#qi-hardware IRC log for Thursday, 2016-03-24

DocScrutinizer05Ist nur ein Zahlendreher: statt 23.03. sollte es heißen 32.03. , also 1.April 16:17
DocScrutinizer05wpwrak: it's like injet printers: the printer is for almost free, while the liquids cost more than gold16:18
DocScrutinizer05also I really doubt they can do proper exact printing from a smartphone display, the distance to the light emitting surface is too high and the directivity of the emitted light too low16:32
DocScrutinizer05just place a sheet of paper on top of your smartphone display and see what you can read16:33
DocScrutinizer05that's pretty much fake, it can produce stuff but it's useless nevertheless, due to lack of resolution and precision16:34
wpwrakall those STL printers use expensive liquids. so that far it would be normal.16:37
DocScrutinizer05sure16:38
wpwrakgood point about blur. they would need a directional filter, further reducing light output.16:38
DocScrutinizer05I got a good misconception as example that you constantly see in movies, where the display of a monitor in front of the actor supposedly gets reflected on his face. Mere idiocy16:39
DocScrutinizer05you physically can't get a resolution much better than thickness of the display glass16:41
DocScrutinizer05unless you'd use a lens16:41
DocScrutinizer05or a parabolic mirror16:42
DocScrutinizer05and without such focusing element you have the problem of a fairly thick layer of half-cured material on the surface. Well you can finish curing at daylight I guess16:43
DocScrutinizer05nah, it's a scam like those 60ct plastic calipers or the 10cm^2 solar panels to charge your phone16:45
DocScrutinizer05it *theoretically* works *somehow*, but in RL it's mostly useless for the purpose you thought it was meant for16:46
wpwrakyeah, i guess we'll have to see what crawls out of this. chances do indeed seem rather high that whatever it does won't be overly convincing.16:47
DocScrutinizer05yep16:47
wpwrakone problem is that all the examples you usually see are "art". so it's hard to tell how good the reproduction of the CAD model actually is.16:48
DocScrutinizer05they do that on purpose ;-)16:48
wpwrakof course :)16:49
DocScrutinizer05honestly, putty is cheaper, faster, and yields better more precise results16:50
wpwrakmicrosoft dabbling at AI (comments are worth reading too, especially the one about the japanese competition): http://www.theregister.co.uk/2016/03/24/microsoft_ai_goes_troll/18:46
wpwraki kinda wonder what they fed that poor both with to make such statements. i don't think even old simpleton eliza could be made to say such things18:47
wpwraki would love to be a fly on the wall in their next project meeting :) here is more: http://uk.businessinsider.com/microsoft-deletes-racist-genocidal-tweets-from-ai-chatbot-tay-2016-318:52
larsctwitter as usual I guess19:12
DocScrutinizer05AI made by fools19:54
DocScrutinizer05"hey, let's mix this flu virus with a hepatitis virus and see what happens"19:55
DocScrutinizer05or "I got a great idea we should see how that pans out: control the 747 in autopilot mode by random signals through a neuronal network and the pilots give feedback to train the NN"19:58
DocScrutinizer05pretty much symptomatic for M$ developments20:00
DocScrutinizer05develop some arbitrary crap without clear design rules or goals, and as soon as it doesn't segfault once every 5 minutes, roll it out as giveaway and see what's the feedback20:01
wpwrakthe engineers that blew up chernobyl must feel pretty good about themselves when reading such stories :)20:01
DocScrutinizer05yeah, sure, that was simple human failure, not utter idiocy like found at M$20:02
DocScrutinizer05WOW,  >> <p>Damit Sie Ihren Benutzeraccount weiterhin nutzen können, bitten wir Sie darum Ihre Daten erneut zu verifizieren.</p> <p><a href="http://paypal.de-ll.cc/deutschland/sicherheit/kunde=237846/" <<20:05
DocScrutinizer05a pity many users will actually not recognize this as the phishing it is20:06
DocScrutinizer05LOL, check the https cert20:08
DocScrutinizer05Police (phonecall): "our internet crime team doesn't work on weekends"  ROTFL  - PayPal at least thanked me for forwarding the mail to spoof@pp20:36
DocScrutinizer05wondering how to take down paypal.de-ll.cc immediately20:37
DocScrutinizer05abuse@de-ll.cc ?20:38
DocScrutinizer05Sponsoring Registrar: BIZCN.COM, INC.   ohmy20:41
DocScrutinizer05https://www.icann.org/en/system/files/correspondence/serad-to-guanghui-08may14-en.pdf20:42
DocScrutinizer05I wish ICANN would force .cc top level DNS to nuke all URLs granted to/by BIZCN20:52
DocScrutinizer05actually the URL is at cludflare.com20:55
DocScrutinizer05cloudflare even20:55
DocScrutinizer05>>Two of ISIS' top three online chat forums are guarded by CloudFlare but U.S. law enforcement has not asked them to discontinue the service<<21:00
DocScrutinizer05>>An October 2015 report found that CloudFlare provisioned 40% of SSL certificates used by phishing sites with deceptive domain names resembling those of banks and payment processors<<21:00
DocScrutinizer05>>On November 2015, Anonymous discouraged the use of CloudFlare's services, following the ISIS attacks in Paris and renewed accusation of providing help to terrorists.[37] CloudFlare responded by calling their accusers "15-year-old kids in Guy Fawkes masks" and saying that they consult "actual anti-terrorism experts" and that they "abide by the law"<< ok, maybe they are at the edge to doom now, let's hope so21:02
wpwrakDocScrutinizer05: how about this ? http://downloads.qi-hardware.com/people/werner/anelok/tmp/time-backup-draft-20160324.pdf21:13
wpwrakvalues subject to more detailed calculation. the 680 kOhm are a common high resistance value in the design, so i try to reuse this21:15
wpwraklarge R3 (you suggested 100 kOhm) is to keep the current through the circuit low. needs experimental verification21:17
wpwrakbtw, i had another idea: if VDD decays quickly, an RC between GND and VDD might be an even simpler and much more efficient solution: decaying VDD would "switch automatically" between charging (rather, holding), and discharging21:19
wpwrak(decays quickly) or, rather, reasonably predictably21:21
DocScrutinizer05yes, but predictability is possibly low21:26
DocScrutinizer05and you can't control the moment of timestamp and writing it to persistent storage - which I'd suggest to do *before* VDD drops21:28
DocScrutinizer05the pdf looks good. I don't get what's Tp or why it comes in there21:32
DocScrutinizer05or rather, Td and Tc are not really needed21:34
DocScrutinizer05yes, when you want to do this calibrate-at-activation-time scheme, you need to store the Vh/Vl/delta-T calculation result and Vs and the timestamp of Vs21:36
DocScrutinizer05and I'd use exponetial curved instead linear charge / discharge21:36
DocScrutinizer05in diagram21:36
DocScrutinizer05this will cause Td to be very short in comparison to Tc21:38
DocScrutinizer05ooh, Charge: ADC H21:40
DocScrutinizer05:-)21:40
DocScrutinizer05yes, the circuit looks decent21:41
DocScrutinizer05((which I'd suggest to do *before* VDD drops)) maybe you got some pre-fail contact that opens a 100mS *before* the battery loses connection?21:44
DocScrutinizer05ideally you use a bayonet peg to close the battery bay, and the pre-fail contact gets engaged as soon as bayonet is turned out of "lock" position21:45
DocScrutinizer05or simply use a metal screw with a spring for minus battery contact in the middle, and a non springloaded pre-fail contact that closes when screw is completely seated21:47
DocScrutinizer05so you have 4 or 5 turns of the screw between pre-fail and loss of Vbatt21:48
DocScrutinizer05plenty of time for doing the charge&timestamp dance21:48
DocScrutinizer05I got a procedure simplification: at pre-fail probe voltage [1] of capacitor and take timestamp1. Then do cyclic ADC until voltage dropped a exactly N (e.g. 10) digits on ADC (should take some 20 to 50ms, adjust N accordingly), then take timestamp2 and store [1] and timestamp1 and timestamp2 (or exact diff to ts1) to storage, keep capacitor discharging22:07
DocScrutinizer05or you assume Vmax to be sufficiently precise after long charging, and you simply take timestamp and stop charging at pre-fail, and the whole calibration thing is a one-time process done at device setup22:10
DocScrutinizer05after all VDD is regulated by the booster, no?22:17
DocScrutinizer05you want a ADC on battery, to find out about remaining capacity22:18
--- Fri Mar 25 201600:00

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