Gone are the days when we used to just slap some 0.1uF capacitors next to the Power Pin of an IC and be done with it. IC's are now very complex little subsystems, Take for example this cool little 18bit ADC I found. Not only does it convert at an alarmingly fast, 1 MSPS rate, but it also has a built-in amplifier/buffer with programmable voltage ranges of 2.5 to 12 Volts p-p and a low drift / trimmed reference all in one IC. The Programmable Gain Amplifier eliminates a whole lot of external signal conditioning circuitry and the 1 MSPS conversion rate makes signal processing easy because of the sheer speed of the ADC.
To do what this little 16 Pin TSSOP IC does would have taken at least 3 IC's and 1 square inch of PCB space just 10 years ago, today this is it (Figure 1).
Complexity = Risk
Naturally, it has to be expected that some mistakes will be made in the design phase, either in conceptual or implementation errors. This is the fundamental trade-off in speed vs. analysis. There is a fine line that one must walk on every project.
No one wants to mess up the implementation so much so the initial breadboard does not work at all and is unfix-able. That would be a bad trade-off with too much speed. Likewise spending too much time on analysis can slow the project down and all that analysis might not uncover many conceptual errors anyway.
So what is the best way to reduce the inevitable risk of complexity? Well for starters, you can look around for other designs to leverage. To a certain degree leveraging working designs can be a great risk reducer.
On this design, however, I had no experience with the IC so what could I do? The data sheet certainly has some application information, but the real information came from the Eval Board. The good news is that with these complex IC's now the manufacturers always have to have an Eval board and these always include the schematic, parts list, and a sample layout.
With the datasheet and the Eval Board, you can at least compare what they both say. For instance, the datasheet showed the decoupling required for the internal reference section of the IC. Simple capacitors of a certain minimum size were recommended, with no mention of ESR, etc. But the Eval board told a different story. The Eval board showed slightly different values and most importantly it showed some small value resistors in series with the capacitors as if to say: "There is a minimum ESR requirement".
Assuming that the Eval board does indeed work (because the manufacturer sells it), this is good information to have. I can add the resistors to the first design and test to see if they are needed. If they aren't needed I can always replace them with zero ohm resistors going forward or remove them from the final layout. That is much easier than hacking in resistors to an already built board. What a drag that is!
Don't Believe Everything That You Read
The Eval board probably has a Material List and that is also a great start. Look up the parts and study their performance, but don't believe everything. In precision Analog Signal Processing you want to use COG dielectric capacitors in the signal path, this included the reference circuit(s). COG types not only have a very low-temperature drift, but they also have almost no capacitance drop with DC bias and most importantly they have no mechanical noise. This 'Mechanical Noise' is often overlooked and it has to do with the piezoresistivity  of the ceramic material. Many capacitor dialectics are Piezoelectric, that is if you induce stress onto the capacitor, it will produce a small voltage on its own. This can sadly often be found by simply tapping a built PCB with a pair of tweezers, or flexing the board, even slightly. As you can well imagine this undoes all the hard work of even the most carefully designed analog circuit.
For power supply rails we can use X7R types, these capacitors have higher Capacitance-Volume (CV) ratio density than COG but have no piezoresistivity effects. X7R types also have less dramatic capacitance shifts with DC bias than other types and have good overall lifetimes. Finally, the COG and X7R types are routinely rated to 125 degrees C.
While there is an alphabet soup of very high CV capacitors available many types have terrible piezoresistivity and exceeding poor capacitance drop with DC voltage, they are many times only rated to 85 Degrees C. And while they are very suitable for digital circuits, all these other ceramic capacitor types should be avoided for any and all precision analog design. Don't say I didn't warn you, be very careful of the capacitor types you choose no matter what the Eval Board uses.
Don't be too cautious or you won't get anything done, also don't be in too much of a hurry or your design won't work at all. Instead, be careful and check all the sources available to you to get as much implementation information as possible to ensure that you can at least get some information out of that first PCB spin.
Bonus "Non-Analog" Idea
This getting a design second opinion also works for digital IC's especially Microprocessors. Read the datasheet but also check the manufacturer's Eval Board. The great thing about Microprocessors is there are so many eval boards and prototyping boards available. For instance, if you are looking at an ST Microelectronics STM32 Microprocessor: ST Micro usually makes several Eval board for each IC and then there is Olemex, Micro-Electronica, Digilent, Addafruit, Sparkfun and other manufacturers that make boards and provide schematics and Material Lists to compare notes with. This is an excellent way to reduce the risk when using a new microprocessor.
Article By: Steve Hageman / www.AnalogHome.com
We design custom: Analog, RF, and Embedded systems for a wide variety of industrial and commercial clients. Please feel free to contact us if we can help with your next project.