I've designed an LED display driver (to drive the SEEED studies 8x8 RGB module>
This is my first attempt at anything PCB, though I built a similar circuit on a breadboard. I say similar because I was missing the TD62783APG, which is required to allow this thing to work at full power without burning out pins.
The schema/board (in PDF) are here:
Eagle files:
Before I order my parts and boards, anyone willing to comment? I'm hoping for both aesthetic comments and technical gotchas.
Thanks, Daniel.
Power and ground look scary; long skinny traces and no bypass caps.
Is this a 2-layer board?
-- John Larkin Highland Technology, Inc jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
As a rule, pay very close attention to your vcc and grounds. Your Xtal caps are going to the power jack. Generally, it's a bad idea to put sensitive stuff in the middle of tracks with transient current. Ground is rarely just "ground".
yeh, it definitely needs bypass caps, the tracks are 10mil they could be bigger especially power and might as well flood front and back with gnd
-Lasse
Usenet tip: don't post the same thing separately in different newsgroups. :) (I replied over in sci.electronics.basics .)
It is possible, and sometimes useful, to post the exact same post to multiple newsgroups at once - there isn't really a technical limit, but three is sort of a usability limit. (Your news provider may limit the number of groups to prevent spam; some individual users may also filter out posts that went to "too many" newsgroups.) If you do this it is considered good form to name one newsgroup as the place where followups should go, using the header "Followup-To:"
Matt Roberds
How would I go about fixing that? It's going to run at 5v DC and < 1 amp.
Yes, it is a 2-layer board. Would that make it harder to solve the issue you're describing?
Thanks, Daniel.
I'm aware of this, though some communities frown on x-posting, and others frown on multi-posting. I will check your reply on the other group. Thanks.
The easy way to do grounds is to make a 4-layer board with one inner layer being a solid ground plane, and generally another internal layer used for Vcc or other power pours. Costs more.
You can certainly do your board as 2 layers, but it would be safer to use fat ground and Vcc traces and try to keep them short. You could run ground all around the outer edges of the board, one or both sides, and tie into that ring. And try to cross-connect grounds where possible, trying for an X-Y grid.
You can expect a fair amount of current spikes, dI/dT, from the LED drivers, and that can make voltage spikes in skinny ground traces and in turn glitch logic chips. It's a real hazard, so good gnd/vcc is worth doing.
Some people copper-pour grounds everywhere on 2-layer boards, which is good as long as the various islands are well interconnected. You probably don't need to do that here.
A few Vcc-to-gnd bypass caps would be prudent. One per chip on a single-sided board is common. 0.1, 0.33 uF, anything like that.
Oh, I don't see any mounting holes. Does it need them?
Add a meatball (pin 1 dot, copper or silk) to each chip and connector.
Are the J1 drill sizes adequate?
There are scores of ways to screw up a PCB layout. We have a pretty extensive review/checklist procedure.
-- John Larkin Highland Technology, Inc jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Lots of people cross-post to multiple groups. Why not?
-- John Larkin Highland Technology, Inc jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
there is a difference between cross-post and multi-posting
With multi-posting you end up with a separate thread in each group
-Lasse
Cross posting is okay as long as it is posted to groups that have some connection to the subject. Take a look at Rec.Crafts.metalworking to see numerous cross posts of topics that have absolutely no connection to metalworking.
Dan
First, I cross-post a lot between here, comp.dsp, and comp.arch.embedded. It seems to be well tolerated, and when the topic is chosen correctly the cross-pollination in threads can be very useful.
Second, there's rarely any point in cross-posting between here and s.e.basics -- nearly all the question-answerers are the same in both groups, its only the question-askers who are different.
Good luck with your board!
-- Tim Wescott Wescott Design Services http://www.wescottdesign.com
I was reading a TI data sheet on a 317 voltage controler. And it said that solid tantalum caps were way better than aluminum caps. Any comments on what types of caps Daniel ought to order?
Dan
This is kind of in order of priority:
1: Don't just consider bypass caps: just put them on. If you leave them off and your circuit doesn't work, you deserve it. Put a 100nF cap from power to ground on each and every chip. If a chip has more than one power pin (real power pin, not just a logic input that happens to be connected to '1') then do a bypass cap for each one. 1a: And a bulk bypass cap for the board, probably a 10uF or 100uF electrolytic. 2: Put the crystal and caps as close to the oscillator as possible. Ideally you'll draw an imaginary line around the crystal, the caps, the crystal pins on the oscillator, and ground, and you won't let any other traces get in there. That super-long ground path you have from pin to feedback caps is just horrid.A physical layout something like this, with NOTHING inside of the circle defined by the I and O traces and the crystal except for the caps, is what you're looking for.
I G O .---' | '---. | | | o--||--o--||--o | | '--(crystal)--'
3: If you can, make the board bigger and put all the chips on top. Trying to debug a board with chips on both sides is a bitch. 4: Where possible, make your power lines bigger. 5: I, too, advocate doing the copper pour thing for ground -- but only if you're willing to spend the time to connect things up with vias. Otherwise the best thing you've done is waste copper. 6: Yes, check the holes on J1 -- what you have there won't work if that's the jack I think it is. You either need to find a board place that'll cut slots and plate them (I don't know of any quick-turn cheapie places that will), or you need to make huge holes. Huge holes filled with solder work well enough, though.-- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Not tantalum. Any old ceramics would work fine.
Some voltage regulators like the ESR of tantalum caps, and might oscillate with all-ceramic loads. But tantalums tend to explode when hung across power rails. Aluminums have higher ESR, especially cold.
-- John Larkin Highland Technology, Inc jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
What!!?? People post off-topic there!!??
-- John Larkin Highland Technology, Inc jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Must be one of those HEAVY METAL groups!
Jamie
Cross-posting (the *same* post in more than one group) can work OK, if the post is relevant in each group. Some people that have useful advice don't read all the groups, so that can be a good way to get responses that you would otherwise miss.
Multi-posting (*separate* posts in more than one group) leads to duplicate discussions in each group, which is not usually as helpful as keeping it all in one place. Interested people have to follow more than one group to see the whole discussion, and it breaks some of the feedback on the quality of answers. (If I told somebody to put 10 uH from +5 V to ground instead of 10 uF, it *might* slide by in s.e.basics, but it *will* get noticed and pointed out in s.e.design .)
Matt Roberds
(Moved over from s.e.basics .)
Daniel Pitts wrote:
A switching 5 V DC wall-wart will probably work OK; these are cheap and easy to buy now. 20 years ago, the average "5 V DC" wall-wart would have had just a transformer, bridge rectifier, and maybe a small capacitor, and would have put out anything from maybe 4.5 to 6.5 V DC depending on load.
You should be able to buy a stout enough 5 V switching wall-wart for reasonable money. Look at the specs to see if it has a *minimum* load - some switching supplies need this and some don't.
Yes, but not just the ripple that might come "for free" from the power supply. Sometimes when the outputs of a logic chip change state, the chip will momentarily draw a lot more current than you would think. Having a small bypass capacitor near the chip lets that capacitor supply the increased load, rather than dragging down the power supply voltage for everybody.
(The current spike amount varies depending on whether the chip is TTL, CMOS, or what... to a zeroth approximation, older chips were worse, but people still put the bypass caps in brand new designs.)
In your case, you also may have big currents being switched to the LEDs, which will tend to change the voltages on the rest of the board.
Now that I have a link to the LED matrix you are using, I think this is an even better idea. The LED driver chip can provide 120 mA, but the absolute maximum peak forward current is 70 mA. Using the fixed resistor makes it harder to burn out the LEDs by turning the pot down too far.
Find out, if you can. Sometimes the list of "what to send" is fixed, but actually getting the silkscreen put on the board costs extra.
On the schematic, I see the connections to SCANSELECT/+UB and SCANSELECT/-UB - those are for the 74238 decoder. I am talking about the power and ground to the TD62783 buffer/driver chip, SCANDRVR, which I don't see on the schematic.
Why? Does it have to fit somewhere? I understand wanting it to be compact and look neat, but when you're new at it, spreading stuff around is easier.
You don't need big traces to the 74238. All it's doing is handling logic-level signals, as you noted.
All the power that goes into the LED matrix will come from +5V and through the TD62783, so it needs big power traces. All the power that comes out of the LED matrix goes through the TLC5916s to ground, so they need big ground traces.
If you turned on all the LEDs at their peak forward current of 70 mA, the matrix would draw 13.4 amps. (It would then start burning out, but it's a useful number to have in your head.) If you turn on all the LEDs at 20 mA each, the matrix will still be drawing 3.8 amps.
It might, if the flood-fill is well-connected everywhere. What I mean is that it's sort of easy to end up with a big flood-fill on the left side, and a big flood-fill on the right side, and just a couple of skinny traces connecting the two big areas - that limits you to whatever the skinny traces can carry.
Matt Roberds
AFAICT that only really applies to folks with manly newsreaders, where with cross-posting you don't see the same post in multiple groups, whereas with multiposting, you do.
In Tbird (and especially google gropes) there's little difference between cross-posting and multiposting.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.