Sunday, March 26, 2006

Problems in Data Acquisition

Digital data acquisition systems have changed the world a bit! Simulated flight training, cars controlled by automatic systems are reality now. Present day research also exercises a great deal of importance on data acquisition systems. When I look around the characterization tools we use, I found most of the machines are computer controlled and I witnessed analogue to digital transition for quite a few equipment. We have our old Scanning Electron Microscope which has all the controls as buttons or knobs. Whereas, the new SEM we got has all these controls but it’s a mouse click away. One of our distinguished faculties had one mentioned that a trained Chimpanzee can operate these machines with its little finger. Surely data acquisition systems has became integral part of characterization, we no longer sit in front of DTA machine and note down the data for hours, it’s a software which does this, but here is the hitch: Data acquisition system (DAS) can easily put you in trouble if you aren’t careful. For sensitive systems which deal with high pressure or high temperature DAS needs to extremely good otherwise it may lead to catastrophe. Let’s first know where we go wrong most of the time. DAS includes four components: transducer, signal conditioner, data acquisition cards and of course the beautiful softwares. Many of us think that transducers are not a part of it and our control will be better if we have a better DAS, but in practice, the whole DAS is as good or as bad as the transducer is hence the first thing that comes into picture

1. Transducers and actuator are integral part of the DAS and a better control is possible only when a good transducer controls the machine.

2. After transducer it is the turn for the signal conditioning unit. One important feature is where to put the signal conditioner? It needs to be near the transducer as the noise is low at this point. In this regard second important point is the cable. The DA board uses similar connector as the some other ports/boards, but the design will be completely different. We have a tendency to use same connector for all application which may at times become very risky. A system should have a shielded cable in order to reduce the noise level. Wiring is important, its better to check the connection twice before starting the acquisition.

3. I will discuss ignorance here. We do data acquisition from both grounded and ungrounded sources. In case grounded source we assume our ground is at zero potential, which may not the case. Moreover, if you use transducers grounded using different ground with different potential, the data which we are acquiring will be highly erroneous. In case of low voltage measurement its quite a problem, your machine may stop working for such small problem. The punch line is use same ground for all the transducers and make sure it’s a zero potential.

4. Aliasing was one of the problems with DAS. According to Nyqist theory if signal has no frequency component above ½ the sampling rate then the signal can be completely reconstructed, and if there are frequencies above that the data will be corrupted. In practice the complete set of frequencies aren’t free from higher frequencies but there are ways to handle it, but if it goes beyond a limit it’s a problem. Suppose you are using a sampling rate of 300, and reconstructing a signal with frequency of 10 Hz, the system is absolutely fine but is its 280 Hz, well it can not show anything, that’s good but you will get a wrong result when its shows a value of 20 Hz. You have a completely wrong info. The soulution is not difficult and most of good DAS systems have this protection, a low pass filter before digitization is enough to reduce the Nyquist frequency effect, and present day most of the companies take care of it

5. Softwares! They are beautiful with all kind of flexibility to analyze the data acquired and present them in beautiful plots. The problems are in disguise, be very careful in using them and don’t belive the information you are getting without verifying it. These programs are good, but many a times they won’t consider the complexity that is hidden in your system, hence analyze the data in a wrong fashion. Its sound very general and more like software haters speak, but it is really important. Now-a-days DAS companies come up with very good softwares, still there are always ifs and buts.

I could remember few of the issues related to DAS, but there may be more and comments regarding this are really welcomed, as far i am concerned, I felt like putdown this issues.

Some interesting old articles are here:

Mistakes in DA

Article by Strether Smith

No comments: