More Thoughts on Via Near Pad

The other day, I wrote about vias near pads. The post got a couple of interesting comments on the Circuit's Assembly blog, where it was also published.

Move via to the leftSome very interesting thoughts there. One of the things Mich said was: "When I was learning PCB design in the 1980's I was taught by a mentor that understood assembly very well." I think that highlights a big component of the problem. I suspect that a lot of folks doing layout today were not taught by anyone but themselves.

CAD packages may have instruction manuals and tutorials, but learning how to use a software package is a lot different than learning how to do the actual process well. It's possible to be very proficient at using a word processor, but still not know how to write well.

It's not an uncommon scenario these days, especially after the economic suckiness of last year, to come in to work expecting to hand off a schematic to the layout engineer only to find that "tag you're it."

Howard, in another comment, suggested that in his experience, filling and plating over vias in pads typically only adds about 8% to the PCB cost. In smaller prototype quantities, it may be a little more then that, but what's the cost of a failed assembly? If you have the room to move the vias off the pads, the only cost may be in layout time. If space is critical or if there are signal/noise/thermal issues that force the vias to be in the pads, then you'll just have to spend the extra to fill and plate.

If you do find yourself suddenly tasked with layout and you've never done one before, find a mentor (or maybe a Minotaur), read up online, call up a manufacturing person, study the Screaming Circuits blog. What ever you do, figure out all these little traps like vias in pad, components library foot print issues, spacing issues, thermal issues, etc. Then dive into the layout and learn from each one. Drink some tea too. It can relax you. Just try to stay away from Oreos and ice cream late at night.

Duane Benson
What's the deal with 1729?

Want Data

RCA12ax7_sq_arms So, I tried to participate in this SparkFun "free day" this morning. They were giving out $100.00 worth of goodies for free per customer (up to a combined total of $100,000), starting at 9:00am MST (8:00am PST here in Oregon).

I was pretty excited about it and had decided to get a new PIC programmer and some pre-assembled jumper wires. I hate crimping those little things by hand. I put it on my calendar for the night before and again for that morning. Then, I found out that I had forgotten a dentist appointment at 8:00 that morning. Bummer.

Just in case it would take more than an hour to burn through that $100K, I went ahead and got ready. I logged in and put the items in my cart. I left the browser sitting there waiting. All I had to do was click the "Place Order" button when I returned after getting my teeth scraped.

But, alas, when I got back, the site was timed out and not accessibly. I refreshed, tried a different browser, refreshed again, etc. I did once get enough of the site to load to see that they had only sold through about $19,000 thus far. Okay. That's not so bad. I could finish making my latte and get in to the office. Maybe try there.

Then, at the office, I was never able to get anything at all from the site to load. All full up. I had to go to a meeting at 10:00 and I thought that if they stayed at around $20K per hour, I might just have a chance of getting through when the meeting was over. But, it was not to be. When I checked in again at 11:30, all $100,000 was sold through. My guess is that so many people were trying in the first hour that the servers only had enough bandwidth to process $20K. After that, enough people gave up trying that the hardware could get the final $80K through in the next 44 minutes ad 50 seconds.

Now here's where my quest for data comes in. I was never able to get more then one click into the process. If all connections were equal, I would presume that everyone would have had the same results. Even if by random chance, someone found a pause long enough to get one page loaded, the chances of each subsequent step would drop astronomically. So, what is it about the Internet that gives some people priority over others? I'd love to see a geographic overlay of the folks that got an order placed combined with their distance from a backbone. Is it distance from a backbone (in hops or in miles) or is it distance from the SparkFun server?

In any case, good for them. It was a fun idea and great gesture of "thanks" Bummer for the inability to handle the load. Here's a Twitter quote from Chris Anderson on the subject: "Google's servers can't keep up with Nexus demand; Free Day brings down Sparkfun. It's 2010--why do we still have these scaling problems?" Ironically, when I first went to Twitter to copy that quote, Twitter was reporting over capacity and I had to wait a while for all the tweets to come back.

Duane Benson
If only my packets were more aggressive...

« December 2009 | Main | February 2010 »