SDN0002

This note describes a laboratory test of the prototype temperature controller developed to provide thermal control of the GNIRS optical bench. The justification for this is presented in SDN007.01, which concluded that thermal control of the bench is preferable to attempting to isolate it. The stability requirement on the GNIRS bench is ± 1 K from the target temperature of 60 K.

The GNIRS Bench Temperature Controller prototype, which is based on an analog-output Omega process controller, was tested using the cryogenic test chamber CROC. The test geometry mimics that which will be used in GNIRS, although the thermal loads are, naturally, much smaller. In GNIRS, the plan is to employ power resistors to inject heat at some point in the thermal distribution system between the cryocooler first stage and the bench, with a temperature sensor on the bench itself providing feedback.

In the CROC setup, the single 130 cryocooler was attached with copper braid to a copper bench suspended within the test dewar (the same setup as used for the Phytron motor tests). A 9.4 kg aluminum block was thermally strapped to the copper bench with a copper strap with an estimated thermal conductance ~2 W/K. The theoretical thermal time constant of this setup is about 3000 s for the initial cooldown, decreasing to about 1000 s at the target temperature of 60 K, due to the decrease in specific heat of 6061 Al at lower temperatures. Two circuits of 15 W 50 W power resistors were mounted on the copper bench and were powered with 0 - 27 VDC from the controller. Two 1N914 temperature sensors, one of which was used as the monitor/control point, were mounted on the aluminum block. Additional sensors were installed on the copper bench and the inner radiation shield of CROC. The test setup is shown in Figure 1.

The Cu plate temp diode is monitored with the NOAO "ADAM" A/D box and Excel software on a PC. A multi-point look-up table serves to convert diode voltage to K. This is the same configuration as used for testing the Phytron motors (SDN0015). Within the GNIRS Temperature Controller the diode voltage is buffered, inverted with an offset voltage and a gain of -1, then sent to the Omega process controller. A 2-point linear approximation transforms volts to K. Needless to say, the temperatures indicated by these two setups differ. The Omega temperature reading is about 7 K greater than the Excel temperature at around 60 K.

We ran three tests, usually starting with the Cu plate at 50 -51 K (Excel) and the Al block at 60 - 62 K (Omega). The control setpoint was either 65 or 70 K. Test 1 had the PID P-term set to 006.0, which had worked well during a pretest with the Al block at room temperature and pressure. When tested with the Al block cold, the temperature overshot by almost 2 K and the control voltage (V-control, the output of the Omega device) had a deep valley and peak before approaching a final value. After an hour the Al block temperature was essentially at its final value and holding to +/-0.1 degree.

For test 2 the P-term was increased to 008.0. The Al block temperature still overshot by over 1 K, V-control still had a valley and peak, although smaller. Final temperature was approached after about an hour.

Test 3 used a P-term of 010.0. V-control had a shallow valley and almost no following peak. The block temperature overshot by less than 1 K and approached its final temperature in less than an hour.

The temporal behavior of the Al block temperature and V-control for the three values of P are plotted in Figure 2. The actual voltage applied to the pair of 15 W heater resistors is 2.8 * Vcontrol, so the applied power is approximately W = 1.04 * (Vcontrol)².

Although the heat capacity of the Al block can be calculated, the thermal conductance of the strapping to the bench is less certain, since it includes the thermal resistance of the mechanical interfaces to both the bench and the block, which cannot be calculated from first principles. The estimated conductance ~2 W/K given above includes no allowance for the interface resistance and is thus almost certainly an overestimate. Nevertheless, one can present several arguments to suggest that the actual conductance was not significantly less than this value.

An extremely rough estimate of the cooldown profile, based on the Cu bench cooling curve during the Phytron motor tests, indicated that the Al block should reach 55 K in approximately 24 hours. Temperature monitoring of the bench and block during the first 8 hours of the cooldown showed the bench to be cooling more slowly than in the Phytron tests (not unexpected, since it is more heavily loaded in the current experiment), and the block to be cooling slightly more slowly than the model predicted. Nonetheless, the Al block had reached ~ 56 K after 23 hours.

During the cooldown phase, when the Al block reached 190 K, it was cooling at a rate of 18 K/hr; using the heat capacity for Al at this temperature gives a heat extraction of 34 W. The measured temperature difference between the block and bench was ~ 47 K (somewhat uncertain because the block was measured using the Omega controller and the bench using the ADAM box), giving an effective thermal conductance ~ 0.72 W/K. This is quite a bit less than expected. Since this is a dynamic measurement, there is room for some uncertainty, particularly in the assumption that the Al block and Cu bench are isothermal.

During the last temperature control test, both the bench and block temperatures were monitored using the ADAM box, removing the relative calibration uncertainty. At equilibrium, a value of Vcontrol of 2.75 v (7.6 W) maintained a temperature difference 5.1 K between the bench and the block, yielding a conductance of 1.5 W/K.

Using the heat capacity of the Al block at 60 K (1880 J/K) and 1.5 W/K for the thermal conductance yields a thermal time constant of 1250 s. This is consistent with the results in Fig. 2, which show that one reaches equilibrium in ~ 1.5 - 2 hr (4 - 5 t).

These tests provide some confidence that this control configuration should work in GNIRS. The temperature control stability achieved in these tests was far better than the GNIRS requirement of ± 1 K. For this experiment, a P-term of 10 - 12 seemed to be appropriate for near-critically damped temperature control. We can start with this value for GNIRS testing. The thermal time constant of GNIRS is sufficiently large (~ 35000 s, or 10 hr) that it will not be as responsive, and that one would be able to damp, but not entirely compensate for, things like diurnal environmental temperature fluctuations. The design of GNIRS is to minimize the effect of such external variations on the bench in any case.

We learned two additional things in the course of this test: the Apex PA-26 power op-amps appear to be more reliable than the original PA-29's used in the prototype. In addition, heat-sinking the PA-26's so that they operate reliably through the entire range of voltages and currents is very important. We will improve the heatsinking of the Apex devices in the GNIRS Component Controller.

Fig 1. The experimental setup in CROC, showing the Cu bench, 9.4 kg Al block, and Cu thermal strap. The Al block rests on two insulating blocks, which provide negligible additional thermal coupling to the Cu bench. The temperature sensors are 1N914 diodes.

Fig 2. Plots of the Al block temperature (red) and heater control voltage (blue) as a function of time after a 5 K change in the temperature setpoint. The temperature/voltage axis scale is relative and is intended to represent the relative change and ultimate stability of the system. The three plots are for P = 6 (left), P = 8 (center), and P = 10 (right).

If you have any questions or suggestions regarding this website, please contact Melissa Bowersock.