
FIFO-based "fast centronics" mode. I wonder if it can do any good here?
We could probably do that with 7I43 = IDcode in Config CPLD
probably no byte-wide input support there
Bye all, time to feed dogustus and bunnies...
heh. see you later. thanks for the info
pcw: nice to talk to you
I'm done for the night too
I wrote one line of code, after all
yep - I'm getting that way too
and 2 lines of commit message :)
(so that I can just unconditionally put 'loadrt probe_parport' in stepconf configurations)
ah- good thinking
yay, nobody has found a new bug in 2.1 yet
it's a new record
has anyone downloaded it? :)
at least one (me)
heh. then I declare EMC2 finished!
it may take a full day for most people's update managers to "find" it
I'm not sure how often those poll
oh, then maybe we can declare it done on Wednesday
daily is the default
I set mine to every 5 minutes
just in case
so THAT's why the bandwidth usage was so high!
wow - bandwidth usage is up a bit. cycle estimate of 531766 MB this month
the two top requesting IP addresses are from Guangzhou and Shanghai
they comprise roughly 2/3 of the total requests, and close to 30% of the total data transfer
that cursed iso!
I don't think so
lots from BDI-related referrers as well
jmkasunich: got the belting, and just got the hektor.pdf. it has the specific math used and good closeups of the mechanisms.http://www.hektor.ch/Book/Hektor.pdf/
maybe alex's trikins can reduce the wobble (instead of hektor bi-kins + gravity )
also found i have illustrator 9.0 so will try the scriptographer ( an ancient version is still avaliable ), meanwhile enlarging parts of hektor photos to see the mechanisms. got a working idea of the gondola and found a disused commercial spray paint gun & pot at work.
just got in
hi jmk. where ya been?
it's past my bedtime (strangely enough) - good night
digital cameras are cool - you can shoot 62 shots of total crap without feeling bad
shooting dim stuff from the deck of a moving boat doesn't work very well
* Skullworks-PGAB knows that feeling - I used to burn 100ft of 35mm Plus-X a week on a slow week.
I had a great opportunity, just not enough light
the rising moon (full, low on the horizon, and rather red) with the top of a lighthouse silloutted in front of it
quick question - which works better under RT - an Nvidia TNT2 M64 or an Ati 9200?
OK - I'll just test each
if you wanted to edit the stupid mistakes you see - otherwise don't worry about it.
are attachments not allowed at all on the emc-developers list, or are short attachments OK?
I know people have attached their configs and pictures..
it must be a size limit, then
jepler: I can check for exact size if you need it
alex_joni: no that's OK
after I got smart, I found that the biggest message in my folder is over 500kb
though that's from 2005, maybe the limit was changed since then
a 250K one from march
you can send attachments, I'm not sure about the size limit
24kB limit for the body
same for emc-devel
can't find anything about attachements atm
thanks for looking
can't see anything else, but I remember getting admin emails if a message is too large
then one of the admins can decide to let it through, or drop it
yeah I guess I can't tell whether a particular message got special approval
out of curiosity, does anyone here have any data about how little memory you need to run an EMC2 system without any swap?
yes, without X
ok, so 512M-1G should be sufficient ;)
I'll probably want to be able to connect a remote display, so the X client apps could be running on the embedded machine
I /msg'd you
yep - saw that
I was looking for reviews :)
plus the discussion today about slightly more expensive ones from Asus (I think)
more expensive - less powerfull/smaller screen
real company ...
although I had gotten some inspirons from dell for $399 (250 dollar rebate.)
I added a section "Patch Submission" to the wiki, page http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?CVS
comments? additions? contradictions?
looks fine to me
jepler: there is support at SF for patches
maybe the "outright rejection" part is a little much
just like the bug tracker, we could set up a patch system, but I think it's overkill
alex_joni: oh? I know some projects have a tracker *called* patches but I didn't see one for our project
jepler: I can set one up, but I don't think we get that many patches to justify it
I think the best way to submit a patch to a project is together with a careful bug report
cradek: "outright rejection" happens -- remember the patch that modified what "G0" did, to change to the final Z before moving X and Y?
a patch is not a substitute for a careful bug report, but it's very nice to get both together
jepler: sure, but he knew that would happen, and wasn't asking for it to be incorporated anyway
question on commit: will cvs commit all changes in the current dir by default, or do you need to specify a file list?
SWPadnos: all changes in the current dir *and under*
if you cvs commit without a -m, you'll see the list in the editor
just wanted to know if the wiki page should have <filename> in the sample commit command
nope probably not
jepler: there are 4 default possible trackers
I'm not sure why we want to maintain our own CVS documentation, aside from what's peculiar/particular to our project
^^ an aside
bugs, support requests, patches, and feature requests
cradek: yeah I'm sure that page could be streamlined
jepler: we only have bugs & feature requests active
having this "outside documentation" in the wiki means people can find out everything about EMC2 and developing it in one place (once we have everything in there ;) )
so I think touching on all aspects is a good idea, with links to external in-depth documentation where applicable
I just would say - I used the cvs server before source forge and after.. other than my denseness.. I could not tell the difference other than my anonymous loggin worked on the new server - on sourceforge it was hit or miss.
I had no suspicion that supporting inflexible windows cvs clients would be important for the cvs of a project that runs only on linux
it is flexible, but hard to configure (go figure)
in fact, I think I can use the Tortoise CVS client with Altium and keep schematics and PCBs (and FPGAs and source code) in a CVS repository
it's really not hard to configure
SWPadnos: I just answered what you need to do
max 2 minutes
heh - thanks :)
get pageant (friend of putty)
maybe I'll update one or two of those dirs instead of deleting them then :)
get the private key safely to the doze box
(your problem how you do that)
convert it to ppk
start pageant, load the ppk
the only issue is that you don't have many possibilities in doze to protect the ppk
I protect it by leaving this unpatched Windows 2000 machine on the internet 24/7, and leaving my administrator account logged in
do you think that's enough?
install some OpenVNC server
that should cut things down to a couple of minutes before you're hacked
ooooh - and pcanywhere client
don't forget sharing C
and make guest an administrative user
yes of course
"Danke an die Programmierer,wenn nur alle Bugfixes bei anderen Programmen auch so schnell
"thanks to the programmers, if only all bugs would be fixed so fast for other programs."
that's from a german guy, who is happy about 2.1.7
was one of the bugs, uh, bugging him?