#qi-hardware IRC log for Saturday, 2011-07-23

kristianpaulwpwrak: nice caps!04:31
kristianpaulyou mean  extrusion  in software?04:32
kristianpaulthats interesting, or you get a personal 3d printer? :-)04:33
Action: roh wants to play around with injection molding. .. JUST need to get my hands onto hw (or build own!)04:33
roh_after_ completing the cases .. and the camp :=)04:34
wolfspraulroh: how is the case work going? got the material back?04:35
rohif it can be done in a vending machine, it can be done in a hackerspace :) http://www.replicationdevices.com/vintage.html04:35
rohwolfspraul: yeah. fetched it today.. payed in cash *sigh*04:35
rohused the day to sort and clean up parts04:35
kristianpaulha, now i cant complain openscad is slow in render after reading results for this cadmium..04:36
rohlooks good so far. will do fit-tests later04:36
kristianpaulfit, *green*04:37
kristianpaulor was grin :p04:37
rohthen i 'only' need to glue another 480 parts together to make up 240 buttons 04:37
rohand package the whole thing.. man .. thats a lot of acryllic04:37
wolfsprauland the logo04:38
rohdidnt get so far yet04:38
wolfspraulsure, the other stuff first. but it soudns good. congratulations you got it this far!04:39
wolfspraulAdam is still in the critical testing phase, some things look suspicious but I have to be patient and wait another 2 days or so before we know more statistics :-)04:40
rohthanks. how should i package it btw? as 'kits' or rather 'bulk' ?04:40
wolfspraulbulk is probably easier for you04:41
wolfsprauland to me there is no advantage of packaging it in kits04:41
rohdoesnt really matter.. need to do qc and counting etc anyhow04:41
wolfspraulup to you, pick the easier one04:41
rohok. we'll see whats easier to package04:41
wolfspraulbut since I hope we do logos, do bulk packing then that's just one pack to go through to add the logo04:42
wolfspraulroh: have you seen an injection moulding process/machine? if you want to go in that direction, I suggest to visit an existing shop first.04:44
wolfspraulI did one or two visits, and I wouldn't know right away how to break this down for a smaller shop, it looks like a process that does need a rather high minimum capital investment, and minimum monthly production quota to make sense04:44
rohwolfspraul: havent seen one in person yet. but simply the point that somebody successfully operated vending machines in the us means it can't be too dangerous ;)04:45
wolfspraulthe problem is the size of the machines and complexity of the overall process04:45
rohtrue. but it seems to fit in a machine big as a vending machine, so i can put it in a hackerspace04:45
wolfspraulsteel tool design, steel tool making, steel tool storage, plastics storage, injection moulding machines04:46
rohi dont expect perfect results. and i got a mill to make my own molds. so the most expensive parts i can make myself04:46
wolfspraulok I went to a big factory, but like I said I have a hard time imagining how to scale it down04:46
wolfspraulthe machines are big, really big, and they need a lot of pressure04:47
wolfspraulwell, you can try. I don't want to stop you :-)04:47
rohit will get much more complicated for bigger parts for sure. i was thinking about starting with something about the size of a box of cigarettes similar to that of the vending thing04:47
wolfspraulin order to make plastic pieces for a few cents, you need a lot of capital invested, and you better have customers ordering millions of plastic pieces from you each month04:47
wolfspraulI think if you scale down, each piece will get more and more expensive04:48
wolfspraulto the point that lasered acrylic doesn't look that expensive anymore04:48
wolfspraulhow about milling or lasering aluminum?04:49
rohthe thing is: even former 'complex' or expensive stuff got really cheap for some weird product in the last 20 years. lots of machines built 30 years and more ago couldnt use the cool manufacturing processes we have now. i can go over the street and do custom lasercutting in a company for an investment of < 100E04:49
wolfspraulyes but that process is very well suited for customization04:49
wolfspraulso there are shops that support themselves with a large number of small orders04:50
rohmilling aluminium i can do. maximum piece our mill can do is ~24x12x20cm (xyz)04:50
wolfspraulthe injection moulding process I have seen is not well suited for customization04:50
wolfspraulit's all fine, but you need to make 50k or 100k pieces, and the factory you will make them at will be _big_, even physically.04:50
rohinjection moulding is only 'cheap' if you do lots of unchanged parts. sure. the 'setup' (milling the molds) is expensive. but atleast the 'using' the mold is much cheaper than what milling or cutting can do04:51
rohmy goal is to make something which can be built for <1000E investment and do parts with better resolution/quality of that a makerbot printer can do04:52
kristianpauli hope <500E ;-)04:53
rohkristianpaul: well.. you will need a mill to make the molds, thats not included04:53
kristianpaulof course04:53
kristianpaulwich material had you tought to inject in the mold?04:55
wolfspraulI think you will be in a long learning process there.04:56
wolfspraulI have only seen a little, but my understanding is the process needs huge pressure. You cannot make good parts with aluminum moulds at all. You need steel moulds, and even then you need to harden the most exposed parts of the surface.04:57
wolfspraulI suggest to go to an existing maker and have them show you the process and some pieces that were made in different ways, including aluminum moulds04:57
wolfspraulbtw, I think you can use silicone moulds as well, with a different plastics material. you just let the liquid run into the silicone mould, something like that.04:58
kristianpaulhum, aluminum moulds indeed sounds challenging..04:58
wolfspraulit's only good for 10-20 pieces though, maybe a bit cheaper than cnc if you need 2004:58
kristianpauli liked werner Counterweight mold process.. and wast so long.. but still limited for some kind of shapes and resolution.. and material04:59
rohkristianpaul: not sure what polymers are useful. will need to research that.05:00
rohhm. need to build another rig to glue the button caps. but i have an idea already05:02
rohand i think its easier for you if i do kits. else you need to pair the buttons to the right panels yourself05:02
kristianpaulmay be do mold in wood, and inject PLA using a gun or something..05:03
kristianpauldamn i dont have CNC machine at home.. and not that i will soon..05:03
rohstart a hackerspace and find other people who want similar toys to share05:03
rohor find a hackerspace05:03
kristianpaulwe have a hacker*lab05:05
kristianpaulbut not spave05:05
kristianpaulnot needed 05:06
wolfspraulroh: pair buttons to right panels?05:06
wolfspraularen't they all the same?05:06
rohwell... tolerances for acryllic are this:05:06
rohToleranz = ± (0,4 + 0,1 x Dicke) mm.05:06
rohmeans with 3mm it can be 2.3 to 3.7mm thick05:07
rohi already what one can do to compensate, but some of that means pairing fitting parts05:07
rohthe material i got this time was inbetween 2.7 and 3.6mm05:08
wolfspraulok but anybody can do the matching05:09
rohi measured every sheet on every corner, and noted the min and max on the sheets and sorted them, the thinner ones became sidepanels, the thicker ones top/bottom05:09
rohif you have somebody to do that puzzling afterwards i can also put them all into one big bag05:10
wolfspraulyour sorting system may be lost in translation anyway05:12
wolfsprauljust 'may'05:12
wolfspraulbtw, if it makes you feel better, matching/pairing of mechanical parts is regularly done at any factory I've seen05:13
wolfsprauleverything has tolerances, even part produced by the most fearful 3m high injection moulding monster with hardened steel moulds05:13
wolfspraulso since the injection moulding is normally done in a separate company/shop, matching/pairing is done as part of the iqc (incoming quality control) of the smt/assembly shop05:14
wolfspraulwhich is probably the place in the process that makes the most sense, because those girls can have piles of parts, and quickly find matching ones. they also decide which ones to return to the plastics shop, which can be as high as 10% regularly.05:15
wolfspraulbecause it's too hard to find a match/pair05:15
wolfspraulso you are right, if you want to do this matching, it's probably best to ship in kits05:16
wolfspraulthen adam doesn't have to figure the whole tolerances thing out05:17
rohi can add a A4 paper sheet ontop of each box 'danger! tolerances inside'05:18
rohbtw.. how big is your army to assemble that pile?05:19
wolfspraul1 man05:19
wolfspraulhe's doing SMT reworks, testing, assembly, packaging, shipping05:19
wolfspraulnice, eh? :-)05:19
kristianpaulon the go? or the idea is to have all the packages ready at once?05:21
roh .oO(this product may contain leftover crazyness of all involved parties)05:21
rohi have some pictures of the process.. will need to collect them all (different cameras) some time afterwards05:22
wolfspraulkristianpaul: it's overlapped, I think we do quite well so far. Adam is doing testing right now for example.05:22
kristianpaul(overlapped) true, i forgot that05:22
wolfspraulthere were some nasty discoveries testing the first few boards, let's see how that develops. But whatever we find, we can only move forward.05:23
kristianpaulyeah, i read backlock. hope that can be fixed..05:23
wolfspraulthe 'normal' OEM would throw a lot more people at this process, say 50, with suppliers even more, hundreds05:23
rohbut its good we moved. i wouldnt have been possible to do so much kits in one batch in the old location. not enough free spare tables one can occupy for days05:23
rohnow i can use the '/tmp/swap' labeld cabin with a big 'danger chemicals' on the door05:24
kristianpaulin rc2 adam was in the SMT personally, this time not needed? or my first asumption is not true?05:24
kristianpaulwell you, and adam i remenber..05:24
wolfspraulI would know how to manage this, I think :-) but I don't have the capital to attack the problem like this. And also we would then be under huge pressure to produce a lot, start the next version right away, and so on. Because all these people need to do something.05:24
wolfspraulwe don't want that either, it would be overwhelming and we would just produced bad quality products05:24
wolfspraulso we find a middle way through this05:25
wolfspraulif we can manage to build traction with users (=sales), we will ramp up the whole thing05:25
wolfspraulif not, at least we are not under a mountain of debt that forces us to do stupid things05:25
rohi am sure we will even find interresting results on the way doing stuff on a much tigher budget compensating in time and reproduceability05:26
rohfor me atleast thats a very important goal when searching for a solution how to do stuff... not 'how do i get it done now' but 'how can i make that possible 10000times the same way.05:27
wpwrakwolfspraul: (1 man factory) ACME = Adam Can Make Everything ;-)06:02
wpwrak(injection molding) from what i hear, aluminium is used quite regularly for low volumes (100s). it probably has some other restrictions (geometry, types of plastics, etc.)06:05
wpwrakthe industrial process probably also aims to have a high open-loop yield, which wouldn't be a big issue for a DIY process06:06
wpwrakalso, the industrial process aims to be as fast as possible. again, DIY cares much less about this. this means that you can try completely different heating approaches. e.g., i would heat the mold and go easy on the pressure.06:07
wpwrakthen, high pressure is not so hard to produce. there are manual bench presses that give you quite a lot of pressure. the challenge is then to combine all this with heating, and make sure things don't slip and splatter hot plastic all around you06:08
wpwrak(well, still better than molten lead. the gravity casting i used for the counterweight is pleasantly safe in that regard ;-)06:09
wpwraknext, if you just want to test drive your mold, you could use a plastic with very friendly characteristics (e.g., a lot meltung point), such as a mix of polyethylene and wax06:12
wpwrakbtw, i don't know if one could also make molds of brass or bronze. according to some sources, these materials may be even easier to mill than aluminium. plus, they have a higher melting point, which means that you could probably heat them on a gas stove06:15
wpwrakregarding molds, the industrial process may also show some co-evolution: since making a mold is expensive, you want to make the most of it, which justifies expensive machinery and complex processes. this in turn justifies expensive molds, and so on.06:21
wpwraknowadays, the machinery to make a low-performance mold can be had for a few kUSD. and maybe even 3D printing a mold would be possible now, which would need a larger initial investment but would simplify and accelerate the process even more.06:23
wpwraknow, if the cost of a mold gets low, you can consider making a larger number of molds, instead of the one super-expensive one of selected exotic steels06:24
wpwrakonce you have this, you can lengthen your pipeline and relax the thermal profile06:25
rohwpwrak: even if a mold is 'hundredts' of euros, its feasible if you want 3-4 digit parts06:28
rohbecause its still loads cheaper than printing or milling06:28
rohi was thinking of custom parts i couldnt yet find anywhere. like a diy case for handlebar mounting (for doing electronics and mount them to a bike)06:30
wolfspraulwpwrak: yes, I think for the 'typical' process nowadays it's multiple factors that drive prices up. but the goal is to bring the per-unit price down, right? so the fact that it evolved towards 50k USD molds and giant injection machines means that that is where the sweet spot is, _IF_ you make hundreds of thousands of parts.06:30
wpwrakyup. well, thanks to apple, there should now be "CNC farms" than do things like magnesium cases by the score in parallel :)06:30
wolfspraulmaybe you first define your economic end goals - how much do you want the case to cost?06:30
wolfspraulthat will narrow down the options a lot06:30
wolfsprauland the path becomes more clear too then06:31
rohtrue. i want machines where the sweetspot is low as 2-4 digits.06:31
wolfspraulcases go from 50 cents to 2000 USD :-)06:31
wolfspraullet's start somewhere. then you need to answer: where in that range do you want to be?06:32
wolfspraulif it must be < 2 USD, there are only a few options/paths06:32
rohof invest per product06:32
roheh sold product.06:32
wolfspraulok good. with that number in mind you can check which process works.06:33
wpwrakwolfspraul: at a few runs of 100k+, the machine cost becomes largely irrelevant. we can't compete with this directly because we have no means to overcome this sort of barrier.06:33
wolfspraulyes but we are discussing processes and materials06:33
wolfsprauland I proposed to first define a goal of the per-unit costs06:33
wpwrakwe're also discussing rapid and low-cost prototyping :)06:33
rohi dont fear some pressured air/valve/arduino foobar.06:33
wolfspraulbecause that will exclude some processes and materials, and as we go and learn about the various processes, we can quickly eliminate some and continue exploring others06:34
wolfspraulkeep the end per-unit cost in mind06:34
wpwrakif you take the DIY route, you need to improvise a low anyway. that's also an opportunity to swap parts of the process.06:34
rohthere is a full box of festo stuff around also. more difficult is the 'how to melt plastic safely' part i guess06:34
wolfspraulbecause that's what all of these processes and materials were designed for (most of them anyway)06:34
wolfspraulsomeone had a certain USD value he had to reach - then a particular process was created06:35
wpwraki wouldn't worry about the end unit cost for now. that's something you can only assess after optimization.06:35
wolfspraulI suggest the thought path the other way round.06:35
rohwell.. time to find out where to get good diy compressors and stuff.06:35
wolfspraulbecause we do not need to duplicate all industry.06:35
wolfspraulso if we define a per-unit cost we want to achieve, we can go through the processes and machines that others have developed already and see which ones fit our needs06:36
wpwrakroh: why not just use a manual press ?06:36
wolfspraulif you start from scratch, I think it will be a long path and in the end you have something that is not economically feasible06:36
rohwpwrak: i was thinking more of the one needed to move the liquid plastic than the one holding the form (that can be manual for now. or pneumatics)06:37
rohi have a compressor downstairs with 80litre tank and up to 8 bar.06:37
wpwrakwolfspraul: i see the current problem in the high up-front cost of injection molding. that's not only a problem for going to production (at which time, you'll already have to have acquired some financing), but it also affects the development06:37
rohwell.. first the acrylic madness06:37
wolfspraulwpwrak: I agree. fully agree. I'm just saying let's define the targeted per-unit cost first, and keep that number in mind throughout our work. that's all.06:38
wpwrakroh: (plastic) yes, there are some designs that have a heated cylinder with a plunger you push down06:38
wolfspraulit will guide us, and help eliminate options06:38
wolfsprauland roh did that, he said 10-20 EUR06:38
wolfspraulso that's good06:38
rohmaybe even a step between casting and molding is useable. use some more plastic and mill it nice afterwards. _much_ less time than using 'full blocks of plastic/metal' and less wear -> cheaper06:39
wolfsprauldon't underestimate how much others have optimized already, the entire industry for decades. We cannot quickly come in with all our genius and find the one optimization that leads to a high-quality low-volume and still cheap case. no way :-)06:39
wpwrakwolfspraul: (unit cost) assume an order size of ten. then we're talking :)06:39
wolfspraulno fix the per-unit cash value first06:40
wolfspraulthen you may have certain MOQ, depending on process06:40
wolfspraulbut I propose to think in that order06:40
wolfspraulnot process first06:40
wolfspraulbut anyway, roh seems to be clear about that (10-20EUR), so that's good!06:41
wpwrakyou can reverse it - take your quantity, then look at what the industrial process costs. that gives you an upper bound.06:41
wpwrakif you're cheaper, you win.06:41
wpwrakif you can make more and are still cheaper, you win too.06:41
wolfspraulwork from per-unit cost backwards, my 2c06:41
rohanyways.. its currently _not_ raining so i will use that opportunity to get home06:41
wpwrakwolfspraul: i think we're approaching this from different ends. i want to extend the range of things we can cover during development. you want to optimize the mass production process.06:42
wpwrakif the development process, it's a bonus if it can also cover part of the mass production domain. this is basically what you're doing with the acrylic cases now.06:43
wpwrakalthough you're probably not very cost-effective. may have been cheaper to put roh on a plane to buga and let him visit that super-cheap laser cutting shop, and let them make the cases ;-)06:44
wolfspraulI think a lot of the machines and processes you find today were created because someone targeted a lower per-unit cost for himself. (including whatever order quantity he had at the time)06:44
wolfspraulah no06:44
wolfspraulI am very clear about material costs06:44
rohon mechanics material isnt the expensive part. its machine cost and runtime (atleast in all my cases)06:45
wolfspraulI work with roh because I also (try to) subsidize his design and openess work06:45
rohthe acryllic doesnt even cost half of what lasering it does06:46
wpwrak(come in with our genius) it's not only that. it's also new tools. for example, low-cost CNC is something relatively new.06:46
rohsame goes for milling06:46
roheven more dramatic when using wood (material worth cents)06:46
wpwrakwolfspraul: (subsidize roh) that's good. but i wonder how much of this really stays in his pockets. seems that most of it goes to the laser shop now anyway06:47
wolfspraulit goes to many places06:47
wolfspraulhis work is also very labor intensive06:47
wpwrakwolfspraul: so perhaps it would have been easier to just take the work to buga and simply give roh some of savings as a token of gratitude :)06:48
wpwraks/some of/some of the/06:48
wolfspraulmaybe in the future we can do that yes, like a design royalty, and all manufacturing happening in China06:48
wolfspraulwe see06:48
wolfspraulI like to work with raumfahrtagentur and roh :-) with all ups and downs :-)06:48
rohyes.. we will hopefully figure something out ;)06:48
wolfspraulgood mechanical work is very important for me in the future, for copyleft hardware06:49
wolfspraulwpwrak: not buga, China (sorry Buga :-))06:49
rohwhereever buga is?06:49
wolfspraulit's a nice town in Colombia06:49
wolfsprauland the laser shop was even in Bogota actually06:49
wolfspraulnothing can beat China on any of those things, if you just want to do it there06:50
wolfspraulmany machines are coming from China now06:50
wolfspraulelectricity is dirt cheap06:50
wolfspraulno environment standards whatsoever, dump anything you have in the toilet06:50
wolfspraulno worker rights whatsoever06:50
wolfspraulgiant domestic market with nearly infinite amount of service providers and shops and companies06:50
wolfsprauland so on06:50
rohgiven the amount of tricky details i encounterd.. i can basically limit the reproduceablity of the product in quality be the grade of detail in the documentation06:50
wolfspraulbut I still need raumfahrtagentur :-)06:50
wpwrak(bogota vs. buga) ah, make more sense. i was wondering how on earth kristianpaul could have such a shop basically right at his doorstep :)06:51
rohwell.. bbl. driving home now.06:51
wolfspraulI work with raumfartagentur _because_ they are documenting well06:51
wolfspraulif that goes away there is no difference to China :-)06:52
wpwrak(china) yeah. depends on the cost of finding a place that'll do it vs. the savings compared to kristianpaul's lucky find06:52
wolfsprauloh, no problem06:52
wolfspraulcome to Shanghai one day, or Shenzhen, or dozens of other multi-million people cities...06:53
wolfspraulbut my goals are different, I want deep documentation, and I want to work with places that can do innovative things, experiments06:53
wpwraksure. just seems to be a bit inefficient to tie up people like roh with chinese manfacturing06:54
wolfspraulyou mean the run of 80?06:55
wolfspraultoo much manual work for roh?06:55
wolfspraulhe needs to speak up himself, that's a problem of optimization that you can only optimize well if you have the problem right in front of you06:55
wpwrakhe seems to grumble a lot. but maybe that's just the noises he makes when he's happy :)06:55
larsci wouldn't count on that06:56
wolfspraulit's also why I say we should not under-estimate the optimizations others have done already06:56
wolfspraullet's get this run of 80 done now, then we try to sell them, then we think about the next steps06:56
wolfspraulI am not particularly proud of a process that involves workers without rights either. But I need to be smart in which problem I attack in which order.06:57
wolfspraulfor example I could spend some money to get rid of that metal sheet06:57
wolfspraulthe metal sheet causes a lot of work06:58
wolfspraulit's something we could get rid of if we would focus on it and spend some serious engineering time06:58
wolfspraultoo many problems all over...06:58
wolfspraulalso right now it seems the rc3 run will give us quite some headache on the electrical side, at least that's what I see in the early test results06:59
wolfsprauljust when I thought the weekend starts! :-)06:59
wolfspraulI don't want copyleft hardware to depend on a process that ruins the environment or exploits workers that have no rights.07:00
wpwrakmurphy never rests :)07:00
wpwrakwell, you always have to choose between mainstream processes and alternative/customized processes. like with injection molding :)07:01
wolfspraulI think we have a chance with that because our stability and value comes from continuity in free software, not in how well we exploit nature or people.07:01
wolfspraulI think you see this too black & white.07:01
wpwrakno, it's all shades of grey07:01
wolfspraulthe 'mainstream' people are very aware of 'alternative' processes07:01
wolfspraulat least the ones I've met07:02
wolfspraulthere is no mainstream or alternative, it's just a lot of processes to pick from, and it depends on the customer which one is best07:02
wolfspraulwpwrak: have you ever met Nico in Taipei?07:03
wolfspraulif we ever have money again, we'll meet up with him and he can take us through a tour of his mechanical supply chain. Very cool!07:03
wpwraki think so, yes. but only briefly and i think it was around the time openmoko went down.07:03
wolfspraulhe loves to take people to 'his' shops that supply him with stuff07:04
wpwrakhehe :)07:04
wolfspraulhe constantly works with 20-30 suppliers, all sorts of different materials and processes, and optimizations07:04
wolfspraulthere's no 'mainstream' or 'alternative' there07:04
wolfspraulit all depends on customer, requirements/priorities, target unit cost, materials, etc.07:05
wolfspraulmechanical supply chain is a very diverse set of companies07:05
wpwrakmaybe. but i'd still expect them to cover only a small subset of all feasible choices. and some of the uncovered ones may be actually work but have no nice migration path from the ones they cover.07:06
wolfspraulbtw, one math works like this: 1/3rd electricty, 1/3rd material, 1/3rd people07:06
wpwrakthat much on electricity ?!?07:06
wolfspraulso when comparing Taiwan and China, for example, material is the same, people cheaper in CHina, electricity cheaper in Taiwan07:06
wpwrakyou have E = m. what happened to E = mc^2 ?07:07
wolfspraulyes, 'one math', probably depending on process (forgot which one exactly this was)07:07
wpwrakmaybe aluminium production :)07:07
wolfspraulhis supply chain of course also extends into China, it's all case by case where the companies are, who is better suited or optimized (invested) for a particular process07:07
wolfspraulno no07:07
wolfspraulask roh about how much electricity costs for the lasering of just those 80 units07:08
wolfspraulkeep in mind: others optimize too07:08
wpwrak(electricity cheap in taiwan) interesting. having seen suzhu at night, i would have thought it had negative cost over there :)07:08
wolfspraulTaiwan gets most electricity from nuclear power plants07:09
wolfspraulChina is very different, electricity prices in CHina are political (different companies paying totally different amounts)07:09
wpwrakyeah, from what else :)07:09
wpwrak(china) nice way of taxation :)07:10
wolfspraulbut from Nico's numbers, he said the companies he looks at pay more in China than in Taiwan. Which sounds plausible to me.07:10
wolfspraulthose are mostly small private companies, they will pay a higher rate to subsidize inefficient state companies07:10
wolfspraullet's ask roh later how many kWh he thinks are needed for the lasering of the 80 m1 cases. I'm curious :-)07:13
rohkWh? no clue07:20
rohbut in money i would say neglible compared to material and manufacturing cost (machine runtime)07:24
rohas soon as we do 80 cases/month i should buy my own bigger lasercutter07:24
rohmaybe we can also retrofit ours with better lenses (seems to be the main issue)07:25
rohour lasercutter needs something like 100-230W of input power07:28
rohi think the big one and cooling is also <500W.07:28
rohif its running for 14hours it will eat 7kwh. but if you take the controlling pc into account and its monitor...07:28
wolfspraulnah, it's all small07:28
wolfspraulno 1/3rd here :-)07:29
rohyou can save more electricity by using a notebook or disabling the monitor than anything else ;)07:29
wolfspraulI would be seriously interested in design improvements that remove the need of the metal sheet, but workload is overwhelming for me right now, so I cannot do it.07:31
roheven if it would be 1kW instead of 1/2 ... i mean.. one kWh is <23cent for '100% regenerative energy (green electricity).. means 2E ;)07:31
rohwolfspraul: lets fix the latest reworks into a rc4 successfully first  ;)07:31
wolfspraulyes understood, insignificant part of the cost in our run07:31
rohthe metal sheet is less headache than the buttons for sure.07:32
wolfspraulbuttons are worst, then metal sheet07:32
rohthird is tolerances on acryllic.07:33
wolfspraulbut getting rid of the metal sheet may be hard work, if it's done inside the pcb, i.e. without insulation patches on the pcb07:33
wolfspraulon the tolerances I think one thing you don't really factor in right now is that in a larger run, you don't need to design in such a way that any set of parts fits07:33
wolfspraulin other words, the tolerances may even themselves out07:33
wolfspraulyou just produce say 100, and then sit down trying to find set that fit, and you accept that you will have a small percentage of leftover parts at the end07:34
wolfspraulthat's a different way to deal with the tolerance problem07:34
wolfspraulin a larger run, the stuff I've seen, you don't produce every part to be within perfect tolerances07:35
wolfspraulthat's just not economical, because even a hardened steel tool will wear out07:35
wpwrakroh: (acrylic tolerances) are they in the raw material or the laser cutting ?07:35
wolfspraulso you just produce, and then sit down to find matching parts :-)07:36
wolfspraulof course that assumes a certain labor cost, often around 1 EUR / hr07:36
wpwraksmall detail ;-)07:36
wolfspraulyes sure, all the optimizations connect to those things though, it's all connected07:38
wolfspraulif you change some costs, suddenly the 'best' process may change too07:38
wpwrakthat much about not wanting a process that relies on exploitation :)07:38
wolfspraulI've just learnt: electricity costs in acrylic lasering are insignificant. need to keep that in mind...07:38
rohwpwrak: raw material07:39
rohwpwrak: http://www.plexiglas.de/product/plexiglas/de/ueber/faq/Pages/toleranzen.aspx07:39
wolfspraulcorrect, that adds another requirement into the process07:39
wolfspraulI'm not saying 1 EUR / hr is exploitation btw, I said what bugs me is lack of worker rights.07:39
wpwrakof course, the alternative is automation, which simply eliminates all those exploiting jobs :)07:39
wolfspraulno, wrong. please think about the tolerance problem more.07:40
wolfspraulyou cannot have it all ways07:40
wolfspraul'optimizations' means to optimize along certain angles. you always 'exploit' something as part of the optimization.07:40
wolfspraulthat's just unrealistic. I think we really want to produce something in our lifetime...07:40
wpwrakroh: surprisingly large tolerances07:40
rohits not only something to counter with time. tolerances produce nonuseable parts if too big07:40
rohespecially as long as we use slots anywhere07:41
rohwpwrak: yeah07:41
wolfspraulif you buy material with 'less' tolerances, you just push the problem out of your sight potentially07:41
rohtoo bad i havent gotten plexiglas XT in colored as default yet. will see if i can get it next time07:41
wolfspraulthey throw away for you, and you pay for it :-)07:41
wpwrakwell, you could buy sheets that are thicker than you need and machine them to narrow tolerances ;-)07:42
wolfspraulassuming your machine can do that, yet07:42
rohfor lasering XT should behave nearly the same. only milling isnt as great07:42
wolfspraulbut if that would be easy, someone else would have done it already07:42
rohwe'll see.07:42
wpwrakwolfspraul: (machining) i think you'd cloud the surface07:43
larscwpwrak: that is an interesting though. the alternative to exploitation would be no jobs at all.07:55
wolfsprauloptimization = exploitation07:57
wolfsprauljust what :-)07:57
wolfspraulI have no problem with low hourly wages, but I want the people we work with to have the right to say 'no', and to have realistic alternative paths for themselves. If that's the case I happily work with someone making 1 EUR / hr, even if that's an exploitation from the perspective of someone making 20 EUR / hr.07:58
wolfspraulthe best is to exploit software of course07:58
wolfspraulwhich is what we mostly try, I think, and why we are optimistic07:59
wolfspraulwho sheds a tear for letting a CPU iterate over a stupid process indefinitely07:59
wolfspraulif I see how Werner exploits his mill, oh my... :-)07:59
larscwell, there is a difference between people and tools08:04
wolfspraulyes, so when we optimize processes, we shouldn't exploit our environment (dump waste), or rights of people. those 2 are off limits, everything else can be optimized, I think.08:05
wolfspraulwe can add animal rights, but I don't think we are planning to use trained monkeys in the mechanical supply chain :-)08:07
qi-bot[commit] Werner Almesberger: cad/test1/Makefile (cadmium.stl): corrected typo in file name (master) http://qi-hw.com/p/wernermisc/7a39c5e08:08
qi-bot[commit] Werner Almesberger: mm1rc3rst/fix2.sch: changed diode symbol to from regular to Schottky (master) http://qi-hw.com/p/wernermisc/fe20d7408:08
larscmaybe we could use them to write code08:08
wpwraklarsc: or solve captchas ;-)08:09
larsci fear that wouldn't get us anywhere08:14
tuxbrain_awaywpwrak: good job with the CAD comparative dude :) good material to make an interesting post :)08:14
wpwraktuxbrain_away: thanks ! :) not quite done yet, though. i found some more interesting details. will update soon.08:15
tuxbrain_awaybetter than update make a second chapter :)08:15
wpwrak(chapter) yeah, and maybe convert to LaTeX ;-)08:16
tuxbrain_away(I hate you)08:20
tuxbrain_awayI will try LyX then , as LaTeX IDE :)08:23
wpwraki have a little perl script that translates TeX (don't remember if it's plain TeX or LaTeX) into ascii ... :)08:25
wpwrakit's LaTeX. quite advanced for its time. i wrote that beastie in 1994 :)08:27
tuxbrain_awayman, there is  anything that you haven't done? How many people compose Werner? I don't believe you are only one person08:29
qi-bot[commit] Werner Almesberger: cad/test2/button.py: Cadmium version without artifact prevention (master) http://qi-hw.com/p/wernermisc/83ddf1108:34
qi-bot[commit] Werner Almesberger: cad/: added description to 2nd series of tests and a pointer from the 1st series (master) http://qi-hw.com/p/wernermisc/8ad6df610:39
kristianpaulwpwrak: bogota, right, but afortunatly mail delivery is very fast :)12:38
wpwrakgrmbl. meshlab crashed by pc :-(14:36
qi-bot[commit] Werner Almesberger: cad/test2/: copied original version of sources from cad/test1/ (master) http://qi-hw.com/p/wernermisc/70eef5614:47
qi-bot[commit] Werner Almesberger: cad/test2/button.py: removed "epsilon" and "noise" (master) http://qi-hw.com/p/wernermisc/429608014:47
qi-bot[commit] Werner Almesberger: cad/test*/button.scad: remove definitions related to the pusher, not used here (master) http://qi-hw.com/p/wernermisc/8ca4b0114:47
qi-bot[commit] Werner Almesberger: qpkg/: committed some old changes (non-functional) (master) http://qi-hw.com/p/wernermisc/326cd3314:47
kristianpaulfrom #homecmos http://colossus.cs.rpi.edu/~azonenberg/images/homecmos/2011-07-22/S7301551.JPG16:39
kristianpaul11:31 < azonenberg> http://colossus.cs.rpi.edu/~azonenberg/images/homecmos/2011-07-22/S7301551.JPG is a 40x die overview after the final KOH wet etch and mask strip (so what  you're seeing is all Si except for a bit of some brownish precipitate in the lower parts of the etched areas)16:39
solrizeyow, shipping 1 nanonote to the US costs almost $40 and thats for the slow option :(17:59
wpwraksolrize: get more than one, then the per unit shipping cost drops ;-)18:02
kristianpaulsolrize: to compesate you i think is 60usd to my country..18:02
kristianpaulso 40 is not that bad..18:02
solrizewpwrak, yeah i asked around a little if anyone else near me wanted one18:03
solrizei suggested them to sparkfun.com18:03
solrizeunfortunately there's no apparent quantity discount18:04
wpwraka distributor in the US would certainly be useful18:04
wpwrakyou'd have to ask wolfgang at what quantities he would consider a discount18:05
kristianpaulask Yi18:05
wpwraksolrize: do you have taxes / customs feeds that get slapped on this kind of imports ?18:05
wpwrake.g., in argentina, taxes/fees usually amount to about 40-50%18:06
solrizewpwrak, hmm, there is theoretically a duty charge of a few dollars but u.s. customs rarely bothers stopping packages18:07
wpwraksome of them are based on the cost of the goods plus the shipping cost. so the real cost of the shipping is actually a bit higher than what gets paid to fedex or such18:07
solrizei buy stuff from dealextreme.com all the time, they just send it in the normal post18:07
wpwraklucky you :)18:07
solrizefedex may be collecting duty charges18:07
kristianpaulsame here (DX)18:07
solrizemight be 10% or so18:07
kristianpauland no TAX?18:08
wpwraksigh. the easy life :)18:08
wpwrakkristianpaul: here in south america, we're wealthy, aren't we ? we can afford high taxes ;-)18:08
kristianpauleven 10% is lower than fedex handling/tax/ here.. wich is kinda 40%18:08
kristianpaulwpwrak: we're rich countries :_)18:08
solrizecustoms might stop a DX package (you get a notice saying you must pay $N duty to get the package) but i've never had it happen18:09
kristianpaulvery rare indeed18:09
kristianpaulDX is develiver by USPS isnt?18:09
solrizekristianpaul, yeah, international post18:10
kristianpaulbtw a bit OT bit how long take to deliver to you?18:10
solrizeDX?  about a week18:10
solrizehmm maybe thats optimistic18:10
solrize1-2 weeks say18:11
solrizeit varies18:11
kristianpaulhere is a month or so,18:11
wpwrakfees/taxes collected by customs are usually collected by the deliverer, be it the fedex/dhl/ups/etc. guy or the postman18:11
kristianpaulsolrize: so you can be a distributor in the US18:11
wpwrakyou'd only have to go to customs if they find some issues or if the carrier doesn't want to prepay the charges18:11
solrizekristianpaul, i'm not set up for that, it's better for a retail operation like sparkfun to do it18:12
kristianpaulsolrize: just sent a couple of small boxes by ground mail from time to time, dint hurt18:12
kristianpaulhad you seen how sparkfun started? :-)18:12
kristianpaulso you asked sparkfun to sell nanonotes? :-D18:13
solrizei dunno but they started by wanting to be retailers, i bet ;)18:13
solrizeyeah i suggested that to them18:13
kristianpauland replied?18:13
solrizeno reply yet, i suggested last night and its weekend here now18:14
kristianpaulsure sure18:14
wpwrakregarding discounts, some margin would already come from avoiding shipping individual units. e.g., if you order 20, you'd probably still pay about the same amount. but you'd then have much cheaper and faster domestic shipping. that alone may yield a margin of 20-30%.18:15
wpwrak(weekend) that's when people work more efficiently, right ? ;-)18:15
solrizeweb site doesn't allow ordering more than 9 ;)18:15
kristianpaulsolrize: ask yi for 10 and beyond :-)18:15
solrizeyeah i saw that.  i could imagine buying 2 but thats about it ;)18:16
wpwraksolrize: they had a 10 pack. maybe that's where the limit comes from. but i'm sure they'll be more than happy to send you up to a few hundreds ;-)18:16
wpwrakthe limit may be around 600, which is when more units would have to be produced :)18:17
solrizethey don't sell replacement batteries either18:17
kristianpaulno need18:17
kristianpaulis generic18:17
solrizethere is a chart of almost-compatible batteries18:17
wpwrakthere's a list of commonly available batteries you can use instead. much more efficient than ordering from abroad.18:18
solrizebut i wish they'd tweak the plastics so a standard phone battery would fit properly18:18
solrizeyeah i saw18:18
kristianpauli buyed a Nokia BL-4C  clone and worked fine18:18
wpwrakyes, the case isn't exactly right for them18:19
solrize"The battery is just a little too short. A small object like a piece of paper is needed to make sure it doesn't disconnect. "18:19
solrizethey should use a bigger battery if there's space in the case for that counterweight thing18:19
wpwrakthat's not easy to solve, though. the case design isn't owned by sharism and in any case, changing the case shape would require a new mold, which is pricy18:19
wpwrak(counterweight) ;-))) that one's a little tricky, though18:20
solrizewe have a molded thing where i work, the mold costs around 7500 usd 18:21
solrizewhich i guess is a lot for a low-qty thing like that18:21
wpwrakyeah. would be difficult to recover such a price18:22
wpwrakhow complex is your mold ? and what is it made of ? steel ?18:22
qi-bot[commit] Werner Almesberger: cad/test2/button.scad: added overlaps to eliminate artefacts at coincident faces (master) http://qi-hw.com/p/wernermisc/521f03118:24
qi-bot[commit] Werner Almesberger: cad/test2/README: added more results and cleaned up the text (master) http://qi-hw.com/p/wernermisc/3f820dc18:24
Action: kristianpaul one day lost because and input that should be output..18:25
solrizepretty simple mold i guess.  it's probably steel, i don't know18:25
solrizecnc machine18:26
solrizewe have a tabletop cnc mill at my hackerspace18:26
solrizei wonder how hard it is to mill injection molds18:26
kristianpaulwich its name is? :-)18:26
kristianpauloh, https://www.noisebridge.net/wiki/User:Maltman23 18:28
solrizeyou know him?18:28
kristianpaulnope :p18:29
kristianpaulbtw you work with FPGA's as well, already aware of Milkymist project i guess18:29
solrizei've been interested in fpga's but no one at noisebridge is using them afaik18:30
solrizelots of AVR stuff though18:30
solrizeyeah i've heard of milkymist18:30
solrizebut it uses a proprietary xilinx chip doing who knows what18:30
solrizei'd like to make a completely free microprocessor18:31
kristianpaulASIC? 18:31
solrizehttp://cmp.imag.fr/  yeah less expensive than i'd have expected18:31
kristianpaulah, checkout #homecmos then :-)18:31
solrizeoh neat18:32
kristianpaulwell, you can implement milkymsit soc, in those ASICs, or atleast the lm32 core.. 18:33
solrizei don't really have any interest in video devices so i didn't look at milkymist very much... i wish qi would make an audio player/recorder18:34
solrizei want to run rockbox on the nano so i can use it as an audio player18:34
solrizeso i can listen to music while i code18:35
kristianpaulneither me, but is a nice project and the SoC works 18:35
kristianpauli think you can run gmu in a separate tty18:35
solrizewhats gmu?18:35
solrizewell anyway there's some music apps18:36
wejpyes, you can, you can start gmu and then switch to another terminal to do other stuff18:36
solrizerockbox is really nice and runs on android and maemo now18:37
solrizeporting shouldn't be too hard18:37
wejpor you could simply use gmu18:38
solrizeyeah i guess so18:38
solrizeis there ever gonna be a YA nanonote?18:39
wejpnoone really knows ;)18:39
wpwrakwe hope but we don't know18:40
solrizeany interest in making an audio player?  all existing ones suck ;)18:40
wpwrakone problem with audio is that you can't support the more popular formats without getting into patent issues18:42
wpwrakthat is, unless you wait a bit longer, until the patents expire18:42
solrizeogg is fine18:42
solrizeit can run rockbox or gmu or whatever18:43
solrizeso people can compile mp3 into it if their jurisdiction allows it ;)18:43
wpwrakpeople compiling mp3 into it is tricky. if you as maker/distributor/etc. much as hint at the possibility, you can still get in trouble18:47
wpwraks/much/as much/18:47
solrizewell rockbox has mp3 and no one has had problems afaik18:47
rohsolrize: nobody makes money with it.18:47
solrizei'm fine with a vorbis-only player18:48
solrizevorbis and flac that is18:48
solrizeactually flac-only is fine18:48
solrizeany mp3 builds would be up to the user groups ;)18:50
wpwrakthe players for which rockbox is written probably already support mp3. so a license is already paid. presumably not one that allows that sort of firmware change, but it may be considered of insufficient interest to argue the fine points in court18:53
solrizetrue... there is a guy making a custom hw rockbox player (that sucks for audio) but he's small enough that no one has hassled him18:54
wpwrakthat's a bit the problem. as long as you're too small and only lose money, nobody will trouble you. but as soon as success seems within reach, the parasites will come ...19:00
solrizehow hard/expensive is it likely to be to make a player? main cost would be molds19:02
solrizethe guy making the custom player uses 3d printing 19:03
wpwrakyeha, i don't know yet how costly that part would be19:03
wpwrakmechanical engineering is something we're still just beginning to touch in qi-hardware19:04
wpwrakso far, the greatest achievements are my wooden molds for casting the counterweight and the laser-cut acrylic case for M1. the latter is actually already almost at a professional level. but it's also a solution with somewhat limited applicability.19:05
solrizea simple box with some connector holes and a battery door would be fine... maybe there's one in a catalog19:06
solrizereprap is ok for cheesy prototypes19:06
solrizehackerdojo.com (another hackerspacei visit) has a professional level 3d printer19:06
solrizebut i've never used it19:06
solrizehttp://www.sparkfun.com/products/10088  hmm19:08
wpwrakyes, if you can find an already existing case, that would help19:09
solrizealtoids ;)O19:09
wpwrakyeah :) also excellent shielding :)19:11
solrizei guess it doesn't need any UI (no screen or buttons). just an on/off switch and bluetooth to control it from a phone ;)19:49
solrizewpwrak, were you involved with openmoko?19:53
wpwrakyup, i was. a bit of hacking and occasional forays into hardware20:08
kristianpaulthere is way to hold the samplin in a scope, after triger the signal right... well may be not in this one..20:13
wpwrakkristianpaul: you mean setting the trigger mode to "single" (as opposed to "auto" or "normal") ?20:20
kristianpaulyes, i'm using single right now20:22
wpwrakso is that what you were looking for ?20:22
kristianpaulyes, ;)20:22
wpwrakproblem solved. that was easy ;-)20:23
kristianpaulnot yet, well the main problem not, here we go again 15m of sinthesis..20:24
viricwpwrak: have you seen fabfi?20:24
wpwrakno, haven't seen that one yet ...20:25
viricopen source wireless with components made from trash20:25
wpwraknice distances - 6 km :)20:26
wpwrakseems a bit bulky, though20:26
kristianpaul6 km, well i can stand on the bulk :-)20:26
viricthey pick access points from the trash though20:27
viricthese afgans are getting too clever. US is not spending enough effort20:28
kristianpaulha, where you can find access points in the trash?..20:28
kristianpauli mean.. not bad for then20:32
kristianpaulone of the missing things in mm soc, for sure a smarter wishbone bus.. ar least one that not freeze the cpu because a delayed or ausent ack signal after non proper termination of a bus cycle..20:38
kristianpauli hope i'm wrong but i'll see20:38
wpwraksolrize: btw, when you make CAD models for printing/fabbing/casting, what programs do you use ?21:09
kristianpaulhell 32° C22:21
kristianpaulwhere is the wind..22:21
wpwrakkristianpaul: can you please send 5 C down to argentina ? :)22:27
kristianpaulhum scope ch2 got biased from a pulse signal in the ch1...22:50
--- Sun Jul 24 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!