How smart SPICE software can save you time & money for thermometer sensor testing
01 October 2019
In today’s digital age, it’s no surprise that thermometers are also making the jump from analogue to digital, promising greater accuracy and wider applications as a result. Since accuracy is vital, it’s essential to ensure the development of thermometers happens in a way that leaves little margin for error.
This tutorial was originally featured in the October 2019 issue of EPDT magazine [read the digital issue]. Sign up to receive your own copy each month.
Doing this while remaining affordable can be challenging. As Bert Weiss, Technical Support for resistors at electronic component distributor, Rutronik, and Alain Stas, Product Marketing Engineer for nonlinear resistors at semiconductor & electronic component manufacturer, Vishay explain here, this is why smart SPICE software for thermometer sensor testing is so sought after right now.
Which is the perfect digital thermometer sensor?
As with most big decisions, in order to find the perfectly suited result for you, it’s important to clarify exactly what you’re after first. So in the case of a thermometer sensor, you’ll need to have the following information to hand:
• Temperature range (someone working in an office won’t need the same extremes as scientists in the Antarctic).
• Accuracy required (medical applications may demand more than a cook)
• Sensor tolerances, based on where it will be used and under how much strain it will be placed (large scale regular readings in a factory will put more strain on the unit than infrequent periodic use in the home).
• Calibration and sampling rates also need to be taken into consideration to accommodate specific needs and individual use cases, in order to maximise accuracy and longevity.
Once you’ve established all this, you can start working towards a decision on which is the best sensor. This largely falls into two types: a sensitive non-linear thermistor or a less sensitive resistance temperature device (RTD), like a platinum sensor.
What affects the choice of the right sensor?
No matter how accurate your sensor is, this will always be limited by the quality of the other components used. If you have a precision thermistor, but only use an 8-bit A/D converter (ADC), then it’s like buying a 4K UHD Blu-ray and playing it on a 720p TV – you’re limited by what it passes through. This works both ways, where a cheaper and less accurate sensor would waste the capabilities of a 24-bit ADC, since it doesn’t offer enough accuracy potential in the first place.
So, if ultimate accuracy is what you’re after, then you’ll be aiming to use a top-end thermistor, paired with a powerful ADC. For the ultimate accuracy, there is the option to go for a Class A platinum sensor, which manages an impressive ±0.15°C at 0°C. The downside? You’ll then need to boost that signal with more hardware, which means greater costs and, of course, more cost-to-accuracy components to balance.
All of that costs time, money and trial-and-error testing. That can be avoided by first of all running software-based tests that show which combination of components offers the best end result. Then all you need to do is make one set of purchases and everything can be built to spec, running exactly as you want. That should mean the perfect balance of offering the accuracy you need, while saving money where possible on component purchases.
This is what SPICE-based software offers. The best part? Many of these (like LTspice from Analog Devices) are actually free and – despite being analogue – can simulate digital thermometer builds too.
How does SPICE work?
Thankfully, SPICE (Simulation Program with Integrated Circuit Emphasis) allows you to run either negative temperature coefficient thermistors or a platinum RTD. The measuring current is a low-voltage source and you also have a voltage divider made up of a thermistor and a fixed resistor. This all allows for digitised voltage to run through an ADC. Using a microprocessor and AFE, the digital data can be calibrated for offset compensation.
This can be tested using a direct transient circuit simulation of a 10 kO NTC thermistor of the NTCALUG series from Vishay and a fixed resistor, using LTspice for example. Thanks to analogue behavioural modelling, it’s then possible to digitise the signal and convert the raw measurement data into a temperature. Then you can vary the A/D converter bits between 8 and 24, and the sample time tone of the hold module can also be varied. This last part can be left out when using the Sigma Delta ADC, since the temperature change is measured in the range of 100ms – presuming you go for a 10ms parameter, as this test did.
Figure 1 shows this in action. Figure 2 then shows how the error function calculation, when varying the ADC, does not decrease further at resolutions of n>16. Then, Figure 3 shows how a series resistor R1 can be optimised to offer a minimum error value.
What happens when tolerances are varied?
The next examples, shown in Figures 4 to 6, clarify what happens when the tolerances of the thermistor and the fixed resistor R1 are varied.
In Figure 4, you get a worst case scenario using a 0.5% thin film TNPW series flat chip resistor from Vishay. This uses an NTC with dR25/R25 = ±1% and a B25/85 tolerance of ±0.5%. The result? The measurement uncertainty increases from ±0.4°C at 25°C to ±1.5°C at 100°C.
The NTC has tolerances of R25 and B25/85, while the tolerances of the fixed resistor R1 is 23 – eight cases are obtained and the white curve is the reference curve here. The result is an even distribution of the tolerance values, which ultimately means that the resistor tolerances have been chosen correctly.
On to Figure 5, where the temperature measurement uncertainty has been halved. So now you have R25 = 0.5% and B25/85 = 0.25% for the thermistor and 0.25% for the fixed resistor.
Finally, for Figure 6, the same simulation is performed with the same values as in Figure 5, but with a B tolerance of ±1.5%. This clearly shows the design is not ideal, as at high temperatures, this results in pretty large measurement inaccuracies, and tolerance values of worst case analysis are also not distributed ideally.
Why is SPICE simulation helpful?
As these simulations clearly show, it’s far easier to gauge a setup using the LTspice system, without the need for pricey components – helping identify inaccuracies before it’s too late. Ultimately, this should help to reduce both time used and costs in the development of the ideal digital thermometer.
Contact Details and Archive...