/diy/ - Do It Yourself

Good luck getting anyone to download the files. You'll have better luck by posting pics of the board layers. Or all layers visible with no copper pours.

I did my first dual layer PCB 2 weeks ago. Follow the sparkfun tutorial on how to use eagle for pcb design, when you are up to routing, use copper pours for ground and autp route for fast design

I'd recommend routing manually so you know where shit goes.

Hobbyist two layer boards are quite rarely complicated enough to use autorouting (of course this only holds if you're not in a terrible rush, then I guess it's fine).
One can learn a lot about layout and get a better feel of the whole PCB planning process by manual routing (even if it can be a bit daunting in the beginning).

>>649657
well I was able to rout it just like you had it but I really think you might want to go over you schematic again...

final size was about 39x42(mm)

>>649657
I didnt bother to look up the chips but some of the connections don't seem to make sense. particularly the power lines

IC2 is getting no power input and then on the segments you have power on pins 3,5 on one then ground for pins 3,5 on the other.

>>649978
hmm jp1 should have been connected to vcc. not sure where the trace went.

Ok OP:
1. more spacing on your copper planes, there is not enough isulation!
2. only use 45° turns, no 90°!
3. always enter or leave a pad at center of a side
4. use thicker copper size for power routes
5. route bypass capacitors correctly, the way you route them they are almost useless
6. use cut outs to net create antennas with copper planes (don't think of planes as easy connectors)
7. use smd caps
8. run a connection check
9. run a design rules check

>>650024
>2. only use 45° turns, no 90°!
Until you get into gigahertz signalling rates, this does not matter one bit.

>4. use thicker copper size for power routes
Also won't be important for a low-power design like this. There are online trace width calculators, and you can carry a decent amount even in a 6 mil trace.

> IC2 is getting no power input and then on the segments you have power on pins 3,5 on one then ground for pins 3,5 on the other.

there are indicators on the drawing telling you to connect pins 8 and 16 to the power rails, so that's OK, but there's a definite error in the LED display. since both displays are identical, they're either both common-anode or common-cathode. they cant be both, so one is wired wrong, and it wont light up. a very quick google search says they're CC, so 3+5 should go to ground. you should double-check that.

>>650056
>Until you get into gigahertz signalling rates, this does not matter one bit.

It's not really about signal quality, usually. 45° traces tend to be shorter and take up less board space. While it may or may not matter, depending on how your board is actually being made (high-quality fab house vs. bargain bin fab/home etching/CNC routing), 45° angles are more consistent than 90° ones. It also, obviously, looks nicer.

>>649978

I'd add more padding around those pours. Looks really sketchy with that tiny bit of clearance in some places.

>>650062
I threw that together for OP as an idea for arranging the parts to minimize size. It's definitely not meant to be a final product.

The padding is just whatever the OSH Park DRC has it at.

>>650062
Beauty is in the eye of the beholder.

>>650070
man this looks like someone took a marker and drew out the board then used that to etch it.

kind of cool in its own way.

>>650074
that can totally be true.

some years ago we use to draw circuits with light proof marker on foils, put them under UV light and etch them in our bathtubs you know.

>>650070
Looks like it was drawn using the tears of Chinese sweatshop children and etched in lead paint.

10/10

>>650070
yeah, that looks like shit.

>>650070
I like it.

>>650070

It wouldn't surprise me to find out this was hand drawn.

2/10 would not solder.

>>650056
90 degree turns can be rude to you if you are etching at home.

>>650070
old boards are very interesting, what were they smoking before cad was invented?
guess it looks much better than trying to make everything strainght by hand

>>650070
That is beautiful.

>>650138
>>650536
Computer dependent youngsters.
My first boards were made with a marker, some acid, a drill bit and some finesse.

> It wouldn't surprise me to find out this was hand drawn.

it wasnt. you can see that there are 3 discrete line thicknesses. that's because they used tiny rolls of tape of three different widths. and the ground plane was outlined in tape, then filled in with more tape.

holyshit. i need this thread. but ofcoarse i only discover it the second i need to leave for work.
dont 404 on me buddy

i needs to make a cpld board

>>650587
>My first boards were made with a marker, some acid, a drill bit and some finesse.

So were mine. Which is exactly why I'd never do it again unless I had to.

>>650759
bumping it only for you my friend

>>650759
curtousily bump

>>651486
thank you sir.

i've been working with electronics for four or so years now. but i've never actually learned to design pcb's. we used proteus in school many years ago but i didn't pay attention and the teachers we horribly inept.
iv'e found sparkfun and hackaday's tutorials for eagle cad somewhat helpful.
you guys know of other good resources for eagle? tuts or libraries?

my fist project will probably a simple cpld breakout 5x10cm board with jtag and regulators.. pretty simple.
thing is that xilinx's own library seems weird and making a device from scratch with the right footprint is something i think i'd fuck up.
couldn't i just copy a device with the default vqp44 footprint and edit its pins and name/value to suit my needs instead of making a footprint from scratch?

just general noob tips would be appreciated also^^
thanks dudes.

also. which service do you use? i was thinking id try dirtypcbs.com

>>651499
>library seems weird
by that, i mean they only have the huge tq144pin ones.
maybe i didn't get the right lib. i'll search more

Some things that have already been stated:

1. ROUTE BY HAND! You should always know what goes on on your board, I've seen some sketchy ass autoroutes, not to mention it looks tons better.
2. 45 degree angles only, it routes a lot easier and looks way better.
3. Do a ground fill, it gets rid of about 20% of your connections, and good god does that help.
4. make your spacing as small as possible while still remaining comfortable. for example, OSHPark can handle 6 mil traces and spaces, so I use 10 mill traces and spaces, this makes it a lot easier to route.
5. Run a DRC, it helps you catch so many errors.

Next post will be about that schematic.

>>651499
>dirtypcbs.com
No. You'd be better off with OSHPark

Okay, now for that schematic.

1. The wires don't actually have to connect, that's just asking for a rats nest to happen. Instead, name them and label them, they'll be connected on the PCB.
2. Use GND and VCC symbols, it makes the schematic easier to read, as well as making it obvious when you missed a connection.
3. C2 doesn't seem to be connected to VCC like C1 is
4. Both of your displays should be CC, so that VCC connection on LD1 has got to go, replace it with GND.
5. Those ICxP connectors don't seem to be doing anything...
6. Those shift registers look really weird, they're using SCL for the VCC symbol?
7. And finally, run a connection check.

On another note, if you want to get rid of one of those shift registers, you can connect both displays to one shift register and break out the cathode pins so that you can hold one high to send data to the other one, and vice versa.

>>651513

Strictly speaking, yes, he'd get better boards with OSH Park (I've used them before; pic related) but you're not going to need that kind of quality for a simple breakout board.(or, honestly, most damn near anything a hobbyist would be designing).

It'll mostly come down to how many he needs. If it's three or less, yeah, OSH Park is probably going to be the better option because the boards are quite nice and it's not that much more expensive than dirtypcbs. However, if you're looking at more than three but less than OSH Park's 150in^2 medium run, dirtypcbs is going to work out cheaper.

>>651538

>pic related

God I am a huge faggot.

>>651540
sweet board
thanks for the feedback. i'll have to chcck oshpark out and compare prices.

but yeah, im learning and i'll probasbly make some useless boards. might as well make them cheap.

>>651540
I just like OSHPark for the superior quality, low price, and free shipping. (Also, because it's 'MURICAN)

>>651542

What kind of trace size are you looking at the smallest? Because the cheapest way to do it, by far, is to just etch them yourself. You're looking at a difference of as much as 99% here. OSH Park prices gets you boards for $5 per square inch. DIY etching can get you boards for $0.05 per square inch, depending on how exactly you do it.

Unless you're looking at making prototypes for a production run (where you actually need to inspect the boards you'll be using), or you're working with REALLY small parts and need insanely small traces (talking like 0201 here; again not something you're likely to be working with for hobby use), DIY etching should generally your first choice.

>>651510
not op, but thanks for the tips.

>>651549
>You're looking at a difference of as much as 99% here
thats only true if you assume my time is worthless^^
but srs'ly, i want to get into fabbing.

>>651544
norfag here,i don't care where they come from.

on the subject of ground planes; how do i make a pad like the left in pic related?
these are hella nice since you can solder without heating the entire ground plane.

you guys are awesome

>>651584
>pad like the left in pic

No idea about Eagle, but those are usually called thermal reliefs. Check the copper pour settings.

Can I use this a thread for bread boards?
I have questions.

>>651600
Considering people are giving this thread keep-alive bumps... Shouldn't be a problem.

>>651606
>>651600
Perfect,
I bought a small bread board, pack of various resistors a toroid bead, 4 220uf capacitors, a pack of 10ish 2N3904 transistors and a couple LEDs, I've done some light switches before and understand how to solve for different values in parallel/series circuits to an extent. I'm building now a Joule Thief but I have no idea what to build after.

Or for that matter how exactly everything connects on the board. You punch it through and run wire connecting them underneath? Some of the holes have copper surrounding 2 or 3 of them, if components are touch that copper grouping, will the current flow through it? Or is it better to wire it form the bottom?

>>651584
You talking about the one with the four connections?
When you make a polygon, you can change the thickness of the line. After re-naming the polygon to "GND" and clicking the ratsnest button, the polygon will connect with all of the pads labeled as "GND", and the connections will be the same width as the initial thickness of the polygon line.

>>651611
Yes, the current will run through the copper. Your best bet is to try to wire it so that as may pins are sharing a connection with as little wires as possible. Wires present complications in most cases.

>>651618
Oh, I forgot to mention, The separation between a pad and a ground pour can be adjusted through the DRC button in one of the tabs (I don't know which because the computer I'm using doesn't have EAGLE), although it should be fairly easy to find using the provided diagrams.

>>651618
i'm slightly confused but i think i'll get it when i have it in front of me.

another thing. i'm unsure what/how to use a proper border in board editor.
i need a frame at 5*10cm to work within. how do i make it metric instead of fucking inches.
and how do i make it so the fab service knows that is the edge or w/e? a specific layer?

sorry for the typical googleable questions.

>>651966
the drc button (bottom right in red) lets you set thermal isolation
and the ratsnest button (top left in red) will fill it in for you

>>651966
as for changing units you want the grid button

>>651982
ok, i though that since most electronics except smd is mils i'd be good to have that while just metric for measureing.
but i suppose switching back and forth isnt a prob. thank you for your help.

i think i might let the cpld wait and just make a standalone atmega just to get the feel of it.

>>651980
>>651982
the screenshots make it way easier to see.
thank you again for the effort.

>>651987
I change units a lot when laying out a board. I like to arrange my components and traces in mils. my connectors in inches (always trying to make it breadboard friendly), and i estimate board cost by calculating area in mm

just remember if you need to move an object after changing you grid hold control then click on it to make it snap to the new grid.... that shit took me forever to figure out.

>>651988
no problem I had eagle open anyway