Take the 2-minute tour ×
Electrical Engineering Stack Exchange is a question and answer site for electronics and electrical engineering professionals, students, and enthusiasts. It's 100% free, no registration required.

For an application a temperature sensor is required. an accuracy of around +-3 degrees is acceptable.

For cost reasons I have chosen for this RTD. The sensor is sampled by a 12 bit ADC of an Microcontroller. The RTD is placed in an wheatstone bridge with a regulated 3.3v supply (looking at the supply with an scope I did not find much noise/ ripple). The Wheatstone Bridge is connected to an MCP6N11 Instrument amplifier.


simulate this circuit – Schematic created using CircuitLab

The RTD is mounted on a separate board with about 50cm of wire, which runs through a noisy environment. Therefore the filter stage is placed between the bridge and amplifier.

The desired temperature range is between 50 and 200 Celsius. The gain of the amplifier is set to 15, and is supplied by a single 3.3v supply.

Unfortunately, the measurements are constantly off by 30 to 50 percent. As well as there is still some noise sneaking through the filter. The amplifier is connected directly to the 12 bit ADC.

What can i do to improve this circuit?

share|improve this question
1. What's the temperature range of interest and the gain of the amplifier? 2. What's the amplifier's supply voltage? 3. How is your ADC set up? 4. Can you post a complete schematic of your analog circuitry? –  EM Fields Aug 25 '14 at 8:45
@EMFields The post is updated. –  Roel Aug 25 '14 at 9:21
Kindly reminded, if you have not yet brought it or it is existing installation, RTD is expensive, big size and slow time response (due to size and heat capacity). Usually used for 0.1 degree centigrade accuracy. 3 deg accuracy can have other easier and low cost choice, like thermocouple, with matched IC chip designed for direct connection. –  EE developer Aug 25 '14 at 9:51
For conducted noise, Coil in series to block high frequency, Cap to ground to short circuit high frequency. Metal shiled to block radiated noise input. Higher frequency includes RF noise and any signal higher than required signal. RTD is slow, as in seconds response, any freq higher than, says 1 to a few Hz, can be filtered out. In rare case, is noise is magnetic, use mag shield, Mu metal or similar with 'low magnetic resistance',as enclosed box,as example in professional audio 600 ohm line transformer.Considee off-the-shelf signal conditioner module (match box size, in out ground terminal) –  EE developer Aug 25 '14 at 10:11
@EEdeveloper I'm aware that a Thermocouple may be a better solution. However, I did not manage to find a cheap thermocouple in some form of stainless steel enclosure. –  Roel Aug 25 '14 at 10:19

2 Answers 2

up vote 3 down vote accepted

Consider getting rid of the bridge and instrumentation amplifier. Feed the RTD (via 10k) from the same reference voltage that your ADC uses - this is called ratiometric measurements and removes one of the big error sources i.e. because RTD and ADC are both ultimately fed from a common reference voltage, it doesn't matter if that reference voltage drifts.

So, from memory a 1k RTD will have a resistance of 1000 ohms at 0 degC and about 1690 ohms at 100 degC. If your ADC reference voltage is 3.3 volts and the RTD is fed from 10kohm, the voltage change between 0 degC and 100 degC will be 0.300 volts to 0.477 volts. This is a change of 177 milli volts for a temperature change of 100 degC

Your ADC is 12 bit, therefore it has a resolution of 3.3 volts/ 4096 = 0.80566 mV.

So, for a 100 degC change, the input voltage changes 0.177 volts with a resolution of 0.80566 mV - this means your ADC can resolve 219.7 steps in a range of 100 degC OR, its actual resolution is about 0.5 degrees C.

This should easily obtain the accuracy you want.

By the way, you need to calculate the full range up to 200 degC but i don't see any problems in achieving that to the accuracy you want.

In short, reduce your errors by getting rid of stuff you don't need.

When it comes to filtering, sure use an external low pass filter but also filter in software by oversampling the signal and averaging.

share|improve this answer
The way he has shown it is ratiometric to the 3.3V supply- can't be otherwise unless there is a second voltage reference somewhere. –  Spehro Pefhany Aug 25 '14 at 11:34
@SpehroPefhany - that's the whole point - his ADC reference may not be the same reference as the analogue circuit. –  Andy aka Aug 25 '14 at 11:39
The circuit runs all on the same 3.3V supply, that includes the ADC reference –  Roel Aug 25 '14 at 11:45
@Roel - thanks for clarifying that but the basic premise of my answer still stands - you don't need a bridge (another source of error) and you don't need an instrumentation amplifier (also a source of error). –  Andy aka Aug 25 '14 at 11:57

If you are using the same supply for bridge energization as the amplifier and the MCU, I don't see that much wrong with your circuit,

However, you must remove C4- it should not be present, or no more than a few hundred pF. The way you've shown it, it is highly likely to cause instability. If you can add maybe 1K between the output and C4 without causing ADC errors, it can stay.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.