#qi-hardware IRC log for Wednesday, 2011-06-08

qi-bot[commit] Xiangfu Liu: gtkguitune, swap left and right interface patch http://qi-hw.com/p/openwrt-packages/221303d00:08
qi-bot[commit] Xiangfu Liu: nanonote-files: reflash_ben.sh using -O in wget http://qi-hw.com/p/openwrt-packages/3ecb2ca00:08
kristianpauloh boy, now are going for nvidia..02:19
xiangfukristianpaul: hi02:26
xiangfukristianpaul: one questioan on verlog , do we really needs this rx1? https://github.com/milkymist/milkymist/blob/master/cores/rc5/rtl/rc5.v#L5902:27
kristianpaulxiangfu: hey 02:40
kristianpauloh sure02:40
xiangfukristianpaul: ok. then I have my first patch on Verlog code, remove rx1 :D02:40
kristianpaulthats used fot clock domain sync lines 59 - 6502:40
kristianpaulno no02:40
xiangfuoh. 02:41
kristianpaulsory i missundertood you02:41
xiangfuI understand now. used for clock domain sync02:41
kristianpauloh sure, we need it ;)02:41
kristianpaulremner verilog is a hdl, try to think cocurrently when checking the code02:42
kristianpauli still did mistakes for let go that..02:42
xiangfukristianpaul: then check the L101 and L106,02:43
xiangfuwe only saved 13bit, not full 14bit? ( the rc5 is 14bits)02:43
kristianpaulrx_reg <= {rx_reg[11:0], rx2}; 02:46
kristianpauli see02:46
kristianpaulwell you know more than me about rc5 :-)02:46
kristianpaulbut rx_reg was originally declared as 13 bits register02:47
xiangfuyes. is that wrong. we should declared it as 14 bits ?02:48
kristianpaulhow that thing will change?02:48
xiangfukristianpaul: hmm.. that make me guess the start bit maybe also wrong?02:51
kristianpaulyou had confirm when you read that from bios you are missing that bit?02:51
kristianpaulmr d that from bios you are missing that bit? [21:52] [kristianpaul(+Zi)] [7:Freenode/#qi-hardw02:52
kristianpaulmr 0xe000e000 02:52
xiangfukristianpaul: (if the start bit is wrong the data will random),02:52
xiangfukristianpaul: let me explain a little.02:53
kristianpaulplease :-)02:53
xiangfu(find some URL first :)02:54
kristianpaulchecking wikipedia article 02:55
xiangfuAdam and me have same IR remote controller,(for test m1 IR receiver)02:55
xiangfuAdam use oscillograph measure the IR data before it goes into FPGA. it's like http://downloads.qi-hardware.com/people/adam/m1/pic/button1_0x0c_wide.JPG02:56
xiangfu3, I also press the button '1', use the mr 0xe000e000 4 in BIOS  read the data. 02:58
kristianpaulhum, but you should loop that read i think02:59
xiangfu4. it give totally random value (in 3~4 values),02:59
kristianpaulthe register seems to change as soon receive data i think. may be wrong02:59
kristianpaulso ramdom data from ir dint sound that bad ;)03:00
xiangfukristianpaul: I run that command many times. not only once03:00
kristianpaulbut how many times, you mean you self?03:00
kristianpaulnot a program?03:00
xiangfumanually. not a program03:00
xiangfukristianpaul: I also modify the test image, there is a while(1) in read the 0xe000e000,03:01
xiangfukristianpaul: it is same as BIOS.03:01
xiangfukristianpaul: so I guess there are some bug in fpga code? that is why I look into rc5.v :)03:02
kristianpaulfrom the scope pic i can count 13 bits03:06
kristianpaulwhere you got the 14 bits?03:06
xiangfukristianpaul: there are 13.5 bits.03:09
xiangfukristianpaul: form right to left, I count 13.5, the last is '0'03:10
xiangfukristianpaul: http://users.telenet.be/davshomepage/rc5.htm03:11
xiangfufrom the document it's 14bits03:11
kristianpaullet me check03:11
xiangfustart[2], toggle[1], address[5], command[6]03:11
kristianpaulah, i see now03:12
kristianpaulgood link03:12
kristianpaulso yes, something is missing03:12
kristianpaulyup, why not rx_reg can be 14 bits, is a shift reg after all 03:15
kristianpauland also rx_data03:16
kristianpaulyou theory looks good for me xiangfu !03:16
xiangfukristianpaul: http://dpaste.com/551777/ :)03:17
kristianpaulalready tried?03:18
xiangfuyesterday I use the 'load_bitstream.sh' under milkymist.git, now I try to flash the modify image to my m103:18
xiangfu(build again...)03:19
xiangfukristianpaul: I don't understand how it find the start bit?03:20
xiangfu(build again, it needs 14 mins in my laptop)03:20
kristianpaulme either.. let me chek03:20
kristianpaulwhat??!! 14 minutes??03:20
wolfspraulkristianpaul: Sebastien said "timing margins may be too tight"03:21
wolfsprauldo you have any ideas how to look into that?03:21
kristianpaulwell, there is a counter wich do the sampling, that need be confirmed as well03:29
kristianpaulrx_count16 <= rx_count16 + 4'd1; // i think this found the T03:35
kristianpaulindeed too tight :-)03:35
xiangfukristianpaul: how to modify it? I can test :)03:45
xiangfukristianpaul: how long you compile bitstream.(soc.fpg)?03:46
kristianpaulxiangfu: from what i understand start bit is found when some timing/period is met in a bit 03:49
kristianpaulmet timing in fpga's is tricky, well.. you need to do some math for rc5.v counters :-)03:50
kristianpaulxiangfu: soc.fpg was 35 minutes with full mm1 soc,03:50
kristianpaulthen i disabled cores i was not using to reduce sinthesis to 10 minutes.. or my patience will blow out..03:51
xiangfukristianpaul: maybe you try 'build_bitstream.sh' next time.03:58
kristianpaulwhat is special on it?03:59
xiangfukristianpaul: just build one file. 03:59
kristianpaulsure me too03:59
kristianpauli hacked makefile  ;)03:59
xiangfukristianpaul: I remember it's 14mins, not test recently. 03:59
kristianpaulno rescue or standby04:00
kristianpauljust soc.fpg04:00
xiangfukristianpaul:  (timing, counters) sounds hard to me. please give me more info. thanks04:01
xiangfukristianpaul: now it's flashing. let's see what happen we change it to 14bits04:02
kristianpaulfast indeed04:02
xiangfuchange to 14bits. same as before. :(04:03
kristianpaulso no need for all bits04:03
xiangfuin fact we only needs the last 11 bits.04:04
kristianpaulsorry i'm off now04:16
xiangfugood night04:23
xiangfunew IR test code04:34
xiangfuhandle repeat values when press button.04:35
wpwrakwolfspraul: hmm, there seems to be some misconception on the list about the ya nanonote schedule ... like "when will it be done ?" where the question should be "will it start ?" :)06:57
wolfspraulI didn't see that. But sure, I am not actively working on Ya NanoNote now.06:59
wolfspraultotally no need, I can see the Ben improving a lot (in software), then ben-wpan, etc.06:59
wpwrakthe "CPU loyalty thread" and now also the USB thread.06:59
wolfspraulI answered the USB, no time for CPU want to focus on m1 rc3 :-)06:59
wpwraki'd quite strongly disagree on the "no need" ...06:59
wpwraki see the "no money, no people" ,though07:00
wolfspraulok, it's a business. the question is would it sell?07:00
wpwrakbetter than the ben :)07:00
wolfspraulI have no data that suggests if a Ya would come out with bigger specs now that it would sell better/a lot.07:00
wolfspraulI think the way this can work is that I show the Ben to customers, and they need to see the need for something bigger and better.07:02
wolfspraulso for the time being, I can only improve the ben, for example in software or ben-wpan07:02
wolfspraulthen use the ben to go for greener pastures07:02
wpwrakwell, the ben has a lot of small issues, too. bad keyboard layout. poor screen size / available area ratio. imbalance.07:03
wpwrakthat's just outward appearance that makes it look bad07:03
wpwrakplus, it's weak on the freedom side. one implication of this is that you can't fix those small problems.07:04
wpwrakthe better specs would be simply benefits you reap as you go. of course you wouldn't design a new product for an obsolete chip, old slow and now expensive memory, etc.07:05
wolfspraulI'm not disagreeing with any of this07:06
wpwraki think the bottom line is lack of money. you can try to argue that there's no need, but it doesn't sound very convincing :)07:07
wpwrakkinda like saying to your friends, after the pretty girl rejected you, that you didn't find her that hot anyway ;-)07:08
wolfspraula business makes money, doesn't need money07:08
wpwrakin hardware, you need money to make more :)07:09
wolfspraulI'm bringing Milkymist One to market now, the second product. That's still a boatload of work. Then I need to work hard on the sales side, for sure.07:09
wolfspraulnah you always make money. make make make. I'm American trained. If you need money, you loan it from someone else. Focus on where you create value, the reap the rewards (=make money).07:09
wpwraki see M1 more like a parallel line. something very different.07:10
wolfspraulyes sure07:10
wolfspraulbut I need to focus on that now07:10
wolfspraulit's out of stock, quite embarassing07:10
wolfsprauland most of the people who bought one are not using it, even more embarassing07:10
wolfspraulI am not worried that my NanoNote sales pitch will get worse over time.07:10
wpwrak)if you need money, you just loan it) so you're saying you had the opportunity to find investment for continuing the nanonote line but rejected it ? 07:10
wolfspraulnot so much "buy Ben NanoNote from me", but "look we have this Ben NanoNote, we could go here and there as the next step..."07:11
wolfspraulno there is no sales forecast, no clear evidence of demand07:11
wpwrakthe value of the nanonote declines since there's no continuation07:11
wolfspraulbuild it and they will come doesn't work, or rather I don't believe in it07:11
wolfspraulnot at all. Software on the Ben needs to stay fresh and clean, always updated. Then the Ben can be the entry point for any discussion about Ya, at any time.07:12
wolfspraulthe market around us is consolidating07:12
wpwrakyou built the ben and people came. despite its poor outward appearance07:12
wolfsprauliTown is being built07:12
wolfspraulAndroid massacre is unfolding as planned07:12
wpwrakyeah, the big ones are slugging it out among themselves quite nicely07:13
wolfspraulSymbian, Nokia, RIM all dead I'd say07:13
wolfsprauljust imagine someone from RIM even, trying to sell their current products as the staging area for something bigger and better07:14
wolfsprauland then tiny Ben NanoNote doing the same07:14
wpwrakwell, RIM is s communication product. ben isn't.07:14
wolfspraulI think we can bring forward some good arguments, and I don't see how they get worse in x months, assuming that we can keep the software fresh, upstream everything, not fork, not create unmanageable maintenance burdens, etc.07:14
wolfspraulI'm talking about Ya07:15
wolfspraulI need to make a sales pitch with the Ben, saying "let's add xyz, and then go after market abc"07:15
wolfspraulthat sales pitch is still intact, and for the most part unaffected by the big movements07:16
wolfspraulbut now I need to launch m107:16
wpwraki'd see two main roles for a ya: 1) a technology/design update. 2) move forward on the freedom side.07:16
wolfsprauldefinitely, both07:16
wolfspraulbut that's just the engineering side. Someone will want to understand why they sell.07:17
wolfspraulWhy people will line up 2 blocks screaming and yelling to get one.07:17
wolfsprauland there we have to wait for the right entrace in the market, the right partner who believes in it and goes with the Ben, turning into the Ya sensation07:17
wpwrakwell, one marketing angle could be to pitch the freedom side07:17
wolfspraulthe pull is limited in my experience.07:18
wpwrakbut then, truth be told, i'd much rather see a sponsor than an investor. don't care whether ya can sell enough to make a profit. be content if it sells better than the ben (as a proof of concept)07:19
wolfsprauland outsider would be outright scared that we think an outside can help create a better freedom marketing story...07:19
wpwrakstrengthen the process. then act from a position of power and preparedness.07:19
wolfspraulI think on the freedom marketing side, we have maxed it out. If someone else proves me wrong, fine.07:19
wolfspraulwe've sold 1200+ nanos on that argument alone, and that's the most anyone could have managed for that spec list at that timing.07:20
wpwrakduno. i think a more free and generally better nanonote would find its customer base. not sure if it would be profitable, though.07:20
wolfspraulso of course I agree the freedom side will improve, that's how we invest, but the sales pull needs to come from somewhere else too07:20
wpwrakgta01 sold about that number. "upgrade" gta02 was 10x ;-)07:20
wolfspraulI wouldn't send anyone down that path :-)07:20
wolfspraulbeen there, done that07:20
wpwrakso there's precedent :)07:20
wolfspraulyes there is a market/size/audience, but it's small07:21
wolfspraulof course with more working features, it grows a bit07:21
wolfspraullet's say linear, every feature another 1000-200007:21
wpwrakagain, i would try to think more about way to decouple the immediate development from profitability07:21
wpwrake.g., i also don't see M1 as a money-maker. of course, it could be a surprise success, but its real value is elsewhere. and it doesn't capitalize on that yet.07:22
wolfspraulbut I think the technology we have created can be used to create value way beyond a pure freedom argument07:22
wolfspraul'created' maybe too much, the technology we have bundled :-)07:22
wpwraki'd say the technology we control. in the end, open source/hardware is all about control. yours or theirs.07:23
wpwrakone problem of the ben is that's poorly suited as an OEM platform. so you can't capitalize on the ability to share the control you've gained.07:24
wpwrakanother problem of the ben is that you don't control all that much. still too many things are closed.07:25
wolfspraulyes yes. The Ya will be much more open/free of course.07:25
wpwrake.g., you can't even just sell a Ben PCB. this pretty much dooms all attempts to position it in the embedded market.07:29
wpwrakOEM looking for a reduced/focused function device would almost certainly want to change the keyboard. nope, can't do that either.07:30
wpwrakmaybe someone would like to change the circuit a little. like adding their favourite foo chip. yet another opportunity you have to let pass helplessly.07:30
wpwrakthere's a ton of things the ben can't do because you don't have enough control. each of them may be an opportunity to make it big. granted, there's no point in expecting each opportunity to turn to gold. but then, some of them do.07:32
wpwrakbut again, i don't even think a ya should be planned as profitable. maybe it would be. maybe it'll need one more spin to hit the spot.07:35
wpwrakand i think the potential investors you talked too understood that too. that's why they didn't believe in your strategy of achieving short-term profitability. maybe they would have liked a promise of long-term profitability even less. but at least it would have been internally consistent.07:37
wpwraks/too/to/ # first one07:38
wpwrakregarding the value of M1, i see its strong point also in the technological portfolio. step 1) make a complex FPGA-based system that can survive in an end-customer situation. that gives you the experience and (some of the) resources to competently use fpga as a platform.07:42
wpwrak2) work on opening the use of fpga. llhdl :) that's the real breakthrough there.07:42
wpwrakthe a follow-on product could focus on doing interesting things with 2). it's a long shot. that's why wouldn't really expect even an M2 to be a commercial success. the whole milkymist line should be the long-term research of qi-hw while the nanonote is the cash cow.07:44
tuxbrainwolfspraul: finish your work with MM, do it well and do it as quick as you can afford, then with same spirit to stay focused go for Ya, having the wpwrak OEM arguments on mind(able to use the board only, accecible gpio for extendability), with all improvements achieved like wpan, and maybe if on time the gps , and also with a more stamina +CPU +ram +res but same size. I think credibility on all qi-hardware stuff is pending on our promise on evolve Na07:45
tuxbrainnonote, lot of our Ben Costumers (if not all) had buyed with that promise in mind, others expect to buy when this evolution becomes a reality because NN is just not enough, I agree you must focus on what you have started on MM but I'm worried about your arguments that there is "no need" of Ya, I thinking more on "we are late" on Ya, but of course our lack of resources is the main reason so I was no worried but your arguments yes makes me worry a litt07:45
tuxbrainle bit07:45
wpwraktuxbrain: fully agreed. there is a sense of abandonment around the nanonote.07:46
wpwraktuxbrain: wolfgang's argument that there's no need to a ya makes perfect sense if you consider him saying "I'm American trained." :)07:48
wolfspraulwhy don't I feel the abandonment?07:49
wolfspraulwhat do people abandon the nano for?07:49
wolfspraulmaybe they will come back :-)07:49
wpwraktuxbrain: US commercial philosophy focuses strongly on making money. that's actually rooted in religion, calvinism to be precise. if you don't have money, you're (morally) bad. nobody wants to feel that way. thus you need to switch to denial, to avoid the emotional burden of accepting the truth.07:50
wolfspraulone big worry I would have is that we can move all of the Ben software forward to the Ya07:51
wolfspraulthat's probably the one thing I would consider the most07:51
wpwrakwolfspraul: (abandonment) perhaps you don't feel it because you're busy with M1 ;-)07:51
wolfspraulyes! maybe07:51
wpwrakwolfspraul: if you look at mailing list or irc traffic statistics, you can see quite clearly a sharp decline07:51
wolfspraulbut my nano serves me well. I just reflash 05-28 and it's cool. I use mine mostly as a music player.07:52
valhallamy opinion from the customer side: I've shown off my nanonote in the local foss community and most of the time the answer was "nice device - oh, no internet - that's useless", and that's from people who like freedom and buy this kind of devices even if they have no real ned07:52
wolfspraulvalhalla: fair enough. so what do they buy instead?07:52
wolfspraulmy guess: android device07:52
tuxbrainwolfspraul: an android phone07:52
valhallacurrently most of them buy smartphones07:52
wolfspraultotally fine, I would never, not in a million years, try to chase that07:53
valhallanot only android, also meego 07:53
wolfspraulnext year all Android then :-)07:53
valhallaand arduinos, on the other end of the spectrum07:53
wolfspraulmy sales pitch for nano should be like this:07:53
wolfspraul1) have you heard of these thousands of different android devices?07:53
valhallathose who don't buy android do it for  the (lack of) freedom, I believe07:53
wolfspraul2) they come in all sizes, all prices, even free if you like07:53
wolfspraul3) they have an infinite amount of great software, also free if you like07:54
wolfspraul4) they are getting better every 3 months07:54
wpwraki don't think the nanonote should be positioned as an alternative to a smartphone. we're nowhere near mastery of that sort of technology. i'm not even worried that much about the technical side per se, but more about access to components (and all that comes with them, support, etc.)07:54
wolfspraul5) you still say no to all this?07:54
wolfspraul6) really?07:54
wolfspraul7) ok, let me tell you about NanoNote...07:54
tuxbrainwolfspraul: don't twist our argument we are not saing to become one more android device we must stay in the linux side but we MUST evolve07:54
wolfspraulyes sure, I was joking07:55
wolfspraulvalhalla: we are well aware of the arduino or android movements, and that's all fine07:55
wolfspraulI cannot add value to either, with this project, or my Ben NanoNote product07:55
wolfspraulneither can Milkymist One btw07:55
wolfspraulso we have to let those play out, whichever way they play07:55
wolfspraultuxbrain: definitely. we need as much power as possible!07:56
wolfsprauland fast07:56
valhallathe fact is, at ~100 eur some of those people who already has both an android phone and an arduino could still buy a nanonote, just to play with it07:56
wolfspraulthis is tech business, not aged whiskeys07:56
wolfspraulwe agree on the goal, maybe differ on the path there. Although I am currently pushing m1, and you are pushing ben-wpan. So we do something :-)07:57
wolfspraulI think Android will take all oxygen out of the room for a number of years07:57
tuxbrainI have invested in all that wpan stuff not just to extend the useful live of Ben, the main reason is to have hardware aviable to let the devels to improve the sofware for the future Ya, but that blurry future on ya makes me worry a lot07:58
wolfsprauloh definitely. we need to make an integrated device.07:58
wpwrakvalhalla: exactly ! the android phone is a daily tool. the nanonote can be something else. just because it's small doesn't mean that it has to be in the same product class.07:58
wolfspraulat least I think that can sell.07:58
wolfspraulvalhalla: which argument could make an android user, maybe even android hacker, consider buying a NanoNote on the side?07:59
wolfsprauland which technical or application feature impresses them the most, when you show your Ben?08:01
valhallawolfspraul: not every android user, just the ones in the area of the foss community I know: the ability to do cool stuff with it, installing ones own linux, do cool small projects (of course that's a niche, you don't get to mass market that way)08:01
valhallathey are impressed from the size, mosty, and the fact that it looks like it is a full featured computer (until they know about no usb host)08:02
valhallaso, something where you can hack as it if was a computer, but pocketable08:03
wpwrakusb host may be actually what breaks the deal in many cases. not because it would be so important per se, but because it's one disappointment too many.08:03
wpwrakso my vision for the ya would be to reduce the number of disappointments :)08:03
tuxbrainI think al lthat 6Lowpan stuff is the main busines oportunity we have, is what make us distiguish from the rest and make us a little avadtage from the others hardware alternatives , talking at the marketing side, on the tech side the fpga to move towards 100% free hardware is what make us as project don't lose the north , 08:03
tuxbrainwpwrak: +1 to kill dissapointment sources08:04
valhallawith usb host it would be more like "no, it doesn't have it... but you can do something for it, even as a customer access just to end user devices"08:05
tuxbrainwolfspraul: I will just happy if in the CPU cicles free of M1 that your mind has instead of the "I'm just happy with ben as is" the idea of "we have to move forward on NN" replace it 08:06
wpwrakwolfspraul: also, the software side on the nanonote isn't all that rosy. true, there is good and continuing work on the distribution side. but there's very little now development that specifically targets the device. e.g., makes use of unique properties or addresses inherent limitations08:09
wpwrakso basically all the sw development resources are spent on porting. that's no bad as such, but what's worrying is that there's little original work in this community.08:10
tuxbrainI think the first step will be to decide the  SOC to build arround it , fix it and move towards, find if we have an devel kit with that soc to start porting the soft to that target anything wrong with that aproach08:11
wpwrakoriginal work is what gets people excited. porting is basically housekeeping. a necessity, but usually not a source of excitement.08:11
wpwraktuxbrain: that's how it's usually done :) important here: don't lose the value that's been amasses on the kernel and distribution side. so it would either have to be a very similar chip (e.g., the 4760), or one with comparably good support08:14
tuxbrainwpwrak: but this needs critical mass and the "what I'm doing will some kind of future" spirit, and with the "ben is good as is" idea we will not achieve it08:14
wpwraktuxbrain: also in terms of a marketing message, "platform continuity" may sound better than "new radically different super-soc". at least for the immediate successor of the ben.08:15
tuxbrainwpwrak:  sure! I don't know the alternative we have but fully agree on the evolution (not revolution) though08:15
tuxbrainwhat devel board are there with the 4760 and what is the list of features of that chip? any link to read?08:17
wpwrak(platform continuity) also one thing that keeps VARs (in a broad sense) away: they often want some sense of stability, be it because they have customers with long-running project (including the support phase), and also to justify their own investment in the platform in terms of technical or business development08:17
wpwraktuxbrain: chip features; FPU, 3 (!!!) MMC/SD/SDIO controllers, DDR2. need more ? :-)08:18
tuxbrainwpwrak: yes, be sure this time the keyboard will not interfere with serial port :P08:19
wpwraktuxbrain: access to documentation is the one major weak point of the 4760. ingenic are very incompatible with our ideals in that regard. e.g., wolfgang's beloved TI would be lightyears ahead there :)08:19
wolfspraulvalhalla: and lack of wifi probably (sorry was away)08:20
wolfspraulbut I think the excitement mostly comes from the 'shrunk notebook' look, like a freakshow basically08:20
wolfspraulthe one argument that may potentially work is 'native mutt client', or 'native emacs' or so08:21
valhallawolfspraul: of course, but usb host gives wifi, with some limitations08:21
wpwraktuxbrain: lars has some devel board for the 4760. don't know how large a choice is there. i wouldn't consider the devel board as an overly critical item, as long as there's at least one that works and where you can copy the design from.08:21
wolfspraulbut in reality the user base for this type of app is very small08:21
wpwrakwolfspraul: native mutt in a non-communication device ? hmm :)08:21
wolfsprauloffline mutt08:21
wolfspraulI'm just saying as a story/argument, not even what works today08:22
wolfspraulthere are many apps that are being pushed into history by the android movement08:22
wolfspraulbut there is little/too little movement to really keep them alive, at least today08:22
wolfspraulthere is some, but we are selling to those already :-)08:22
wpwrakwolfspraul: you sound like sean, centering his PR effort for wikireader about boldly claiming that you don;t need communication ;-)08:22
valhallanative serial console with a real keyboard08:22
wpwrakheh, the retro angle has something :) maybe retired engineers are the way to go08:23
tuxbrainwpwrak: (sean comparation) wow that was hard :P08:24
wpwraktuxbrain: (keyboard) oh yes. the 4760 has more I/Os than the 4720 for sure.08:24
wolfspraulI want connectivity, I think Ya must have connectivity. and I think it should be wireless right away, forget the wires...08:25
wpwrakwires are good for embedded use. i would see embedded as a potentially very strong point.08:26
wpwrakyou could even do something nice with the ben. but of course, people will bark at having to buy all that plastic.08:27
tuxbrainyes OEM guys are a real part of my serious costumer base, they go for arduino, so that's why my strong efforts to relate both worlds, they are serching for open and easy to integrate solutions08:27
wpwrakand USB host is less about connectivity-for-communication (you're thethered for power anyway) but for peripherals. e.g., a usb keyboard can make a world of difference.08:28
tuxbrainbut let's not lose the focus, wpwrak do you agree to start a lobby to push NN forward? :P08:29
wpwraktuxbrain: have you been able to talk any embedded developers into using the ben for a project ?08:29
wpwrakyeah :)08:29
whitequarkwolfspraul: maybe just integrate what is currently called ben-wpan to next NN?08:29
tuxbrainwpwrak: yes but lack of accecible gpios don't make it atractive and they finally go to arduino08:30
tuxbrainwpwrak: also as arduino terminal is not complete we can just flash it now08:31
tuxbrainthat's why I'm also intereste on achive wpan with arduino08:31
wpwrak(usb keyboard) i had an eye-opening experience in the early 1980es. back then, i had a sinclair ZX81 at home. it was slow and had a horrible foil keyboard.08:31
tuxbraina NN as data colector /analisys of an arduino based system will be also "easy" to sell08:32
wpwrak(usb kbd) then there was a consumer electronics trade show in zurich. there, one company showed a real keyboard for the zx81. they actually just put the zx81 pcb into the keyboard. it was exactly the same slow hardware. yet, with real buttons under your fingers, not only the typing was much better, no, the whole machine suddenly felt snappier.08:33
tuxbrainwpwrak: (ya lobby) we must find a also 4760 based devel board affordable ,already the market and reproducible as first step, then colect as much doc as posible about it.08:34
wpwraktuxbrain: (NN+arduino) i see than as a strong point too. let a custom peripheral (arduino or such) worry about the bits and voltages, then take the processing to your linux system (ben), where you're in a familiar and powerful environment08:34
wpwraktuxbrain: (nn+a) one problem is the growing competition of usb-capable mcus. a lot of people are still afraid of usb, but it's getting easier. (and then there's of course always ftdi)08:36
tuxbrainwpwrak: mcus?08:36
wpwraktuxbrain: MCUs. on the small end, silabs c8051f3xx, atmel amega*U*, etc.08:37
wpwrakmicrochip also have usb-capable PIGs08:38
tuxbrainwpwrak: another argument to justify NN +(arduino or whatever hardware DIY) is that you can simplify Human interface (buttons, LCD, etc..) you have a full keyboard and a color screen to interact with it, 08:39
tuxbrainwpwrak: (usb capable) yes but we can offer a wireless solution on it.08:39
wpwraktuxbrain: yup, good point. if you need an interface anyway, that's a lot cheaper than building your own.08:39
wolfspraulok I read this later, too many things going on!08:40
tuxbrainwolfspraul: remember..... NN has to move on... NN has to move on... NN has to move on...08:41
wpwraktuxbrain: (usb/wireless) sure. what i'm aiming at is that the problem is more general than just wireless. also, if you're already wireless, you may find it more convenient to use a real pc.08:41
wpwraktuxbrain: (ben as a human interface subsystem) sadly, the lack of design and manufacturing control over the keyboard is probably an issue. e.g., if the system needs to look "nice", you'd probably want a different layout. but nobody can deliver that.08:42
tuxbrainwpwrak: well depend on the aplication or the need of portability but whatever we have another product if you need a full laptop on your app, us AtUSB :) 08:42
tuxbrainus-> use08:43
wpwraktuxbrain: i've actually been thinking about ways to turn the ben into a component of an embedded system. i don't see a way to do this nicely with the keyboard. but just the two pcbs would be possible.08:43
wpwrak(atusb) yeah, there's no escaping us ;-)08:43
wpwraktuxbrain: i think, from the engineering side, we're well-set for working towards a ya that's an incremental improvement of the ben. we have the ben to use as a basis. we have development expertise with the ingenic cpus. as far as i can tell, these cpus aren't particularly nasty to integrate.08:48
wpwraktuxbrain: we probably (*) have a working open hw design process. (*) we'll find out when you get to test the wpan boards ;-)08:49
tuxbrainman need to go now, regarding the open hw desing yes, let's see if our friendly relationship not end with rage and fury when the fab panels arrive :P08:50
wpwraktuxbrain: we also have precedent for fairly successful community-driven design process, namely with gta02-core. wolfgang has good contacts to ingenic and probably other vendors that are useful to know.08:51
wpwraktuxbrain: sharism also has proven manufacturing contacts which are probably less expensive than your fabs. did you compare notes on that with wolfgang ?08:51
wpwraktuxbrain: what's the ETA for the panels ?08:52
tuxbrainwpwrak: today or tomorro08:52
wpwrakwhee ! :)08:53
wpwrakmake sure you have that oscilloscope ready :)08:53
tuxbrainwpwrak: also think I will be easy to find investors from my part with the promise I will mantain the production locally .. this is a point that if reach the production ready status must have in account08:54
tuxbrainwpwrak: osciloscope... this will be complicated, but I will try to do my best08:54
wpwrak(local production) hmm. well, you don't have to explain to your investors too clearly what copyleft hardware means when it comes to keeping production local ;-)08:56
wpwrak(scope) yo must know *someone* who has one ;-) or maybe find one on ebay for 10 EUR ;-)08:56
wpwrakheh, that would be about the minimum that's probably sufficient ;-) http://articulo.mercadolibre.com.ar/MLA-115269097-osciloscopio-funcionando-belgrano-_JM08:58
wpwrakthat's 1000 argentine pesos. about EUR 16008:59
wpwrakwell, 167. and here such things are at least 50% more expensive than elsewhere. so you should be able to find something for around EUR 100.09:00
tuxbrainwpwrak: beleiveme when I said Tuxbrain right now is on the edge and we can afford even such cost... I have to contact some friends to see their disponibility09:05
wpwraktuxbrain: inviting friends with scopes to share the production test experience is also a good idea :)09:14
whitequarkwpwrak: won't 2.4G require quite an expensive scope?09:16
tuxbrainheheheh yeah but I think the procces will be as follow, me doing the the poor man test with atben/atusb form fab and discard those not covering the minimum crossing fingers that they are just a few, and taking this ones to my friend house and analize it.09:17
wpwrakhttp://www.gocomics.com/nonsequitur/2011/06/06 ;-)))09:17
wpwrakwhitequark: it you want to measure the RF, yes :) what i'm after are things like the clock, which runs at 8 MHz09:18
whitequarkwpwrak: ah yes, that's simpler then09:19
wpwraktuxbrain: yeah, people who have scopes may not want to take them out of their lab.09:19
tuxbrainI have this one http://www.syscompdesign.com/CGR101.html but you said is not enough09:20
wpwraktuxbrain: hmm. it's very borderline. the theoretical minimum for the sample rate to detect an 8 MHz signal would be 16 MHz. that one does 20 MHz. normally, you want something like 4-5x.09:22
wpwraktuxbrain: but if you have it, you may as well give it a try. plug in an semi-working atusb and see if you can measure the clock.09:23
wpwraktuxbrain: details are here: http://downloads.qi-hardware.com/people/werner/wpan/prod/analysis.html#clock09:24
tuxbrain(4760) I have found this devel board, but not find a place to buy it ftp://ftp.ingenic.cn/2soc/4760/RD4760_LEPUS/RD4760_LEPUS-HW%201.3_EN.pdf09:24
wpwraklarsc: do you know where one can get 4760 devel boards ?09:25
wpwraktuxbrain: (scope) if the scope shows a stable 8 MHz, you're good. if it shows major amplitude variations or a different frequency, then your sample rate is too low09:26
wpwraktuxbrain: an 8 MHz signal should look roughly like this on your 20 MSa/s scope: plot "<perl -e 'for (0..20) { print sin($_*3.1*2/20*8),\"\n\"; }'" with linespoints, "<perl -e 'for (0..200) { print $_/10, \" \", sin($_*3.1*2/25), \"\n\"; }'" with lines09:56
wpwrak(for gnuplot)09:56
wpwrak(assuming no interpolation)09:57
larscwpwrak: i guess from ingenic10:49
wpwraklarsc: ah, the obvious source then :)11:01
larscdon't know, but maybe you need to order a few thousend SoCs to get one11:26
larsc"alsamixer: function malloc failed: Success" ...13:24
lunavorax_frizzlHi all14:56
rjeffrieswpwrak are you awake?15:45
wpwrakof course :)15:46
rjeffriesthe man who never sleeps15:46
rjeffrieswish to share a factoid re cost to go from FPGA to asic15:47
rjeffriescurrent cost is $20,000 USD. less than what I guessed15:48
kristianpaulwhere you get that number?15:48
rjeffriesI am working with a startup that is almost ready to do firt asic run. that is a real number15:49
wpwrak20k is remarkably cheap indeed. do you have any idea how complex the chip is ? (e.g., what fpga they used, the structure width in the asic, etc.)15:51
wpwrakah, and clock speed. very important ;)15:52
rjeffriesthe design is fewer than 10K gates. the process is an older node, similar to what Ingenic uses.16:26
rjeffriesclock speed is less than 100Mhz16:28
wpwrakokay, that's a pretty simple chip. clock speed is a bit less that what we'd want, too.16:29
kristianpauljust 10K ? what is that asic designed to do?16:29
kristianpaulyeah, looks better having a fpga with that speed..16:30
wpwrakkristianpaul: if they need lots of them, an asic would be cheaper. the price (20 k) is certainly attractive16:31
wpwraknow we need another data point. see how it scales ;-)16:32
rjeffriesyes need LOTS of them. the simplicity is a design objective. 16:49
rjeffriesone wonders what the fisrt run of asic from Milkymist might cost. 16:51
wpwrakrjeffries: digitalizing pop corn ? one tiny little chip per ... duh, fluffily exploded kernel ;-)16:51
rjeffriesthat FPGA is only $4016:51
wpwrakthe asic will be a lot cheaper then16:52
rjeffrieswpwrak when the time comes you'll understand16:52
rjeffriesdo you think MM will cost $100,000 for the first pass to asic16:52
wpwraki don't really know how to calculate those prices. M1 is certainly a lot more complex. also with a huge number of I/Os.16:53
rjeffriesM1 supports some interaces that would not be needed in a general purpose SOC. 17:01
wpwrakoh, it doesn't matter what function the pins have. it's just their number. and memories tend to be the main consumer there.17:08
Fusinhi qiots ;)17:48
rjeffriesFusin  qiots -- what a lovely term. ;)18:10
Fusinit's (c) by Fusin18:11
Fusinwhich means: it's CopyLefted by me :P18:12
wpwrak(„) ?18:15
Fusingood idea wpwrak  :D18:15
Fusinmy old motto was (and still is) share, chat and have fun18:16
wpwrakah, in the good old times, it was sex, drugs, and alcohol. not too different, if you think of it ;-)18:18
Fusinthose 3 are included in 'have fun' :-D18:19
Jay7sex, drugs, rock'n'roll :)18:19
Jay7good old times ;)18:19
Fusinr'n'r is included in 'share'18:19
Action: kyak is back18:48
wpwrakseems that they cleaned up the plane a bit. before, it looked like this: http://bucket.lanacion.com.ar/anexos/fotos/82/1389682w346.jpg18:51
wpwrakthe test points under the battery must have been designed by a sadist. there are so many interesting things you could almost do. but then there's just a signal missing or two test points are some tenths of a millimeter too close to each other. grmbl.19:10
wpwrakdo we have something like this already ? if not, then we do now ;-) http://downloads.qi-hardware.com/people/werner/tmp/tp-map.png19:15
tuxbrainwpwrak: what do you plan with this "nightmare made Testpoints"19:24
wpwraktuxbrain: there are two ideas circling in my brain: 1) a "docking station" that would connect via the battery compartment and bring out a few GPIOs, USB_BOOT, plus VGA (stolen from the LCD controller this time)19:25
wpwraktuxbrain: 2) more radical, strip the ben down to the bare PCBs, then add a contraption that covers the component side of the board and brings out various interesting signals (the ones from above, plus the remaining KEYxxx and such. this should also include one real SPI.) place the LCM on top of all this, add some screws and an acrylic cover. voila, an "industrial" PC.19:28
wpwraktuxbrain: in scenario 2) you'd of course lose the keyboard19:28
tuxbrainwpwrak: serial keyboard? start to think in a wpan keyboard?19:30
wpwrakmaybe your industrial application doesn't even need a keyboard ;-)19:30
paulrojaswpwrak: the other pcb side have intersting TP too like the unpolulated IC for i2c i think19:32
wpwrakand yes, you could of course plug in atben and go wireless. maybe even integrate an spi-based wpan into the board. should be easy.19:32
wpwrakpaulrojas: i think all the tps are on the non-keyboard side. so one board could get them all. would have a weird shape, but should be possible.19:33
wpwrakwell, almost all. unfortunately, the spacing between TPs is sometimes too narrow to reach them with test pins.19:34
paulrojasbut if you stole vga from LCD controller, you loose dma right?19:35
wpwraktuxbrain: (test pins) the kind you have on atusb-pgm. there are also some with a center pin that allows for more accurate positioning. the ones on atusb-pgm are a bit of a nightmare when it comes to precision.19:35
paulrojasso it will be kinda reduced on resolution19:35
paulrojasbut sound good idea anyway :-)19:35
wpwrakpaulrojas: (lcd) the idea would be to generate the full video timing with the lcd controller. so it would still use DMA.19:35
paulrojasi see19:36
wpwrakpaulrojas: one limitation would be that the LCD controller only does up to 800x600. so no 1024x768 :-( ironically, ubb-vga can do more in this regard. (but you'd have more colors, more stable timing, and the 8:10 card slot still available if using the LCD controller for VGA)19:39
paulrojasfor industrial ethernet is a must i think, but with free gpio that will not be a problem to attach19:41
wpwrakah, i thought the nick sounded familiar ;-)19:42
kristianpaul_lapsorry, another power failure at home, so i lost my screen sessions again :(19:44
krispaul_laptopof course i miss the real SPI part :-)20:16
wpwrakTP69, TP70 .. should be feasible in scenario 220:23
wpwraknot an ideal situation, though. tp70 is very close to the shield of the 8:10 card holder.20:24
wpwrakand tp69/70 are too close, so the pins couldn't attach to the center, forcing the one of tp70 even closer to the shield20:25
qi-bot[commit] Werner Almesberger: tp/: map of test points on the Ben (battery area for now) http://qi-hw.com/p/wernermisc/8ccfd2b22:42
DocScrutinizerso what? this coming *now*22:58
DocScrutinizernot going to brag, but that's been bare minimum on the "open" devices I contributed23:01
wpwrakDocScrutinizer: yeah, i'm a bit surprised myself. of course, the information has been around. just not as convenient to access as this picture: http://downloads.qi-hardware.com/people/werner/tmp/tp-map-ba.png23:09
DocScrutinizeryeah, otoh the testpoints actually seem designed by a sadist. I completely agree here23:11
DocScrutinizernot really meant to tease hw-hackers23:12
wpwrakthat 1.39 mm spacing is particularly interesting. i think it pretty much guarantees that you won't be able to find any spring-loaded device that connects to it23:13
DocScrutinizercould've used some 2--4h paid professional review ;-)23:13
wpwrakah well, qi-hw and money ...23:13
DocScrutinizerI'm more and more averse to doing anything unpaid. Usually anybody short of those few $$$ isn't serious anyway23:15
wpwrakyou should consider your contribution as an investment. see, you don't have to be a financial genius - even a lowly engineer can be an investor ;-)23:16
DocScrutinizerand those preferring to spend $$$$ on testruns rather than $$$ on review or some tasks outsourced, well...23:16
DocScrutinizereven *I* an good enough a financial genius to know this isn't an investment ever going to pay back, neither on monetary nor on any metrics of satisfaction or whatever23:18
wpwrak(just checked spring-loaded connectors) as i expected. all the regular ones have different spacings. 50 mil, 100 mil. 2.5 mm, 3 mm, 4 mm. never 2.0 mm, 1.4 mm aka 55 mil, ...23:18
wpwrakwell, if you hate it, then i can't help you23:19
DocScrutinizeryou could as well have saved your time for doing the lookup ;-)23:19
wpwrakwell, every once in a while you find something you didn't expect. alas, not this time.23:20
DocScrutinizeryou as well could try to find proper ZIF sockets for 2.5mm 64pin DIL23:20
wpwrakhmm, can you bend those pins about 0.6 mm ? ;-)23:21
DocScrutinizermaybe in some russian surplus from an ancient lab23:21
DocScrutinizersure, but at pin 32 it gets nasty - "who stole that pin?!??!!!"23:22
wpwrakwouldn't the russian variant be something like 1 cm pitch, 5 mm diameter solid copper wire? :)23:22
DocScrutinizernah, they really used 2.523:22
wpwrakif you can bend them up to 0.6 mm, you'd just be able to fit those 32 pins :)23:23
DocScrutinizererr 2.5023:23
DocScrutinizerhuh? 32 * 0.06mm?23:24
wpwrak0.04*16 = 0.64 mm23:25
DocScrutinizeryour last pin is 1.8mm off23:25
wpwrak32 pins each side, 16 to bend to the left, 16 to bend to the right23:25
wpwrakvoila, fits into a 64 pin 100 mil socket :)23:25
wpwrakwhere do i pick up my mcguyver price ? :)23:26
DocScrutinizerwhatever, /me away for another pint - got my "MEH!" day23:26
wpwrakDocScrutinizer: cheers ! :)23:27
DocScrutinizer51wpwrak: got your MISO fixed?23:33
wpwrakDocScrutinizer: that's tuxbrain's bean paste :) (jap. "miso")23:38
DocScrutinizer51I know (both)23:38
wpwrakDocScrutinizer: and no, not fixed so far. seems to be mechanical. perhaps i cooked the chips too well.23:38
wpwraki had a bunch of such failures around the transceiver. the problem there is that i need to work with a lot of heat because of the nearby ground. perhaps that's pushing the chip too close to the point of failure.23:39
wpwrakwith proper SMT, this shouldn't be an issue23:40
wpwrakan alternative explanation would be that the board bends too much and damages the chip. we'll find out ....23:40
wpwrakmay also be a combination of the two: chip is already weakened from overheating and then the mechanical stress kills it.23:41
wpwraki tried that. 0% yield.23:44
DocScrutinizer51still searching for a proper low temp jointing method23:46
DocScrutinizer51glue doewn't work23:46
wpwrakthe problem with preheating may be that i can't work on the hot plate for long. so all the placement gets down off-plate. which increases thermal stress.23:46
DocScrutinizer51SMT is an art23:47
wpwrakin any case, the ground should be perfect with proper SMT ;-)23:47
wpwrakyeah. with non-RF designs and thus less aggressive grounding, i have no trouble23:48
wpwrakbut RF is picky ...23:48
DocScrutinizer51you forgot the thermal fireall trenches?23:49
wpwrakthe what ?23:49
DocScrutinizer51the gap with 2..4 thin traces around each gnd pad23:49
DocScrutinizer51meant to reduce thermal coupling to the large gndplane23:51
wpwrakoh, ground pads should be fine. http://downloads.qi-hardware.com/people/werner/wpan/tmp/atusb-110330-overview.png23:51
wpwrakbut the traces and the proximity of ground areas still sucks away too much heat23:51
DocScrutinizer51yeah, see it done right on CN0123:54
rjeffrieswpwrak a few weeks ago I mentioned a $199 or so android capacotive touch tablet that uses Ingenic SOC can't remember the brand it had pretty good hardware specs23:55
wpwrakmy trouble component is U2 ... that's the one beleaguered by ground on almost all sides23:55
wpwrakrjeffries: may be 4760-based then. that chip looks pretty decent. well, even the 4720 we have could do better. e.g., the DRAM and NAND bus are only half as wide as the chip could do. this has a cost ...23:57
--- Thu Jun 9 201100:00

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