My First PCB designed, comments requested.

When you populate the board, you can leave the bypass caps off and see if i t will work. But have the pads for the bypass caps on the board. It is a whole lot easier to add caps to existing pads than to add pads and caps.

When I first started designing circuits , the tech would pick the smallest board possible to build the circuit on. So I learned to put everything tha t might be needed on the schematic. And the circuit usually worked. Then I would remove the parts that I thought might not be needed and see if it still worked.

Dan

As was taught to you by Professor Muntz.

True, it does tend to keep the discussions in sync better, and of course nowadays the difference in disk space is a nit.

Cheers

Phil Hobbs

Depends. I found that there is little difference in ESR and ESL between Ta and radial-leaded Al caps of similar capacitance and voltage ratings. Axial-leaded caps have a lot of ESL and small-valued Ta caps (

1. The board is bigger then the display. Do you intend to place two displays side-by-side ? You can not place two displays over/under for multiple rows.
2. How do you intend to mount the board/display ? How is the board/display going to be captured ?

Kind of like Muntz, but really as taught to me by Richard Sonquist. He was trying to create work. If the breadboard was so small that nothing could be added, then he got to build a new breadboard.

Dan

I have one actually. I don't recall where I picked it up or how expensive it is, but its specs look good.

Yes. The duty cycle though is far less than 12.5% most of the time (only full bright will be 12.5%). Unless of course my MCU stops running. Hmm, yep, I'll put in a resistor to get it max 70mA, though might even do 50mA, as that's the max continuous current.

Fare enough, thanks for that hint.

Ah right. It's one of those implied things. The device has a symbol called "PWR" as an "Add: request".

Two reasons. One, I want to keep it under 5cm*5cm, since that's the cheapest option from the PCB providers I've seen. Also, I want it no larger than the LED display module itself. Later down the road, I may want to design them to be chained together. At that point though, power will become a much much bigger concern.

That makes sense. I'll see if I can get something worked out with larger traces thereabouts.

The 74238 (actual part is 74HCT238 BTW) will only allow 1 eighth of the LEDs on at any given time, so the max power at 50mA each is 1200mA. I was actually planning on limiting the current to less than peak (its bright enough around 20mA). So the whole thing will be < 1A. This is just as well since that's the rating of the wall-wart I have.

Noted.

I'll see if I can apply your advice to my constraints ;-)

Actually, the outer outline is a (rough) placement of the display, not the board. The display is bigger than the board. At some point, I may try to design this to work with multiple modules, but being my first project of this scale, I thought I'd avoid the headaches of powering dozens of amps.

I don't know what you mean by "captured". For my first prototype, it'll probably just sit on my desk at work plugged into a wall. I'll either add an encasement or at least feet to the design later.

Thanks, Daniel.

add other bypass capacitors. At this point, I'm going to just say its good enough for a prototype.

I chose 360?, since that is the minimum required for 52mA.

I've added a mark.

I've increased the size of the all the traces that you've suggested. The only traces that are < 16mils are the ICSP and the 74238.

Okay, it was much harder to route with your advice, but I think I've done it. I've updated the files on my server with the latest schema/board.

Thanks for all your suggestions. They were quite valuable, and forced me to learn a little more about how to do routing in EAGLE.

How about 3.2 x 1.6 mm

Let us know how that works out for you. :)

Dan had a good suggestion - put the pads and traces for the bypass caps on, even if you don't think you need them. You can always not solder the caps in.

Well... not just me. Several people commented on this.

Yeah, the logic-level stuff doesn't need to have huge traces.

You have just discovered why tiny boards have multiple layers. The original iPhone 4 uses a 10-layer board; it wouldn't suprise me if newer ones have another couple of layers. Cross-section picture at

Well, a few of mine, and many from people that are better at this than I am.

Standard disclaimers apply: I don't get money or other consideration from any companies mentioned.

Matt Roberds

Can't tell from the blue area of your drawing, but some general comments might be in order. Blindly pouring ground may not help you. You need to know where every current goes.

Here's one way to look at it. Imagine a capacitor that can supply all the current you ever need...just go with it.

For every current on the board, map where it goes. Some go thru that capacitor, some go off board, around some other thing before they get back to some source/gnd. Include the VCC and GND and external load paths. Those connection paths have inductance and resistance. There will be drops, sometimes huge transients generated along those paths. If you have ANYTHING tapped off ANYWHERE on that path, those voltages will be induced into the thing tapped.

Pouring ground helps if that reduces the inductance or resistance along the path that the current flows...for each net on the board. If the current doesn't flow along that path, it doesn't help.

One mistake people make is to use a ground plane on one side and traces on the other. You need to get a trace across another, so you cut the plane and put a trace thru it. You just transformed a perfectly good transmission line into an inductor. You forced the ground currents to go around a different path and couple into stuff clear across the board. And this can have considerable bearing on third party certifications for EMC.

And you don't need to be designing microwave stuff for this to happen. I OEM'd a power supply from a respected vendor. They exploded randomly. When I complained, I got the runaround..."we've been doing this for 20 years...it works fine".

I built a transient load synchronized to the switching frequency. Invited their engineers over..."bring some supplies." I exploded one. Showed them how the switching waveform caused transformer to saturate. I exploded the second one. I asked them if they wanted me to explode another. They got the point. Turned out to be a ground loop that coupled the load transient into the switcher, changing the duty factor, saturating the inductor, exploding the fets.

But they still couldn't fix it.

Now, it's just a funny story... Fired the company and hired another 500 miles away. As luck would have it, the old vendor had a surplus employee. The new company needed an employee for the task. Guess who they hired! I went ballistic when I found that out. Good times.

A frame would capture the outside edges of the PCB or display to hold them in place.

That way you would not have fasteners showing.

hamilton

In general cross-posting is much better tolerated than multi-posting. Usenet is not blogland, nor is it like many other forums, it predates almost all of them thus has its own rules.

?-)

I just wanted to say thanks to everyone who gave me suggestions. I've sent my design in today, and if it fails spectacularly, then its a $14 lesson that is well worth it ;-)

I did manage to "fix" an issue with my prototype, thanks to suggestions here. My prototype worked fine if I used 3 AA's or 1 9v and a 7805, but failed to run when I used a 5v power wall-wart or 5v adjustable power

working flawlessly! Imagine that.

Originally I thought it was failing due to too much current going through my 74HCT238 (the prototype doesn't have the Darlington array), but my new theory is that the power draw of the LEDs caused too much ripple on the input. I suspect I'll get a lot more brightness out of this thing once its all soldered.

Thanks again everyone! I really appreciate it. Now lets see how the board turns out.