2019/11/15 (Friday)
| Setup | DIB1/2/4 → 3.2 MHz filters → cable 1/2/3 → gender changer 1/2/3 → movable source controller/board with external Feedback resistor/nothing → DCRC55(1)/DCRC33(2)/DCRC50(3) |
|---|---|
| 0838 | Tower/MC/Coldplate = 34.7/30.4/192 mK, Still = 1.54 K/ 0.130 mbar / 0 mW, G1/G2= 207/182 mbar, 1K Pot = 1.66 K / 30±2 mbar. MB= 35.0% |
| 0840 | Fill LN2 cold trap bath. |
| 0847 | Wrote and uploaded a new version of the arduino stepper code which allows the user to wake/sleep the driver board manually. This means we can try leaving the motor powered, but not stepping, to find if we can operate at a low enough drive current so there's minimal heating between stepping actions. The first few steps of the test at 1704 yesterday indicate this might be possible. |
| 0852 | Power and connect to arduino. Set current step to 862 (though this is likely meaningless). Set autosleep to 0 with wr autosleep 0. This controls whether the driver board will wake/sleep automatically when we call a move. With it set to 0, we can wake/sleep the board ourselves with the commands wake and sleep. |
| 0905 | Turn on motor power supply to 4.2V, this is the lowest setting at which the warm motor would run. Wake driver board with wake command. Using scope, measure drive voltage on coil A got to 3V, and trim pot voltage start at 3V, but fall to 0V (indicating resistance increase) within a few seconds. Tower/MC → 1.5 K/400 mK. |
| 0909 | Put board back to sleep with sleep command. Drive/trim pot → 0/0 V, tower starts cooling |
| 0914 | Tower = 412 mK |
| 0930 | So what is causing this heating? The NbTi critical current should be ~few Amps, so something else must be warming it up. Could be the short pigtail of copper wire from the top of the tower to the motor coil. That should only be ~10 mOhm which, with say 50 mA dumps 25 uW at the tower. But our fridge has 55uW at 100mK cooling power, so that seems to not be enough. Maybe eddy currents cause heating when first turning on the power, and once any part of the coil starts to go normal the heating increases and we get in a positive feedback loop. Can try just applying a DC current to one coil and slowly increasing it until there's heating. |
| 0935 | Disconnect from and shut down arduino and motor power supply. |
| 0936 | Tower = 49.8 mK |
| 1056 | Tower = 35.7 mK. Connect DC power supply across Motor coil A leads. Monitor this voltage as well as the trim pot voltage as before. We'll slowly increase the power supply voltage, and thus the DC current going to the coil, and watch for heating. |
| V_tot (V) | V_trim (V) | I (mA) | R_fridge (Ohm) | Joule heat (uW) | T (mK) |
|---|---|---|---|---|---|
| 0 | 0 | 0 | - | 0 | 35.4 |
| 0.042 | 0.001 | 0.01 | 4104.6 | 0.4 | 41.4 |
| 0.3 | 0.0028 | 0.03 | 9904.6 | 8.9 | 263 |
*Based on measured trim pot resistance of 95.4 Ω
This is confusing. It seems like we're just sending down a very small current (less than QET biases) and getting a ton of heating and getting coil resistances higher than their normal values of 7.2 kΩ
| 1134 | Re-arrange circuit to measure current with Fluke 87 meter and repeat above test. |
|---|
| V_tot (V) | V_trim (V) | I_trim (mA)* | I_fluke (uA) | R_fridge (Ohm)‡ | Joule heat (uW)‡ | T (mK) |
|---|---|---|---|---|---|---|
| 0 | 0 | 0 | -4.2 | - | 0 | 43.6 (cooling) |
| 0.029 | 0.001 | 0.01 | 227 | 32.4 | 1.67 | 41.5 |
| 0.336 | 0.0015 | 0.02 | 2530 | 37.4 | 239 | 290 |
*Based on measured trim pot resistance of 95.4 Ω
‡Based on I_fluke
We're using the oscilloscope to measure V_trim, so maybe it's just not accurate at these low values. The fridge resistances are now just about what we expect from the striplines. And are consistent with measuring the resistance directly (26.2 Ω). The power we're dumping now seems more consistent with the fridge cooling power. However, the heating seems to indicate that the power is being dumped at the MC stage which is inconsistent with the 30 Ω coming from striplines alone.
Actually, if we assume the coil and stripline resistance is the measured 26.2 Ω when cold, then with V_tot = 0.336 the additional 37.4-26.2=11.2 Ω is responsible for 72 uW of heating. So this could be located at the MC.
| 1206 | Commence LHe transfer, MB = 29.8% |
|---|---|
| 1240 | Cease LHe transfer, MB = 92.9% |
| 1248 | Tower = 38.3 mK. Power on DCRC1, others still powered. |
| 1255 | Start up DCDC programs and do an Al flash for 90s using the sequence Al_flash.msl Tower spikes to 1.46 K |
| 1423 | Tower = 38.2 mK |
| 1424 | MIDAS has crashed. Restart MIDAS. |
| 1433 | Most channels look good. DCRC3 pB is misbehaving. It rails way low when put into FB mode. |
| 1436 | Zap SQUIDs. |
| 1452 | Channels are all back to good, DCRC3 pB is back to working. |
| 1458 | Start 1x S1/S2=+4/-4,-4/+4 V loop of 30 min Bkg data series, Driver gain = 1, with 1000 initial randoms, with 30s flashes and 30 min cooldowns trigger on DCRC2pA/B/C = 8/8/8 ADC using sequence qbias_pm4_mp4_loop.msl |
Measurements of a small test segment of NbTi connected to two pieces of normal wire in the same manner as in the motor assembly, estimate that each such joint could be as high as 2.7±0.7 Ω. If these heat, they may drive some of the NbTi coil normal as well. The total normal coil resistance (900 turns @ 1.5–1.75 cm diameter) is 7.1–7.5 kΩ. To add the ~10 Ω we observed, we'd only neat to drive L_coil*10 Ω / R_coil = 6.4±0.7 cm of the NbTi normal (~10 K). This is less than 2 turns of it.
| 1827 | Fill LN2 cold trap bath. |
|---|---|
| 1829 | DCRC3 pB is railed again, all other channels look ok. Resetting the feedback fixes it, seems to be in a good spot. |
| 1835 | Start 4x S1/S2=+4/-4,-4/+4 V loop of 30 min Bkg data series, Driver gain = 1, with 1000 initial randoms, with 30s flashes and 30 min cooldowns trigger on DCRC2pA/B/C = 8/8/8 ADC using sequence qbias_pm4_mp4_loop.msl |