« first day (915 days earlier)      last day (4023 days later) » 

12:03 AM
@ThePhoton I am
I know ON Semiconductor has a location down in RI
and I think I've seen a Fairchild building up here before
never seen an AD building, though
 
Norwood Mass
 
yeah, it's close distance wise, but I never travel near it
 
 
3 hours later…
3:26 AM
@AnindoGhosh You still there?
 
@angelatlarge Kind of. Stuck in some software learning adventures
 
@AnindoGhosh Turns out that I am even dumber than I thought: someone I misread/overlooked the fact that TLC5940 can only sink 120mA maximum. Duh.
 
@angelatlarge But that's pretty evident from all the literature
 
@AnindoGhosh Is there a cheap-o PWM LED driver that's has a better current drive/sinking capability?
@AnindoGhosh Like I said, "dumber than i thought"
 
@angelatlarge Yes, I had mentioned it in an answer. 144 channel. Hold on.
 
3:33 AM
@AnindoGhosh Hmmm... I vaguely remember this...
 
3
A: Series of illuminating LEDs...sort of

Anindo GhoshHere are two of many possible approaches to the problem, one using a microcontroller, one without. The microcontroller approach is simpler to implement, the second option is merely included for completeness. At the outset: The physical layout (two concentric circles, et cetera) is irrelevant ...

3
A: Best way to control 130 - 140 LEDS with Arduino Uno 32/64 bit led drivers?

Anindo GhoshThere is actually at least one 132-LED driver IC, the AS1130 132 LED driver with PWM ($2.99). This precisely matches your application requirement at least on the LED count. AMS has a versatile set of LED drivers: For instance their 144 LED driver (AS1119) with its integrated charge pump for dri...

Wait, those have many more channels per part, but not sure whether current is higher
Oh, and @Angela, one reason I recommend the AS1119 all the time is, it allows offloading of microcontroller data traffic - it has a frame buffer for short sequences (6 frame) or repeating animations.
 
@AnindoGhosh Yes, that's cool, though I2C is a problem: too slow, I think
BTW, I've been sinking about 300mA into TLC5930: do you have any experience with that?
 
If 1 MHz I2C sustained bursts are too slow, then I have a sneaking suspicion that there's something wrong with your architecture. Please relook at the problem.
 
@AnindoGhosh Yes, 1Mhz is too slow
 
@angelatlarge I have seen at least one design where someone has a thermal fast-fuse mounted on each TLC, and the whole driving power line cuts out on overtemp.
 
3:43 AM
@AnindoGhosh :) Is that the main issue, overtemp? Because our device will have excellent cooling characteristics.
 
@angelatlarge People are running a full blown discrete-LED based television off of AS1119 devices, at railway stations here in Mumbai.
 
@AnindoGhosh I am building a persistence of vision display that needs to update 120 RGB leds. So speed is important. I guess your TV screens farm out the data pushing to different uCs, but that's a situation I really want to try to avoid.
 
@angelatlarge Look, each AS part handles 144 or I think 160 leds at tv refresh rates.
64 LEDs, with 8 bits for each, at 60 Hz, is 30kbps of data. The 8-bit address, 8-bit register code, and various ack/setup bits cramp the speed a bit, but only by 70% or so. I2C can usually handle 400 kHz, but that NXP chip is capable of 1 MHz "Fast Mode Plus" I2C and should therefore be able to handle over 1000 LEDs on a single bus. Don't worry about the speed difference in I2C and SPI for this use case. — Kevin Vermeer Sep 21 '11 at 17:11
@angelatlarge I've seen several display boards using a variety of LED drivers driving everything from 64 to 160 LEDs, at POV rates with video output, on single MCU (for the output), or MCU + DSP (for incoming TV data shtream processing).
 
@AnindoGhosh My calculations: I need to load all LEDs in .75mS. I2C can do it 3.24mS. That's why I worry about the speed difference between SPI and I2C.
The 3.24mS is generous for I2C, but whatever. Still takes too long.
@AnindoGhosh Maybe I need to make this into a question. Damn it, I thought I was all set with TLC5940s
@AnindoGhosh I'd almost try the WS2801, but they have no latch.
 
4:10 AM
@angelatlarge But those are 1 LED per IC, right? 3 channels, so 1 RGB LED.
 
@AnindoGhosh Yeah. I'd need 120 of them :) No latch makes it impossible to use though. But they can sink a lot of current :)
 
@angelatlarge Those are meant to be used for cascades, so no latching needed. Look up their "free-running" mode, I think they do latch - just in a different way.
 
@AnindoGhosh Yeah, the way I need it to work is to be able to load all the data in all the leds (in .75mS) and then flip a switch to latch the new values on all LEDs at once.
 
@angelatlarge Hmm, then this won't do, unless you drive at 25 MHz or so
 
@AnindoGhosh What is "this"?
 
4:15 AM
@angelatlarge ws2801
 
@AnindoGhosh Yeah, exactly.
 
4:48 AM
@AnindoGhosh Unfortunately they are selling for $1/piece on Ebay... That's a bit much for 120 LEDs
 
@angelatlarge Which ones?
 
@AnindoGhosh LPD6803
There are some for $1 for 2, but that's still $60 worth of drivers
They don't seem to have them in large quantities
 
But yeah, makes no sense to go with either of those (i.e. that or the WS)
@angelatlarge How come my search throws up 10 for $4 straight off, and yours doesn't?
 
The WS is available in quantitty discounts. What was th pjrc link for?
 
@angelatlarge FREE Standard Int'l Shipping it shows me. Where's the $1 from?
@angelatlarge That was with reference to your earlier remark - someone else facing the same challenges.
 
4:53 AM
@AnindoGhosh Shipping :)
@AnindoGhosh Ah.. yes.
 
@angelatlarge That's my point: WHAT shipping? It's FREE International Shipping.
 
@AnindoGhosh I guess international shipping to India is free, but shipping to the US isn't?
 
@angelatlarge Although I like this deal better - controller + LED in nice little modules, less than $1 each: ebay.com/itm/…
@angelatlarge Since I only search by the freeship filter, you should try the same.
 
@AnindoGhosh Cute!
@AnindoGhosh: I am sort of tempted to go with TLC5940's and only over-current them only 2x... I mean so far they've been fine with 300mA!
 
@angelatlarge It'll fry. :-)
 
4:56 AM
@AnindoGhosh 2x would mean 16mA/channel if you are only using 15 channels/chip
@AnindoGhosh You think?
 
@angelatlarge I'm a firm believer in thermal sensing / preemptive shutdown
 
@AnindoGhosh Been OK so far... Actually... hmm... I've been powering them off USBtinyISP, which is a regular USB device, so the three TLC5940s cannot have been drawing more than 200mA, right?
 
@angelatlarge USB without strapping resistors = 100 mA (but many USB sources do not honor that). With the right resistors, it is 500 mA
 
@AnindoGhosh Hmmm.. I guess I need to make this whole issue a question....
 
@angelatlarge Also, if the TLCs are not drawing 200 mA, then they're probably muxing in some way - Even at 5 mA per channel of LED, a white display would be 5*3*160 = 2.4 Amps.
BTW I don't want to answer questions today :-) I have an Autodesk Inventor model to send out for a project, today.
 
5:01 AM
@AnindoGhosh What are all those numbers? 3 = # of drivers, 5 is the current/channel, but 160?
 
@angelatlarge Number of LEDs.
 
@AnindoGhosh If we already have 3 = # of drivers, then # of channels is 15 /chip (16 max /chip)
 
@angelatlarge 3 was my calculation of number of channels (not drivers necessarily), R, G and B
 
Ah
@AnindoGhosh Is that for the hydroponic Arduino thing?
 
@angelatlarge Also, I would invariably put each color on its own TLC in your situation - That way I can tweak supply rails for R G B separately and save some heating
 
5:04 AM
@AnindoGhosh Hmmm... that's an interesting idea, but makes loading the data potentially trickier. Interesting thought, I need to think about that.
@AnindoGhosh Thanks!
 
@angelatlarge No, I'm not doing that one, a first-timer's contract offer on any of those sites invariably leads to them trying to get out of paying. They think it's all a joke. I don't deal with people who have not done some business before on a given site, unless it is a $5 effort.
@angelatlarge LUT the data on a per-color basis
@angelatlarge I tend to go into auto-solution mode on LED stuff, it's always been a big client pleaser. Give them blinky lights, they pay you money fast!
@angelatlarge So the Inventor model is for yet another chemical reactor. This time with multiple fluidic processors in parallel sleeves. Messy.
 
@AnindoGhosh Interesting. There was someone who wanted to use Arduinos for bitcoin mining :)
@AnindoGhosh I didn' know you did reactors
 
@angelatlarge Yeah, they need to get Stellaris instead.
@angelatlarge I have a patent on a flow cell reactor for pharmaceutical industry.
 
@AnindoGhosh Oh, neat
 
hi all
 
5:11 AM
@Rick_2047 Hiya
 
@Rick_2047 hallo
 
I think I have my next big project
I know its a tad cliche but I think the Cortex M0 can handle it, I am going to voice automate my room
 
@Rick_2047 Cool
 
hah
@AnindoGhosh are there any big electronics or hobby DIY conventions in India ?
 
@Rick_2047 no
 
5:21 AM
@AnindoGhosh any interesting conventions in recent future?
actually my college just ended and I am taking a year off
 
@Rick_2047 Umm "recent" and "future"?
 
to prepare for admissions next year but that does not stop me from going places
@AnindoGhosh stop playing with semantics, you know what I mean
 
@Rick_2047 "Near future" works.
@Rick_2047 My brain makes sizzling sounds when I read "recent future".
 
@AnindoGhosh now that i think about it, it should actually
but do answer my question
 
@Rick_2047 No, none that I know of. If you hear of any, I'd be keen to attend.
 
5:30 AM
i will look around
 
@AnindoGhosh Have you seen WS2803? It claims 30mA/channel, and it has 18 channels and built-in temperature protection. At 30/chan 18 chan/chip, this would be 1/2 A if all channels were on, and I don't see a max I/chip in the datasheet.
 
@angelatlarge I thought those were unrealistically priced in retail. I've used strips with WS2803 built in.
 
@AnindoGhosh ...and also no latch pin
 
@angelatlarge Umm that series of ICs is designed for cascading data, they all latch on clock.
 
@AnindoGhosh Yeah, it is problem for me
 
5:42 AM
@angelatlarge It is not the product for you, basically. Or at least not for the architecture you are trying for. My way of doing a PoV would be to use dynamic real-time-output-refresh data streams and not bother to latch at all.
 
@angelatlarge you are doing a POV? propeller based?
 
@Rick_2047 Propeller = parallax chip, or propeller = fan?
 
fan
 
@Rick_2047 Trying to :)
@Rick_2047 No, bike spoke based
 
i made a rudimentary one few years back. they are not fun to time
how many LEDs are you gonna use?
 
5:44 AM
@Rick_2047 Wanting 120 RGB leds
 
and @AnindoGhosh is right, do not use latch, its really unnecessary
 
@angelatlarge Did you not mention 160 earlier?
 
@Rick_2047 I haven't had timing problems, actually.
@AnindoGhosh If I did it was a typo. 120 is the goal
@Rick_2047 Why do you say that?
 
Let the eyes do the latching. That's what PoV is about
 
@AnindoGhosh this @angelatlarge
you will spend too much time doing the logic and will not make any difference
you just have to switch them fast enough
 
5:47 AM
Yummy, prawn sausages. I've never had this before. Never knew they existed.
 
@AnindoGhosh @AnindoGhosh I don't understand what you guys have in mind.. basically you need to create virtual pixels, for this you need to switch your led colors when the led is positioned at the pixel boundary. This is what the latch is for: since it takes a while to load 360 channels, you need a latch to switch on the entire pixel column
@AnindoGhosh Wrong window?
 
No, I multiposted on the 4 different chats I am active on at the moment.
 
@AnindoGhosh Didn't know you can do that.
 
@angelatlarge It might behoove you to look at some of the more ambitious PoV spinner projects out there.
@angelatlarge Manual multi-post, and the other 3 chats are not SE sites
 
@AnindoGhosh It might. I mainly looked at Adafruit's code, not much else
@AnindoGhosh But, honestly, I am having hard time seeing a non-latching design that can approach the switching speed needed for a 30mph bike spoke.
 
5:51 AM
@angelatlarge I think it's the other way around - A latching design brings in overhead that a non-latching one does not.
 
@AnindoGhosh When you say "non-latching" you mean the World-semi style latch on clock?
 
@angelatlarge Yes
 
@AnindoGhosh That's the adafruit design, which is latching. They use 74HC595s, which I've used for this too.
@AnindoGhosh The problem with that design is that you can't get more than 3 greyscale values: just not enough time
 
@angelatlarge I didn't say it was or wasn't. I just started youtubing SpokePOVs and wanted to share.
Ohhh mindblowing: Any GIF or JPG to a spoke PoV:
 
@AnindoGhosh The problem with non-latching design is as follows: at 30mph, your 10mm pixel lasts 0.75ms, which incidentally is about the fastest your can load pixels from an ATmega at 20Mhz. If your pixels latch automatically, then your columns are no longer columns, and you cannot control your image.
@AnindoGhosh That's the goal, except dynamic.
 
5:58 AM
@angelatlarge Did you notice the optimization that last guy does? 3 different spoke light strips, one R, one G and one B.
 
@AnindoGhosh Interesting idea, but there is a problem with that at slow speed. When you are going slow, there is no persistence. Actually for a single spoke you need at least 15mph for full persistence.
 
@angelatlarge pedal faster, Gilligan!
 
@AnindoGhosh That's one solution to an engineering problem...
 
@angelatlarge Hehehehehehe
And no more youtube, or I will never get my work done.
 
@AnindoGhosh (thank god :)
@AnindoGhosh That car thing...oh, I guess that's not surprising... they need only have 30 leds, and if we want to do 120 at 30mph, then they can do their 30 at 120mph. So that makes sense.
 
6:07 AM
Back to prawn sausage. Very important to prioritize.
 
@AnindoGhosh It is 39 LEDs I guess for them.
 
user61389
Good morning all
 
Hiya @CamilStaps
 
It is still a work in progress, but yes, be my guest.
https://github.com/angelatlarge/motionlog
 
user61389
@angelatlarge cool, thanks :)
 
@CamilStaps I am confused... I don't see the error in the "full" log
 
6:35 AM
Hi!
 
@abdullahkahraman hiya
 
user61389
@angelatlarge yeah it's on the end of a very long line...
 
@AnindoGhosh That one is from Turkey
 
user61389
@abdullahkahraman morning
 
@abdullahkahraman Nice toy.
 
user61389
6:36 AM
@angelatlarge that's a little bug in MPLAB I guess
 
@CamilStaps Right, so it's a linking error, not a compiler error, i think
 
user61389
@angelatlarge I don't know, where does the compiler stop and the linker start? I think the problem is those .c files can't be found
 
@CamilStaps I think it can't find the library files
 
@AnindoGhosh No, I think only the uploader is Turkish :)
 
user61389
@angelatlarge what lib files?
 
6:38 AM
hlink "--edf=C:\Program Files (x86)\Microchip\xc8\v1.12\dat\en_msgs.txt" <-- linking starts here
 
user61389
@angelatlarge yes but where in theory? What does the compiler output?
 
@CamilStaps Well, presumably the libs containg _putch(server.obj) and all the other PIC vomit are already precompiled for you. Somewhere. But the hlink can't find them.
 
@CamilStaps Compiler generates .o or .obj or .lib from the .c files. Linker resolves addresses between these binary object files, inserting actual addresses as will be deployed by the object loader, into the appropriate pointer locations, and generates a final binary. This binary is loaded by the object loader at runtime, which may further adjust address offsets depending on architecture.
 
@CamilStaps Or what the reactor engineer said.
Indeed, I think it is looking for server.obj...
 
OK, that's very funny - why would anyone want to star a geek databurst?
 
user61389
6:41 AM
Ah I see, I'll have a look where those lib files are. But the XC8 comes with C files, so I don't think those are precompiled?
 
As a test, find server.obj in your PIC crud directory, copy it to where your source files live and see what happens.
@CamilStaps And then fix the paths and linker options if it works.
@CamilStaps Unless your source file is called server.c... is it?
 
user61389
@angelatlarge It is, why?
 
@CamilStaps They may still be precompiled: Two reasons, 1. All library sources may not be shipped with the compiler, and 2. There may be platform-specific compilation variations - typically only lower-level functions need that, while higher level ones just use stubs to refer to these lower-level labels.
 
@abdullahkahraman: Gak! YOU are the person starring all sorts of unfunny things!
@CamilStaps Shoot. Well then that error is not telling us much.
 
@angelatlarge Nah, I give out my stars carefully..
@angelatlarge Unfortunately, I cannot star my own posts.
2
 
user61389
6:45 AM
@angelatlarge the server.obj is in my project dir
 
@angelatlarge Wait a sec, hold on - If a compiler finds a pre-existing object file, it will not recompile the C file, it'll just ignore it.
@CamilStaps Do a make clean
 
@CamilStaps Right, nevermind... let me use my Googling powers...
 
user61389
@AnindoGhosh as an automatism I always do a rebuild instead of a build, so I think a clean is included, but let me try
 
@CamilStaps Even if your object file is in the project folder, the linker won't find it unless the project root is in the linker path variables: Check where the compiler pukes out its object files from any one of your C files. That's where the linker expects to find all OBJ files.
 
@CamilStaps Just as expected, your libraries are precompiled: Library files contain precompiled code, typically functions and variable definitions
 
6:47 AM
good morning @all
 
user61389
@AnindoGhosh but I always had my object files in the same dir :P
 
Hey @jippie-man!
 
Ah.. This is really unfair.. I have taken my bachelor's level courses as English.. Now I have to do my homework, but I don't know what is Turkish translation of "Marginally Stable"..
@jippie y0y0
 
user61389
@jippie morning on this glorious day :)
 
@CamilStaps glorious? Not here apparently.
 
user61389
6:48 AM
@AnindoGhosh nope, I get the same errors, even with clean
 
user61389
@abdullahkahraman we get a king today.....
 
@CamilStaps ?
 
@CamilStaps Check timestamps on all required OBJ and LIB files. Are they all current to the compile time?
 
user61389
@abdullahkahraman we had a queen, as of this afternoon we'll have a king
 
@CamilStaps Are you using the IDE?
 
user61389
@AnindoGhosh I can find server.obj, but I don't see any lib files, what do you mean?
 
@CamilStaps Wow, monarchy still there?
 
user61389
@angelatlarge yep
 
user61389
@abdullahkahraman they don't really have teh powerz
 
@CamilStaps I know, lol. But they have the moneyz :)
But I thought this system was only valid for UK
 
6:50 AM
The compiler will automatically include any applicable standard library into the build process when you compile, so you never need to control these files. So you are obviously not having a problem.
@CamilStaps Faker!
 
"My oldest son will now take over a beautiful task which is filled with great responsibility," the queen said. "I am absolutely convinced that Willem-Alexander will be committed in faithful devotion to be a good king and to do what a good King is asked to do."
 
user61389
@angelatlarge make that an answer! :P no but seriously there are no lib files in the XC8 folder.
 
@CamilStaps What is the king asked to do?
 
@CamilStaps can we see the xc8 command line, please?
 
@CamilStaps Since I don't know your project structure, I don't know what object files are needed by it. If you only need server.xyz as a source for some function invocations, then check whether the server.c is being compiled afresh or not. Also, some compiler / linker combinations build a "library" differently from a regular "object". Not everyone does.
 
user61389
6:52 AM
@abdullahkahraman sign the bills, wave at the children, go on holiday
 
@CamilStaps lol.. Btw your new king and your old queen looks better and warmer than UK's
 
@abdullahkahraman Wikipedia is your friend. Homework: Find out how many figurehead monarchy systems are still active.
 
@AnindoGhosh Yeah, sure, as soon as I finish my homework for Active Filter Synthesis class
 
user61389
@AnindoGhosh I have server.c in my project dir. I'm using the standard libs for SPI which are included with #include <spi.h> and a homebrew library included with #include "../../inc/enc28j60.h (the path is correct)
 
@abdullahkahraman Prioritizing is important.
 
user61389
6:54 AM
@abdullahkahraman we're less happy with them than they are in the UK
 
@CamilStaps have you seen this:
http://www.microchip.com/forums/m357096.aspx
 
user61389
@angelatlarge how do I do that?
 
@CamilStaps The suggestion is that it is a DFU error, but no specification of which one.
 
@abdullahkahraman The old queen was always Freddie Mercury. All the rest were just his courtiers.
 
user61389
@angelatlarge what's DFU?
 
6:55 AM
@CamilStaps You don't want to know until you solve this problem. Trust me.
@CamilStaps Just read the linked to page.
 
user61389
@angelatlarge okay :)
 
@angelatlarge Funny, the first response on that forum thread is by our own "Nothing to do with Electronic Design" heller.
 
@AnindoGhosh Yeah, I saw that :)
@AnindoGhosh He's got perky ears in that photo too.
 
user61389
@angelatlarge I don't have a server.h, would that be a problem? I never used one for my main file.
 
@CamilStaps I am not sure that would affect the perky ears
 
6:57 AM
@angelatlarge ROTFL
 
user61389
@angelatlarge then I don't understand the solution.. could you explain it?
 
@CamilStaps You have to write your own putch.. I think. Still reading the compiler manual
 
user61389
@angelatlarge really? Why? I thought they would do my job.
 
@angelatlarge No, you have to grab the right putch and #define it in. That's because putch() can be for serial, or for any other character output method supported by a given platform.
 
Yes:
The putch() function must be written as part of each project, as well as code to initialize any peripherals used by this routine.
@AnindoGhosh Or written, as the manual specifies. But yes.
@CamilStaps Maybe your putch needs to release a carrier pigeon.
@CamilStaps Or if you are Leon Heller, a carrier cat?
 

« first day (915 days earlier)      last day (4023 days later) »