#qi-hardware IRC log for Tuesday, 2013-02-26

cpgnot sure if this is apropriate & i'm looking for advice on how to use this power mosfet: BSS138LT103:51
cpgor similar03:51
cpgto drive LEDs with pwm03:51
wpwrakwolfspraul: ron dug out this overview paper, on new programming paradigms (C and such) for FPGAs: http://queue.acm.org/detail.cfm?id=244383614:55
larscso true15:00
lekerneljust use migen. their FP temperature conversion example could be nearly the same as in C with the appropriate library.15:02
wpwrakheh :) well, i'm hoping for an approach with slightly deeper abstraction. i.e., go from regular C to whatever low-level. may well be that bits of migen would come in handy at some point of such an exercise, though15:14
lekernelmigen already does python subset to straightforward FSM. increase the subset and optimize (eg. modulo scheduling) and you should get approximately what you want...15:21
kristianpaulwpwrak: hi :)15:27
kristianpaulwpwrak: do you a good source (book?) for learning string/text manipulation with C?15:27
kristianpaulor basically you would use a library like pcre? ..15:28
wpwraknaw, no idea. i tend to just do such things on the fly.15:28
lekerneldid you just use "good" and "string manipulation with C" in the same sentence?15:38
larscwell that's easy: "string manipulation with C is not a good idea" ;)15:40
wpwrakhmm, how about "strings manipulation in C is just as good as you are" ? ;-)15:42
Fallenoustring manipulation in C is the best backdoor ever15:43
paul_boddielekernel: Did you look at MyHDL? I did look at your thesis, but I can't remember if you mentioned it.16:14
lekernelyes - but everything in the thesis is pre-migen16:14
rohhey.. when you get confronted with a war-drone from the other side.. you'll be happy about broken string routines on network sockets, i assure you16:16
paul_boddieMyHDL seems to be of the PyPy school of code generation: load the system, let it warm up, read out the result. Although MyHDL seemed to actually have the system run and simulate the hardware before generating VHDL/Verilog, as far as I remember.16:16
lekernelpaul_boddie, huh16:17
paul_boddieMaybe I should look at it again.16:18
lekernelthe only thing related to pypy in myhdl is they recommend using it to make their simulator faster16:18
lekerneland you can also run simulations from migen ...16:19
paul_boddieYes, I know about the benefits of PyPy running MyHDL, but what I meant was that the way the PyPy interpreter is built also involves it being imported into Python and then the relationships between the units of the interpreter being investigated.16:21
paul_boddieIn other words, neither PyPy nor MyHDL use static analysis.16:22
kristianpaulwpwrak: yeah as good as i'm not hehe thats why i asked16:23
wpwrakFallenou: add valgrind and C string manipulation suddenly looks much better. e.g., a lot of bugs show up quickly. in other languages, it's all silent failure.16:38
wpwrakkristianpaul: you can always make regression tests. run them under valgrind and - if your tests have reasonably good coverage - it'll find pretty much all atrocities that may be hiding there.16:39
Fallenouwpwrak: yes, use valgrind, use ssp, -fstack-protector, compile as PIE to allow ASLR of .text, don't use lazy binding for dynamic linker resolution (in order to set .got section read only at startup)16:48
Fallenouand use -Wall and -Wformat16:49
Fallenouand use -D_FORTIFY_SOURCE=216:49
Fallenou(and a PaX kernel)16:50
wpwrakheh, and i thought i was paranoid ;-)16:52
larscFallenou: you forgot -pedantic17:16
Fallenouand -Wformat-secutiry17:17
larscthat should be included in -Wall17:18
larscok, the documentation says I'm wrong17:18
paul_boddiewpwrak: I read that you were ramping down your qi-hw activity levels. Is that true?17:45
rohhm. sharism.cc down18:21
wpwrakpaul_boddie: yes, the time of being able to act as if i was independently wealthy has come to an end18:23
rohwpwrak: sigh. same here18:30
paul_boddieIt's a shame that it hasn't proved to be a sustainable way of life, I guess.18:37
larsctime to win the lotto jackpot18:46
wpwrakpaul_boddie: yeah, it's for lack of trying. but nothing worked :-(18:58
paul_boddiewpwrak: I think you've done some very interesting projects, and I still think that there are lots of nice applications for the Ben and the UBB, but I don't know how you get people to support these things, really.19:00
paul_boddieThere seems to be a lot of economic activity around stuff like Arduino and the many other systems out there. Maybe people don't perceive the Qi-Hardware offerings as systems and so they don't pick up much momentum.19:09
rohpaul_boddie: in the end it boils down to sales of products cross financing development20:03
wpwrakyeah, things like UBB were meant to make the Ben more attractive. but the ben would have needed a facelift, which never happened.20:35
wpwrakperhaps the mistake was to switch too soon to M1, which looked attractive and exciting, but never found a market20:37
wpwrakbut then maybe the nanonote never had a chance, and even with everyone on that, we wouldn't have made it20:37
paul_boddieI think the Ben still has an interesting role to play for electronics experimentation. You see people messing around with the Raspberry Pi and saying that it's more usable than Arduino for flashing LEDs and all that kind of stuff, but the Raspberry Pi isn't a true portable solution.20:42
paul_boddieHaving a device with a screen, as opposed to needing a HDMI-equipped monitor (or whatever), allows all sorts of things to be done on a fairly casual basis: you can just take the Ben along and it provides the whole package.20:44
whitequark"more usable than arduino for flashing led"20:46
paul_boddieThe argument for Raspberry Pi in bringing electronics into the classroom would also melt away substantially if someone just made a simple and cheap USB GPIO box, and that probably already exists, although I don't have any links to hand.20:46
whitequarkarduino is a product entirely designed around flashing leds :S20:46
whitequarkpaul_boddie: it's not that simple. you can't reliably bitbang something over USB if the timing requirements are strict and freq is relatively high20:46
paul_boddiewhitequark: I'm not advocating flashing LEDs. I know that even mentioning it to you is like waving a red rag in front of a bull. :-)20:46
whitequarkyou can, of course, if that's a synchronous bus20:46
whitequarklike ft2232 jtag flashing works20:47
whitequarkbut you won't really be able to send or recv something with that box over a 115200 UART, or at least I think so20:47
paul_boddieYes, but nobody is doing classroom electronics interfacing with really high frequency stuff.20:47
paul_boddieAt least not the kind of classroom that the Raspberry Pi people are catering to.20:47
whitequarkraspberry pi advocates on my blog were saying that rpi isn't really designed for meddling with electronics at all20:48
whitequarkjust a cheap computer for developing countries and otherwise economically constrained environments20:48
paul_boddieThose people regard it as everything to everyone and yet nothing to anyone in particular. They could run for public office.20:48
whitequarkin which case it even somewhat makes sense20:48
whitequarkhaha that's excellent20:49
paul_boddieWell, it's actually for the UK education market. Given that cheap stuff is king in Austerity Britain, they'd be onto something, but then they'd need government grants to upgrade the "monitor stock".20:50
whitequarkyeah, it's quite insane to require hdmi there20:50
paul_boddieI actually perceive it as being a product designed for one thing and justified by another argument entirely.20:50
whitequarkwell, if that "one thing" is "Broadcom promotion" then indeed20:51
whitequarkI'm not sure what else RPi is actually good (opposed to "usable") for20:51
paul_boddieBy the way, I wrote a blog entry about this which referenced yours: http://blogs.fsfe.org/pboddie/?p=11120:51
whitequarkquoted by something.fsfe.org. Not sure if good or bad :)20:51
paul_boddieI could blog elsewhere, but this was convenient and pertinent.20:52
whitequarkpaul_boddie: are you from the UK?20:53
paul_boddieBut I don't live there.20:53
whitequarkyour article is written much better than me. lots of interesting context I didn't know about.20:54
paul_boddieThanks for the compliment! I was expecting flames. ;-)20:54
whitequarkwell, one does not exclude another20:54
paul_boddieActually, I learned a few things while preparing that post, like the bit about the BBC Micro being made in India, although I don't think they managed to make the whole thing there in the end.20:55
paul_boddieAs in, I don't know whether they managed to do licensed 6502 CPU production.20:55
paul_boddieThere's a ton of interesting stuff you can dig up if you start digging.20:55
paul_boddiewpwrak: The M1 seems very interesting and it has immediate applications, but then there seems to be a lot of FPGA-related hardware out there, and I get the feeling that it mostly appeals to people who know what they want to do with such stuff. For everyone else, it's a bit mysterious.20:58
paul_boddiewhitequark: Not that anyone will have read it, but I wanted to tease people a bit about the Raspberry Pi and the BBC Micro because there's a kind of advocacy from RPi people that really trades on the BBC legacy, and I wanted to point out that they don't necessarily have legitimacy in building on that.20:59
whitequarkin a certain sense, they do build on the BBC legacy the way it has happened21:00
whitequarkit's just that these days, computers are much more complex21:00
whitequark(and license-encumbered.)21:00
paul_boddieWhat they forget is that there was a vast amount of material produced before, during and after the BBC Micro shipped. What they've done is to ship the RPi (eventually) and then start talking about the materials. My impression is that Google has done a lot more in the past year or so to get educators involved in the new curriculum.21:03
wpwrakpaul_boddie: the nanonote is indeed very good for tinkering. but it would need stronger marketing for that. also, the ben's feature set is outdated, it has some annoying quirks (like a bad keyboard layout), and - worst of all - we never had full design control.21:15
wpwrakpaul_boddie: so the ben was very successful for getting something usable with an impressively small initial R&D effort, but it would have needed a follow-in product to capitalize on that achievement21:16
paul_boddiewpwrak: I guess we'll have to see how that handheld Ingenic-based console (whose name I don't remember, but it was related to the Dingoo) works out.21:17
wpwrakpaul_boddie: M1 is too complex as a pure FPGA experiment device, but it seems that the competition in the intended field outclasses it in terms of effect quality and overall usability (including compatibility with state of the art infrastructure).21:18
wpwrakpaul_boddie: so main feature the M1 had over the competition was integration. but that's probably just not enough. perhaps it could have been more successful with better marketing, though.21:19
whitequarkmarketing is too often overlooked :/21:20
paul_boddiewpwrak: Yes, I think that people who know what they want probably just go straight for whatever that is, and people who aren't sure probably don't get an M1 just in case they don't get into FPGA-related stuff but need a VJ station.21:20
paul_boddieWell, I think the M1 is a success if you compare it to that project whose goal was to make an open graphics card.21:20
wpwrakwhitequark: yeah. and we're not good at it, so we'd need external help. help that is likely to cost money.21:21
wpwrakpaul_boddie: the M1 is a nice demonstration of technological ability ;-) even if it's far from on par with any PC or even a smartphone, it accomplishes a lot considering the ridiculously small resources available for its development21:23
wpwrakpaul_boddie: but of course, that doesn't mean that it's actually a viable product21:24
paul_boddieI do think that the Arduino market is interesting in this regard, though. We're not talking about mass-market stuff, but there are a lot of vendors selling boards and accessories, although I would also point out that many of them probably don't really add very much value, even the ones making their own versions of boards.22:08
wpwrakthe market seems to be interestingly large. proves that such a thing can be successful.22:10
paul_boddiePeople do seem to be willing to spend money on things that interest them and are presumably put to good use, but I think that there's probably a lot of trade on the basis of people liking the idea of doing a project and maybe having the time and energy to complete it, but also maybe not managing to do so.22:15
paul_boddieThe idea of unfinished "do it yourself" projects is a recurring joke in popular culture, and I think that it fits in quite well with that kind of hobbyist culture.22:16
wpwrakyeah, a nanonote may look too much like a "serious" investment for that segment of the market22:17
paul_boddieThere's probably an easily explained correlation between price and out-of-the-box usability. So, someone might be happy to spend $10 on a breakout board that may or may not give them an interesting project, but things like GPS boards for $50, say, and 802.11 on a board for $100 really need to be almost ready-made solutions (and have clear purposes).22:23
wpwraknot sure if it's so much the price or the appearance22:28
wpwrakan arduino looks clearly rough and unfinished. a ben looks pretty and shiny.22:28
paul_boddieThe other thing that the electronics hobby vendors are good at is selling you stuff that you didn't know you needed, and by the time you do know you needed them, you feel committed and won't just drop everything and take up something else instead. ;-)22:29
wpwrakheh :)22:29
wpwrakof course, there would be something to be said for nice packages. e.g., nanonote plus UBBs plus components for experiments plus a little booklet explaining the experiments plus software to drive the things22:30
wpwrakall this could still have a very low hardware cost. BUT ... you'd have to put time into creating the experiments.22:31
paul_boddieEven if you started out with the Sparkfun starter pack (or whatever it's called), there's always going to be something else, although some of those packs are quite comprehensive.22:31
paul_boddieLots of LEDs in the box with whitequark in mind. ;-)22:32
wpwrakoh sure. you can always add some special interest packages22:32
paul_boddieExample of LED flashing plus Android: http://blog.tkjelectronics.dk/2013/01/bluetooth-controlled-rgb-light-strip/22:33
wpwrakadding wireless is a problem22:35
paul_boddieI think a lot of the sales are generated on the promise of interesting projects, but these need people to step up, and you do see a lot of unhappy people asking questions that the vendors don't answer.22:36
larscmaking the device wireless is not a problem, adding connectivity add the same time is ;)22:36
paul_boddieAn example of this is the Arduino shield I'm playing with for USB Host. It's a Sparkfun board derived from some guy's board where they point everyone to his site to get the library. I think I mentioned this before.22:37
paul_boddieSo everyone looks to him to troubleshoot their projects despite them buying Sparkfun's board.22:38
wpwraklarsc: yeah, a nice little decorative antenna is easy :)22:39
wpwrakpaul_boddie: the joy of outsourcing :)22:39
paul_boddieSparkfun are making their money on the "this will be so cool" factor without really saying what that might be.22:39
wpwrakas long as it works ... :)22:40
wpwrakonce they have a reputation, they can start selling black boxes. no description of content, just a price tag.22:40
wpwrakactually, apple could do that :)22:41
whitequarkit looks the other way for me22:44
whitequarksparkfun is selling components and tools, not finished, polished products22:45
whitequarkwant an accelerometer? sure, here's one on a convenient board, with docs layout and such.22:45
whitequarkthey aren't going to convince you that you indeed need an accelerometer for whatever22:45
paul_boddiewhitequark: Right, they make it easy to start but don't give you the whole product.22:45
whitequarkthe "wow factor" from a polished look and nice description still persists, but that's just due to inexperience of an average buyer. I don't think this was deliberate on sparkfun's side and/or their fault22:46
whitequarkpaul_boddie: digikey doesn't either!22:46
wpwrakthe joy of DIY is providing that missing link :)22:46
whitequarkand I would argue that it's actually very good!22:46
whitequarkfor a DIY shop, of course, as wpwrak says.22:46
paul_boddieIt appeals to people like me who think that I might have fun playing with an accelerometer (which I have, actually), and can be tempted to buy the least inconvenient thing to get me started.22:47
whitequarkwell, that, and because when people resort to DIY it's often because such a product does not exist or is too niché22:47
whitequarkpaul_boddie: well, they for sure sell them cheaper than vendor development kits22:47
whitequarklike an order of magnitude, even if you consider shipping22:48
paul_boddieIf Sparkfun only offered the surface mount component, I wouldn't go back to their site, so they are providing a service that justifies the price tag.22:48
whitequarkand soldering a 3x3mm SMT accelerometer is anything but "fun" for me22:48
paul_boddieWell, vendor development kits are probably just a way of you getting to know the vendor's salesforce while paying their salaries.22:49
whitequarkI mean, I know a guy who can attach *wires* to it22:49
whitequarkI have no clue if he had to do human sacrifice to achieve that ability or what22:49
whitequarkhow is that even possible at all22:49
wpwrak3x3mm is pretty large :)22:50
whitequark.5mm pfc-coated wires22:50
whitequarkor even .3mm22:50
whitequarkwpwrak: but you can't heat one pad without heating the entire device22:50
whitequarkand you are stretching it quite a bit ;)22:50
paul_boddieHe should be on that Stan Lee show featuring people with "superpowers".22:50
whitequarkit's indeed very small, even in the current industry status22:50
rohsparkfun lives off big numbers22:50
rohas in amounts, not money22:51
rohlots of cheap devices built even cheaper.22:51
wpwrakwhitequark: why would heating the device a little be a problem ?22:51
whitequarkwpwrak: because the other wires will get desoldered22:52
wpwrakmaybe you shouldn't use the large flamethrower then ;-)22:52
whitequarkand as .5mm pfc-coated wires are multiple-conductor ones, they have some internal tension22:52
whitequarkwhich moves them off the position they were in, detaching them from the pad or even worse22:53
wpwrakdoesn't sound too messy. just a question of finding the right amount of heat.22:54
whitequarkI've had a 30W or so iron then, temperature-regulated22:55
whitequarkprobably it's too much22:56
whitequarkwell, understandably so, that's almost 3W per square millimeter :D22:56
wpwrakthe main factor should be the size of the solder drop you're going to deposit. the iron doesn't need to make contact. the thermal energy in the solder will do all the work.22:57
whitequarkwpwrak: not sure. surface tension forces resulted in the entirety of the solder drop of any size to be deposited all over the tip of the iron22:58
whitequarknote that a pad is something about 0.8x0.4mm there22:59
whitequarkyou really need a very small amount of solder, and it's not easy to deposit it at all, much less cleanly22:59
wpwrakwell yes, just a tiny drop :)23:00
wpwrakand if even this is too small, pre-tin the wire and touch it onto the pad. then timing and aim become important.23:01
wpwrakof course, if you're really having such a hard time, then you may be better off making a pcb. soldering wires directly is meant to make it easier, not harder ;-)23:02
paul_boddieI sense a challenge in the making.23:04
Action: paul_boddie is, in any case, off.23:05
whitequarkwpwrak: it was much, much easier to just make a pcb.23:07
wpwrakah well. problem solved then :)23:07
whitequarkI mean, I just *love* slapping things on a masked, pre-tinned PCB and applying a hot air gun23:07
whitequarkit almost always works (except for when I'm too impatient or fail to distribute heat evenly.)23:08
whitequarkeverything else--wires, laser-printer-gone-pcb-maker stuff--always failed me23:08
whitequarkthe article http://queue.acm.org/detail.cfm?id=2443836 is funny23:10
whitequarka view of FPGAs by an ignorant C developer, may I say23:10
whitequarkwhy exactly do we need chars, ints and longs (ints!) in an FPGA if we have integral types of parametric width?23:11
whitequarkalso the suggestion to convert celsius to farenheit with floating point is hilarious23:11
whitequarkwhat a waste of resources23:12
whitequarkstop slapping C everywhere already.23:14
whitequarkit's correct in that the current EDA tools are crappy, though...23:15
larscstuck in the 90's23:17
whitequarkI guess we really need at least wolfspraul's work.23:19
whitequarkand, possibly, one more iteration after it...23:19
whitequark(as with proprietary cc's/gcc/clang)23:20
--- Wed Feb 27 201300:00

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