The CARMENES Spectrograph at the CAHA observatory in Almeria, Spain needed a temperature stabilised Fabry-Perot Cavity wavelength reference source. The source was designed and built by Ernesto Sanchez Blanco at Optical Developments, Sevilla. I was contracted to build the control system and user interface.
The system is shown schematically below. The temperature noise was around 300micro-Kelvin RMS (Note that this is not the same as absolute temperature stability, since the temperature sensor is enclosed within a servo loop).
The control system was based on a 16-bit PIC microcontroller. As well as controlling two nested thermal control loops, a vacuum valve, a vacuum sensor and a vacuum pump were also controlled. Remarkably, the refractive index change due to residual air within the pressure vessel, even at the 10-2 mbar level, would be sufficient to push the wavelength stability of the Fabry-Perot Cavity outside of requirements so the control system also had to maintain a stable vacuum, in addition to a stable temperature.
The electronics case contained an oven section to further stabilise the preamplifiers but this was not really needed given the high intrinsic stability of the components. The oven section is covered by a bare copper rectangle of PCB material in the image below.
The absolute temperature stability of the system over long periods could not be measured due to the lack of an absolute temperature reference. The Fabry Perot reference used diode temperature sensors and no way could be found to simulate these sensors using non-temperature sensitive components. The controller did, however, have two Pt100 temperature sensor inputs. Since these sensors are simple resistive elements, they could be easily replaced with fixed value high stability resistors to simulate an absolute temperature reference. Any drifts in apparent temperature over long periods could then be attributed to drifts in the electronics of the controller. Two high stability resistors of 100 and 50 ohms were used to allow both offset and gain drifts to be measured. The equivalent temperature of these two resistors were measured continuously for 4 months. For extra precision the resistors were glued to the PCB inside the preamplifier oven. This then effectively removed their, already very low, temperature coefficients.
The data for the 50 Ohm simulated temperature sensor is shown below between July and November. For the first month the indicated temperature dropped by approximately 10mK after which the system stabilised and had a temperature noise of about 3mK RMS for the following three months.
Stable temperatures are particularly important for high-resolution RV spectrographs. From the literature it is not clear how the absolute temperature stability of these instruments is measured. I favor the use of Pt100 sensors in such spectrographs. Although their sensitivity in terms of mV/Kelvin is lower than competing diode temperature sensors, they can be substituted with fixed high-stability resistors. A stable simulated temperature reference can then be switched-in to reveal any drifts in the temperature-sensing electronics.
Some more thoughts on temperature sensors below:
