#qi-hardware IRC log for Saturday, 2011-08-27

Action: DocScrutinizer feels like eventually having a glance at m1 schematics00:01
DocScrutinizerif anybody would bother tossing over a URL of a pdf00:02
wpwrakDocScrutinizer: here it is: http://milkymist.org/mmone/rc3_schematics.pdf00:08
DocScrutinizercoming to think of it, I'd probably nevertheless design this circuit with a beta>150 NPN emitter follower, 4k7 from emitter*pin23 to GND, and a 2 pcs 220k divider for base, from AVDD to AVSS00:08
DocScrutinizeroh, and of course anothe decoupling C from emitter to pin2300:09
wpwrakphew. getting complex :)00:09
DocScrutinizernah, bom=5 pcs00:10
DocScrutinizeryou save one R of your divider though00:10
wpwraki think the ~50 pF cap is something we could get into rc4. anything else is scary.00:10
DocScrutinizermake 400:10
wpwrak*grin*00:10
larscDocScrutinizer: http://www.analog.com/static/imported-files/rarely_asked_questions/groundingADCs.ppt00:11
DocScrutinizeradvantage: idependence of actual Z if chip'00:11
DocScrutinizers input00:11
DocScrutinizersingle point of failure (transistor)00:12
DocScrutinizercheap to fix00:12
DocScrutinizerinherent ESD-prot00:12
DocScrutinizerhigh effective input Z allows for clean design of the input devider and input Z, and also allows proper RF rejection00:14
DocScrutinizerjust IF I were to design this detail ;-D00:14
DocScrutinizerwtf is this URL above?00:15
wpwrak"AGND and DGND [...] should be connected together [...] Even if the data sheets suggests otherwise!" touche ! ;-)00:18
wpwrak(ADCs ppt) good stuff00:19
DocScrutinizerwell, I've never seen a datasheet suggesting to use a bead to connect AGND and DGND. But it's also probably not correct to simply use one plane for AGND and DGND00:23
DocScrutinizerraher use 2 planes and connect them at one point, next to power supply00:23
wpwrakDocScrutinizer: (the beads) they were basically a misunderstanding that propagated into various subsystems. we've already fished one out of the video section a while ago. now there's another one in audio.00:23
DocScrutinizerouch00:24
DocScrutinizernever place beads into ground (exception that confirms the rule: headset jack, for antenna purposes)00:24
DocScrutinizeranyway for your usecase you shouldn't bother at all about noise00:26
wpwrakyeah, it did interesting things to the respective codecs. luckily, no permanent damage.00:26
DocScrutinizerso for this particular usecase a common groundplane is just fine00:26
wpwraki have my doubts about all the ground going to peripherals. M1 connects to a lot of stuff. i think it would be safer to have things like MIDI, DMX, maybe also audio galvanically separated from the rest.00:29
DocScrutinizeranyway, time for a burger. cya00:29
wpwrakfood sounds like a good idea :)00:29
Action: wpwrak contemplates his options00:29
DocScrutinizeryeah, for a stage device connected to a lot of peripherals, galvanic separation is a must, at least for audio00:30
DocScrutinizerstage environment can do to a device without bead in gnd what even your home stereo did to a device with bead ;-D00:31
DocScrutinizerbuzzword ground loops00:31
wpwraki think a few unpleasant lessons will be learned00:32
wpwrakbut right now, nobody is in the mood to do major changes00:32
DocScrutinizeryou'd either use the floating gnd+opamp design I initially mentioned here, or you (better) use a real oldfashioned trafo00:33
DocScrutinizermantelwellenfilter00:33
wpwrakare there actually galvanic audio separators available as separate devices ?00:33
DocScrutinizeraudio separator trafo00:33
DocScrutinizeryes sure00:33
steve|mgood ol' DI-boxes will work as well, everyone on stage has them ;)00:34
wpwrakheh ;-)00:34
DocScrutinizerDI-box *is*  a audio separator trafo00:34
wpwrakgood. maybe wolfgang can source one in china for a few femtocents00:34
DocScrutinizersteve|m: and everyone on stage knows why :-D00:34
steve|mDocScrutinizer: sure00:35
DocScrutinizerground loops are ubiquitous00:35
wpwraki also wonder how things like DMX (lighting control) treat their ground00:36
wpwrakmy guess would be "not well"00:36
DocScrutinizeraudio via DI, all else is one digital GND level00:36
DocScrutinizerfloating00:36
DocScrutinizerfor DMX devices - uneducated guess of me00:37
wpwrakon stage (well, VJ desk), you may have video in, video out, MIDI, and DMX connected00:38
DocScrutinizerbtw seems nobody knows what "DI" means00:38
wpwrakthere's also ether, but that's probably rarely used there. USB is for things like keyboard and mouse, so no ground issues.00:38
DocScrutinizerwpwrak: yeah, and each of those possibly has its own mains fuse and another phase00:38
wpwrakso what does it mean ? :)00:38
wpwrak(other phase) exactly :)00:39
DocScrutinizerI dunno :D00:39
steve|mdirect injection00:39
wpwrakespecially light should be able to get pretty distributed00:39
DocScrutinizervideo out equipment is light here too00:39
wpwrakyeah, also goes a distance00:40
DocScrutinizervideo in probably less of a problem00:40
wpwraksomething must be :)00:40
DocScrutinizeraudio yes. Another system with antenna on the roof and looong speaker cables, and lots of power pulling up the neutral wire of mains00:41
wpwrakwell, let's hope people approach the big and complex installations slowly and with care :)00:42
DocScrutinizerso connect GND of all these systems and feel happy when there's not Ampere flowing thru it00:42
DocScrutinizerthen delta-I protectors kick in and everything dark :-D00:43
wpwrak*grin*00:44
DocScrutinizeror you got no delta-I breakers in your fuse box, and see some of your equipment literally sending smoke signals00:44
DocScrutinizer(I witnessed such mishaps)00:45
wpwraksweet. let's add some extra ground planes ;-)00:46
DocScrutinizer6mm^200:46
DocScrutinizer;-P00:46
DocScrutinizerwire of course00:47
wpwrakcopper bars linking all the external grounds :)00:48
DocScrutinizerthat's what I meant by 6mm^200:48
kristianpaulhttps://secure.wikimedia.org/wikipedia/en/wiki/All_Watched_Over_by_Machines_of_Loving_Grace  also video link http://www.youtube.com/watch?v=C17zbTTYVME00:49
DocScrutinizerwell a 6sqmm wire is not yet a bar, but almost00:49
Action: DocScrutinizer waves, for good00:49
Action: wpwrak hopes you'll return one day :)00:50
DocScrutinizerpost me a URL linking to m1 schem eventually, please00:50
wpwrakDocScrutinizer: i already did: http://milkymist.org/mmone/rc3_schematics.pdf00:50
DocScrutinizerhope I'll get to move my arse one day00:51
DocScrutinizerooh00:51
wpwrakDocScrutinizer: a view of the pcb is here: http://downloads.qi-hardware.com/people/werner/m1/tmp/front.png00:51
DocScrutinizeryou did?00:51
wpwrakand here: http://downloads.qi-hardware.com/people/werner/m1/tmp/back.png00:51
DocScrutinizerindeed you did00:51
wpwrakhehe :)00:51
DocScrutinizerWTF L1 ?!00:55
DocScrutinizeryeah, and L300:56
wpwrakseems that L1 has already become 0R00:56
wpwrakmy L3 is now, too :)00:57
wpwraksome 80+ will follow soonish, i guess00:57
DocScrutinizerWTF R1100:57
wpwrak;-)))))00:57
wpwrakand C1000:58
wpwrakhah, but that's what wolfson do, too00:58
DocScrutinizerC2400:59
DocScrutinizerC10 is kinda OKish00:59
DocScrutinizerR11 is NOT00:59
DocScrutinizerC10 is actually for creating some sort of floating GND for CD input01:00
wpwrakno clue what R11 does01:00
DocScrutinizersame a chokes in GND do01:01
wpwrakC24 is do-not-place01:01
DocScrutinizerI see01:01
DocScrutinizernevertheless, the pin is NC01:01
wpwrakCD in has probably never been tested01:01
DocScrutinizerI bet01:01
wpwrakyeah, dunno what C24 was supposed to do01:02
DocScrutinizeras this R11 spoils stereo separation, at very least01:02
wpwrakah, that would make sense ;-)01:02
DocScrutinizerL1 (without a buffer C at chip side) is *evil*01:03
wpwrakwell, if the codec has very constant consumption ... ;-)01:07
wpwrakanyway, i think L1 is 0R01:07
DocScrutinizerC122 R47 next button1,2,3 and no ESD prot on signal lines. I'm concerned01:08
wpwrakhmm, fun items01:09
DocScrutinizerI'd make D4 a zener maybe, 2V201:10
DocScrutinizer\o/ two zeners in parallel: D14, D1501:12
DocScrutinizerusually a sure blow01:12
wpwrakhmm ?01:12
wpwrakdunno what the idea is. maybe one as backup for the other. in case there's some problem (soldering or burned-to-open)01:14
wpwrakof course, if the plan is that they share the current, then that probably won't go so well ..01:15
wpwrakat least not in a parallel way :)01:15
DocScrutinizerC168 C170 and the Rs, are they needed?01:16
DocScrutinizeror can C167 C169 take over01:16
DocScrutinizerVBUS on J16 J20 could probably use a fuse01:18
wpwrakat least C168 looks a little odd indeed01:19
wpwrakVREF in an SDRAM ? wow. never saw that.01:19
wpwrakhmm, dunno what to make of it. the data sheet doesn't recommend anything like this01:23
DocScrutinizerI'd spend some OVP/ESDprot for VGA RGB01:24
DocScrutinizerL19 01:26
DocScrutinizerOK, done01:27
DocScrutinizerwhat was that, a burger? ;-D01:27
Action: DocScrutinizer heads out01:27
Action: wpwrak grabs empanadas, beer01:35
rjeffriesfrom my perspective, therte's a lot to like about this board. I agree, it is not "real" yet, but seems to be getting there. If they can sell it for z$50 it will be quite attractive. I do wish it had a small AVR on board to provide a bunch of i/o05:17
rjeffrieshttp://hackaday.com/2011/08/26/raspberry-pi-might-not-be-vaporware/05:17
rjeffriesthe other missing piece of the puzzle is a microsd (aka 8:10) slot. It REALLY mneeds removable memory.05:18
wolfspraulkyak: I read about your misfortune with a 'g' key - too bad. how is the key behaving now?08:09
lekerneldocScrutinizer: L1 (without C) and R11 come from the LM4550 "typical application" schematics from National Semiconductor09:08
lekernelas surprising as it may seem, NS actually recommended L1 be added without capacitor for the first revision of the LM455009:10
lekernelthey rolled it back for the B revision (that's why it's 0) and other codec manufacturers do not recommend it either09:11
lekernelhttp://www.national.com/an/AN/AN-1528.pdf09:11
lekernel"12) L1 was required with the LM4550, but is not required or recommended with the LM4550B."09:11
lekernelwpwrak, all modern SDRAMs (2000+) have VREF *g*09:15
lekerneland it needs to be precise and free from noise; some sources recommend not sharing it across chips09:28
lekernelanyway we'll definitely not risk a SDRAM disaster on 160 boards just to spare 2 resistors and 2 capacitors09:28
lekernelI don't think the zeners in parallel are that bad: their response curve is not square, they have positive temperature coefficient (so the zener that dissipates the most has its voltage increasing until the other zener takes over), and the circuit needs to operate for only a few seconds (and when the user has done a mistake) until the fuse trips09:40
lekernelFYI those zeners can take 6.5A for 1s each09:41
kyakwolfspraul: (the g key) it's not so good.. i haven't tried Werner's suggestion to use a piece of paper (didn't have time and a proper screwdriver)09:55
kyakwolfspraul: speaking about hardware issues, the USB port has become pretty loose over the time09:56
kyaki guess i use it much more frequently than average Ben user09:56
wolfspraulmost likely :-)09:57
kyakwhen i take Ben in hands while it's connected to USB, it is a common that it would disconnect09:57
wolfspraulprobably just mechanically wearing out09:57
kyakyep. And a lot of blindly sticking at night :)09:58
wolfspraulthat's why we need to go all rf in the future. not much we can do about it.09:58
kyakif you are talking about some external rf card, this would cause problems, too10:00
kyakbecause one would have to put it in and take it out every time he puts Ben in a pocket10:01
wolfsprauloh no, of course I mean integrated10:05
wolfspraulpeople underestimate how bad connectors are10:05
wolfspraul'bad' in terms of life expectancy, for example10:05
wolfspraulwell, keep us posted, even if your Ben shows signs of aging now. what can we do. 'g' key, usb connector, hinge probably also getting loose over time...10:06
kyakwolfspraul: i read about the problem with hinge, but luckily it works well for me so far10:22
kyakanother thing i have in mind are silicone "legs" of Ben10:23
kyakthere is only one left out of four :)10:23
kyakeven though i glued it myself some time ago10:24
kyakthey get ripped off - i guess they just have a small area of connection10:24
kyakand probably the last thing are those dust particles under the screen glass - but we discussed it already10:25
Jay7compare to 'legs' of Zaurus SL-C3x00: http://www.penguin.cz/~utx/zaurus/bottom.jpg10:36
wolfspraulyes the feet, true. nice zaurus picture btw!10:49
wolfspraulthe zaurus feet are a little longish, not sure whether that helps or not. I think all feet that are only glued will show the same problem.10:49
wolfspraulas usual Apple seems to be the one with the most thoughtful approach, I have to take my old MacBook apart to see how they did those roundish feet so well...10:50
wolfspraulmy Asus notebook is loosing the glued feet left and right10:50
wolfspraulI'm worried the same will happen to the 4 feet on m1, although roh chose a very expensive 3M thingie10:50
wolfsprauljust checked my 2.5 yr old Asus notebook - 1 out of 4 feet left :-)10:51
Jay7wolfspraul: at least I've not seen reports about lost Z's feet :)10:58
Jay7lost stylus is more frequently reported :)10:59
Jay7btw, that's not my photo10:59
Jay7:)10:59
Jay7I'm looking at Z's feets now11:00
Jay7they are at least half inside back panel11:00
Jay7looks like just glued11:00
wolfspraulsure, glued11:00
wolfspraullemme try some surgery on my old macbook :-)11:01
wolfspraulfor one the feet on the macbook are curve-shaped so they provide very little attack surface from the side11:01
Jay7that's the point (little attack surface from the side)11:02
wolfsprauleven better11:04
wolfspraulthere's a hole in the plastic, below you can see some round piece of metal, probably part of what holds the inside together11:04
wolfspraulthen the rubber/plastic is melted into the hole, so it doesn't sit on a flat surface but first fills the hole, and then only the topmost part stands out (and is curve shaped all around)11:05
kyakwolfspraul: seems like you will be sued for the feet of Ya :)11:07
wolfspraulnah11:07
kyaki think the key is the surface area11:07
kyakZaurus feet's surface is larger than Ben's11:08
kyakso it holds glue better due to a bigger point of contact11:08
Jay7kyak: this is true too12:00
Jay7lower attack surface, larger glue surface, deeper hole12:01
DocScrutinizerlekernel: (NS actually recommended L1) maybe that chip version had a buffer C on another pin? I don't understand the rationale anyway || ( just to spare 2 resistors and 2 capacitors) full ACK, I just asked IF it can be shared. If it's deprecated, then OK || (Zeners) If those Zeners have positive coefficient then it's semi-safe. I'm not sure if some Zeners have negative coefficient though. For the purpose I had no doubts. Anyway if the 12:25
DocScrutinizerpower dissipation is the limiting factor, 2 Zeners in series would be the better choice: half the voltage, double the max I. || Thanks for comenting on it :-D12:25
wpwraklekernel: (sdram and vref) yeah, we live in crazy times :)12:25
lekernellower voltage zeners have negative temperature coefficient12:25
lekernelthe temperature dependence of zener diodes is actually made of two semiconductor effects, one with a positive temperature coefficient and one with a negative coefficient12:26
lekernelthe effect with the positive coefficient takes over when the zener voltage increases12:27
DocScrutinizerwell, I learnt it doesn't always work, the hard way :-D12:28
DocScrutinizerbut yes, low voltage zeners are usually made of simple diodes in series, while 'real' zeners are a special diode operating in inverse direction. Only those have a virual diode antiparallel so that they work as zener or as diode depending on direction of current. So for my suggestion regarding D4(?) - the one antiparallel to optocoupler's LED - to use a zener there was implying a higher voltage zener properties that's probably not 12:36
DocScrutinizeravailable for 2V212:36
DocScrutinizers/virual/virtual/12:41
DocScrutinizeryes, according to datasheet a 1B5339 5V6 Zener shall have a positive coefficient of ~0.5..2mV/°C. Even for 5V0 Zeners of that series (if they existed) the diagram (figure 2) seems to suggest they could have negative coefficient. Sideeffect of protection acainst inverse polarity aka true Zener with low Vf confirmed12:58
lekernelwhy do you want to use a zener for MIDI? I don't understand ...13:00
lekernelin one direction, the optoisolator conducts, in the other direction (which shouldn't happen anyway) the diode conducts13:00
lekernelI actually don't know why that diode is there, but the MIDI standard recommends it... for optoisolator protection maybe13:00
wpwraklekernel: for overvoltage, it may make sense. so you burn the diode, not the coupler.13:01
lekernelboth are cheap and easily replaced13:02
lekernelalso, you'd need to dissipate a bit in the resistors for such things to happen :-)13:03
wpwrakyes, the R may burn first. dunno how "quick" LEDs are in comparison. i've had both go up in flames quite rapidly. (in controlled experiments :)13:05
DocScrutinizera LED will degrade while a parallel Zener of correct voltage could dissipate a lot more energy and keep realy high Vf away from the optocoupler. It was a suggestion if we are free to reconsider and source a nice matching part for D413:10
DocScrutinizerand yes, purpose is OVP from midi, obviously13:10
DocScrutinizerlekernel: midi standards suggest this D4 protective diode for a good reason: LEDs are *very* sensitive to both ESD and OV in reverse direction, and they have a really poor Vr_MAX13:13
DocScrutinizerthe more expensive high-power LEDs and laser LEDs even come with built-in ESD13:15
DocScrutinizerwhich - as always - isn't sufficient for "real life", when e.g. the LED is driven via a connector to the outer world that can inject all sorts of surges13:16
DocScrutinizeron-chip ESD frequently is only for safe handling in fab, not to meet and ESD model13:17
DocScrutinizers/and/any/13:17
DocScrutinizerin general I agree with the rationale that those optocouplers are cheap and easy to replace. I've done this so many times :-D And I always appreciated when they were DIL-8 and on a socket13:20
DocScrutinizer(btw same for the TL074(?) OPAMPS frequently used in consumer grade audio mixer table inputs. Blow frequently due to missing DI box ;-P and often are on socket from fab)13:31
qi-bot[commit] Werner Almesberger: labsw/: first draft layout complete (master) http://qi-hw.com/p/wernermisc/80084df16:18
lekernel_DocScrutinizer: thanks for your comments.17:25
DocScrutinizeryw18:59
DocScrutinizerlekernel_: one more question: why are the SDRAM ADR lines terminated by shared resistor networks (R71-90), while DQ lines have individual termination (R93-133)?19:02
lekernel_they're all terminated with the same resistors19:03
lekernel_I didn't use resistor packs because when soldering them manually it's easy to short circuit pins if you are not very careful19:04
lekernel_ah, no, sorry19:04
DocScrutinizerthat's not the point of my question19:04
lekernel_that's just because address pins and command pins are shared on both SDRAM chips, and the simpler single-resistor solution met signal integrity requirements19:05
DocScrutinizerr71-90 are used shared while R93-112 is left chip, R114-135 right chip19:06
lekernel_the whole thing acts like a single 32-bit SDRAM chip19:06
DocScrutinizeroooh, the data lines differ on both chips?19:06
lekernel_yes19:06
lekernel_the point of having two chips is to get more bandwidth19:07
DocScrutinizerindeed, didn't notice - nevermind19:07
wpwrakDocScrutinizer: Q: assuming an opto-coupler i want to protect against a negative input (i.e., on the diode side), if i don't want to use a diode to short it, like M1 does, but instead have it fail open, what can i use ? one objective is that the minimum voltage for operating the optocoupler should be as low as possible, i.e., about 1.4 V.21:38
wpwrakDocScrutinizer: i thought of using a P- or N-MOSFET, but their gates don't handle much abuse (only up to +/- 20 V) in the ones i found21:39
wpwrakDocScrutinizer: i think i can operate the opto-coupler up to 30 V (with series resistor) before anything blows21:40
wpwrakDocScrutinizer: ah yes, other goals are very little activation current and generally a wide input voltage range. (this is for lab use - shouldn't be too picky about what it's being fed, be able to resist some abuse, and of course not be too complicated/expensive :)21:42
wpwrakDocScrutinizer: any ideas ?21:42
DocScrutinizersorry, I missed to understand your fist post21:45
DocScrutinizerwhat's fail open?21:45
wpwrakDocScrutinizer: open = no current21:45
wpwrakDocScrutinizer: as opposed to the diode in M1, which just shorts the input if reversed21:46
wpwrakDocScrutinizer: e.g., if someone connects something fragile, i don't want to fry it21:46
DocScrutinizerso you want an input that has galvanic separation, operates at <=1.4V, at 1mA, doesn't draw any current when reversed, and is immune to OV up to 30V any polarity?21:47
wpwraki think it will actually work already at ~100 uA :)21:47
DocScrutinizerwpwrak: well, it not shorts the input, as the diode is behind the 22R21:47
DocScrutinizer220R21:48
DocScrutinizersorry, I don't get it21:48
wpwrakyes, but at ~30 V, that would still be a significant current (in my case, I'll use 1 kOhm, but still)21:48
DocScrutinizerif you say "it will actually work already at ~100 uA" it occurs to me you already got a solution and are asking me to make me happy 21:49
wpwrakno no .. my solution has no reverse voltage protection yet21:49
wpwrakthe 100 uA path would be forward21:49
DocScrutinizersorry, you completely lost me21:50
wpwrakmy opto-coupler should transmit enough current for the MCU input on the other end if driven with ~100 uA at ~1.5 V (if my calculations are right)21:51
wpwrak(this is not M1, it's something else)21:52
DocScrutinizeryou sound like those "how could I write a letter using firefox" lusers. Either you got a solution that fulfills your requirements, or you asking for a better solution then I don't see how you can say "it works with 100uA already"21:52
wpwrak"already" referred to the low current, not to the point in time21:52
wpwrakyou wrote 1 mA, i mentioned that 100 uA should do21:52
DocScrutinizer"should" != " i think it will actually work"21:53
wpwraki think "i think" conveys the concept of colloquial "should" quite well :)21:54
DocScrutinizersorry, no idea21:54
wpwrakso, the circuit is as follows: input goes through 1 kOhm/1 W, and into the opto-coupler's LED. on the other side, i have a reasonably sensitive MCU input with a weak pull-up.21:55
wpwrakall want is a means to cut current to, say, <= 1 mA when the voltage is reversed. at up to ~30 V21:56
wpwrakcut current through the opto-coupler's LED21:56
wpwrakfirst thought was to add a schottky in series, but that would increase the minimum voltage at which all this works. and i want it to be sensitive21:59
wpwrakthe objective of the whole exercise is to have a galvanically-separated input i can use to probe digital signals in the lab. that may be a power rail, some 3.3 or 5 V logic, or maybe something 1.8 V-ish. the more i can do, the better22:01
wpwrakit may also be something not exactly at the respective rails, e.g., an output that's already loaded22:02
DocScrutinizeruse an optocoupler opamp, get a 2nd PSU for the input22:02
wpwrakthat brings me to yet another objective: simple and cheerful ;-)22:03
DocScrutinizeryou can do sth like 0.1..200V with it22:03
wpwrakyes, that would be great. but too complex.22:03
DocScrutinizerI want a pony22:03
wpwrakindeed :)22:03
DocScrutinizeropto-opamp us as simple as it gets, for your requirements22:04
DocScrutinizers/us/is/22:05
DocScrutinizeryou can have Z of several 10..100kR, and after that it's easy to clamp any OV22:05
wpwraki want the input side to be unpowered22:06
DocScrutinizerwon't fly22:06
wpwrakso there's no component that can simply cut if the voltage is reversed ? sort of like an ideal diode22:06
DocScrutinizerno22:07
DocScrutinizernot if you also want OVP22:07
wpwrakokay, two components. one to cut, the other to protect it from harm. in fact, three in total: the series resistor to limit the current, the X to cut reverse voltage, and maybe a third to protect X from ESD22:10
DocScrutinizerOV not ESD. For "ideal diodes" OV starts at maybe 10V22:11
DocScrutinizerit won't fly22:11
wpwrakif you go over the maximum design voltage (~30 V in my case), all bets are off anyway. but 30 V should be enough for this kind of use.22:11
DocScrutinizerI already pondered multi-component solutions22:11
wpwrakhmm, so the shorting diode is the only reasonable choice ?22:12
wpwrakor a big warning label that you'll get to swap the coupler if voltage is reversed and above TTL levels (the coupler can handle 6 V)22:12
DocScrutinizereven then your claims of 0.1mA@1.5V and safe up to 1mA@30V is hard to achieve22:13
wpwrakthe MCU needs only about 50 uA. the current transfer ratio of the coupler is 50% or better. thus 100 uA should do, no ?22:14
DocScrutinizerno, as it needs a certain minimum current to operate afaik22:14
DocScrutinizereven then I don't see how you get from 0.1mA@1.5 to 1@3022:15
wpwrakno no, 30 mA at 30 V22:16
DocScrutinizeralso if the coupler needs 1.5V Vf, how much is your series R then?22:16
wpwrak1 kOhm. the coupler needs 1.4 Vf(max)22:16
wpwrakthe data sheet only shows If down to 1 mA :-(22:17
DocScrutinizertoldya22:17
DocScrutinizerthat's what I had in my storage about optocouplers22:17
wpwrakif i extrapolate (looks linear in the area), then i may be good up to ~200 uA22:18
wpwraki'll have th measure what it really can do22:19
DocScrutinizersounds like a "let's see if it works with components of this production week" design22:20
wpwrakwell, i have a bit of tolerance. of it works reliably at 0.5 mA, i'm happy22:21
DocScrutinizerbtw I've actually seen couplers that produced false positives when you opened the case and light fell on them22:21
wpwrak;-)))22:21
wpwrakyes, i've heard of that22:21
wpwrakmine have black plastic, that's already a bit reassuring22:21
DocScrutinizerthose were in a design similar to yours, ultralow currents, very high impedances22:22
DocScrutinizerthose had as well black plastic22:24
wpwrakdon't buy those made by murphy inc. ;-))22:26
DocScrutinizernow imagine you got them on a frequency generator for a 2MW rotary current motor22:26
wpwrakwell, the thing will have a reasonably light-tight case22:26
wpwrak;-)))22:26
DocScrutinizerwill make it safe until you walk by with a gas lantern or petromax, anything that has a incandescent mantle22:28
wpwrakits output should eventually be able to handle mains (not in this version yet, though), so a bare board wouldn't be so nice ;-)22:28
wpwrakok, 5 mm aluminium bar over the couplers ;-)22:29
DocScrutinizerthe more I'd try not to drive it to the limits of the components, rather make it safe and with sufficient headroom in all parameters22:29
wpwrakand an emergency ration of liquid nitrogen somewhere :)22:29
wpwrakyes, of course22:29
DocScrutinizerlimits to the microscopic end are as unsafe a ground to stand as are ABS MAX rathings22:30
wpwrakthe intended voltage range for the couplers is about 1.8 V - 15 V. minimum current at 1.8 V about 0.5 mA. i'd like to be compatible with 1.8 V logic.22:31
DocScrutinizerhmm, that sounds somewhat more sane22:32
DocScrutinizerstill I don't know of any couplers designed for <1mA22:32
DocScrutinizerdoesn't mean you might not be able to find some22:33
wpwrakthe relays (two of them) will be able to do more. eventually, i want to be able to switch mains. but i need various different components for this to be even remotely safe.22:33
DocScrutinizerthe sanest method is to exploit random 50kHz and a small siferrit transformer to get sth like 3V@1mA galvanically separated power supply for the input stage 22:35
wpwraki'll just measure what the coupler can do. if the curve can be extrapolated ... fig. 5, page 11 of http://media.digikey.com/pdf/Data Sheets/Lite-On PDFs/LTV-8x6.pdf22:36
wpwrak... then i should hit the zone where If*CTR is too low around 200-300 uA22:36
DocScrutinizeryou quite usually mustn't extrapolate curves of datasheets22:36
wpwrakit looks nice and flat for almost half the range before ;-)22:37
wpwrakbut yes, there can be surprises :)22:37
DocScrutinizerthere WILL be surprises, as the fab doesn't trst for those extrapolated component properties22:38
DocScrutinizertest*22:38
wpwrak(transformer) waaaay to complex :)22:38
wpwrakof course. but that doesn't mean they're per se unusable. lot of things are operated at points well outside what the data sheets specifies. sometimes, there's not even a sane way to avoid this ...22:39
DocScrutinizeryeah, that's what they told me about getting and keeping a pink pony22:39
wpwrakyou should have asked for a turquise one. or just a standard unicorn :)22:40
wpwrakanyway, measurements will show if extrapolation looks reasonable or not. and no reverse voltage protection then.22:42
wpwrakthe chip are cheap anyway :)22:43
DocScrutinizerI'd go for a max373 or whatever they are called, and then build proper signal converter in front of that22:43
DocScrutinizereven more "complex", but proper22:43
DocScrutinizerin this context signal converter could mean sth as simple as a transistor with a huge resistor on the basis22:44
DocScrutinizerto protect the whole thing against OV22:45
wpwraknaw, disposable optocouplers will be the solution22:46
DocScrutinizerif 100k or 1M is still too low an impedance for reverse polarity case, you can use a schottky then22:46
DocScrutinizermeh22:46
wpwrakokay, thanks ! now i think i know what i'm up against22:50
wpwraknext stop: sockets :)22:50
--- Sun Aug 28 201100:00

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