<tuxbrain>
I also want a "make all" that draws the pcb :P
<wpwrak>
having said all this, engraving would still be nice, because the machine could then do everything without even moving the board (well, one side). but it's difficult.
<wpwrak>
(the other side would be a pain, though, because you'd either need to position the board _exactly_ or adjust the machine coordinate system)
<wpwrak>
(make all) yeah, plus an automatic tool changer ;-)
<tuxbrain>
ok , for this I need an automatic tool changing but well I can live just screewing idle
<wpwrak>
get a 6 axis mill while you're at it :)
<tuxbrain>
yeah in conbination with a plastic extruder, and I can even make it a box :P
<kristianpaul>
you need a full line
<kristianpaul>
conveyors every where
<kristianpaul>
small boxes may be, like for atben/atusb. but for big things naw
<roh>
milling makes a lot of dirt. have a extra room for it or buy something with enclosure if not possible
<tuxbrain>
now seriously, there is no more solution to what I'm searching for than DIY, or pay a windows licence?
<kristianpaul>
you know wine? :_)
<tuxbrain>
yeah I know it but, If I pay 6k¬ I don't want to be the first to try if it works... and end buying a window licence :P
<kristianpaul>
askfor linux support,yuo-re the guy with the money, well, at least worth the try no?
<roh>
tuxbrain: usually if its parport controlled and runs with mach3 you can assume emc2 is possible by just adding config
<wpwrak>
roh: does anyone still have a parallel port these days ? ;-)
<wpwrak>
roh: (enclosure) DIY it !!!
<wpwrak>
roh: i make the comfty box my mill sits in with my own bare hands. no noise gets in, no dirt gets in or out :)
<roh>
usb isnt realtime enough usually. atleast not without extra hardware (counters/pwm gens etc)
<kristianpaul>
becasuse not use proper buffering..
<kristianpaul>
screw*
<kristianpaul>
wpwrak: there are PCI cards and such for parallel adding port i think
<roh>
and pci parport cards are cheap and easy to add 2 more. (we have and use 3 parports on ours
<tuxbrain>
mmm roh... how much to mount, and test a machine with such characteristics: not very big, accurate (enough to do let's say atusb?:P), no need to be fast, drill surface about DINA3, with a dedicate computer with EMC2 installed and tuned ready to accept GCODE or Gerbers through Ethernet port, via ftp or whatever....
<tuxbrain>
and including shipping cost to spain :)
<roh>
heh.. no clue
<kristianpaul>
you lost a bussiness ! ;-)
<wpwrak>
roh: hmm, i'm a little afraid when you mention the parallel port and "real time" in the same sentence ;-)
<roh>
wpwrak: with a rt kernel it is. every other port on a pc isnt really (besides some chipset vendow dependant gpio maybe)
<wpwrak>
tuxbrain: din a3 is quite large. the small affordable ones usually don't do much more than ~20 x ~20 cm
<wpwrak>
roh: still seems hairy. i think i'd take usb + a microcontroller :)
<roh>
and emc2 forces you to run a rt kernel anyhow (and brings its own drivers for parport) .. there is also the possibility to use external or internal fpga cards if the frequencies get too high for the parport
<roh>
wpwrak: more latency
<wpwrak>
kristianpaul: (circuit) sweet :)
<roh>
even the fpga is pci or parport via pci in the best case.
<wpwrak>
roh: put the control loop into your controller :)
<tuxbrain>
ok 20x20 is enough is a quotation inquiry no need to be answered in a second of course but a wild guess less than 6k¬ if the anwers is yes I can  busines is not over
<tuxbrain>
the business windows is not over
<wpwrak>
tuxbrain: if all you care about is drilling and cutting the PCB, but not engraving, then you have lots of choices and you can expect most of them to work without too much pain
<wpwrak>
if you require engraving, things get tricky
<kristianpaul>
chemical, chemical,chemical, you need dream with it tonight tuxbrain :)
<kristianpaul>
even with windows and that 6keur machine, i dont think that will be a one button to push solution..
<kristianpaul>
also you need to consider suplies for the head, replacement part for the future
<wpwrak>
oen button .. let's see ... 4th axis, tool changer, ... hmm, maybe 20 kEUR on a good day :)
<kristianpaul>
calibration over time
<kristianpaul>
yup wpwrak
<wpwrak>
will be noisy too, with the vacuum and such
<wpwrak>
also, once you have all the chemicals, you're also halfway there on your way to a respectable drug lab :) always a good options if those electronics don't sell too well.
<wpwrak>
(drug lab) of course, it seems that pretty much any chemical can somehow be used to make drugs. so this isn't saying much :)
<tuxbrain>
common guys, no need to be a one button push solution, manuall tool changing is acceptable, and some maintinace also, noise is not a problem then enclosure will amortiguate the sound and due we will drill small pcb not wood panels there ar quite house vacuums not too noisy tan can make the work
<tuxbrain>
the enclosure will amortiguate partially the sound.
<wpwrak>
yeah, cutting pcbs doesn't make so much noise. metals are worse :)
<wpwrak>
of course, the sound-proofing box is still nice if you decide to make a bunch of boards at 3 am ;-)
<tuxbrain>
no one catch the glove to mounting such machine?....
<tuxbrain>
any one knows this software (Circuit CAM, BoardMaster)?
<wpwrak>
i see a lot of "click" for a single button press solution ;-)
<tuxbrain>
(again) I'm (we are) not searching for a on click solution.... we are searching for an non chemical solution.
<tuxbrain>
LPKF ProtoMat E33 is sooo pretty
<tuxbrain>
just what we need but... window!! arg!.
<tuxbrain>
we have passed 2 years without need it an any part of the our job...
<roh>
tuxbrain: relax. maybe you should attend the camp and stay in berlin a few days.. then we can show you around and you can see what solution would be useable for you.
<roh>
cnc is always a custom solution imho. atleast if its nice to use in the end
<tuxbrain>
I can do a lot of things but travel :(
<roh>
oh. why?
<kristianpaul>
family guy :-)
<tuxbrain>
this is one, and other is a lot of task here requires my presence from now to the end of the year
<kristianpaul>
okay, i'm not going to write a pull-in algoryth.. whatever osgps does should work..
<roh>
i see
<roh>
maybe to congress or so
<tuxbrain>
well ce ya :(, I will go to sleep without my magic pcb creator machine.
<wolfspraul>
wpwrak: I don't understand your question
<wolfspraul>
yes gta02 had mp3 installed, but after legal consultation we were told the infringement could most likely also be claimed if it were merely installable from servers operated by om, or if om would give instructions on how to do it
<wolfspraul>
now, if the advertisement saying that it can play mp3 is a complete lie, then I'm not sure, if that's the case you want to ask about
<wolfspraul>
that's a pretty bizarre problem though, since you are making a false statement to your customers. sisvel may bring forward a case of unfair competitive behavior, after which you would have to stop saying (lying) about the mp3 capability.
<wpwrak>
(false claim) probably not really worth their time to go after you in that scenario
<wpwrak>
(origin of that infringement interpretation) thanks ! i was asking because i realized that the only source for this concept is "wolfgang said so". and i realized there must be a little more to it :)
<kristianpaul>
arghhh spam
<kristianpaul>
wolfspraul: after last mediawiki update i cant see a human verificaion method for public editing btw
<wolfspraul>
wait what are you trying to tell me?
<wolfspraul>
we get spam? the math captcha is missing?
<kristianpaul>
yes
<kristianpaul>
main page got spamed about a load modification,in the orange follow box
<kristianpaul>
reverted now..
<kristianpaul>
captcha missing yes wolfspraul
<wolfspraul>
oh nice
<wolfspraul>
yes I will look into the captcha
<kristianpaul>
yeah, bot like that :)
<wolfspraul>
kristianpaul: I just tried to edit anonymously, but the math captcha came up
<kristianpaul>
humm
<kristianpaul>
ah, yes
<kristianpaul>
it appear before save..
<kristianpaul>
okay, that wasa  smart bit
<wolfspraul>
yes
<kristianpaul>
bot*
<wolfspraul>
so before you just thought it was missing?
<kristianpaul>
sorry for the noise :)
<kristianpaul>
yes
<wolfspraul>
no problem
<kristianpaul>
biased by the upgrade ;-)
<wolfspraul>
thanks a lot for helping me watch out and do the maintenance work!
<wolfspraul>
oh sure, absolutely
<wolfspraul>
on guard
<wolfspraul>
I don't want to go sleep and then have to fish 100 pieces of bot junk out of the wiki.
<kristianpaul>
np, newsbeuter do all the dirty work, i just read my news :)
<wolfspraul>
the upgrade went very smooth
<wolfspraul>
took me 30 minutes maybe
<wolfspraul>
now we have a shiny new mediawiki :-)
<roh>
is that still untuned or will it stay that low?
<kristianpaul>
i think you can reacj 100Mhz, disabling some non needed cores  like tmu may be
<roh>
hm. so fpga simulated cpus are like 4-8 times slower than native ones given the same level of miniaturisation?
<roh>
or is there a fpga able to do something compareable to 5-800mhz arm cores in performance?
<kristianpaul>
it depends, what you want
<wolfspraul>
rejon: 80 mhz
<wolfspraul>
sorry I meant roh :-)
<kristianpaul>
and want you may want is not generic stuff i think :)
<wolfspraul>
roh: I think it's more like a factor of 20-30 compared to an asic produced with the same process technology
<wolfspraul>
maybe even more
<roh>
sigh
<roh>
and the same in cost eh? ;)
<wolfspraul>
raw production cost is always the same, and it's in the pennies
<kristianpaul>
fpga getting faster too, well xilinx s7 looks promising
<wolfspraul>
costs are determined by the business models of the companies making chips, the value (and investment) of their IP etc.
<wolfspraul>
I think the main thing we should look for is continuity. the rising tide.
<wolfspraul>
if we cannot achieve continuity, we will fail (with copyleft hardware as a concept)
<roh>
true. i still believe its the tools.
<wolfspraul>
roh: so it's even worse :-) our spartan-6 costs ca. 40 USD (other models go into the hundreds), whereas a 1 GHz ARM chip maybe 15-20 USD
<roh>
i for example still havent gotten used to binary tools and maybe never will again ;)
<wolfspraul>
yes correct, that's how I look at it. where's the continuity, how can we reuse free tech in the next improved product version, or a new product.
<wolfspraul>
if we get that done right, it will work
<wolfspraul>
the speed at which our ecosystem (any company) can make great performing _new_ products will be the differentiator
<roh>
i would be happy with something which competes against stuff on the level of a freescale imx 28 or so.. arm9, 450mhz, ethernet, nand/nor/spi flash and ~7usd$/chip
<wolfspraul>
I think we should look at products, including software. not compare chips.
<roh>
any slower and one only gets left boring products mostly or needs to compete on power savings/special features
<wolfspraul>
I don't understand the point of comparing chips. Are we sourcing?
<wolfspraul>
no not at all, I don't agree. you don't need a lot of 'mhz' for great products.
<roh>
wolfspraul: depends. on 100mhz i cannot even run a decent browser or ui which gets a vga screen driven
<wolfspraul>
(well, that's the point we are trying to proove. we better have sales to back it up :-))
<wolfspraul>
yes but no browser on a video synthesizer
<wolfspraul>
use your iPad :-)
<roh>
true. still you earn a very limited market by that.
<wolfspraul>
yes and no. indeed we need to find paying customers.
<wolfspraul>
but it's a big world.
<wolfspraul>
very big. there are many opportunities. we need a good product, good marketing, good channel. then a little luck. but there is no reason you cannot sell 10,000 Milkymist One, theoretically.
<wolfspraul>
or 100,000, for that matter
<roh>
i mean.. i guess there are more armzone develboards sold per type atm
<wolfspraul>
what's your point?
<kristianpaul>
cheap and fast :-)
<kristianpaul>
it seems
<roh>
well.. i would like to do projects based on free hw. but atm its just not feasible if the end product is as complex and expensive to make as a mm.
<wolfspraul>
sure so maybe we can branch out something much smaller and cheaper
<kristianpaul>
complex as any other hardware i guess
<wolfspraul>
there was the Xue plan for a while, in the attic now :-)
<wolfspraul>
we could make a board with the target being i/o, dsp, rf
<roh>
and while i got loads of hw around me i would like to make new ones sometimes. for me that means i need to limit myself to manually or hand solderable packaes tqfp and similar, no bga. same goes for most people (means boards on mills/etching by hand/pcbpool or similar outsourcing)
<roh>
which after some research came down with 'no bga means no soc with ethernet'
<wolfspraul>
what's your goal actually?
<wolfspraul>
to solder things together yourself because you like the experience?
<wolfspraul>
to make a high-performing product?
<roh>
a universal reuseable free cpu to run linux on. currently people use arm9 or faster for that.
<wolfspraul>
well you cannot have it all ways I think
<roh>
something versatile as a linux based arduino
<wolfspraul>
I like high performance because it enables/unlocks software.
<wolfspraul>
you try to have it all ways again :-)
<wolfspraul>
if you insist on tqfp package vs. bga, you may still have a 90nm semiconductor inside, right?
<roh>
nope. i neither asked for high multimedia performance or such.
<wolfspraul>
depends on what you want to run. normal people will mostly compare with what is available today.
<wolfspraul>
and since everybody can buy/use the same state-of-the-art, you do have to make a decision about performance early on
<roh>
but running linux on <400mhz is pain (speaking from experience) if you want to do any one of these things: drive a vga display, use wired ethernet not completely low end, play audio or such.
<wolfspraul>
what device are you trying to build? which function should it have?
<kristianpaul>
all of then  :-)
<wolfspraul>
I think we have a pretty good understanding of the 'hardware world', and improving.
<wolfspraul>
so for my part, I will just continue with that. and eventually we have the magic sauce together to make great competitive products.
<roh>
mostly 'connected electronics' .. e.g. a cpu board to put into your hifi amp, or a board to use with a tft as a home-control terminal with a touchscreen. or simply a device with ethernet which has some hands of gpios.
<wolfspraul>
I still have blind spots, and tools/processes I think can be vastly improved (use more boom for example).
<roh>
the whole point would be to make a board to sell and maybe single chips and or breakout boards to use on a breadboard
<wolfspraul>
there's tons of that stuff on sparkfun/adafruit/seeedstudios/arduino/etc
<wolfspraul>
nothing there that matches your requirements?
<roh>
well.. nothing has a free cpu.
<roh>
open
<wolfspraul>
puh
<wolfspraul>
if I were them I would even dismiss the idea :-)
<wolfspraul>
because I wouldn't understand the point
<kristianpaul>
it is, papilio board roh , i you like free arduino cpu :-)
<wolfspraul>
early on when I started with copyleft hardware, we tried quite hard to get in touch with the Arduino folks
<wolfspraul>
not easy because they carry their nose quite high
<wolfspraul>
but eventually we made it, we got a serious email conversation back and forth going
<roh>
currently freescale is what makes the stuff most interresting, but everything with ethernet is bga again also there. only the variants without are hand-soldering compatible
<wolfspraul>
and at the end of the conversation, the bottom line was that they are not interested in the entire concept/idea of 'free/open cpu'
<roh>
kristianpaul: arduinos are too low end for serious ethernet devices
<wolfspraul>
they are thinking exactly in the opposite direction "we have to dismiss this kind of idea" "we have to become better Atmel customers"
<wolfspraul>
and that makes perfect sense to me
<wolfspraul>
so they are focused
<wolfspraul>
they want to become better Atmel customers, over time. that's what enables them to make their products.
<kristianpaul>
hah (atmel loyalty)
<roh>
well, from their pov there is no way to sell something much more expensive that what they have, since customers dont really care.
<wolfspraul>
will Atmel start to work on a 'free cpu'?
<wolfspraul>
that would be a very strange thing to expect from them
<wolfspraul>
no it's all fine, it's just different things/concepts
<wolfspraul>
it makes perfect sense to me
<kristianpaul>
sure
<wolfspraul>
they weren't even interested if we would do all the work of 'freeing the cpu' for them :-)
<wolfspraul>
THEY WANT TO BECOME BETTER ATMEL CUSTOMERS :-)
<kristianpaul>
but no make sense for a open hardware project over the time..
<wolfspraul>
not in my book, yes
<kristianpaul>
it will be just arduino, nice trademark :)
<wolfspraul>
we need a platform that will allow any company to come in, take our starting point, and be able to make a world-class leading performance piece of hardware fast
<wolfspraul>
the key 'bottle opener' is free software in that case. no need to sign ndas, to ask questions, to go through lengthy licensing talks.
<wolfspraul>
just take the stuff, customize, make, sell
<wolfspraul>
either that idea works or it doesn't work :-)
<roh>
well.. a proper working develboard with openwrt port which sells for <200$ would be something i would buy immediately
<roh>
i mean.. even with 200mhz or so for the beginning that would be great
<wolfspraul>
he
<wolfspraul>
we had some crazy plans to sell Xue for 99 USD
<wolfspraul>
but 199 USD would have been possible for sure
<roh>
currently people buy routers or so to do such stuff or use $arm develboard from olimex. but the ports are often a pain
<wolfspraul>
I think as we continue to build our free platform, the opportunity to build this kind of thing will emerge.
<wolfspraul>
without large investments
<wolfspraul>
we just have to continue steadily to build a truly free and (re)usable platform
<roh>
true.
<wolfspraul>
all the way to KiCad, boom, missing MMU, missing this and that
<roh>
what would be the minimum bom to run a nanonote cpu? (crytal, ram, flash? can it boot from sd?)
<wolfspraul>
so someone can realistically branch out and make something new and improved and focused on a new use case with a few thousand USD in cash
<wolfspraul>
you mean the nanonote board?
<wolfspraul>
maybe 30 USD or so
<roh>
put only the essentials on a board, wire the rest on pads in a square or like dil (breakout) .. sell it for low 2 digit euros
<kristianpaul>
99 ;-) ?
<wolfspraul>
oh it's all dirt cheap, but that skips over all the hard parts
<wolfspraul>
the hard parts are the tools, processes, testing, software ports, kernel drivers and stability, actual performance of key features, and so on
<roh>
true :) just thinking what the next step could be and how to attract lots of multiplicators
<wolfspraul>
an entire Android smartphone with gps and wifi and touchscreen can be made for ca. 30 USD now
<wolfspraul>
can and is
<roh>
i want to make something i always missed and which would make my own prototyping easier and nicer. stuff i like so much to work with that i would use it even in bigger projects
<wolfspraul>
maybe you need to be a little patient and hopefully Milkymist SoC and tools and ecosystem over time will become your starting point for such projects
<wolfspraul>
I think it's moving in the right direction
<wolfspraul>
but hasn't taken off yet :-)
<roh>
sure. as soon as there are no evil tools neede anymore ;)
<kristianpaul>
or start learning verilog or some HDL now :-)
<kristianpaul>
roh: alwayls will be *evil
<roh>
kristianpaul: i would if that wouldnt mean wasting another 2gig diskspace on sick shit
<wolfspraul>
10gig
<roh>
wtf?
<roh>
seriously?
<wolfspraul>
the xilinx ise download is already 3.7gb
<wolfspraul>
oh sure
<wolfspraul>
after installation it's over 10 gb
<wolfspraul>
kristianpaul: true?
<roh>
and one needs all that? isnt there a script to throw all not essential stuff away?
<kristianpaul>
hey,you're pickier than me btw roh :-)
<wolfspraul>
it's one huge thing, very hard to take it apart. I mean there's work everywhere in tools etc. but few people doing that, really. from a free software perspective.
<kristianpaul>
yes
<wolfspraul>
so it's like 10 people worldwide doing that
<roh>
kristianpaul: well... i just know from experience what motivates me and what not. gcc and makefiles always were my friends.
<kristianpaul>
good choice :-)
<wolfspraul>
and Xilinx alone probably has hundreds that continue to add more bloat to the prior bloat, every day.
<wolfspraul>
roh: it's going to take many more people to dive in there and take it apart and open it up.
<kristianpaul>
wolfspraul: 4.7Gb last installer for version 13.2
<roh>
even when it explodes or i switch platforms.. its always there. using binaries always makes me remember the countless times of evil hacks, workarounds, broken stuff i couldnt debug or fix even if i knew how or what it was... that demotivates more than my nosyness which is one of my primary drives.
<wolfspraul>
Sebastien himself has made a significant start with the llhdl project, but even that is at best in alpha state, and only addresses a part of the process.
<wolfspraul>
ah 4.7 already
<wolfspraul>
there you go
<wolfspraul>
maybe they have 1000 people adding every day, Monday to Friday :-)
<kristianpaul>
ans 12G installed
<wolfspraul>
we are trying all we can to free this
<wolfspraul>
really
<wolfspraul>
but it's little baby steps
<roh>
i know. maybe i am impatient ;)
<wolfspraul>
we can reflash boards over jtag now without the Impact tool (xilinx tool)
<wolfspraul>
that was a lot of work
<wolfspraul>
we can write the eeprom of the jtag-serial boards without the ftdi windows software
<wolfspraul>
that was a lot of work
<wolfspraul>
but there are things 100 times bigger than that waiting in line
<wolfspraul>
kristianpaul knows better than me
<wolfspraul>
kristianpaul: how about debugging and simulation tools? You use a lot of ISE stuff, or free tools?
<wolfspraul>
kristianpaul: which ISE tools do you frequently use nowadays? and which would be easy to replace with free tools?
<kristianpaul>
free tools for simulation
<kristianpaul>
well,the only ISE thing i need a replacement is sinthesizer :-)
<kristianpaul>
debugging with scope ;-)
<kristianpaul>
and a cheap logic analizer with my avnet board
<kristianpaul>
also using milkymist bios, very usefull for debug too
<wolfspraul>
which simulation tools do you use?
<kristianpaul>
icarus verilog
<roh>
hm. how many chips would one need to sell per revision of a soc to make any sense?
<kristianpaul>
cver for verification
<roh>
i mean... how is the cost of asic production in detail?
<kristianpaul>
and gtkwavefor the vhd out coming from verilog testbench
<wolfspraul>
oh that's very difficult to answer
<wolfspraul>
first - everybody is moving forward, so you have to factor that motion (and the related investments) into your plan
<wolfspraul>
there may be large investments going on around you that you can piggypack onto
<wolfspraul>
when you go to asic, you loose the programmability (of course :-))
<wolfspraul>
and the ability to bugfix/update your product
<wolfspraul>
so you need to factor that in and express it numerically somehow
<kristianpaul>
(bugfix) oh dear..
<wolfspraul>
there are steps between fpga and asic, 'structured asic'
<roh>
sure. thats why it shouldnt be too expensive. say.. 10-20E/each chip
<wolfspraul>
they have more unique pros and cons to be considered
<wolfspraul>
any optimization will make you dependent on the technology around you, so your design may depend on something in the fpga that is hard to move to another process, even another fpga
<wolfspraul>
finally when you do start producing, you may need to license/become part of proprietary IP of the foundry, even if you manage to not need any IC design IP
<wolfspraul>
there may be different models there with upfront payments, or royalties, etc.
<wolfspraul>
as a 'Chinese rule of thumb', you don't need to consider any move away from fpga until you need at least 10,000 pieces of that chip
<wolfspraul>
that's what I was told by very experienced IC designers who make these kinds of decisions all the time
<wolfspraul>
and like I said, even then you need to carefully consider the various 'move to what' options
<wolfspraul>
because there are many, and the costs differ widely
<roh>
so there is no way to build a non-fpga asic when the market isnt something like 20000 or more per revision?
<wolfspraul>
once you get over all this, if you remove all investment/one-time/IP/royalty costs, the wafer is always dirt cheap. figure 50 cents / chip (say for a 5x5mm die)
<wolfspraul>
he
<roh>
or how does the initial/running cost ratio stand?
<wolfspraul>
I just tried to tell you there are several factors to consider
<wolfspraul>
it's an equation with a number of unknowns
<roh>
what does a new 'mask' cost etc
<wolfspraul>
so I cannot answer your question
<wolfspraul>
because every move you make has pros _and_ cons
<wolfspraul>
short-term and long-term
<wolfspraul>
yes, the films are the most expensive thing
<wolfspraul>
once you have a film for a wafer, and all IP/foundry issues sorted out, you can make chips dirt cheap
<wolfspraul>
with my current knowledge, Milkymist the ecosystem would not benefit from any move away from fpga for several years at least
<roh>
true. and since its a small market maybe never
<wolfspraul>
it won't help us accelerate the speed at which the free CPU can grow
<wolfspraul>
ah I wouldn't say that
<wolfspraul>
we just have to go step by step
<wolfspraul>
let's look at a smaller problem - when/how/if to consider Xilinx -7 series, switch to Altera, etc.
<wolfspraul>
even that decision is hard to make, and for now I agree we should first max out what we can do on a Spartan-6
<wolfspraul>
our resources may not be best spent switching now, but it's a tradeoff...
<roh>
true.
<wolfspraul>
so if even that decision is hard to make, you can imagine that a decision whether to 'do asic' is 100 times harder to make still, it's totally speculative at this point.
<wolfspraul>
let's just learn more things and move step by step, then slowly the options that make sense will become more clear.
<wolfspraul>
on that end I am also following Andrew's homecmos project with great interest lately
<wolfspraul>
not that I think we will make our own asic or fpga in that process anytime soon :-)
<wolfspraul>
kristianpaul: or are you starting to setup your semiconductor home lab already? following Andrew's lab notes?
<roh>
hrhr. sure not.
<roh>
(own fpga soon)
<wolfspraul>
roh: oh, it may emerge as another options, Andrew seems very determined.
<wolfspraul>
not fpga, asic first.
<wolfspraul>
cmos
<kristianpaul>
wolfspraul: not yet
<wolfspraul>
but if anything it may become an option for very small specialized chips
<roh>
the homecmos is interesting for sure. but i am not sure how far thats useable for people besides learning
<wolfspraul>
that remains to be seen ;-)
<wolfspraul>
Werner is making home PCBs and that does serve a very useful function for him.
<wolfspraul>
Andrew's process may very well one day become something that does serve a function somewhere in our processes or products.
<wolfspraul>
'one day' can be years out though, and he or someone else needs to invest a lot of time and money in between
<roh>
true. after all opensource also means a 'truth by somebody tried and documented the result before' lots of learning by (sometimes) systematic testing possibilities
<wolfspraul>
if he can make a working Intel 4004 one day, that'd be awesome. lots of things you can do with a 4-bit cpu...
<wolfspraul>
roh: yes, his lab notes are excellent. did you see them?
<roh>
not really engineering all the time. or maybe a more.. 'special' kind of engineering (opensource in general)
<kristianpaul>
Andrew is very responsible about that
<roh>
wolfspraul: some
<kristianpaul>
is not that wild hacking that get lost somwhere..
<wolfspraul>
correct
<wolfspraul>
so it may add to our 'continuity', though not now in any practical way, for sure
<roh>
proper opensource always has continuity. since its documented and can be reproduced. havent seen than in the commercial world really
<wolfspraul>
continuity is on cash flow there
<wolfspraul>
(seriously, no offense I like this perspective but that's how it is)
<wolfspraul>
and that's not a bad continuity imho :-)
<roh>
well.. imagine all free gpios would be available directly in linux via the gpio api and some easy to use lib exposes that arduino style
<roh>
you could even have the gcc on the device to compile the 'apps' people hack up to 'wiggle the 4 pins they use'
<roh>
or maybe even code in scripting languages via some web-ui
<wolfspraul>
you are talking about the m1 expansion header? or nanonote/ubb?
<roh>
any platform one would want to make popular for the 'swiss knive tool' jobs arduino is not mighty enough
<wolfspraul>
the swiss knife. I don't even have one.
<wolfspraul>
don't know how to make a good swiss knife. seems a very niche idea to me.
<wolfspraul>
I'm not even sure Arduino is or wants to be a swiss knife. The one word that pops up repeatedly around Arduino users is 'interactive'.
<wolfspraul>
and interactive can be done without high bandwidth/high mhz
<wolfspraul>
some Arduino folks like to say 'physical computing', which in my understanding is similar to 'interactive'
<wolfspraul>
swiss knife, don't know. don't know how to make one, don't need one.
<wolfspraul>
if we do well Milkymist One can be known to be a good platform for visual things.
<wolfspraul>
that would be a starting point imho
<wolfspraul>
visual swiss knife :-)
<roh>
interactive as in development. the '20 minutes to blinking led' is _very_ important to give people the learning curve and satisfaction to keep them interrested and continue the learning
<wolfspraul>
for myself, I want more software
<wpwrak>
catchig up ... (ftdi a lot of work) hehe. the joy of undocumented internals and the resulting more than dubious drivers ;-)
<wpwrak>
roh: so make an arduino ursurper with some low-cost linux-capable cpu ? you don't need an FPGA for that.
<roh>
wpwrak: that would be nice.
<wpwrak>
roh: for the FPGA, if you want to sell it to the nerds, you probably want to emphasize what cool little special-purpose hw you can synthesize. basically what kristianpaul is doing, just 1000x times easier and faster :) like your 20 minutes to LED
<roh>
wpwrak: i mean.. even the ingenic would do. just make it boot from sd and add some ram, crystal and thats it. rest can be external
<wpwrak>
(fpga) and of course, the tools need to be free
<roh>
wpwrak: for fpga to 'take off' it needs much better and open toolchain first imho.
<wpwrak>
(ingenic) the newer ones can boot from SD. so ... do it :)
<roh>
else people like me will always use something else (simpler, less hassle)
<wpwrak>
if you look at "linux en caja", thet pretty much have what you want. except that they still have NAND, i think
<roh>
wpwrak: are there some with eth mac and in tqfp?
<roh>
nand is pain and nothing i would like to burden somebody with ;)
<roh>
and it eats pcb space
<wpwrak>
hmm, tqfp may be. the ethernet could be your first "shield" (-:C
<roh>
then no.
<wpwrak>
please. you're seriously trying to convince ->me<- that NAND sucks ? ;-)
<roh>
i want internal mac. something proper.
<roh>
not that spi or dm9000 play-doh crap
<wpwrak>
the linux en caja board is open, designed in kicad. so you can make you own variant. it's two layers.
<roh>
well.. then imx28 or so would be the target.. but that excludes all hand-soldering again *sigh* (bga)
<wpwrak>
so you can just add your ether to the design
<roh>
wpwrak: there is no really nice 10/100 nic which can be added without a messy waste and a full databus
<roh>
thats WHY i want internal mac :)
<roh>
keep the bus clean for nice stuff
<wpwrak>
linux en caja is basically the sakc/sie minus the fpga. but i think you could still simplify it
<wpwrak>
hmm, usb to ether dongle ?
<roh>
nope. too low end
<roh>
have you ever tried using all that crap?
<wpwrak>
you're after GE ?
<roh>
no. but serious working 100mbit with working hw filters.
<wpwrak>
hw filters for your own MAC ? or for firewalling ?
<roh>
means something like 6-10mbit real multicast load with small packets on the same port should not make it go down like e.g. all enc28xx or dm9000 do
<roh>
wpwrak: hw mac filters like every half way sane mac does
<roh>
so if you dont subscribe to a mac group matching in hw, you dont even get an irq
<wpwrak>
btw, does thiemig.de have a web shop or a price list ? i wonder what those adhesive mats cost (or where to source them properly, without going through a .de importer)
<wolfspraul>
ah the NAND deniers again
<wolfspraul>
good thing someone invented SD
<wpwrak>
roh: (multicast) is it still not dead yet ? it's been smalling funny for a very long time now ;-)
<wpwrak>
s/small/smell/
<wolfspraul>
I just wish more people move from talk to action. I've probably worked with more SD cards than anyone else here.
<wolfspraul>
they are pretty much having the exact same issues as NAND. ah no, wait. It's SD, I forgot :-)
<wolfspraul>
I'm not against SD btw, but if you have trouble with NAND, you will have the same trouble with SD.
<wpwrak>
wolfspraul: an idea for the marketing department: since linux "BogoMIPS" are self-declaredly bogus, why not left-shift them by, say, 3 bits ? :)
<wolfspraul>
I'm fine with either
<wpwrak>
naw, SD troubles are different. easier software. more graceful aging. more obscure and harder failures.
<roh>
wpwrak: mail him.. or call. its a really small business
<roh>
but quality goods
<wpwrak>
nand can also be faster, although surprisingly often it isn't
<roh>
wolfspraul: the point with sd is: i do not sell it. its 'user replaceable'
<roh>
and its not 'expensive' as low amounts of nand and makes is 'my fault' to get badblock-foo working but somebody elses (sandisk? etc)
<wolfspraul>
well then. I have a lot of practical experience with both and I'm saying they are equal.
<roh>
in short... it makes the thing 'permanent storage' in general somebody elses problem and also 'easier to swap' than bga nand ;)
<wolfspraul>
sourcing is the same, tech difficulties are the same, all the same
<wpwrak>
"Handhebelschere (Papierschere)" the translation into normal language is nice ;-) "can i have a glass of 200 ml concentrated hydroxyl acid (water), please ?"
<roh>
wolfspraul: alone the fact that one can remove it, dd a image to it, plug it back in and go on working is worth it for me.
<wolfspraul>
my feedback is on record :-) go for it!
<roh>
sure... i will put it on my list ;)(
<wolfspraul>
the good news is - your life will not be harder with SD than with NAND
<wolfspraul>
it pretty much evens out, especially at low volume stuff
<lekernel>
i'm not sure
<roh>
but i have another (1 device for now) project before that.. which will use a pcengines board as linux hw
<lekernel>
sd cards are extremely pesky when it comes to being compatible with all of them
<wolfspraul>
in higher volume you will start to pay an increasing premium for anything SD
<lekernel>
thank their crap overengineered protocol
<roh>
because sadly,, its so cheap and easy to work with... its a pc for 70E
<wolfspraul>
lekernel: yes, one of many factors
<wolfspraul>
I agree
<wolfspraul>
difficult beasts, just like NAND
<lekernel>
I guess they want to keep the IP vendors happy and their design comittees well paid
<wolfspraul>
like any hugely successful standard there's a lot of pulling by many parties, extensions, etc.
<wolfspraul>
the only times you don't have that is with the unsuccessful standards, and then nobody wants that either, right? :-)
<lekernel>
and make people who paid for the spec feel they got their money worth
<roh>
hm. weird.. my flash somehow works basically allways.. maybe i just buy too expensive ones
<lekernel>
they would look stupid asking people $30k for a document that basically says "here's the block read command: send command, send address, get data back.  here's the block write command: send command, send address, send data, get ack."
<lekernel>
that's however the correct way to make a memory card, technically
<roh>
lekernel: sd isnt that far from that ;)
<lekernel>
lol
<lekernel>
roh, get it to work on milkymist :)
<lekernel>
this will make you swallow those words
<roh>
eheh.. let me guess.. you believed any of the specs?
<roh>
maybe we need to sniff some of my sandisk readers timings. it never failed me so far
<lekernel>
I don't think it's a timing problem
<lekernel>
more various options of the overengineered protocol
<roh>
on the moko also most of our sd issues were timing crap or similar near-electrical details
<roh>
basically no card really does 25mhz e.g.
<roh>
anyhow. werner knows more ugly details there than i do ;)
<wpwrak>
the main SD issue we had in openmoko was EMI
<roh>
gladly forgot some of the horrors of engineering abyss we went through back then
<wolfspraul>
I have seen and worked with hundreds of cards, dozens of manufacturers. Can only repeat it. Go and do more of those things, then let's compare our experiences.
<roh>
wpwrak: that was the 'drive strenth' thing right?
<lekernel>
wpwrak, but you used the SoC's controller and off the shelf software, right?
<wpwrak>
the good thing with SD/MMC/etc. is that most of it is a path well traveled if you use regular hardware
<wolfspraul>
lekernel's observation makes full sense to me. lots of work waiting for SD lovers in Milkymist, btw
<wolfspraul>
the first 'if', Werner hedging his bets :-)
<wpwrak>
of course, on the M1, things are a bit different, because you start from scratch with everything
<roh>
wolfspraul: moko also was a bad example... not the nicest soc, not the best engineering.. basically a chain of spofs all adding some extra bugs for werner and harald to stumble upon
<wolfspraul>
first of all everybody agrees that sd is a layer over nand, I would hope
<wpwrak>
but even the M1 should be fine once you overcome that hill
<roh>
lekernel: the mainboard of the moko was a horror of manual routing madness
<wolfspraul>
and from then on there are _many_ pros and cons for each one, that's why both are still so prevalent
<wpwrak>
roh: drive strength and "superstition caps" :)
<roh>
lekernel: we had evil problems from emi to speakover from line to line
<wolfspraul>
the Palm Pre had an actual card on the board, right?
<wolfspraul>
something like that
<roh>
magnetic and or capacitive coupling over multiple subsystems
<wpwrak>
lekernel: yes, soc's controller and linux :) that doesn't solve all the problems, but quite a lot of them
<lekernel>
in milkymist, linux _adds_ quite a lot of problems too
<roh>
wolfspraul: dunno. there is something called 'movinand' from samsung which is basically a sdcard as bga (saves pins on the soc)
<wpwrak>
lekernel: so for the board roh wants, that's by far the easiest way forward
<wolfspraul>
wasn't the most successful product :-) (I'm not saying that one tech detail, if it's even true, was decisive for the market failure)
<wolfspraul>
yes I can imagine crossovers
<wolfspraul>
connector alone is a headache, for manufacturing. well, there are many pros and cons, really.
<roh>
wolfspraul: i see 'soldered' or even plugged fixed assembled sdcards more often lately. especially on new android devices from asia
<wpwrak>
lekernel: c'mon ! linux now runs now in a usable form for for what, about 3 whole days, and you're already complaining that it adds a lot of problems ? :)
<roh>
easier to manufacture. you can just flash the sdcards en batch and simply let the last stage flip it in and clip on the cover. done. no complex flashing with connectors and such
<wolfspraul>
I don't think you can imagine how efficient NAND flashing is
<wolfspraul>
there are special fixtures and PCs just for that, and you will _NOT_ be able to beat their performance easily. trust me.
<lekernel>
wpwrak, still lacks a lot of drivers. plus we still did not benchmark performance.
<roh>
i know whats possible. and even gang programming doesnt save you from reflashing hw with a custom rig when you got some hw on stock and release a new revision
<wolfspraul>
they are shuffling the megabytes onto these chips in parallel, unbelievable. at high volume you can even do that higher upstream, right where the NAND chips are made (same for SD cards I would think, but NAND is probably more common in very high-volume applications).
<roh>
the replaceable way means you stock un-populated hw, readily tested and only add the sdcard shortly before sale.
<wolfspraul>
most stuff nowadays is over-the-air updatable, or moving there. cards are still a logistical nightmare.
<roh>
wolfspraul: i dont speak about 'high' volumes .. rather what we see.. few to maybe 4 digit sales
<wpwrak>
wolfspraul: (sd connectors) i see two types of uSD connectors: the simple and cheap ones i get from digi-key, which work great. and then the ones people put into certain products, which don't :)
<roh>
you know how it is.. whats on the box is what the customer has as first experience. better make it count
<wolfspraul>
in low volumes sure, you may come to different conclusions
<wolfspraul>
wpwrak: on M1 we have the best we could find, from Molex :-)
<wpwrak>
lekernel: (lots of drivers, benchmarks, ...) 3 days !!!!
<lekernel>
3 days?
<lekernel>
almost 2 years no?
<wpwrak>
3 days since the the damn thing actually booted into something you would call usable, even if barely
<wpwrak>
seems to me that since larsc and mwalle got their hands on it, it moved forward quite well
<wolfspraul>
that's not entirely fair
<wolfspraul>
it has quite a long history, but of course Lars and mwalle really kicked things up a few gears in recent weeks
<wolfspraul>
there was another burst last December/January
<roh>
well.. i will watch some tv and go to bed now.. it seems it will rain for the next week or so anyhow
<wolfspraul>
and all of last year (2010) and before, Takeshi Matsuya was quitely chugging away and upleveling it, and it booted as early as 2009 I believe?
<wolfspraul>
but of course there can be a world between 'it boots' and 'it boots'
<wolfspraul>
and yes, progress lately is FANTASTIC!
<wpwrak>
wolfspraul: (memory card holder in M1) hmpf.
<wolfspraul>
hmpf? not good?
<wolfspraul>
we chose the molex one after actually comparing different ones, the Molex one was worth its extra price tag
<wolfspraul>
I think if you do a connector, you have to do it right
<wolfspraul>
but a 'right' connector will cost
<wpwrak>
the thing with such projects is that they only move quickly while the people doing them feel an itch to scratch. and they're hard enough that you don't have a lot of people who can do them. and of course, the platform is obscure enough that few feel that itch.
<wolfspraul>
huh? and? what's your point? yes sure, that's what we need to break through...
<wpwrak>
(molex) i'm just surprised it's still so bad. at least the one in rejon's M1 felt rather poorly designed
<wpwrak>
(point) sure. and that's what's happening now. i'm just answering lekernel's "2 years"
<wolfspraul>
maybe you demand the impossible sometimes
<wolfspraul>
a mechanical connector with those kinds of precisions is _HARD_
<wolfspraul>
and even though it's 'expensive', it still sells for cents
<wolfspraul>
best is no mechanical connectors at all, all RF
<wpwrak>
of course, once linux runs properly, with an RT core allowing flickernoise to perform smoothly, rtems will probably forgotten within half a week or so ;-)
<wolfspraul>
possible but even in that scenario it may re-emerge later in whatever other Milkymist SoC based product
<wolfspraul>
(and I don't think this scenario is likely anytime soon)
<zulu7>
I just bought a TP-Link TL-WR1043ND (AR9132 chipset) and am now wondering what firmware to use... is, for example, libreWRT known to work on it?
<wolfspraul>
if I understood things correctly, even Harald switched away from Linux to a microkernel called 'nuttx' right now, for osmocombb
<wpwrak>
(connector, impossible) dunno. i find it discouraging when, after finally prying it open and inserting a card, the lid comes off when trying to close it. luckily, nothing broke. so at least the "fail safe" part of the design worked ;-)
<wolfspraul>
if that allows him to make more progress in the short-run, I think it's the right decision. Linux doesn't run away...
<wpwrak>
(connector) but perhaps the difficult to access location also played a role there
<wolfspraul>
it will all move to RF
<wolfspraul>
connectors are a pain
<roh>
wolfspraul: not switched away. the hw isnt feasible for linux and nuttx is simply much less, but useable on arm7
<wolfspraul>
yes so I understood it correctly :-)
<wolfspraul>
it's the right decision, it's not 'anti' Linux, but in favor of PROGRESS, which can later applied to other environments, including Linux
<wolfspraul>
that's how I see it
<wpwrak>
(rtems vs. linux) oh sure. for development whatever help you keep going is good ;-) but eventually, the higher layer driver monster will raise its ugly head beyond the point where you can just deny its existence
<wolfspraul>
nobody would dispute that
<wolfspraul>
also Linux strength in networking, for example
<wpwrak>
yeah, that too
<wpwrak>
but anyway, that's still a while away. first more drivers need to work, then someone has to figure out the MMU, then some decent RT core needs to be integrated, and then comes the tuning for flickernoise. the good thing is that the road is pretty clear ;-)
<larsc>
well, the problem with drivers for the more fancy peripherals is, that without flicknoise there would be no user of drivers
<wpwrak>
roh: there are several bus variants. look for kzs8841 and kzs8851
<roh>
hm. but all need lots of pins
<wpwrak>
larsc: X ? :)
<roh>
thats why i wanted the internal mac.. to save pcb space.
<wpwrak>
roh: then take SPI ;-)
<roh>
wpwrak: or a freescale in bga
<wpwrak>
(internal mac) then i wish you luck. you'll need it, considering all your other constraints ;-)
<wpwrak>
and what's so bad about, say, 8 data + a few control pins ?
<larsc>
wpwrak: X can't make much use of it either
<roh>
wpwrak: it need _lots_ of pcbspace and eats gpio
<roh>
would make a '40pin dil type' breakout bigger
<roh>
think arduino with ethernet and linux
<wpwrak>
are the arduinos dil40 compatible ?
<roh>
juzt a userspace process as 'app' instead of a flashed 'app'
<wpwrak>
larsc: hmm, chicken and egg then. well, just makes the pill a bit harder to swallow, not less tasty :)
<roh>
wpwrak: nope.. but breadboard-compat
<wpwrak>
roh: so it seems that space isn't a concern if you want to compete with arduino
<wolfspraul>
bga is fine, just find a vendor to do soldering for you, can't be that hard
<roh>
wpwrak: its also about drivers and design simplicity
<wpwrak>
roh: i mean, they're using technology made in the 1990es, that tried hard to uphold the virtues of 1980 nostalgia ...
<wpwrak>
(simplicity) how is requiring a chip you can't solder and that forces you to go multilayer simpler than two chips that don't have these properties ? :)
<zulu7>
where do I ask GPL related questions?
<wpwrak>
wolfspraul: unless you value your own work at ~$0/h (and china is far away :)
<wolfspraul>
ok I cannot give reasonable input without understanding the end goals, true
<wolfspraul>
why not 2 simpler chips? I don't know
<wolfspraul>
if we are just discussing to find the one and only eternal world truth, we will never find it anyway :-)
<wolfspraul>
zulu7: ask here, we are all friendly people :-)
<wpwrak>
i think that's pretty much what roh is looking for ;-)
<wolfspraul>
zulu7: but sorry I have to run, he he. dinner time. check back later, GPL questions are always a good start here...
<zulu7>
I heard that Android 3.0 and 3.1 are closed source so far...
<wpwrak>
roh: kick out the nand and put the ethernet chip in its place. done.
<larsc>
zulu7: they don't have to publish the sources until they publish the binaries
<roh>
wpwrak: nice board. do you know how many layers it has?
<zulu7>
larsc, tablets and stuff has been released featuring Android 3.0 and 3.1
<roh>
or how much it costs in bom?
<wpwrak>
roh: 2 and it's all kicad
<wolfspraul>
the (forked) Linux sources have been published afaik
<wolfspraul>
they go by the book, not by the culture. was never meant to.
<wpwrak>
(bom) that you have to ask those guys
<roh>
wpwrak: actually i wasnt sure if i want to put the phy on the same pcb and not only do a module which could also be a 5-8 port switch instead a single phy
<larsc>
zulu7: according to the internet they released the gpl'ed parts of 3.0
<wpwrak>
roh: get some wlan box supported by openwrt and rip out the pcb ?
<wpwrak>
add a passive auxiliary board with the DIL40 header ;-)
<roh>
wpwrak: thats what i do atm.
<roh>
but you know how that is. difficult to get the same hw reliably again and again
<larsc>
roh: what do you want to build?
<roh>
larsc: nothing specific right now. i just wished for a cheap enough single board linux computer with lots of gpio and proper ethernet
<wpwrak>
larsc: in german ist's called "eierlegende wollmilchsau". and it also has to be a dinosaur ;-)
<larsc>
i see
<roh>
wpwrak: no. much less low end that stuff like the beagle
<wpwrak>
roh: and an efficient multicast router :)
<roh>
wpwrak: not neccessary. just not crashing or dos-ing by something like that would be nice.
<wpwrak>
i think you're falling victim to the "there's no tomorrow" syndrome. you shouldn't try to design every feature you'd eventually like to have into it. make something basic that fulfills your minimum needs. then see how to go from there.
<wpwrak>
otherwise, you end up with an ever-growing design you'll never quite finish
<wpwrak>
(and it'll be expensive, too)
<wpwrak>
particularly if it's a DIY-able board, the cost of making a variant is marginal anyway
<DocScrutinizer>
honestly, using such a batch for "ticket" of a 4 day meeting is so extremely geek, I love it. Schematics also look good on a first cursory review
<DocScrutinizer>
I wonder though if "Transfer data to other badges via radio: /firmware/tbd" means ToBeDone or TransferBinaryData
<DocScrutinizer>
s/batch/badge/
<wpwrak>
(tbd) probably both :)
<DocScrutinizer>
>>High-current sink drivers (20 mA) on two I2C-bus pins in Fast-mode Plus.<< WTF?
<DocScrutinizer>
is this meant to establish a 1Mbaud 20mA current loop bus, or what?
<DocScrutinizer>
I thought 20mA tty was obsolete even back in 1985
<wpwrak>
20 mA * 3 V (?) = 60 mW tx power :)
<kristianpaul>
why radio is 2.4Ghz... 433hz and alike, are evry common for rocket telemetry, at least in the UK i had seen