Back
[04:17:33] <rue_bed> I wonder, how many positioning PID loops can an AVR can do at once...
[08:25:39] <twnqx> sorting chips and downloading their data sheets is so much fun...
[08:25:58] <twnqx> so much stuff i never knew about.
[08:42:14] <twnqx> ah, datsheets... where would i get a 10µF _inductor_...
[09:24:12] <rue_house> 10F inductor, hmm
[09:24:20] <rue_house> thats a lot of self-capacitance
[09:28:20] <twnqx> it's a typo, sould have been 10µH
[09:29:10] <rue_house> oooooooh!
[09:29:11] <rue_house> :)
[09:29:47] <rue_house> hmm, 2.2F caps are about $2 ea
[09:30:17] <twnqx> also, does your clinet not display the "µ"?
[09:31:03] <rue_house> ?
[09:31:13] <twnqx> mu symbol. µ.
[09:32:01] <rue_house> ?
[09:32:02] <rue_house> that one?
[09:32:10] <twnqx> i only get a question mark :P
[09:32:21] <twnqx> seems we use incompatible encodings *g*
[09:32:23] <rue_house> huh
[09:33:13] <twnqx> ... datasheets. i really wonder if this opamp will do anything with +vcc and -vcc both at +7.5V, as shown in the reference circuit.
[09:33:29] <rue_house> dk goes down to $3.1
[09:34:07] <rue_house> as a kid, I always thought it was my fault the circuits in books didn't work
[09:35:01] <rue_house> even when I managed to find the obscure parts like the infamous 2.0k resistor
[09:35:35] <twnqx> don't think 2x 1k is that obscure...
[09:35:43] <rue_house> its hard when you need to go thru old circuit boards till you find the value your looking for
[09:36:07] <twnqx> and you didn't consider going for 2 1k resistors?
[09:36:37] <twnqx> but yeah, sometimes resistor values are... unreal
[09:36:49] <twnqx> i like the 12.7k i needed for something
[09:37:06] <rue_house> not always, before I was used to the standard values some resisotrs were just REALLY hard to find
[09:37:07] <twnqx> or 4.42k
[09:37:31] <twnqx> so yeah, i have 4.42k 1% resistors in my stash.
[09:37:46] <rue_house> I have a program I wrote that goes thru standard series and parallel combos to find a set that makes what I want
[09:38:22] <twnqx> i don't see the point any more, with resistors of arbitrary values in cent prices
[09:38:36] <twnqx> 4.47k
[09:38:52] <twnqx> eh... wrong keyboard
[09:39:30] <rue_house> lowest error:
[09:39:31] <rue_house> hit, error = 0.139417% vals = 5600.000000 and 22000.000000 at 4463.768066
[09:39:47] <rue_house> in parallel
[09:39:56] <twnqx> obviously :)
[09:40:30] <twnqx> funny, mouser has 336 hits for 4.42k resistors
[09:40:53] <rue_house> just sayin, standard 5% values
[09:41:16] <twnqx> 0.1%, single piece, 0.503 euros
[09:41:29] <rue_house> today in going to be awefull
[09:41:41] <rue_house> I didn't actually fall asleep last night
[09:42:18] <twnqx> oh look, single quantity 0.181€
[09:42:50] <twnqx> maybe still usefull for breadboarding with leaded resistors
[09:43:37] <rue_house> I wonder what update rate I need a PID position controller to operate
[09:43:44] <twnqx> hm. 0.02% precision gets expensive :D
[09:43:53] <rue_house> I wonder if 300 - 600Hz is enough
[09:44:15] <twnqx> i... don't even know what that is :P
[09:44:22] <rue_house> hmm, hobby servos are 50-60Hz
[09:45:15] <rue_house> I wonder if an atmega32 can handle 8 PID loops under serial control
[09:46:27] <rue_house> the trick to today will be NOT thinking about work
[09:46:46] <twnqx> woah, an SGS Thompson LM741CN
[09:46:51] <twnqx> to they even still exist?
[09:46:58] <rue_house> hey, the 555 exists
[09:47:08] <rue_house> good luck finding an LM37 tho
[09:47:14] <rue_house> 37?
[09:47:25] <rue_house> or was it an LM3
[09:47:31] <rue_house> HV op-amp
[09:48:18] <twnqx> i'm just sorting chips :P
[09:48:29] <twnqx> and after 4 drawers with voltage regulators, opamps are nxt
[09:48:36] <rue_house> ah
[09:48:47] <rue_house> I been there
[09:48:54] <rue_house> "this part number looks intersting..."
[09:49:49] <twnqx> i wonder if i can store a TI LM741CN datasheet and eventually use it for n SGS LM741CN :P
[09:50:44] <rue_house> I made a php page to search for pdfs on my fileserver
[09:52:05] <rue_house> zippo:/morfiles/doc/pdf# find |grep pdf|wc -l
[09:52:05] <rue_house> 6329
[09:52:12] <rue_house> :/ not much, but there is a lot there
[09:52:32] <twnqx> grep | wc :(
[09:54:09] <rue_house> |emacs ##ugly LISP single-liner ##
[09:55:51] <twnqx> find|grep -c pdf!
[09:56:27] <twnqx> (or find -iname "*.pdf" | wc -l
[10:01:49] <twnqx> oh. opto couplers.
[10:07:20] <twnqx> more opto couplers Oo
[10:10:01] <twnqx> finally, a chip mouser doesn't know...
[10:10:13] <twnqx> does OKI even still make ICs?
[10:11:22] <twnqx> hm.
[10:14:50] <twnqx> ok. it's a 64kx1 dram
[10:14:59] <twnqx> compatible to 4164...
[10:22:29] <twnqx> ... a wiki from... 1985? i kinda doubt that
[11:09:26] * twnqx stares
[11:09:40] <twnqx> i apparently just found something that looks like a bus pirate clone
[11:09:46] <twnqx> that i didn't even know i own?
[11:35:19] <julius> hi
[11:35:22] <julius> in this mosfet vgs vs ids graph:
http://postimg.org/image/x7okv05xn/ is v_gs 0.15 for 175C ?
[11:35:32] <julius> for 10a?
[11:35:46] <julius> v_ds i mean = 0.15 for 10a at a gate voltage of 4.5v
[14:19:40] <z3t0> hello, I am trying to figure out how long ADC conversion should take on an atmega328p, currently it takes about ~120us. Is that an acceptable speed or can I expect to optimize that further
[14:20:21] <z3t0> I would prefer the fastest possible speeds, can I maybe reduce resolution to 8bits to increase the read speed? Speed is my main priority as this is a moving robot, accuracy is not nearly as important
[14:21:31] <z3t0> I used ADLAR and ADCH to read only the 8-bit result, however the conversion seems to take the same amount of time regardless of resolution.
[14:22:25] <z3t0> http://maxembedded.com/2011/06/the-adc-of-the-avr/ This suggests a 13 - 260 us... so I am wondering How i can get it closer to the 13 xD
[14:26:20] <z3t0> Hmm I managed to half it to 50us by changing the prescaler from 128 to 64, resulting in 250Khz, but my understanding is that that is too fast..?
[14:29:33] <z3t0> Ok now I have it at 13us at 1Mhz which itself is unrecommended, but keeping in mind that accuracy is really not important I'll take a look at what sort of results I seem to be getting
[15:15:36] <hetii> hi
[15:30:11] <hetii> e
[15:30:13] <hetii> re
[15:31:01] <hetii> Anyone use Saleaellogic analizer as a Jtag interface with Urjtag ?
[15:41:53] <eszett> hi
[15:42:32] <hetii> ;)
[15:43:36] <eszett> I have a little problem with soldering. My normal temperatur (with solder iron & with hot air gun) ist 300 Celsius. With this temperature the solder wick doesnt suck any solder. I tried different wicks, no effect. I tried to increase temp to 400 degree, and then the solder wicks soaks, but my chips, pcb etc. get destroyed.
[15:44:02] <eszett> 300 should be plenty enough for solder wick to work, isnt it?
[15:45:13] <hetii> suppose 300 should be fine but i`m not an expert
[15:45:33] <eszett> for soldering a TQFN44 i absolutely need a solder wick, since my solder sucker doesnt work on this small scale ..
[15:46:41] <eszett> with a solder oven, people use something like 260 (or something?) which is even less than 300
[15:46:53] <twnqx> i don't really need wick for tqfps
[15:47:06] <twnqx> what you really need is less solder
[15:47:10] <twnqx> and more flux.
[15:47:51] <eszett> twnqx: i know that. but i tend to put too much solder on. and this for a reason. When there is just right enough solder on it, i cant see anything, and put more on it. and then there is too much.
[15:48:27] <twnqx> well, here's something i do if i really need wick: drown the wick in flux first.
[15:48:35] <eszett> oh..
[15:48:37] <eszett> good idea
[15:48:39] <eszett> i will try that
[15:49:43] <twnqx> i also use higher temps, around 360°C for lead-free solder
[15:49:44] <eszett> And i guess my sight is too bad. I need better lighting and magnifiers
[15:50:05] <eszett> 360Degree on a atmega32u4?? wouldnt it destroy it?
[15:50:24] <twnqx> more a factor of time than anything
[15:50:36] <twnqx> i usually take a wide flat tip, 2-3mm
[15:50:45] <eszett> im using that tip too
[15:50:55] <twnqx> and put solder only on the front, narrow tip
[15:50:59] <eszett> my solder is lead-free balver zinn
[15:51:07] <eszett> hmm i see
[15:51:13] <twnqx> and then put that just on the pins and let the flux move it were it should be
[15:51:33] <twnqx> also nicht eine flache seite, sondern nur die schmale kante vorne :P
[15:51:52] <eszett> das ist alles so winzig..
[15:52:09] <twnqx> "geht so"
[15:52:14] <eszett> furchtbar..
[15:52:18] <twnqx> try 0.4mm pitch vssop :P
[15:52:40] <eszett> ne muss nicht sein. was benutzt du für eine lichtlupe wenn ich fragern darf
[15:52:49] <eszett> gar keine??
[15:52:51] <twnqx> zum löten: keine
[15:53:02] <eszett> im keller ist bei mir viel zu dunkel
[15:53:04] <twnqx> zur nachkontrolle: so eine von reichelt
[15:53:30] <eszett> hmm sowas werd ich mir mal holen
[15:53:39] <twnqx> haben aber sicher schon wieder das angebot gewechselt.
[15:53:52] <eszett> wenn das Lot ne andre farbe hätte wäre es für meine augen soviel einfacher..
[15:54:31] <eszett> aber silver auf silber ist halt scheisse. daher kam mir die idee ich könnte ja vielleicht die PCB mit goldüberzug bestellen, das kostet nur paar euro mehr und dann hätt ich nen guten Kontrast von Lot auf Lötstelle
[15:54:57] <eszett> Silber auf Gold kann man mit blossem auge besser sehen, so dass man auch beim auftragen von nur wenig Lot sofort weiss: da ist Lot drauf
[15:54:58] <twnqx> http://www.reichelt.de/Werkstatt-Lupenleuchten/LL-FP-45273/3/index.html?&ACTION=3&LA=2&ARTICLE=151188&GROUPID=4515&artnr=LL+FP+45273 in der art, mur noch mit leuchtstoffröhre
[15:55:06] <twnqx> und ja, nur noch ENIG für mich :)
[15:55:18] <eszett> aus diesem grunde den ich beschrieben hab??
[15:55:23] <twnqx> nö
[15:55:26] <eszett> hmm
[15:55:32] <eszett> also bringt es was mit dem kontrast oder ned?
[15:55:47] <twnqx> enig lagert besser. wenn ich in nem jahr nochmal eine platine bestücken will ist das mit hasl oft schwer weil schon oxodiert
[15:56:11] <eszett> kannste doch mit flux entoxidieren wieder
[15:56:33] * eszett boah 77 euro für die lupe..
[15:57:13] <eszett> naja ok ich weiss dass ENIG allgemein besser ist.
[15:57:19] <eszett> aber bringt es auch mit dem kontrast was zum lot?
[15:58:00] <twnqx> hab ihc noch nie drüber nachgedacht... aber würde ich schon sagen
[15:58:01] <eszett> für 77 euro hol ich mir nen gebrauchte s mikroskop n:-)
[15:58:10] <twnqx> gibt auch fpr 47 eine
[15:58:15] <eszett> dann ist es das was ich brauche ENIG
[15:58:52] <eszett> hm ja
[16:35:45] <lorenzo> I think I've written the simplest library ever for ds1307 :D
[16:35:50] <lorenzo> and ds3231
[16:36:03] <lorenzo> https://github.com/lcafaro/avr-ds3231/blob/master/libs/ds3231.c
[19:56:43] <Casper> hmmm look like the clock jitter on the tiny85 is very negligeable...
[19:57:16] <Casper> to the point where I wonder why people even mention it... or maybe it is because the supply is stable and nothing else happend?
[20:54:50] <cehteh> Casper: Jartza did his vga stuff with that an the display was very fuzzy, for myself here i am more concerned about drift which ist quite a lot, but can be somewhat compensated
[20:56:20] <Casper> I was under the impression that the clock was very unstable
[20:56:37] <Casper> speed drift isn't an issue for me, but "lots of speed change" is
[20:56:59] <Casper> now, need to check the temperature effect on the RC clock
[20:57:21] <Casper> I decided to go PWM for data transfert, like 1wire does
[20:57:57] <Casper> so need to count the cycles... and wanted to know what to expect
[20:59:39] <cehteh> depends on what you are doing, jitter and drift are quite noticeable
[21:00:43] <Casper> yeah if I go to the second smallest time division possible on my scope, I can see about a 2 pixel instability on the side of the screen
[21:00:57] <cehteh> i parsing cppm signals here (rc transmitter sum signal) which are some kind of PWM and the default clock was quite inprecise
[21:01:07] <Casper> so it is noticeable, but way less than what I expected, so I wonder if my test is valid
[21:01:20] <cehteh> but with syncing to the frames its quite good enough
[21:01:23] <Casper> I need about 15kbps
[21:02:21] <cehteh> cppm is pulses from 1000-2000 microseconds length
[21:03:21] <cehteh> the better you measure them the more exact is the result
[21:04:03] <cehteh> that's not *that* much
[21:04:53] <cehteh> but ideally you want it 1µs precise ..
[21:05:04] <cehteh> 8mhz tiny84 here
[21:07:02] <cehteh> timer /8 would be 1µs (thus according to niquist law, i can only measure 2µs) .. i am using timer /1 here now in test config, later /8 is ok because the project doesnt need full resolution
[21:27:29] <Casper> timer is unavailable
[21:27:51] <Casper> unless I... hmmm software pwm...
[21:37:22] <Casper> hmmm using software pwm might actually be simpler... if there is an input capture avail...
[21:38:09] <cehteh> tiny85 has no input capture
[21:38:23] <cehteh> but why dont you have any timers available?
[21:39:05] <cehteh> i mean even if you use both for pwm, you can still use one as clock
[21:39:43] <cehteh> so far i am using only one timer for anything, intentionally to keep as much timers free for other purposes
[21:40:23] <Casper> I use 3 PWM channels
[21:42:15] <cehteh> as long you dont modify the pwm frequency, or even better stay at max, it should be easily possible to use one timer for multiple purposes
[21:42:54] <cehteh> even with modifying pwm freq its possible but a bit more complicated to count on overflows
[21:43:37] <Casper> the code will be quite simple actually...
[21:43:46] <cehteh> ok mine is not :D
[21:43:48] <Casper> it's an addressable RGB led controller...
[21:44:04] <cehteh> you can use the usi too
[21:44:05] <Casper> get data via a 1 wire like signal
[21:44:21] <Casper> the usi on the 85 is useless
[21:44:26] <Casper> it don't do async
[21:45:03] <cehteh> i am planning for some pwm like bus too
[21:45:32] <cehteh> maybe with different implementations, PWM, USI, bitbanged, input capture
[21:46:59] <Casper> the code will be: mesure pulse width. if > x then reset else if > y then 1 else if > z then 0 else ignore. if 32 bits has been received, read first 8, does it match device ID or is ID == 0? if so set the 3 PWM to the 3 other bytes
[21:49:23] <cehteh> maybe with some tricks one can use the usi, running at high speeds, captureing pulse length, not bits
[21:49:50] <cehteh> 10000000 11000000 11100000 ....
[21:50:08] <cehteh> thats what i have in mind at lesst
[21:50:17] <Casper> I wonder if it have hardware 1 wire hmmmm
[21:50:25] <cehteh> no avrs have that
[21:50:43] <cehteh> iirc
[21:51:06] <cehteh> do you 'like' 1-wire or more 'exactly' one wire?
[21:51:17] <Casper> I have whatever I'll code
[21:51:33] <Casper> and it support 3 wires and 2 wires
[21:52:00] <cehteh> my plans are to make a 'like' protocol, 1 wire like bus, no master, token passing
[21:52:12] <Casper> same, a like
[21:52:22] <cehteh> transmitting clock and 2 bits per clock
[21:52:42] <Casper> idle at 0 instead of 1 (save the wire from damage/wear since it will be burried)
[21:52:51] <Casper> unidirectional
[21:53:07] <Casper> the unidirectional simplify alot the stuff
[21:53:29] <cehteh> leading edge is clock, duty cycle encodes 2.5 bits (one symbol for error correction)
[21:53:57] <cehteh> ah i want completely n:m anyone can talk to anyone
[21:54:03] <Casper> I most likelly will transmit at 19V instead of 5V, divide at the avr, so will help against offset in the ground wire too
[21:54:33] <cehteh> for longer lines you may consider differential signals
[21:54:47] <Casper> I'm also considering to add a 1 bit checksum
[21:54:52] <cehteh> higher voltage isnt always good i think
[21:55:08] <cehteh> better decouple DC with some capacitor
[21:55:19] <cehteh> but mhm not that good either
[21:55:31] <Casper> up to 100ft of power wire
[21:55:55] <cehteh> i think you really want differential signaling
[21:56:13] <Casper> 1 wire is said to work to over 1000ft
[21:56:48] <cehteh> depends on noise on the environment
[21:56:54] <Casper> what I'll do is: take about 100ft of wire, drop in water, ground water and check what I get on the end
[21:57:17] <cehteh> i am already scared then that it may catch induction and noise destroying the inputs of the avr's .. add some protection
[21:57:59] <cehteh> my plans are like 1k at the input pin, and 4.7 zener to ground at least
[21:58:07] <cehteh> dunno if thats good enouh
[21:58:32] <cehteh> looked into bus isolators but they are expensive and hard to get
[21:58:40] <Casper> that's also why I plan on using the divider
[21:58:41] <cehteh> esp when you want bidirectional ones
[21:59:08] <cehteh> and how do you drive the voltage?
[21:59:45] <Casper> the divider may be a 15/5k divider
[21:59:55] <cehteh> the higher the voltage the more capacitance problems you get
[21:59:59] <cehteh> well .. just try it
[22:00:08] <Casper> push-pull driver
[22:00:13] <cehteh> but also add some zener to protect the input
[22:00:38] <Casper> there is some input diodes in the avr
[22:00:51] <cehteh> if you need a driver on the line then you could use some ready to use rs485 transmitter already
[22:01:09] <cehteh> then you have differential signals and proper protections
[22:01:17] <cehteh> and higher voltage levels
[22:01:33] <cehteh> the input diodes on the avr dont protect much
[22:01:35] <Casper> and that add money :D
[22:01:49] <Casper> they can handle a few mA
[22:01:51] <cehteh> what push/pull driver do you use?
[22:02:20] <cehteh> i wont be surprised if you can get a serial transceiver for comparative prices
[22:02:41] <Casper> a npn/pnp
[22:02:56] <cehteh> just try
[22:03:03] <cehteh> may or may not work :D
[22:03:24] <Casper> yup
[22:03:53] <Casper> may actually go 12V for the driver, just enought so there won't be issues with the voltage drop/raise of the ground
[22:05:00] <cehteh> i am more worried about noise and signal quality than level
[22:05:06] <Casper> I even considered pure PWM
[22:05:15] <cehteh> ok you increase the snr with higher voltage
[22:05:22] <Casper> send 8 bits address, then send PWM :D
[22:05:33] <Casper> but would be a pita to deal with
[22:06:10] <cehteh> so analog :D
[22:06:39] <cehteh> but might be not that bad
[22:06:56] <cehteh> how fast do you want to updates?
[22:07:17] <Casper> 20Hz
[22:07:32] <cehteh> have you looked into the ws2812 protocol?
[22:07:41] <cehteh> its really pita
[22:07:58] <Casper> I might decide to drop to 8 devices and go 30Hz
[22:08:00] <Casper> nope
[22:09:26] <cehteh> http://hackaday.com/2014/12/09/digging-into-the-apa102-serial-led-protocol/
[22:10:15] <cehteh> send shitloads of data, time critical, no sync .. no checksum
[22:11:30] <cehteh> ah bits are encoded as diffent length pulses, clk+bit per cycle
[22:11:56] <Casper> that's what I'm doing, different pulse length for reset/start, 0 and 1
[22:12:16] <Casper> and of course, I'll also use the WDT to turn all off if something happend
[22:13:15] <cehteh> cant you just use the USI and normal RX .. somehow sync clock every now an then
[22:13:38] <Casper> no pin available
[22:14:06] <Casper> usi require a minimum of 2 pins
[22:14:08] <cehteh> i pin only
[22:14:10] <cehteh> RX
[22:14:22] <Casper> wait.. I think I do have 2 pins avail
[22:14:26] <cehteh> no you can generate the clock internally iirc
[22:14:49] <Casper> I DO have 2 pins! o.O
[22:15:02] <cehteh> the trick is to sync the clock (at boot listen to the port and tune osccal)
[22:15:35] <cehteh> you only want to receive .. no tx needed
[22:15:45] <cehteh> can you disable TX? dunno
[22:15:56] <Casper> seriously... why did I fucked up so much on planning?!?
[22:16:08] * Casper goes check to be sure he didn'T fucked up on the part too...
[22:16:47] <cehteh> hah
[22:16:53] <Casper> tiny85, check, they are 8 pins, not 6
[22:16:59] <cehteh> lol
[22:17:10] <cehteh> you can use a bootloader and resetdisable :D
[22:17:25] <cehteh> 6 pins for io then
[22:18:18] <Casper> man you would be surprised to see how much I screwed up...
[22:18:28] <Casper> my first math said I needed about 50kbps
[22:18:36] <Casper> recent math say about 16kbps
[22:18:45] <cehteh> lol
[22:19:09] <Casper> initial math: about 100 cycles per bit, new math: at 20kbps it would be 800
[22:20:13] <Casper> pin available? 6 pins, 2 power, 3 pwm, 1 data in (and where is the reset???), reality? 8 pins, 2 power, 1 reset, 3 pwm, 2 data
[22:21:03] <cehteh> ah just looked at the datasheet
[22:21:20] <cehteh> usi can have software defined clock, but it is pita
[22:21:50] <cehteh> but on the other hand your PWM's run freely, you have the whole cpu to handle communications
[22:22:13] <cehteh> so you can generate the usi clock in software
[22:22:41] <Casper> YUP
[22:22:52] <Casper> but 2 wires, even better
[22:22:54] <cehteh> you have the timers overflowing at MAX?
[22:23:00] <Casper> and my scope can decode 2 wires
[22:23:12] <Casper> at max or 255 yes
[22:23:27] <cehteh> then you may even trigger usi clock on the overflow interrupt
[22:24:08] <Casper> but 1 or 2 wires, it do not matter, I tought I had no spare pin
[22:24:17] <Casper> since I have one, better use it!
[22:24:37] <Casper> (I think the initial idea was to use the ADC to set the address... will set in eeprom)
[22:24:41] <cehteh> just think carefully how you use it
[22:25:14] <cehteh> like can the analog comparator shift into the USI?
[22:26:01] <cehteh> no
[22:27:14] <cehteh> using the comparator and 2 pins for differential driven bus would be my choice on long cable
[22:27:37] <cehteh> comparator interrupt registering the timestamp for every edge
[22:27:47] <cehteh> and the rest is history
[22:39:00] <Casper> if only high voltage programming was as easy as putting 12V on the reset instead of grounding it...
[22:39:11] <Casper> I could have yet another pin...
[22:39:18] <Casper> I don't have an high voltage programmer
[22:44:59] <cehteh> bootloader ftw
[22:45:11] <cehteh> then you can disable reset
[22:45:13] <Casper> no official bootloader support on this one
[22:45:22] <cehteh> tiny85?
[22:45:25] <Casper> yup
[22:45:37] <cehteh> https://github.com/micronucleus/micronucleus
[22:45:43] <cehteh> i am using that :D
[22:57:13] <Casper> look like I can expect a -3% to +1% in speed due to the voltage and temperature, not bad
[23:06:54] <cehteh> huh
[23:07:08] <cehteh> datasheet says 10% ... and 1% for user calibration
[23:07:37] <Casper> for calibrated I mean
[23:07:52] <cehteh> and OSCCAL is quite coarse, a step there takes almost the whole 1%
[23:08:33] <cehteh> my first try to calibrate was trying to get close to 0 drift which resulted in a lot jitter
[23:09:24] <cehteh> now i much simplified the code and added a deadband around the target frequency, that works like a charm but i have about +/- 1%
[23:09:32] <cehteh> often better but not always
[23:10:05] <Casper> I still think to go with my original idea... the 1 wire..
[23:10:11] <cehteh> do that
[23:10:23] <Casper> I'ld like to do data over power... but that would be hell
[23:10:39] <cehteh> depends
[23:10:48] <cehteh> FM A/C
[23:11:08] <cehteh> you rectify it for power and measure the frequnecy for signal
[23:11:49] <cehteh> but you drive leds, so you need to transmit a bit power
[23:12:06] <cehteh> alternatively you can modulate some signal onto vcc
[23:12:06] <Casper> 3x PWM... would be noisy as hell
[23:12:29] <cehteh> then need some LDO for vcc .. and use the analog comparator to derrive the signal
[23:12:46] <cehteh> ah forgot you need it fast
[23:12:51] <Casper> possibly more than 10 light... each at 1-1.5A
[23:13:10] <cehteh> huge filters :D
[23:13:28] <cehteh> well when you have signal and power in the same wire you get similar noise problems
[23:13:52] <cehteh> put some big caps at the led drivers
[23:14:11] <Casper> I'll just run another wire, will be cheaper :D
[23:14:25] <cehteh> you need caps there anyway
[23:34:13] <Casper> I was for a while wondering about using a full bridge driver, and play with the power polarity for the signal, but... switching 10+A at 15kHz... :D
[23:42:02] <cehteh> how about using glass fibre for signal :D
[23:44:50] <Casper> cost :D
[23:45:53] <cehteh> you dont need carrier grade .. there are this cheap cables for audio links
[23:46:06] <cehteh> havent checked, but possibly as cheap as copper if not cheaper
[23:47:33] <cehteh> huh even the real stuff is not *that* expensive anymore
[23:47:42] <cehteh> 20Eur for 20m ...
[23:47:50] <Casper> but now... I have a question: do I upgrade my computer now? or later this year...
[23:48:03] <Casper> so pricey now :(
[23:48:18] <lald> Build your own PC from an array of ATtinys.
[23:49:07] <cehteh> haha
[23:49:23] <cehteh> why do you need to upgrade it anyway?
[23:49:43] <cehteh> i almost only replacing parts which are broken
[23:49:46] <Casper> games, core2 start to be slow...
[23:50:12] <cehteh> mhm
[23:50:20] <cehteh> dont play games :D
[23:50:26] <cehteh> or only play older ones
[23:50:47] <cehteh> only linux here and only intel graphics
[23:51:30] <cehteh> btw toslink cables are also about 20eur / 20m ... maybe you find them cheaper somewhere
[23:51:32] <Casper> cpu: i7-6700k 475$, cooler about 50$, motherboard: 325$, memory 380$ :(
[23:51:58] <cehteh> was just an idea, but actually using light might be feasible
[23:52:10] <cehteh> high bandwidth, no noise
[23:52:15] <cehteh> and you only need one direction