---++Tc with Carrier The idea behind measuring the Tc of a bolometer with a carrier is to apply a carrier, vary the temperature of the bolometers in a controlled manner and watch for the point where the current through them changes. If the carrier has sufficient amplitude then it will provide Joule heating to the bolometer, raising its temperature above the bath temperature. This would cause erroneous readings, e.g. measuring a Tc lower than the actual Tc.

We therefore apply a carrier with a small amplitude. (79 nVpp) [carrier gain = 0, amplitude = 0.1] Assuming a mid-transition resistance of 0.5 Ohms this is 12.5 x 10^(-15) Watts of power on the bolometer. We record the temperature as a function of time and current as a function of time and then plot the current vs. temperature. We take measurements both cooling down and warming up to look for systematic effects such as a hysteresis in Tc temperatures for cooling and warming. The hysteresis is caused by a thermal lag between the bolometer temperature and the temperature readout on the cernoxes attached to the bolometer housing.

When the bolometers go superconducting the carrier is amplified tremendously. The resulting signals could have a number of effects like causing the SQUID to flux jump, railing the 2nd stage amplifier on the SQUID control board or railing the ADC on the demodulator. We can survive with one, maybe two, bolometers superconducting, but as more bolometers transition the Tc data rapidly becomes unreliable. Therefore, we only measure the Tc for one bolometer in a comb at a time. We overbias seven of the eight bolometers so that they don't transition, leaving one bolometer with a reliable Tc measurement. Below is a plot of the data produced. In this plot the average rate of cooling and warming is ~1mK/1min. We do not see jumped SQUIDs in the data we take. The drop in the center portion is most likely due to railing of the 2nd stage amplifier, though we have not concentrated on understanding it. In these measurements we are most interested in seeing the transition clearly and not concerned with instability in the system once the bolometer has gone superconducting. However, the latest firmware builds do have the ability to monitor the ADC for railing. When railing occurs a flag is raised and will persist for 1 second. (This time is likely to be increased in the next build to something slightly longer.)

In this plot we see four transitions. Two of them are real and two are produced by the railing of the 2nd stage amplifier, which causes the drop in signal. The two real transitions are the ones second from the left and the right most. We can see that they slowly turn on, climbing off the lowest current level before transitioning.

Problems with this measurement come from a few sources:

• Time lag between the cernox and bolometers. We suggested last week in telecon.
• Joule heating could become large low in the transition. (V^2/R)
• Only one bolometer in a comb can be done at a time.

We have addressed these problems by doing two things:

• We have decided to send the "probe" signal through the nuller instead of the carrier, and use the imperfect non-zero input impedance of the SQUID. This makes the current through the SQUID go down instead of up when we go through transition, eliminating the problem of squid jumping and ADC railing.
• We have also implemented a software PID that controls our QMC fridge controller. This allows us to cool at a pre-determined rate. Please note that the PID adjustments introduce some wiggle in the cool down curves, as can be seen in the red plots.

Tc with Nuller

We apply a nuller with a small amplitude. When using a carrier the current has to flow through the bolometer to the SQUID. With the nuller, however, current can flow directly into the SQUID because the nuller enters the circuit after the bolometer comb. When the bolometer is normal there will be a minimum of current flowing through it and conversely a maximum of current through the SQUID. As the bolometer transitions the lower resistance allows more current from the nuller to flow up the bolometer RLC leg, through the bias resistor and to ground. There is a corresponding drop in the current through the SQUID. The result is that we now look for the temperature at which the current drops significantly. The main benefit is that the current through the SQUID is at a maximum while the system is warm. This allows us to make sure that we will not overload the SQUID, and therefore we should be able to measure Tc on all 8 bolometers in a comb at once. Despite the label of the plots below they are of data from seven bolometers.

The upper and lower plots were taken on April 3rd and 6th respectively with PID thermal control. On the 6th the average cooling rate at transition was ~1mK/5min and the average warming rate at transition was ~1mK/1.3min. The cooling rate on the 3rd was faster (~?mK/?min). Note that the cooling Tcs (the transitions at lower temperature) on the 6th are higher by ~7.5mK than the data from the 3rd (from 517.5 to 525 mK). The cooling and warming rates on the 6th are also significantly slower than the data taken with carriers.

Our results are explained well by assuming a thermal lag between the thermometers and the bolometers. As the rate of temperature change decreases we see the cooling Tc rise and the warming Tc fall. We see no evidence for heating low in the transition, which points to heating being a minor effect w.r.t. thermal lag. We are not likely to see the effect of heating in our data because the thermal lag effect is much stronger. To see evidence of heating we would need to cool and warm sufficiently slowly that thermal time constant between bolos and thermometry becomes subdominant to the time constant of the temperature change. At that point we would expect the warming Tc to be at a lower temperature than the cooling Tc. However, it took 5 hours to cool and warm in our last measurement and it seems likely that it would take days to take the measurement sufficiently slowly. At this time we don't believe the heating is a strong enough effect to commit to the time and effort needed to characterize it.

-- PeterHyland - 06 Apr 2009

---

This topic: BolometerTesting > WebHome > TcNuller Topic revision: r5 - 2009-04-08 - PeterHyland

---