No Title
TOFr5 Temperature Test
(system test #1)
Sep 28, 2004
bill, jianhang, and jing for the TOFr5 Construction Team
(Lloyd Bridges, Geary Eppley, Jerry Hoffmann, Justin
Kennington, Jing Liu, W.J. Llope, Ted Nussbaum,
J. Schambach, and Jianhang Zhou)
[ PDF Version ] [ HTML Version ]

Contents

1  Introduction
2  Equipment
3  Temperature Calibrations
4  Temperature Asymptotes
5  Power Estimates
6  Cold Water Test
Index

1  Introduction

The on-board electronics for TOFr5 drop ~ 140W, implying a full system (120 trays) of this design would drop ~ 17 kW into the very tight cylindrical space between the STAR TPC and BEMC systems. This heat must be removed from inside the pole tips as efficiently as possible in order to limit side-effects in the neighboring subsystems. Radio-frequency (and mechanical) shielding for the TOFr5 electronics using an outer aluminum skin is necessary, Such a skin retains heat at the electronics. The HPTDC chips begin to fail to operate sensibly if their local temperature reaches ~ 60 °C. Efficient metallic thermal paths between the ground planes of the electronics and the aluminum mechanical structure of TOFr5 imply some heating of the MRPC gas box, which is known from Run-4 to significantly increase MRPC noise rates and currents. The actual magnitude of all of these effects for the final configuration of "final" TAMP and TDIG boards (8 each), mounted on the final mechanical structure (gas box, plus mechanical support for all MRPCs, electronics, and outer cover) for Run-5, and powered up, were thus of special interest. A "System Test" of the final components of TOFr5 was performed in order to investigate these power, heat flow, and temperature issues, and the results are described below.

2  Equipment

The TOFr5 electronics layout returns to the Run-3 design of two layers of circuit boards, where the first layer has comparatively few components. This first layer, the TAMP layer, closes the gas box and provides pre-amplified and input-protected MRPC signals to a second layer of TDIG boards which do the discrimination, digitization, trigger matching, hit counting, and data buffering. Compared to TOFr' (Run-4), the TOFr5 design includes a more tightly-integrated water path that uses square 1/4" copper tubing that runs along the two long lengths of the tray and between the two layers of electronics boards. A view of the h ~ 1 end of the TOFr5 tray, cooling loop, and electronics at an intermediate stage of the TOFr5 fabrication is shown in figure 1.
./plots/tofr5_ttestendview2.jpg
Figure 1: End view of the TOFr5 tray at an intermediate stage of the fabrication. The top layer of electronics are the TDIG boards. The square copper cooling loop is just under the TDIG layer, and just under the cooling loop is the TAMP layer which closes the gas box. [JPEG] [PDF]
The vertical gap between the TAMP and TDIG layers is set by connectors at 7/16". The cooling loop is 1/4"-thick, leaving 3/16" which is made up of two aluminum pieces (1/8" and 0.05") and thermally conductive plastic shims (7 and 11 mils). The hope was that this water to electronics thermal path would be efficient enough to bring the majority of the electronics' total power out of the STAR pole tips via water and hence keep the local temperatures comfortably low. We tested the efficiency of this path with the set-up shown in figures 2 and 3.
./plots/TtestSetup.jpg
Figure 2: Schematic view of the equipment for these tests. [JPEG] [PDF]

./plots/Ttest_photos.jpg
Figure 3: Shown on the left is the TOFr5 bottom and top assemblies and the thermocouple wiring for the interior of the gas box. In the center, TOFr5 reassembled and running during these tests, and on the right is the heat exchanger. [JPEG] [PDF]

The water source was simply Rice tap water, available at the faucet at ~ 2.3 Gpm and 31 °C. Approximately 10' of 1/2" vinyl braided hose brings the water to a heat exchanger (an ice cooler holding 40' of coiled 3/8" Copper tubing), then ~ 30' of hose connects to the input of tray cooling loop, and ~ 40' of hose returns the water. The flow rate through the heat exchanger and the 30' hose was 2.2 Gpm, while the (more relevant) flow rate through the entire system was 1.36 Gpm. The water flow rate in STAR is locally adjustable in the range from ~ 0.5 to 2 Gpm.
The full complement of TAMP and TDIG pairs (8 pairs) were installed on the final TOFr5 mechanical components [1]. Low voltage for these electronics was provided by the Kepco supplies to be used in Run-5. The low voltages, currents, and total electrical power for TOFr5 are summarized in Table 1.
Table 1: TOFr5 low voltages, currents, and electrical power estimates.
V (Volts) I (Amps) P (Watts)
+4.2 16 67
+4.2 11 46
-8.5 3 26
(total) 139±10
The temperatures at a number of locations inside the tray and the electronics volume are obtained from a Kinetics Systems 1992 Thermocouple Termination Panel and a 3516 Scanning A/D [2]. The total number of temperature measurements is 32, one of which is internal to the 1992 (the "reference temperature"). The remaining 31 channels read remote temperatures over type TT thermocouple wire. The Kinetics units are read-out via CAMAC over a GPIB interface to a PC with a custom DAQ code. This code reads all 32 temperatures continuously with a 4 second interval between reads, and stores each 32-channel read with a time stamp to an ntuple which is analyzed in ROOT. The thermocouple locations on-tray, on the ambient air, and in the water path, are summarized in Table 2. The calibration of the data from thermocouple system and the resulting temperature resolutions in °C following these calibrations is discussed in section 3 below.
Table 2: The thermocouple locations for the present tests.
T/C Location T/C Location
0 Reference Inside KS1992 16 TDIG Regulator pos6
1 Tray Inside Air h ~ 0.2 17 TDIG pcb near sensor pos4
2 Tray Inside Air h ~ 0.5 18 TDIG pcb near sensor pos4
3 Tray Inside Air h ~ 0.8 19 Air Gap Above TDIG pos1
4 Tray Side Wall Inner h ~ 0.2 20 Air Gap Above TDIG pos4
5 Tray Side Wall Inner h ~ 0.2 21 Air Gap Above TDIG pos6
6 Under TAMP near sensor pos4 22 TDIG pcb top pos0
7 Under TAMP near sensor pos4 23 TDIG pcb top pos7
8 Air Gap Below TDIG pos1 24 Lab Room Temperature
9 Air Gap Below TDIG pos4 25 Lab Room Temperature
10 Air Gap Below TDIG pos6 26 Water At Tray Input
11 HPTDC1 Chip pos6 27 Water At Tray Input
12 HPTDC4 Chip pos6 28 Water At Tray Output
13 HPTDC2 Chip pos6 29 Water At Tray Output
14 HPTDC3 Chip pos6 30 Water In
15 TDIG Altera Chip pos6 31 Water In

3  Temperature Calibrations

A simple calibration of the thermocouple system is required. The first step is the calibration for the reference temperature, cf. figure 4. In the upper left frame, the black histogram depicts the reference temperature in °C versus the read number over a several day running period. An approximately 1 °C periodic (daily) variation in the reference temperature is observed. Also shown in this frame as the cyan histogram is one (of two) ambient air temperatures measured over the thermocouple wire. The upper middle frame shows the same reference temperature as the black histogram, and the second ambient air temperature is the cyan histogram. The strong correlation between the reference temperature and the two ambient air temperatures throughout the running period is clear. Shown in the other four frames are the same black and cyan histograms as in the upper middle frame. Also shown in these four frames as the blue histograms are the four water temperatures. When the water is not flowing through the system, these temperatures are also tightly correlated with the ambient temperature. When water is flowing these temperatures are off-scale vertically in this plot.
./plots/t01.gif
Figure 4: The correlation of the 1992-internal reference temperature and other ämbient" temperatures measured over the thermocouple wire. See the text for the details. [EPSF] [PDF] [PS]
The first calibration then simply involves correcting for the reference temperature. This is done by calculating a running average of the reference temperature over the last two reads, and forming the difference between this running average and the initial average of the reference temperature. This provides a read-dependent offset that is used to correct all 32 temperatures for the variation in the reference temperature.
The second step takes care of channel-dependent offsets in the 1992+3516 system. This is done by forming the average values of the temperature read-out by each channel over ~ 200 reads when all thermocouples should be reading the (same) room temperature. These offsets can reach ~ 0.5 °C. They are tabulated and then used in subsequent analysis passes to remove these offsets.
./plots/t02.gif
Figure 5: The temperature difference resolution following the calibrations. [EPSF] [PDF] [PS]
The single-channel resolution of the system is indicated in figure 5. Shown in each frame is the difference between a water or ambient air temperature and the first ambient air temperature, i.e. frame i shows t[i+24]-t[24] where 0 £ i £ 5. The standard deviation of these difference Gaussians is 0.13-0.15 °C, implying a single-channel resolution in single DAQ reads of approximately 0.14*sqrt(2) ~ 0.2 °C. A typical calculation of an average temperature for some thermocouple at some point during the data-taking involves no less that 15 reads ( ~ 1 minute of real time) so the statistical uncertainty on one-minute averages is ~ 0.2/sqrt(15) ~ 0.06°C. We see slight drifts in the temperatures beyond that expected after the calibrations that are of order 0.02 °C (seen in section 5 below). The present calibration scheme could be augmented to suppress this drift, but since the drift is so small in magnitude we choose to simply assign this as systematic error. The total single-channel resolution in single reads is thus estimated to be 0.2Å0.02=0.2 °C, while the total resolution on one-minute averages of single channels is 0.06Å0.02=0.063 °C. The total resolution on the difference of two two-channel averages is the same as the total single-channel resolution, or 0.2 °C for single reads and 0.063 °C for one-minute averages. Certain estimates made in section 5 are based on averages of this difference of single-read averages over a few hundred reads, for which the total error is taken to be the 0.02 °C systematic error.

4  Temperature Asymptotes

These tests consist of a number of different phases, each with a specific configuration of the system (LV on or off, water flow on or off, and so on). Each phase was generally run long enough for asymptotic temperatures to be reached. A graphic depiction of the entire ~ two day experiment is shown in figure 6. The various phases are marked by the different colors and correspond to the configurations shown in the accompanying table.
./plots/t03.gif
Figure 6: A single temperature in the tray vs. the read number. [EPSF] [PDF] [PS]

Phase Configuration
0 LV On, Water Off
1 LV On, Water On (31 °C)
2 LV On, Water On (colder)
3 LV Off, Water On
4 LV Off, Water Off
5 LV Off, Water Off
6 LV Off, Water On
7 LV On, Water On
8 LV On, Water On, Blanket
Figure 7: The different data-taking phases.

The asymptotic temperatures reached in the various phases of the data-taking are depicted in figure 8. The black points (phase 0) correspond to no water cooling and indicate the highest temperature reached in the system is ~ 53 °C at a LV regulator on the TDIG boards. The temperatures reached on the casing of the TDIG chips themselves is ~ 48 °C for the high-resolution HPTDCs (stops data), and ~ 45 °C for the low-resolution HPTDC (ToT data). The tray interior (inside the gas box) reaches ~ 38 °C, while the air gap between the TAMP and TDIG layers reaches ~ 45 °C.
./plots/t21.gif
Figure 8: The asymptotic temperatures reached in the various phases of the data-taking. [EPSF] [PDF] [PS]
These asymptotic temperatures without water cooling are comfortably below the specified maximum tolerable levels of 60 °C for the HPTDC chip temperature and 80 °C for the TDIG regulator temperature. Thus, the TOFr5 system should operate correctly in the absence of cooling water. However, water cooling would still minimize the heat radiated into neighboring detectors, mininize the MRPC noise rates, and maximize the service lifetime of the electronics.
Flowing water is seen to reduce the TOFr5 temperatures dramatically. This is seen as the red squares and green triangles (phases 1, 2, and 7), which depict temperatures that are 10-15 °C lower than when water is not flowing (black points, phase 0).
The input water temperature in these tests was 31 °C, while the temperature of STAR 'chilled water' is ~ 24 °C. Phase 2 thus includes the operation of the heat exchanger by filling it with ice. That the asymptotic temperatures in this phase are very similar to those at the end of phase 1 simply indicates that phase 2 was ended at a point when all the ice in the heat exchanger had melted and the tray input water temperature was back to that of the water as it comes out of the wall. Estimates for the temperatures that would have been reached if the cooling water would have been kept at ~ 24 °C indefinitely, as in STAR, are thus not shown in figure 8 (but are estimated as described in section 6 below).
Phase 8 is similar to phases 1 and 7 in that both LV and 31 °C water flow were on, but includes a covering of the perforated aluminum cover layer on TOFr5 with several layers of cellophane and a blanket. This phase was done with the hope of estimating the amount of power being radiated through the perforated aluminum cover. With the perforated cover blocked off, the TOFr5 temperatures reach asymptotic values that are 2-4 °C higher then when the cover is not blocked.

5  Power Estimates

The TOFr5 electronics drop 140W per tray. The central goal of the present tests is an estimate of the fraction of this total power that is removed by the water path. The remainder would unfortunately be radiated into the STAR TPC and BEMC. In this section, we estimate how this 140 Watts/tray of total TOFr5 power is shared amongst the possible heat removal mechanisms - water flow in the TOFr5 geometry, radiation, and convection.
The power removed by the water path is a function solely of the flow rate and the difference between the tray output and input water temperatures, according to
Pwater (Watts) = 1.15 * FlowRate (liters/hr) * DT (° C).
(1)
For the 1.36 Gpm flow rate used in the present tests, this equation becomes
Pwater (Watts) = 355 * DT (° C).
(2)
The two pairs of thermocouples at the tray water input and output are averaged separately in each read, and the difference between these two averages (output-input) is plotted versus the read number over the entire running period in figure 9.
./plots/t44.gif
Figure 9: The average output water temperature, (t[28]+t[29])/2 after the calibrations, minus the average input water temperature, (t[26]+t[27])/2, versus the read number (4s/read). [EPSF] [PDF] [PS]
In phase 0, water is not flowing, so the average water temperature difference, output-input, (black points) remains consistent with zero throughout this phase, as expected given the KS1992+3516 offset calibration described in section 3. The initiation of the (31 °C) water flow at the beginning of phase 1 (red points) causes a dramatic drop in the on-tray temperatures and an increase in the water temperature difference (tray output-input) of ~ 0.295±0.02 °C. This implies that the power removed via the 31 °C water path in phase 1 was 105 W, which is ~ 75% of the total power dropped by the electronics (140±10 W). This configuration was retested in phase 7, and the water temperature increase observed in this phase is consistent with that observed in phase 1.
Thus, 31 °C water flow removes 3/4 of the total power, and the remaining 1/4 of the total power is carried away radiatively or convectively. This is tested in phase 8, where the cellophane and blanket wrapping outside of the TOFr5 tray strongly suppresses both the radiative and convective heat removal paths. In this phase, the asymptotic water temperature difference was observed to be 0.385 ±0.02 °C. The total power being carried away by the water path is thus 136W, which is consistent with 100% of the 140W total power being dropped by the electronics. This result implies that the efficiency of the water path is not limited by its design, but rather by the fact that heat removal via the water flow is competing with radiative and convective heat loss mechanisms at the same time.
The total radiative and convective heat loss of 355*(0.385-0.295) = 32±7 W is made up of the following pieces. We now attempt to estimate the relative importance these three mechanisms to provide a full accounting of the total TOFr5 power.
We can dispense with the convective process immediately. The room air is "calm"; it is not being forced across the tray in any way. The power removed by air flow is given by,
Pairflow (Watts) = 0.125 * FlowRate (CFM) * DT (° C).
(3)
As the air flow rate through the cover, over TDIG boards, then out the cover again could not be larger than 0.1 CFM at any time during the present tests, the heat removed through the cover by convective air flow can not exceed a fraction of a Watt.
We thus assume we are left with only the two radiative processes. We can check that this assumption with the phase 3 results. In this phase, the electronics are unpowered but 31 °C water is still flowing. Here the temperature rise is negative, as in this phase the tray is serving as a "radiator" and removing heat from the water, rather than in phases 1, 2, 7, and 8 where the water is removing heat from the tray. While the distribution of heat sources over and inside the tray is different in these two cases (all local heat generated on the electronics versus all local heat generated on the cooling loop), the efficient all-metal paths between the electronics and the water loop implies that the data from phase 2 should lead to an approximate but direct estimate of the total radiative heat loss. In this phase, the observed water temperature difference is -0.06 °C, which corresponds to 20±7 W. This is not so far off from 32±7 but again there are caveats for this comparison. We note for fun that a more dramatic example of the tray acting as a radiator to remove heat from the water is seen at the beginning of phase 6. In this case the entire tray is at room temperature. At the initiation of the 31 °C water flow, the water temperature difference (output-input) is ~ -0.14 °C, implying the (cold) tray is initially removing 50 W from the water path!
Given the caveats in the estimates for the total radiated power obtained from the phase 3 data, we also estimate the radiative power directly. As power is being radiated at different rates both by the outer aluminum skin of TOFr5, and from the TDIG boards directly through the cover, we need to consider these two processes separately. When 31 °C water is flowing, the aluminum skin temperature is approximately 30 °C, while the TDIG board temperature is ~ 38 °C. The radiated power is given by,
Prad (Watts) = e s A (Th4 - Tc4),
(4)
where e and A is the thermal emissivity and the area of the surface, respectively, Th and Tc are the elevated and room temperatures, respectively, and s=5.67×10-8 W/m2/K4. The thermal emissivity [3], e, of commercial sheet aluminum (i.e. unpolished and not heavily oxidized) is 0.09, while the thermal emissivity of FR4 (i.e. the electronics boards) is ~ 0.8. The total surface area of the TOFr5 outer aluminum is 10.8 ft2, while the total area of the upper surface of the TDIG boards is 4.5 ft2.
The resulting radiated power estimates are thus 3.3W from the aluminum skin, and 30W from the TDIG boards. As the perforation holes total 50% of the tray cover, approximately 3.3+(30W/2) ~ 18W would be expected to radiate off TOFr5. This is consistent with the estimate of 20W obtained from the phase 3 data.
Note that this 29.9W estimate of the power radiated by the TDIG boards ignores the components. While these are small in area, they can be significantly hotter. The net result would be an increase in the estimate for the total power radiated by the TDIG boards of a few Watts. This is at a level that's small compared to the measurement errors in this analysis. Improving these estimates would require a more complete treatment of all of the mechanical components in the system, and their interactions via all possible thermal paths, using a full thermal simulation package such as is available commercially.
Nonetheless, the facts that the water loop improves in efficiency in phase 8, and that these power estimates imply that most of the radiative power is coming off the TDIG layer and not the aluminum skin, seem to imply that a perforated cover would not be the optimal approach for the full system in STAR. With a solid cover, the heat radiated by the TDIG layer would be captured on the cover, which has an efficient (all-metal) thermal path to the water. In this case ~ 130±10 W would be removed by the water path and the rest, ~ 5±10 W, would be radiated into the inside of STAR. The power estimates in kW for the different heat exchange processes and for a full system of these two designs are given in Table 3.
Design Power/tray Power total
(kW) (kW)
TOFr5, Water .105 12.6
TOFr5, Radiative .035 4.2
TOFr5, Total .140 16.8
TOFr5 (solid cover), Water .135 16.2
TOFr5 (solid cover), Radiative .005 0.6
TOFr5 (solid cover), Total .140 16.8
Table 3: Estimates for the total power in kW for the TOFr5 design as constructed and for a TOFr5-design but with an unperforated cover, for the different heat removal mechanisms of the water flow and the total radiative processes.
Subsequent rounds of temperature tests are planned for the near future which will basically retake these data in order to double-check the various power estimates discussed in this section. Beyond these retests, it will also be important to compare the power, heat flow, and temperature properties of a TOFr5 tray with a solid cover to that with the present perforated design. Also, tests could be also be performed at lower water flow rates, for which the water temperature rise (output-input) should be larger than that observed for the 1.36 Gpm flow rates used in the present tests. The total power removed by the water path should however be a constant versus the flow rate, so collecting the data from such a flow rate tests should improve the estimation of this important constant.

6  Cold Water Test

The water available for these tests comes out of the wall at a temperature of 31 °C, which is approximately 7 °C hotter than that available at the tray when TOFr5 is installed in STAR. In phase 2, we added ice to the heat exchanger in order to cool the water that was input to TOFr5. The results from this phase are discussed in this section.
./plots/t03c.gif
Figure 10: Calibrated temperature versus the read number ( ~ 4s/read) during the cold water test of phase 2. [EPSF] [PDF] [PS]
Shown in Figure 10 are the on-tray and water temperatures versus the read number ( ~ 4s/read) for this phase. The water input and output temperatures are the two lowest curves, while the on-tray temperatures are the rest shown. For a short time ( ~ 20 minutes) it was possible to keep the water temperature at the tray input near ~ 23 °C. Twenty minutes is not a sufficient amount of time for any on-board temperature to reach an asymptote, as is visible in this figure.
We thus fit the functional form, par[0]+par[1]*exp(par[2]*t), to these temperature versus time curves over the region in time where the water was near ~ 23 °C. The values of par[0] extracted from these fits are an estimate for these asymptotes, and are shown versus the on-board thermocouple position as the green triangles in figure 11. The same procedure was also applied to the data from phases 0 and 1 and the asymptotes extracted in this way for these phases are shown as the black circles and the red squares, respectively.
./plots/t51.gif
Figure 11: Asymptotic temperatures for phases 0 (black), 1 (red), and 2 (green), extracted by the fits of a constant plus an exponential to the temperature versus time curves for each thermocouple. For phase 2, the fits were performed only over the limited time range where the water temperature at the tray input was near 23 °C. [EPSF] [PDF] [PS]
According to this figure, the asymptotic temperatures for ~ 23 °C water are approximately 5 °C colder than when the water is at 31 °C. The green points are thus the expectation for the on-board temperatures when TOFr5 is installed in STAR. The asymptotes shown for phases 0 and 1 agree very well with those shown in figure 8, which were obtained by simply averaging the last 2 minutes of temperature values. This indicates that the fitting procedure used to extract the asymptotes for phase 2 is reasonable.

References

[1]
http://wjllope.rice.edu/ ~ TOF/TOFr5/Documents/tofr5.pdf
[2]
http://www.kscorp.com/camac/1000/1992.html
http://www.kscorp.com/camac/3500/3516.html
[3]
http://www.electronics-cooling.com/html/2003_august_techdata.html

File translated from TEX by TTH, version 3.38.
On 28 Sep 2004, 14:15.