First Time Designing a PCB

[OP No.108920]

Hello /diy/,

Here's the results of a high schooler's first time trying out PCB design. It's the control board for a quadrotor that has an XMEGA microcontroller, accelerometer, gyroscope, and CC1101 radio. I'm still new to this kind of stuff, so I would appreciate any sort of feedback on my design. Any electric engineers?

Schematic: http://dl.dropbox.com/u/538662/schematic.png
Schematic file: http://dl.dropbox.com/u/538662/xmegaboard.sch
Board file: http://dl.dropbox.com/u/538662/xmegaboard.brd

Things I'm worried about:
-The rather random routing of the power supply lines
-Running power and other lines below ICs
-Running traces between pin holes for the servo and ESCs
-Placement of crystals and antennas causing interference

Thanks in advance,
Anon

[Anonymous No.108937]

Your schematic tells me that you're doing a double-sided, surface-mount PCB &, by the looks of it, it's a good routing you've done.

There are a couple of things for you to note:

1. Just because you can automate the routing process, it won't stop you from being able to manipulate the result on your own.

2. Do your best to avoid right-angled traces on a circuit board; try to use diagonal traces in place of right-angled joints.

3. Wherever possible, try your best to make a one-sided PCB that's as condensed as possible. Otherwise, go double-sided.

[Anonymous No.108942]

if you plan on hand soldering it, i would do something about those caps on the top left.

[OP No.108944]

>>108942
I'm planning on using solder paste and a frying pan to solder all the components at once. Many of the components at the top left are inductors for the radio, and I'm wary about letting the trace lengths get too long....

[Anonymous No.108957]

What process are you using to fab the PCB op? Im asking this because if your not sending it to a board house then your going to have trouble with via's under surface mount IC's, also if your hand soldering vias you'll want to fatten them up a bit (2mm vias are piss easy to solder, >1mm is a pain in the ass)

Also eagle is fucking retarded for a lot of things so you just kind of need to get a feel for it, and what processes your using. DRC is a great tool if it's set up properly but it's not gospel (I use an isolation router for most of my prototyping and know exactly what I can get away with)

Also as a general eagle tip, when routing things like SMD IC's with tight pin pitches, always rout from the pad out (at least for the final connection to / from that pad) because eagle will center the trace even if it doesn't align with the grid (eagles internal grid is in mils like a fucking retard)

As for bunching your shit together like that, in my experience the 1206 pads have miles of room to spare, 0805 not so much.

Id probably be inclined to rotate Y4 and move it a bit closer to the uC too, just as best practice.

[OP No.108968]

>>108957
I'm going to have it sent out to be made. Probably FusionPCB.

Thanks for the advice. I have no idea how to use DRC, though. There's a first time for everything, I suppose...

[OP No.108969]

Anyone have any tips on ground planes? Should I just put one down over both sides of the whole board, or should I try to isolate the radio circuit?

[Anonymous No.108971]

>>108968

Your fab will give you some numbers, typically just minimum track width, minimum clearance and minimum drill size. When you set up DRC it will ask you for about a million numbers but they will just be variations on the above (minimum track to track, track to hole, hole to via etc) so just punch in the same numbers, you can change the units by typing them in so type say 0.1mm rather than converting to mils.

Then when you run it it'll bitch about everything, just click through the list and it'll take you to each "problem." Clearance and dimension are things you'll want to pay attention too, stop mask and other bollocks you can ignore.

[Anonymous No.108976]

>>108920
That's not a bad layout, but some suggestions:

Those SMD components in the top left are so close together you will have trouble hand soldering them. You need to get the pad hot, and you can't do that with other parts in the way.

Use a ground plane, and a VCC plane too if you want. It will prevent crosstalk between the signals; especially if you hatch it.

There are vias everywhere. I can see a few unnecessary ones just by looking at it. Auto-routing has that problem.

Add some drill holes for mounting.

[Anonymous No.108979]

>>108976
The whole board is routed by hand, not auto-routed.
Also, about the components in the top left, will it be alright if I use an oven/pan to solder with paste instead of hand solder?

[Anonymous No.108983]

Couple of tips from another high schooler still learning (sorry if I'm repeating stuff from the thread already)

With your schematic:
What protocol are you going to program the AVR with? Is that something unique to the XMega? I don't recognize it based off of the data lines.

I like to break out my extra GPIO's and such if I have the time and space. Its nice to be able to add on in the future, or just mess around.

With the board:
You should try and put your crystal for the AVR as close as possible to the avr to prevent parasitic capacitance of the traces.

If you're going to be sourcing the power for the servos off that board I consider widening the power traces just a hair.

You're too close to the edges of the board in a few places.

I think with a ground plane or a power plane you could reduce the number of traces and vias.

Don't worry about traces under IC's, I do it all the time. Just remember to pay attention to the board manufacturers limits and stay above them.

Construction:
I recommend a hot air gun, I've never tried the oven/pan method. Also those radio chips may be hard to solder do to their package.

As a first try you did pretty damn good.

[OP No.108985]

>>108983
>What protocol are you going to program the AVR with? Is that something unique to the XMega? I don't recognize it based off of the data lines.
It's Atmel's somewhat-new PDI interface.

As for extra GPIOs, I do have one row of headers breaking out some stuff, but I removed a few because they made routing more difficult.

[Anonymous No.108986]

>>108985
Oh ok, I haven't looked into it.
How big will the motors be? Also, are you the anon who posted the picture of the quadrotor on diy earlier?

[OP No.108997]

>>108986
I'm probably the same guy. The quadrotor is fairly small, about 7" long.

[Anonymous No.109006]

>>108997
Oh ok. That will be fun to program! Are you planning an autonomous design?

[OP No.109021]

>>109006
I'm hoping to have the basics down first, with a kalman filter filtering the accelerometer and gyroscope data, and some balancing algorithms for manual control.

[Anonymous No.109073]

> -The rather random routing of the power supply lines
> -Running power and other lines below ICs

Ignoring the radio, not really a problem with that design.
However, since you have two sides and a radio, use the bottom side for ground plane.

> -Running traces between pin holes for the servo and ESCs

Not a problem.

> -Placement of crystals and antennas causing interference

THIS can be a significant problem. You have traces directly under antenna. Not only this can (and likely will) ruin the performance of the antenna, but it also induces shit on the bottom side traces.

The radio placement/routing also looks like your own. Instead of trying your best with it, I seriously recommend copying the manufacturer's reference design as-is. Also, I see that you've left the radio's thermal pad floating. Check if that's really ok. Keep everything else as far away from the antenna as you can. Avoid long traces near the antenna.

If you're worried about the crystals, you can ground their case. It helps a bit. The crystal oscillators rarely cause problems, though.

[OP No.109911]

>>109073
Alright, thanks - I'll try moving around the traces and antenna.

As for a reference design, I can't seem to find one on TI's site: http://www.ti.com/product/cc1101.. Maybe I could lift one off one of their products that incorporate a CC1101...

[Anonymous No.109933]

I wouldn't use auto-routing for anything simple like this. It can save a lot of time on big projects but holy fuck can auto-routers be retarded. You'll spend more time fixing it than if you'd simply laid it out yourself.

[OP No.110535]

>>109933
refer to
>>108979

It's not autorouted ;-;

[Anonymous No.110585]

Hello /diy/,

Here's the results of a high schooler's

Stopped reading right there. The only way you could've sound more like a faggot is if you were a femanon and somehow managed to get half of your body in the screenshot. No body gives a fuck, this is 4chan. Don't grow up to be a tripfaggot.

It does look good though. fag.

[OP No.110605]

>>110585
love you too