#homecmos IRC log for Saturday, 2011-12-10

--- Sat Dec 10 201100:00
bart416azonenberg, did you ever take part in a "design a DSP system in 4 hours" contest?13:06
bart416We did yesterday13:06
bart416Worked amazingly well :S13:07
azonenbergDesign a DSP system in four hours? No13:07
azonenbergI'm currently in the middle of a self-driven "design a pipelined superscalar DSP softcore in 4 months" contest :p13:07
bart416We used a 8 bit flash ADC heh13:07
azonenbergoh, check out this board http://i.imgur.com/sZng0.png13:07
bart416Not much that can go wrong there13:08
bart416And DACs generally well behave13:08
azonenbergmy first design in kicad13:08
bart416Not bad13:08
azonenbergback side has an RJ45 on it13:08
azonenbergthe TQFP is the ethernet PHY/MAC13:08
azonenbergthe MCU is the little QFN at left13:08
bart416I figured out it was RJ45 due to the connector pattern :P13:08
azonenbergjack is 10/100 with integrated magnetics13:09
azonenbergso two center-tapped transformers13:09
azonenbergall contacts are SMT13:09
azonenbergThe only through-hole component on the entire board is the JTAG header13:09
azonenbergthe footprint at left (no 3d model) is a mini USB13:10
bart416You sure?13:10
azonenbergYes, but i just realized i did osmething dumb13:10
bart416Looks an awful lot like the footprints for network connectors lol13:10
azonenberg*reopens kicad and fixed bad footprint*13:11
azonenbergits small :p13:11
azonenberglet me put it this way, the giant header at the bottom is 0.1" pitch13:11
bart416Let me put it this way, my latest board is A4 sized xD13:11
azonenbergHow many parts on it?13:11
bart416roughly 200 BJTs13:12
bart416700 resistors or so13:12
bart4162-3 inductors13:12
azonenbergis this your RTL monster?13:12
bart416part of it13:12
bart416testbench board13:12
bart4167 capacitors13:12
bart41610 voltage regulators13:12
bart4161 opamp13:12
bart416(relaxation oscillator)13:12
azonenbergalso... LOL good thing this is a batch board13:13
azonenbergthere's time to change it before it goes out :po13:13
azonenbergjust realized my pinout on the usb connector was wrong13:13
azonenbergpins 4 and 5 got swapped somehow13:13
azonenbergNothign i couldnt fit with rework but it'd be annoying13:13
bart416btw, yellow at the left is a capacitor right?13:13
bart416yellow-beige I meant13:14
bart416or is it a resistor?13:14
azonenbergtantalum cap13:14
bart416your track to it is rather small13:14
bart416Might cause problems if it's for decoupling like I assume it is13:14
azonenbergYes, and also good catch13:14
azonenbergAll the 3.3V stuff is on a filled plane13:15
azonenbergsomehow the 5V rail got forgotten lol13:15
bart416you getting the 5V from the USB?13:15
azonenbergNothing high powered13:15
bart416be sure to add diodes to protect the usb port from reverse current flow13:16
azonenberghow would that happen?13:16
azonenbergthere's no external power source anywhere13:16
bart416It's standard procedure on good designs13:16
bart416It just happens13:16
bart416I've gotten into the habit of adding diodes to protect my power rails recently13:17
bart416And it has saved me a lot of trouble13:17
azonenbergThat'll go in the next board rev, i think13:17
bart416There was a while I constantly had issues like my 0V going above my 5V in potential :S13:17
azonenberghow is that even possible?13:17
azonenbergI just have a linear reg and then everything else is 3.313:17
soul-di hope  n4148 orsomthing will do nice for that13:17
bart416it happens if you operate a switched mode topology13:18
bart416you build up charge13:18
azonenbergOh, i see13:18
azonenbergYeah, no SMPSes here13:18
bart416And suddenly it ends up in places it's not supposed to go13:18
bart416It's just, the cost of the diodes is so low that there is no point in really adding it for small home designs13:18
bart416If you're making a run of 10000 then indeed it's worth checking if it's necessary or not13:19
azonenbergWell, its more that i dont have any diodes handy :p13:19
azonenbergSo i think i'll leave them out this revision13:20
bart416I bought this reel of diodes a while back cause it was really cheap and would push my order over the shipping fee limit13:20
bart416So free shipping13:20
soul-dbought 500 smd ones on last order should be enough to protect me from my stupidity13:20
soul-d3,25 euro for that13:21
bart416And that was actually cheaper than the original order + shipping fee lol13:21
soul-ddid come from china though13:21
bart4162000 1Nsomethings lol13:21
azonenbergA big load of schottky diodes is on my next parts ordrer13:22
azonenbergbut not in time for this board13:22
bart416What's really funny is the cost to make an A4 sized board at university13:22
bart41610 euro >_>13:22
bart416I can't etch it myself for that price13:22
bart416can't order it myself either for that price :S13:22
bart416order it at morning, have it in your hands by late afternoon :S13:23
soul-dhow many layers13:23
bart416double sided13:23
bart416Didn't believe it myself at first either :S13:23
soul-dso you pretty much pay cost price of pcb13:23
bart416ofc you don't get a soldermask or anything13:24
azonenbergbart416: yeah13:24
bart416just an etched pcb13:24
soul-dnot drilled either i geuys ?13:24
bart416But that's not really all that bad all things considered13:24
bart416Nope, not drilled13:24
azonenbergi'd homebrew then13:24
azonenbergi send out boards for things i cant do at home13:24
soul-ddo you predrill azonenberg  ?13:24
azonenbergLike PTH13:24
bart416You can throw it in the CNC drill in the lab azonenberg13:24
bart416Give it 5 minutes...13:24
azonenbergsoul-d: I havent gotten plating to work yet13:24
azonenbergbart416: its not the hole drilling, its the plating13:25
azonenbergIn any case the best price i've seen yet for full-service fab is dorkbot's batch service13:25
azonenbergthere is long lead time13:25
soul-ddrilling is nasty work to do manual13:25
bart416For through hole I can't say their board quality is bad13:25
azonenbergBut $5/in^2 for three units13:25
bart416We've used it for loads of projects13:25
azonenbergor $10/in^2 for quad layer13:25
azonenbergthree units13:26
soul-dyou checked out seeedstudio azonenberg  ?13:26
bart416power or signal layers?13:26
azonenbergbart416: fully routable13:26
bart416soul-d, seeedstudio is good if you do larger batches13:26
bart416I intend to do it for my final design as the core components will always be the same13:26
azonenbergsoul-d: we're talking prototype volumes here13:26
soul-dok  yeah i would like a palce for that13:26
soul-di usaly need 2 boards at most13:26
bart416The thing is, you can't homebrew an A4 sized board for that price13:27
bart416I often etch PCBs for other people13:27
bart416can't make it for that price13:27
azonenbergseeed doesnt seem to provide quad layer13:27
azonenbergI actually want to try to find a cheap place for prototype volume 6 or 8 layer boards lol13:27
azonenbergBecause when i get into designs that use stuff like 484-ball FPGAs and multiple banks of DDR313:28
azonenbergThey get pricey :p13:28
soul-dplus you probably want to outsource soldering  id would find that tricky :P13:29
azonenbergNot sure about that yet13:29
azonenbergDepends on how well my first BGA board goes13:30
azonenbergOnly two metal layers13:30
azonenberg6x6mm CSBGA CPLD, 56 balls13:30
azonenbergbroken out to DIP-pitch headers and a JTAG port13:30
azonenbergFor scale the caps on the underside are 040213:30
azonenbergAnd the traces are 6 mil / 150um13:31
bart416soul-d, do you etch your own pcbs?13:33
soul-did be happy if i design somthing with 0805 and some tqfp      do have some  qfn's around13:33
soul-ddo have stuff laying around13:33
azonenbergsoul-d: I used to do 0805 and 0.8mm TQFP, moved very quickly to 0603 and 0.5mm13:33
soul-dbut not complete drill set :)13:33
azonenbergBut now i want to take the big leap13:33
azonenberginto more "hardcore" packages13:33
azonenbergBGA, QFN, 040213:33
bart416bga is annoying13:33
azonenbergI know it is13:33
azonenbergBut a lot of the really cool parts dont offter any alternatives13:34
azonenbergthis CPLD does come in both TQFP and QFN13:34
bart416yeah :(13:34
bart416It's a pest13:34
azonenbergBut i got the BGA version because it's $1.60 a chip13:34
azonenbergso i can afford to kill several while refining my reflow process :p13:34
bart416I'm looking for some time on the university's pick and place machine at this point actually13:34
azonenbergAnd not kill a $25 FPGA13:34
bart416The machine only has an error rate of 1 in 1000 or something like that heh13:34
bart416Can't do that at home13:35
bart416azonenberg, why don't you use the university's machines over there to do the BGAs?13:35
bart416I'm sure they have some13:35
azonenbergbart416: I'm not trying to build a board per se13:35
azonenbergi'm trying to develop a process that i can take with me after i graduate13:36
azonenbergI may use their gear the first time, to get started13:36
bart416Here they keep doing it for old students13:36
azonenbergbut i want to avoid being dependent on them13:36
azonenbergI may not be living anywhere nearby13:36
bart416Some guy consulted a professor this week (was working in the same lab at the time) on how to build a computer using vacuum tubes btw :')13:37
soul-dyou did that bga in kicad to ?13:38
bart416The professor started laughing at the guy lol and told him to go and get a life13:38
bart416(keep in mind this is an old guy from the vacuum tube era that is an expert at high frequency design as well, so he's used to things being hard or near impossible to make)13:38
azonenbergsoul-d: yes13:39
bart416How good is kicad at bgas?13:40
azonenbergbart416: So far? Decent13:40
bart416You sure?13:41
azonenbergI need to play more, obviously13:42
azonenbergBut it handled this little guy just fine13:42
bart416The only program I've found good at auto routing them is orcad/allegro at this point13:42
azonenbergOh, this was hand routed13:42
soul-ddoes anybody acrutaly do autoroute ?13:42
bart416Altium and Allegro can do it13:43
soul-dyeah started that stuff up once i din't realy get it :P13:43
bart416You can write an entire book on how to use those auto routers lol13:43
bart416They're also capable of auto placing components taking into account design rules13:43
bart416another thing eagle fails horribly at13:44
azonenbergi've never tried doing something quite that complex lol13:44
azonenbergmy biggest board to date was in the 200 component range13:44
soul-di hate eagle every minitue i use it lol13:44
soul-dim just stocking up on basic components for now13:47
soul-dnow im broke ofcourse13:47
bart416I still find myself often just grabbing transparent paper and a marker and drawing the routes by hand...13:47
soul-di probably make to may mistakes for that13:49
bart416It's good enough for small boards13:51
bart416You should try it once azonenberg13:51
bart416works great for high frequency board as it's pretty much impossible to make sharp corners by hand :)13:51
bart416azonenberg, you have some experience with this14:02
azonenbergwith what?14:02
bart416What do you think about level triggered flip flops in digital design14:02
azonenbergI've always worked with synchronous design so far14:02
bart416Yeah, but I meant triggered on clock level instead of edge14:02
azonenbergi'm so used to doing edge triggered stuff14:03
Action: bart416 too14:03
azonenbergnever had the chance to fool around much14:03
bart416I mean, I sometimes dare to use a regular RS flipflop in async design14:03
bart416But I'm talking about an entire register bank this time14:03
bart416The risk of unwanted triggering seems to be fairly high14:04
azonenbergYes, it does14:04
azonenbergYay glitches14:04
bart416glitches on the clock are unlikely but yeah, glitches when clock is high...14:05
azonenbergYou'd be much more sensitive to skew14:05
azonenbergif data starts changing before your clock falls14:05
bart416it's risky14:06
bart416I'd have to buffer the input to the flip flops somehow14:06
bart416I can use a S/H circuit for that I guess14:06
bart416but that just seems stupid14:06
bart416Btw forgot to mention, we had to make a guitar distort with our dsp14:13
bart416so the 8 bit wasn't much of an issue14:13
bart416the distort already destroyed so much information anyway that using 8bit wasn't a problem lol14:13
berndjbart416, not really tubes, but http://www.ese.upenn.edu/~jan/eniacproj.html15:15
bart416berndj, not of much interest to me to be honnest15:17
bart416I don't have this much space btw: http://upload.wikimedia.org/wikipedia/commons/4/4e/Eniac.jpg15:18
--- Sun Dec 11 201100:00

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