Avoid data-acquisition mistakes

While data-acquisition systems offer plenty of power, you have to use them properly to get the most out of them. Yet, many users make mistakes that prevent them from taking full advantage of their systems. Here's how you can avoid the most common mistakes:

Brent Boecking, a data-acquisition product manager at National Instruments (Austin, TX), says that while bit resolution is the most important spec, you must consider many specs when choosing a data-acquisition board. Nonlinearity and noise errors can reduce the usable resolution of a 16-bit board to 14 or even 12 bits.

Boecking says factors that are external to the ADC will affect a board's accuracy. He notes that many customers choose boards with multiplexed inputs because they cost less than ones that use a converter per channel. Unfortunately, the nonlinear resistance and capacitance of the multiplexer circuitry can decrease the accuracy of a board accuracy and increase the settling time.

When purchasing a data-acquisition board, pay close attention to the absolute accuracy and settling time. You should probably avoid boards that do not provide these specs.

Arun Sheth, president of DaqScribe Technology (Centennial, CO), says that many customers understand the computers that run the data-acquisition systems his company provides, but they don't know enough about how the analog components work. "They're quite computer-literate," he says, "but not 'analog-literate.'"

Because of this, customers commonly choose transducers that don't meet their needs. They may buy an inexpensive transducer, only to find that it does not have enough resolution or accuracy for the application.

Sheth observes that this lack of knowledge also leads to customers making calibration mistakes. They may use instruments without the proper accuracy to calibrate transducers or not rigorously follow a manufacturer's recommendations for calibration.

Grounding a shield at both ends creates a ground loop that makes measurements less acurrate.

Typically, SCSI and other common cables have multiple unshielded twisted pairs and possibly one shield around the entire cable. This works fine when all the signals are digital, but when the signals are a mix of analog and digital, crosstalk can occur that will affect your measurements. A well-designed cable for data-acquisition will have separately shielded analog and digital sections, individually shielded twisted pairs, individually shielded analog outputs, and a double ground shield to maintain signal integrity.

Daqscribe's Sheth adds that you must connect the shields properly, too. If the shields are connected incorrectly, they will provide no shielding at best, and may make the problem worse. For example, to protect against RFI from 60-Hz power, the shields should be grounded at only one point (see the figure) to avoid ground loops.

According to Gary Marker, an application engineer for Heim Data (Shelby Township, MI), the biggest mistake most people make is that they do not spend time learning the instrumentation before using it. Reading and understanding both the user and technical manuals will help you obtain better measurements and waste less time.

Steve Ross, tech support manager at Mars Labs (Laurel, MD), has seen customers try to measure a current with a voltage card or try to use an input designed for a full-bridge strain gage with a quarter-bridge strain gage. This latter mistake can be fixed by adding a resistor, but the extra hardware can cause problems and introduce errors.

Sometimes sensors get connected to incorrect inputs. Depending on how fragile the sensor is, this mistake could damage, if not destroy, the sensor.

Another common mistake is incorrect sensor wiring. Users may connect grounds to the wrong place or signal lines to excitation signals. To prevent these mistakes, Ross advises, "Double-check your wiring every time, and once you are sure it's right, check once more."

I've often seen engineers not give themselves enough time to program, integrate, and check out a system. The problem seems to result from hardware engineers who aren't familiar with software development, so they provide time estimates that are too low to get the job done properly.

Following good project management practices can prevent this. Break down the work to the smallest tasks. Estimating the time required for small tasks is easier than it is for large tasks, so the final estimate should be more accurate.

Also, plan for a calamity to occur, such as an important data-acquisition board going bad or a forklift running over and destroying your cables. Murphy's Law is not just a suggestion, and if you plan for something bad to happen, you'll be able to recover quickly. And if nothing serious does go wrong, you'll finish your project ahead of schedule.