Share |
Login Form
Newsletter(s)



Receive HTML?

Latest Members




Caveat emptor: Seven warnings to heed while reading datasheets

 
User rating
 
4.0 (1)

Datasheets, particularly in the world of analog and mixed-signal IC, are often fascinating bits of technical literature. On the one hand, they are engineering documents that specify what an OEM designer can expect from the IC and, more often than not, include application information to jump-start the design-in process. On the other hand, they are marketing documents that attempt to capture a customer’s notice and inspire interest. For most beasts, two hands are enough, but here we have a third.

Datasheets are also contractual documents. They stipulate what customers are buying and what the manufacturer is obligated to supply in response to a purchase order.

Balancing the demands of the hands [sic] is a delicate craft that, more and more, renders datasheets examples of documents that you must study with care because, all too often, they allow, if not invite, misinterpretation. Here are seven issues to keep in mind while perusing datasheets for analog or mixed-signal ICs:

  1. Don’t ascribe significance to qualitative descriptions such as low-noise, low-power, high-speed, or high-linearity until you’ve examined the relative specifications. What is attractively low or high in one application is pedestrian performance or out-and-out insufficient in another. Such phrases do serve to identify the manufacturer’s intent and direct your attention to the relevant parts of the spec table, but until you see the numbers and are convinced by them, don’t put much stock in the marketing pitch.

  2. Design to minimum and maximum specifications—they represent the manufacturer’s commitment. Specifications marked typical may give guidance but you can’t bank on them to support your design’s specified performance. Depending on the vendor, they may represent the true statistical average of samples drawn from some small number of lots early in a products lifetime—usually immediately before the datasheet’s initial release. Manufacturers rare confirm them after that point. If a particular spec is important to you and only available as a typ, ask the manufacturer how the typ spec was established and ask for what the data’s standard deviation. Also ask how often (if ever) they reconfirm that production parts conform to the original samples’ statistics. You may or may not like the answers but at least you’ll know what to expect from the ICs you purchase.

  3. Read and understand all test conditions before evaluating the relevance and significance of any specification in the context of your application. Beware of cases where test conditions differ significantly from the operating environment your application will supply. If test conditions aren’t given or aren’t clear and complete, ask the vendor to supply more information or move on to a different part.

  4. In a similar vein, beware of specifications given under operating conditions that run counter to the device’s normal operation. My favorite recent example is an amplifier spec sheet I reviewed recently that made its output power claim under conditions that drove the amplifier into clipping. That is to say, the manufacturer specified the output power of a supposedly linear transfer function outside of the device’s linear operating range. The manufacturer in this case reports an unclipped output power 80% of nameplate claim, and looking at the characteristic curves, I wouldn’t trust the part much beyond half of that. I consider such specsmanship to be disingenuous at best, and suggest you treat such instances likewise.

  5. Beware of numeric specifications that represent spot measurements of nonlinear parametric behaviors. Use these number in combination with characteristic curves that indicate how the part behaves as the driving parameter varies. For example, an amplifier’s datasheet might give a distortion spec under specific conditions of supply, signal amplitude, and frequency. Is that number likely to be representative of the amplifier’s performance under other conditions? Check the characteristic curves to see what you can expect from a typical part under conditions your application imposes.

  6. Remember that the characteristic curves represent typ, not min or max performance. As such, they are guides, not guarantees. Pair them up with min max numerical specs and you’ll have a pretty good idea of what kind of ride you’re in for with a given part.

  7. Don’t forget to examine performance over temperature. There are precious few applications that operate at 25 °C ± some small number. Particularly with so many portable products, you need to be sure that the ICs in your design will satisfy your requirements over the full range of anticipatable operating conditions including temperature. Got a great little ADC for your next design that guarantees DNL (differential nonlinearity) of ±1/2 LSB at room temp? How does it do at 50 °C or -25 °C? Still no missing codes? Great! DNL decay to ±5 LSBs? Maybe you want to keep looking if your application depends on this spec.

User reviews

Average user rating from: 1 user(s)

To write a review please register or login.
Overall:
 
4.0   (1)
 
 

Both sides of the fence

Overall:
 
4.0
0 of 0 people found the following review helpful

I started out my technical career designing DRAMS at Intel and we certainly wanted to have the most impressive datasheets in comparison to our Japanese and American competitors. The temptation was to cherry-pick the ICs used to characterize our new DRAM in the hopes that the impressive specs could be backed up by a fab getting enough yield to make money.

Intel eventually left the IC memory market for DRAMS because they couldn't stay number one or number two.

In EDA I've worked in marketing and created glossy brochures extolling the virtues of my software tools, making them sound overly impressive and powerful.

You're right, buyer beware when reading vendor information.

 
 
Written by :
Joshua Israelsohn