I like to look at a data acquisition system as a measurement chain.
Each link in the chain is responsible for taking some information and transforming it in some way. Why a chain? Every link has a responsibility to maintain the quality of the stage before as far as possible. One strong link cannot make up for weak links.
For example, if your sensor is only accurate to 1V – the best DAQ card in the world won’t make it good to 100mV so you must consider it as a holistic system.
This article lays out the components of a data acquisition system and how they affect the quality of the measurement chain.
With that in mind you must make sure that each stage can maintain the desired:
I will talk in detail about these in a future article. As with all things though, the better you need, the more you pay! But equally, there is no need in spending money on a DAQ card with 10x the precision of your sensor so we can use this to ensure we don’t overspecify as well as underspecify the system.
The sensors are where the physical phenomenon is first turned into something electrical. There are too many types to go into detail here but the key concerns tend to be accuracy, stability and repeatability since there is no digitisation yet (normally, I expect this will be more common going forward).
As well as the electrical characteristics, the mechanical solution must also be considered. For example how you mount temperature probes or strain gauges can make a big difference in how effective they are.
Signal conditioning is present in all these systems, but perhaps less obvious in some. Signal conditioning is electronic circuitry whose purpose is to ensure that the signal we get from the sensor is suitable for input to an analogue to digital converter. Typical examples include:
Signal conditioning concerns tend to depend on the sensors. We are still in the analogue domain so generally, the concern is accuracy, repeatability and stability. A concern can also be frequency response at this stage, especially with filters. Does it treat different frequencies in the same way? If not, how will that affect our measurements?
The ADC is where we convert from the analogue to digital domain. It will sample the analogue signal at specific intervals so that a computer can use the data further. There are a number of key specifications here:
A benefit to some of NI’s hardware which integrates the ADC and signal conditioning into a single device is that you have a single specification covering this and the previous stage. This can improve understanding of your system and make sure your components are well matched.
Now we have the data in the digital domain, it needs to be transferred to our computer. Because it is digital things like noise and accuracy are no longer a concern (most communication links have a higher resolution than ADCs).
The key thing here is speed. Can we transfer data fast enough to keep up with our ADC’s sample rate?
Common options are USB, Ethernet, PCI(e) which give different length and speed options. PCIe and highly integrated systems like PXI can provide the highest bandwidth (4GB/s+ in some PXIe systems) but for many applications, USB provides enough bandwidth and additional flexibility since it can be routed up to 10m.
The final step we consider is the software. Whilst this doesn’t inherently suffer from problems with measurement concerns we can make design decisions to make sure we are sensitive to the rest of the chain.
For example, how many decimal places do you save or display? If you have a temperature sensor that is only accurate to 0.5 degrees, then using 3 decimal places may confuse the user about the accuracy.
Equally, if it is a voltmeter accurate to a millivolt then saving to CSV with one decimal place is reducing the accuracy of the information you collect.
There is also an opportunity for software to compensate or enhance other elements. For example, averaging can reduce noise, or we can use software filters to adapt to variable noise levels.
So, when you design your next system, make sure you know which concerns or specifications are most important to you. Then check these across the full measurement chain to ensure that the data you get is reliable and informative.
We will be publishing more articles in the future to look at specific concerns and some of the mitigations we can put in place to improve them. Subscribe below to be informed when they are published.
Of course, if you need some help now, get in touch.