PCB design best practices & critique (RFC)

A

ags0000

New member
I'm about to submit my board for fab, and would appreciate input on how to make it better. I have one 50MHz clock, several 5MHz clocks, and many 2.5MHz signals. I've gone to SMD components (even 0603 size - I hope it's as easy as I'm told) other than RJ45 jacks.

1) I have tried to maximize signal integrity by maximizing ground plane by filling (copper pour) both sides aggressively. Is there any downside to this?
2) In some cases I dropped an "orphan" via in an area that wasn't filled (on one side), added it to the GND net (the other side is GND) and the pour then covers both board sides, including the empty "island". Is there any reason not to do this?
3) I noticed that on terminals on the GND net, after copper pour I see thermal relief as expected, but the traces I had to connect the terminals to the GND added more copper. I realize that (in KiCad - but this didn't seem to work with DesignSparkPCB) I don't have connect all the GND nets together if the copper pour covers the terminal. I've removed all the GND traces from my design where this was the case. I can then add copper fill and run DRC to make sure there are no unconnected GND terminals. Is there any downside to this? (Is this what is normal and/or accepted best practice, I just didn't figure it out initially - or a bad idea?)

I'd like to post the board layout for others to review. What's the best way to do this? Individual Gerber files for each layer give full detail without one layer obscuring the other, but it's harder to see how the whole board connects. With copper pour or without? Would the silk/mask layers also help?

Thanks.

Edit: Added PDFs since Gerber (*.gbr) files would not upload. Separate Front & Back Copper (w/pour) and all layers w/out pour.
Edit: I should add that design grid is 0.025", signal traces are 6/6 and power is 32/6. Default (signal) vias are 10mil drill and 5 mil annular ring, power vias are 16mil drill and 8mil annular ring. All SMD on bottom, and 2 SMD LEDs on top (for obvious reasons) as well as THT RJ45 jacks and 0.1" headers.
 

Attachments

  • Test-F.Cu.pdf
    23.7 KB · Views: 28
  • Test-B.Cu.pdf
    29.6 KB · Views: 18
  • Test.kicad.pcb
    322.9 KB · Views: 11
  • Test-pcb.pdf
    95.6 KB · Views: 26
Last edited:
I don't have a lot of help for you (sorry), but I wanted to comment on the 0603 parts. I don't solder daily, but I will touch up or build small boards. If I am making requirements and I know that I might have to swap some things, 0603 is as small as I want to go. I wouldn't want to do a ton of them on a board, but a handful would be fine. I am not sure if you are planning to populate these yourself, or if the end-user will, but I do not think that 0603 is for the faint of heart. At a minimum, I would make sure you have plenty of spares. As the Old Man in A Christmas Story said, "be careful, they run all over"
 
@toozie21 I have exactly 10 caps, 2 diodes and 2 resistors sized 0603. Also other SSOP/SOIC packages and the pitch is pretty tight, but at least I can find them if they fall on the floor. I'm tempted to get a stencil made, but I'll still be hand soldering, and this is the first prototype, which may have errors before I even get to soldering.
 
Personally I would not bother with a stencil. 0603 are not bad, but then again I do 0402 in the lab as well. Just get a good magnifier. My process is to drop a bit of solder to one pad, tack the parts down, and then go back and do the other pad. Get some fine solder wick, for the TQFP and SOIC parts, makes it easier to get rid of the solder bridges. You will also want to clean the board after, depending on the solder used. I prefer organic so I can wash them in hot water, and dry in the oven.

For review of the board, gerbers work for me. I use gerbview to review everything. I ALWAYS review the gerbers before sending the board out to make sure the solder mask looks OK and the silk screen isn't doing anything unexpected.
 
Tried to upload Gerber files but no luck. Uploaded PDF files instead, but now it looks like they only download, don't display in the post...
 
Last edited:
My point was only that if this board is being designed so that the final version is to be built by an end-user, it might be daunting for someone getting started or has a modicum of soldering skills. If you are going to do all the soldering, or have it eventually farmed out, my concerns are moot.
 
@toozie21 I understand your point - but I've never done 0603 before so it be beyond *my* skills as well. Only one way to find out though... :)
 
ags0000 said:
@toozie21 I understand your point - but I've never done 0603 before so it be beyond *my* skills as well. Only one way to find out though... :)
Click to expand...
No worries. Just take your time, use a nice thin solder tip, and follow Ultimgr's advice about doing one pad and then the next, you will be golden.
 
I was thinking I might use solder paste. Probably not with a stencil since this is a prototype (that might not work) and I'm making only one. Just using a syringe/dispenser and covering multiple pads and a soldering iron with a fine point (as you recommended above).

Is it easier to use thin solder and the "tack one corner first, then other side, then all pins" method with a soldering iron, rather than just placing solder paste and then reflowing with an iron?
 
ags0000 said:
I was thinking I might use solder paste. Probably not with a stencil since this is a prototype (that might not work) and I'm making only one. Just using a syringe/dispenser and covering multiple pads and a soldering iron with a fine point (as you recommended above).

Is it easier to use thin solder and the "tack one corner first, then other side, then all pins" method with a soldering iron, rather than just placing solder paste and then reflowing with an iron?
Click to expand...

I would say it likely depends on the size of the board and number of components. For this board, I think the method you described with thin solder will work well. Also liquid flux can be your friend.
 
0605 is not unreasonable, the annoyance is that a while ago they stopped printing values on 0605 under the guise of environmental concerns, another reason to throw away any you drop.( If you have ones with markings be glad it is old stock).

Everyone will work up their own technique, I tin one pad and push or hold the part with high quality curved tweezers ( cheap ones don't work) , curved allows me to rest my wrist on the bench for stability.

We use a guy that happily hand builds prototypes down to 0201- and he is over 70.
 
I'm a bit late to the game, but what do the various layers represent? In particular, which layer(s) have the ground pour on them? I'm assuming both Test-B.Cu.pdf and Test-F.Cu.pdf have ground pour on them.

Also, take a close look at the ground connections of the bottom three of the 20-pin chips. They do not appear to be connected to anything. Unless you've fixed that on later revisions of the board that might be the cause of the problems that you've been discussing in the signal integrity thread.
 
Last edited:
As 6mil seems to be the limits of many vendors I would only get that thin where I had to, seems like you have plenty of space on most of the board.
 
@P.Short Yes, front and back copper pour are ground. While it's not shown very well, the ground pins are directly tied to the copper pour (through thermal relief). Those .PDF files I had to upload are (for me) almost useless. Can't really see what's going on. If anyone knows the magic to make KiCad print out what I'm viewing in pcbnew I'd appreciate the tip. Screenshots are the next best thing I can think of.
 
You must log in or register to reply here.
Back
Top