Build a constant current source for PT1000

Hi Jan,

Q1: Are the recommended steps for common-mode and differential filtering enough to reduce emitted interference  

PT1000 RTD is typical thin film type with R at nominal 1kΩ at 0C (.2-wire RTD is mandatory). What kind of emitted interference are you concerned about? RTD is typically driven with constant current source. If the driven current is low, it will use less power with less self-heating issues. For 4.5m-8m RTD length, PT1000, 2-wire type is good options. What is the temperature accuracy that the application is designed for?  

There are many design approaches for the topic. I am enclosing a design method with Instrumentation Amplifier(IA), see the application below. If your application is required to keep the high temperature measurement accuracy, IA is one of the best approach. for the design requirements.  

https://www.ti.com/lit/an/sboa425/sboa425.pdf?ts=1616778419923&ref_url=https%253A%252F%252Fwww.google.com%252F

If you are able to keep length of PT1000 sensor matched (same R at either side of PT1000 input)  IA should be able remove and reject any common mode noises at the analog sensing front end. Since the rate of temperature change is fairly low, you can place LPF to eliminate unwanted high frequency noises at the analog front end, see the attached application note. 

If you are able to tell us the available power source (voltage level, single or dual supply rails)  in your application, we can recommend IAs for your application. 

Q2: Are there special circuits for the current source which have a good immunity against interference.

REF200 is dual package current sources that are good for RTD application.. It will be better to use the IC package rather than design on your own, unless your temperature operating conditions is outside of a current source.

Please tell me more about your application. What kinds of immunity are imposed in the design requirements? What kind of environment that the sensor is required to be operating under.

IA has differential input sensor stage. If you are concerned about external noise interferences, you may consider differential input to output through out the design approach. 

If you are going to use ADS119, it is likely that you only have +/-5V available in the design.  INA333 may be an option. . 

If you the design is required to meet and/or pass particular test compliances (automotive, military or aerospace etc.), please let us know. We can select a part that is able to address or meet these design requirements. .

I am enclosing 3-wire RTD application note below (it is 3-wire design, but it is also applicable for 2-wire application).

https://www.ti.com/lit/ug/tidu969/tidu969.pdf?ts=1616776511847&ref_url=https%253A%252F%252Fwww.google.com%252F

https://www.ti.com/lit/ug/slau520a/slau520a.pdf?ts=1616777373742&ref_url=https%253A%252F%252Fwww.google.com%252F

Best,

Raymond

Hello Raymond,

thanks for your reply. As I represent the mentioned customer, I will communicate directly in this forum.

Q: What kind of emitted interference are you concerned about?

The sensor lines run in dedicated cables, but are most likely unshielded.
The sensor cables are embedded in a larger cable, with DC lines that carry up to several hundred amperes of DC current with up to 1000V DC coming from a DC-DC-converter. Allowed disturbance voltage limits for these DC power cables are:

In addition, another parallel signal cable is embedded in this larger cable to transmit PWM in the kHz range and HomePlug PLC modulation as well. As far as I can see, all of this lines are unshielded, so I'm concerned mainly of interference coming from the described other parallel cables.

RTD is typically driven with constant current source

Preferred is a 300uA current source.

Q: What is the temperature accuracy that the application is designed for?

- Full ADC measurement range is from -30...+100°C
- Used PT-1000 has a sensor error of +/- 1.5°C @ 90°C
- 300uA leads to self-heating of max. 0.1K @ 90°C

The application does not allow an offset calibration.
90°C is a particularly important temperature value, since 90°C means an overtemperature situation that must be detected.
I.e., with a sensor reading of 88.4°C it must already be assumed that in reality 90°C have been reached.

On the other side with worst case negative sensor tolerance, a sensor reading of 88.4°C could be 87°C in reality.

Regarding accuracy of the circuit, up to additional +/-0.5°C @ 90°C introduced by the circuit are considered as acceptable.

Q: If you are able to tell us the available power source

Available power sources: +3.3V, +5V, +15V, -15V

Q: What kind of environment that the sensor is required to be operating under.

The sensor is operated in an outdoor application.

Some additional information:

There are two PT-1000 sensors in the application. While the ADS1119 supports reading two sensors in series in a ratiometric measurement with a single current source, we have to use two ADS1119, each with one sensor and current source.

The reason is that we have to realize an additional, purely hardware-based overtemperature detection in SIL-1. For each channel one dedicated circuit.
Ratiometric measurement is preferred because then the SIL-compliant circuit simplifies (deviation errors of the current source do not change the measurement accuracy).

When using the approach with the IA's, we would lose this circuit structure. So for clarification - do you advise against ratiometric evaluation?

Kind regards, Oliver

Hi Oliver, 

Thanks for your design requirements. 

I took your design requirements as a reference and implemented in the following Tina model. As suggested previously, I continue to use INA333 instrumentation amplifier for the RTD (PT1000) application. If you want to use higher voltage IA, we have that as well (which is comparable to INA333, such as INA821, INA818 or similar).

The image below is the designed references that I used for the simulation. 

In the simulation below, I placed input differential LP filter at approx. 245Hz. The temperature change or response rate can be readily measured and monitored up to 25Hz. In addition, I assumed 28 gauge RTD and lead wire up to 8 meter in length. The RTD cable bundle will have better electrical and/or magnetic attenuation, if it is twisted and matched pair. The RTD simulation is driven with 100uA in constant current. If you want to drive PT1000 in 300uA, we can modify the simulation accordingly (also means that we may need use high voltage IA parts). 

The RC LPF filter has attenuation range between -40dB to -60dB. INA333 has additional CMRR up to 110dB, therefore, I did not spend a great deal to take care of simulated input noise, which I assumed 1Vpk. The RTD input circuit will have input noise attenuation up to -150dB to -170dB range, and the coupled common mode noises or interferences at the analog front end of INA333 will not be significant due to the design and the part selection. 

Enclosed is a ratiometric RTD design application for your reference. 

https://static5.arrow.com/pdfs/2013/11/24/3/19/23/744/txn_/manual/slau520.pdf

INA333 PT1000 03292021.TSC

Assumed this is a proposed draft design. Once you have decided what you want to do, I can modify the schematic and/or make further improvements in the design. 

Please let me know if you have any questions. 

Best,

Raymond 

Hello Raymond,

thank you very much for your support and the valuable information.

One question regarding the RC LPF filter you implemented in the simulation. You used the PT1000 resistance itself as filter resistance, right?

Are there any disadvantages if additional filter resistors in the LPF stage are used and the filter capacitor values are reduced accordingly?

In the simulation, the current source actually drives the RTD only, but not the wire resistance. To be correct, the current source must flow through the RTD and the wire resistances. The sensor cable actually is AWG20 (0.5mm2, stranded wire), so the wire resistance can be assumed smaller, approx. 0.27 Ohms per line.

Kind regards, Oliver

Hi Oliver,

Q1: You used the PT1000 resistance itself as filter resistance, right?

Yes, this is my intention. I am trying to keep the input RTD leads in low impedance. Since the pair of wire is inside of a cable bundle, low impedance will have better noise immunity against magnetic and electrical induced noise and/or spike induced interference into the RTD sensor/cable. I am hoping that it will have better radiated and conducted susceptibility at high frequency as well.

(The differential LPF is configured at approx. 250Hz, From 250Hz --> 2.5kHz -->25KHz --> 250kHz etc, you have a minimum 60dB attenuation up to 250kHz. If you are talking about MHz signals, the -20dB/decade roll off will continue to attenuate significantly. INA333's 100dB CMRR (DC-60Hz) works well at lower frequency, though it will start to diminish at higher frequency, see Figure 19 of the datasheet), but differential LPF will attenuate more significantly as input noise increases above MHz range.). 

The overall INA333's transfer function looks like this: Vout_RTD = Gain(V+ - V-) + Gain*Vos_max + Vos_DA + Vref, where Gain = 11, Vos_max = 25uA and Vos_DA = 75uA (2nd stage of Differential Amplifier in INA333), Vref = 1.25Vdc (say 200uA constant I is used, see the attached Tina simulation). In this case, I changed the constant current source to 200uA, where REF200 is the constant current source (combine 2x100uA into one 200uA current source). 

If you are able to increase Gain(V+ - V-) signal amplitude, the overall errors will be considerably small. As long as input signals or noises are behaving in common modes at the analog front end of INA333, it will be rejected by the part. 

INA333 PT1000 03312021.TSC

Q2: Are there any disadvantages if additional filter resistors in the LPF stage are used and the filter capacitor values are reduced accordingly?

The answer is partially replied above. If you insert small filter resistors (matched) within the RTD input signal, it is likely that IR voltage drop errors will be small or insignificant. If there is an error in the inserted filter resistors, the differences in voltage (200uA*ΔR) will convert from some of common mode into differential mode at the INA333's input, which it may affect the total temperature measurement errors.  

In additional, the temperature accuracy requirements are "PT-1000 has a sensor error of +/- 1.5°C @ 90°C", so I think that the  measurement errors mentioned above may be insignificant, if your filter resistors (low values) and RTD wire length are reasonably matched. Perhaps you need to reduce the differential capacitance somewhat, which it will increase the BW of the INA333 circuit or temperature response somewhat. (I do not have information about your temperature response characteristics of sensed object. Typically the rate of change in temperature is not very fast; and1Hz in rate of change in temperature is already very fast). 

Best,

Raymond

Hi Jan,

I would do it this way:

No current source, just one precision resistor and a 24bit delta-sigma ADC.

To keep the self-heating of PT1000 minimal choose a suitedly low reference voltage REFP and/or a suitedly high ohmic precision resistor Rref.

Kai

Hello Kai,

thanks for your reply.

How does this circuit behave regarding the coupled common mode noises / interferences at the PT1000 and its sensor lines?

Kind regards, Oliver

Hi Oliver,

there's enough room for adding filtering caps. To avoid rectifying effects across nonlinear cap characteristics, I would use high quality plastic film caps (polypropylene or similar). Ceramic caps should be avoided here, unless NP0 is used.

Kai

Hello Kai,

thank you for your answer and suggestions.

Kind regards, Oliver

Hello Raymond,

thank you for the further explanation regarding the filter circuits.

Kind regards, Oliver

Good luck

Kai