After my embarrassing mistake by having the component in backwards, I'm wondering if anyone has advice for the newbie (or expert) on reading datasheets.
What conventions are there which aren't obvious.
What's the first thing you should check when something's not working.
What can I safely ignore and what should I worry about?
Anything else?
Read the datasheet carefully. Details matter.
Be Humble: Don't assume there is something wrong with the part right away, assume the problem is in the way you used it.
Frequently, I find that when things aren't working, I've either not read the manual well-enough, or I did something wrong.
Look hard at where pin one is. Make sure you understand if the diagram is a top or bottom view (top is I think most common).
To put this another way, some datasheets give you the dimensions of the part itself, and some datasheets give you the recommended land pattern (aka footprint, solder pads).
If they give you the dimensions of the part, they usually give you a top view and then bottom view. When you're looking at the bottom, imagine you flipped the chip over with your hand.
Terminology There is industry standard terminology that has evolved for electronic components. It is important to understand the definitions.
Absolute Maximum Ratings Understand the limitations of the component.
Specification Conditions Although the terminology is standard the specific test conditions for each specification are not. It is important to account for these differences when (1) determining the proper component and (2) when comparing devices.
Application Notes Most datasheets have a variety of application circuits.
Reference Designs A good reference design will demonstrate the proper application of a component and the proper circuit layout to achieve the specified performance.
Component Package Specifications Review the component packages and the recommended PCB land patterns.
I have placed some additional details at http://wiblocks.luciani.org/FAQ/faq-reading-datasheets.html I will try to add some details about the datasheet specifications that deal with thermal calculations -- thermal resistance, transient thermal response and switching losses. There seems to be quite a few thermal questions (on a variety of sites).
Have another engineer review your work. You will not catch everything.
One thing with data sheets is that the manufacturer usually goes out of the way to make everything obvious. This means that often times you have to look around to find a specific parameter.
Another thing about data sheets is that they usually only put the things that actually are important on them (unlike marketing material!). Point being that it's usually worth reading the entire thing through, at least once.
If you really want to prioritize, check the 'absolute maximum electrical ratings' first. This helps keep from frying the part. You might defer wading through tables of timing parameters or expected waveforms, or look at selected ones first that relate to whatever basic function you're trying to get up and running. But, even the application notes are often useful.
I like to print them out, or at least print the useful pages. Scribble in the margins.
best advice i can give you is to read them carefully but take them with a grain of salt. Don't just blindly do what they tell you, even with it comes to example circuits. I've seen multiple datasheets with blatant errors in example circuits. one famous one i remember called for a 470uF cap instead of a 470nF cap in a reconstruction filter, which the designer(me) followed without questioning resulting is a massive aluminium cap on the rev 1 pcb instead of a 0402 SMD package cap.
Also keep in mind that they are written by people who work on specific parts in specific sectors and usually have a distorted view of what the overall system design calls for. A common example is devices that do A to D or D to A. They will almost always call for separate and isolated digital and analog ground. While in theory this is fine but in practice designing a PCB with segmented ground planes is almost always a terrible idea unless you have massive amounts of board space to play with.
All of the above answers are good and valid. I especially like the point about being humble and looking hard at what you might have missed.
Often we read datasheets and specify parts for our design, quite early on in the project.
It can be a month at the very least before we actually have a fully populated board. :) During such time, often chip manufacturers find slight issues with their parts, and update the Errata for these parts. Most often than not, they usually specify a workaround for the problem which we might need to incorporate in our design.
Unfortunately, I learnt the hard way, after months of trying to debug an issue, read the datasheet over and over, and then found that I had not read the errata. :)
