---++Temperature Readout System(Under Construction)

The temperature readout system has two main parts, the drive circuit and the sense circuit. For the sensors to do a good job, it is essential that they be excited at recommended levels of voltage or current. Thus the function of the drive circuit is to produce those recommended levels such that the resulting signal to noise ratios are high enough for an accurate readout. The purpose of this document is to layout the sensor specifications and link that to the justification of the components selected.

Two types of sensors will be used: diodes and Cernox resistive thermometry devices (RTDs), described below. We plan to minimize the design time and complexity of our board by having a single hardware circuit for both diodes and RTDs (though the user will have to change the gain of the sense amplifier by inserting a simple header for RTD channels). We also plan to minimize the hardware complexity by accomplishing all of the RTD drive sine-wave synthesis, lock-in detection, and post-lockin low-pass filtering in firmware. Diodes temperature measurements will invoke a different firmware module that produces a constant drive current and direct voltage measurement.

RTD signal flow:

• Firmware produces sine wave at 100 Hz
• bit shifter allows amplitude changes by factor of 2
• 16 bit D/A produces analog voltage at 500 Hz sample rate
• analog drive circuit provides current through RTD
• instrumentation amp senses voltage across RTD
• 24 bit ADC digitizes the signal at 500 Hz sample rate
• firmware lock-in mixes signal down to baseband
• firmware low pass filter acts at 50 Hz
• output is tagged with GPS timestamp and made available over ethernet.

Diode signal flow:

• Firmware produces programmable amplitude output voltage. Default is the voltage corresponding to 10 μA through diode.
• 16 bit D/A produces analog voltage
• analog drive circuit produces current through Diode
• instrumentation amp senses voltage across Diode
• 24 bit ADC digitizes the signal at 500 Hz sample rate
• firmware low pass filter acts at 50 Hz
• output is tagged with GPS timestamp and made available over ethernet.

Description of Diode Sensors

• Recommended excitation 10μA ±0.1%
• Max current before damage 1mA continuous or 100mA pulsed
• Thermal response time. <10ms at 4.2K, 100 ms at 77K, 200ms at 205K
• Range of use 1.4K-500K
• Range of voltage across sensor for a 10μA excitation is 0.2V-1.8V

The range of voltage drop is clearly shown in the following graph (from Lakeshore webpage).

Description of Cernox (Resistive Thermometry Device) Sensors

• Recommended excitation 20μV(0.1K-0.5K), 63μV(0.5K-1K), 10mV or less for T>1.2K
• Range of use = 0.1K to 325 K
• Thermal response time. 1.5ms at 4.2K, 50ms at 77K, 135ms at 273K
• Range of resistance values 20K at 0.1K to 20 at close to 500K

The data sheet tells us the recommended voltage levels at different temperatures. Since the drive circuit is meant to act like a precise current source, these recommended voltage levels need to be translated into current levels. A ball park estimate of the current can be made by checking the resistance of the RTD at different temperatures and using the recommended voltage levels at those temperatures.

For the Cernox, the temperature range 0.1k-1k is important as it will be used mostly in that range. Therefore the current levels can be estimated as follows

• (0.1k-0.5k) - recommended excitation 20μV. resistance from the graph observed to be of the order of 10K. consequently current 2nA
• (0.5k-1k) - recommended excitation 63μV. resistance from the graph observed to be of the order of 1K. consequently current 63nA

Amplifiers component selection

Differential amplifier: THS4131, U101

• Selection criteria:
  • fully differential. Prefer a differential amp in a single package, rather than building one out of discrete op-amps, so that we don't have to worry about the matching of feedback resistors.
  • amplifier will run effectively open-loop at low frequency. We need high gain through a bandwidth of about 1 kHz - easy to obtain.
  • white noise needs to be in the order of nV. This is because the least voltage the amplifier will have to produce is 420μV, calculated below, and so the variation should not be more then ±0.1%. *1/f noise needs to be sufficiently small at 50 Hz to allow us to modulate our carrier there, again should be in the order of nV.
• Description
  • High Speed
  • Low noise
  • Fully Differential
  • Absolute maximum's
  • Supply voltage = +-33v
  • Input voltage = ±Vcc
  • Output Current = 150mA
  • Operating free air temp = 0-70 C
  • Low white noise 1.5nV/sqrt(Hz) at 1 KHz.
  • Cut off frequency from bode plots, 10 Mhz.

• Voltage levels to be produced by the op-amp U101.
  • Diode circuit. Constant excitation current, 10uA. Voltage drop ranging from 0.2-1.8v across the diode. Voltage drop across the R104 and R105 fixed to be 1 V each. Voltage supplied by the op amp will lie between 2.2v to 3.8v.
  • Cernox Circuit. Varied excitation across the cernox. Varied current. *(0.1K-0.5k) 2nA current for this range. Voltage drop across R104 and R105 200uV each. Voltage supplied by the op-amp is 420μV. *(0.5k-1k) 63nA current for this range. Voltage drop across R104 and R105 6.3 mV each. Voltage supplied by the op amp = 6.3+6.3+0.063=12.663mV Voltage levels will vary from 420uV to 3.8v.
• Noise calculations. The following graph shows the noise characteristics of THS4131.

For the diode, voltage supplied by the op amp will vary from 2.2V to 3.8V which is high enough for any noise introduced into a DC signal by the op-amp. For the RTD we will need to use a sine wave at around a 100 Hz, with sidebands of 20 Hz in total. This will introduce a total of around 13.4nV peak to peak noise ( (√20)*3, 20 is the total sidebands of the signal we are interested in and 3 is the spectral noise density read off at a 100Hz from the plot) . This is much lower then the least voltage that might be required of the op amp for the RTD circuit, which is 420μV. Also the least voltage across the RTD will be 20μV which is again around a 1000 times larger then the noise.

Instrumentation amplifier: LT1168, U102 and U103

• Selection criteria
  • low bias current(!)
  • high gain (1000) out to a bandwidth of about 1 kHz.
  • low white noise and 1/f noise, allowing us to see a tiny RTD signal modulated at 50 Hz
  • stable and low noise for small gain=2, for diode operation.

• Description
  • Single Resistor Gain Programmable
  • Low Power
  • Precision Instrumentation Amplifier
  • Gain equation: G = (49.4k/RG) + 1
  • Typical Input Bias current 40pA
  • Noise Spectral Density at 10Hz is 10nV/√Hz
  • Slew rate 0.5V/μs
  • Noise levels are depicted in the diagram shown below:

It has higher noise levels then the THS4131. However, If we are using carrier signals at around a 100 Hz with a total of 20 Hz in sidebands for the RTD, then for LT1168 we'll have to worry about a 44.7nV peak to peak noise, (sqrt(20)*10)--> (sqrt(f)*(noise spectral density)). The lowest possible input that U102 or U103 will receive is 20uV (in case of U103 measuring a voltage across the RTD), which is higher then the noise by an order of a thousand.

Specification of values for the passive components

• R101 and R102. A high value of a 100kΩ. Highest power dissipation is V2/R = (5*5)/100k = 250μW
• C101 and C106. The purpose is to act like low pass filters against the signals being amplified by U101. They work in correlation with resistors R104 and R105 to attenuate signals over 1600Hz. (f = 1/(2πRC))
• C102, C105, C107, C110, C114 and C115. 10μF polarized capacitors acting as large reservoirs of charge.
• C103, C104, C108, C109, C113 and C116. 100nF capacitors acting as immediate and efficient reservoirs of charge at high frequencies.
• Low pass filters
  • R104 and C111. A resistance of 100kΩ and a capacitance of 6.8nF, combining to give a single pole low pass filter with a cut off frequency at 234Hz (f = 1/(2πRC)).
  • R105 and C112. A resistance of 100kΩ and a capacitance of 6.8nF. Single pole low pass filter with a cut off at 234 Hz.
  • R110, R109, C117 and C118. Resistances and capacitances of 100kΩ and 6.8nF respectively. Forming a double pole low pass filter with a cut off at 234Hz.
• R103 (gain resistor for U102). Resistance of 24.95kΩ. As per the gain equation of the op-amp LT1168, this results in a gain of 3. This is convenient because U102 will be amplifying the voltage across R105 which in the case of the diode will be a constant 1V. In case of the RTD, a jumper will be placed in parallel to R103 with a resistance of a 100Ω, resulting in close to a 100Ω of equivalent resistance. This results in a gain of 500 as the maximum voltage across R105 in case of the RTD can be around 6.3mV. Amplified by a factor of 500, this is close to 3.15V, thus below the maximum voltage U102 can provide.
• R108 (gain resistor for U103). Resistance of 49.9kΩ. As per the gain equation of the op-amp LT1168, this results in a gain of 2 which is convenient for U103 in the case of the diode. The maximum voltage it’ll have to amplify is 1.8V, making it 3.6V and thus below the maximum voltage that can be provided by U103. A change for the RTD circuit is to place a jumper in parallel with R108 with a resistance of 49.9Ω, resulting in a gain of 1001. The maximum voltage across the RTD is going to be 63μV, amplified by a factor of around a 1000 makes it 63mV going to the ADC.
• R106 and R107. 10MΩ resistance each. High resistances to provide a path to ground for the input bias current.

Noise considerations

Noise sources (Applicable in a bandwidth of 200Hz)

• R101 and R102. 100kΩ resistors.Voltage noise generated = √(4kTRf). Here f is the bandwidth and if taken as 200Hz, this value comes to 566nV for each resistor. Converting that into a Norton equivalent, the current noise generated = V/R = 5.65pA. Furthermore, as a current of 5/100K = 50μA through these resistors corresponds to a current of 10μA through the sensor, a current of 5.65pA would cause a current roughly around 1pA in the sensor. The least current expected through the sensor is 2nA giving an SNR of 6odB.
• R104 and R105. 100kΩ resistors.Voltage noise generated = 566nV. Therefore current noise = 5.65pA. Even with a zero impedance sensor, this noise is negligible to us in the Diode channel however for the RTD where the recommended current level can go down to 2nA, the noise will still be lower then the signal by a factor of around a 1000.
• Digitization noise incurred from the DAC = 0.5μV/√Hz. This is a ceiling ball park of the noise. This is much higher then the noise produced by U102 within the bandwidth of the circuit operation and so will dominate once added in quadrature. This noise at the DAC corresponds to a 5pA/√Hz at the sensor. This can be concluded from a simple ratio computation(5v--->10μA, 0.5μV/√Hz--->5pA/√Hz). Once calculated for a 20Hz sideband signal this noise can go as high as 20pA, which is a factor of a 100 times weaker then the least value of current through the sensor, 2nA.
• Noise produced by U101. As mentioned earlier this noise is around 13.4nV in the bandwidth of the signal for the RTD case. As the voltage produced by the op-amp itself will not go below 420μV, this noise is negligible.

Mechanism

Diode Current Drive

• For diode current drive, we set the jumper as follows
  • P101---> Open. This leaves the gain resistor value to be 24.95kΩ resulting in a gain of 3 for U102.
• This results in a transimpedance of -3x105 V/A for the drive circuit, where the voltage is being specified at the DAC output and the current is through the sensor.
• Output is sourced by a constant voltage of -3V at the DAC. (This is too big, we don't want the DAC to be running on its rails. Re-specify the passives so that the DAC is running in the range of 1-3V for 10uA please.)
• Feedback network transforms this into a fixed current of 10 uA through the diode.
• The primary source of noise is the 1/f noise produced by the op-amp U101, however it is far below the signal that will be used to drive the diode circuit. The data sheet for the op-amp does not specify its noise within the 0.1Hz-10Hz range, however looking at other op-amps with higher noise levels it can be concluded that the noise cannot go over 1μV peak to peak within a 50Hz bandwidth. The least signal produced by the op-amp for the diode circuit would be 2.2 V, which is more then 120dB over the noise.
• We don't need to worry about the THS4131 bias current as an analysis performed at the input nodes of the amplifier tells us that the current will be divided among R101 and R102, producing opposing voltages, which will in turn cancel each other out.

Diode Voltage sense

• For diode voltage sense, we set the jumper as follows:
  • P102---> Open. This leaves the gain resistor to be at a value of 49.9kΩ resulting in a gain of 2 for U103.
• R104, R105, C111 and C112 form single pole low pass filters with a cut off at around 200Hz (calculated earlier). These filters throw away a major proportion of the noise being generated by the drive op-amp's total bandwidth (the bandwidth of THS4131 is 150Mhz) and are located right at the output of the drive circuit. Add a section here describing the low pass filter at the output of the drive. What frequency? etc. What is the bandwidth of the drive amplifiers?
• Diode voltage will be in the range of 0.2V-1.8V.
• The readout noise will be dominated by the first stage instrumentation amp, which has a noise of around 2μV peak to peak between 0.1Hz and 10Hz for a gain of 1(in our case it is a gain of 2 but the noise characteristics would only be better for a higher gain), and then from 10Hz to 50Hz it can be approximated to be another 1μV peak to peak. Added up in quadrature, this leaves us with a 2.23μV peak to peak (a sum of 1/f noise and the input referred noise of the op-amp) corresponding to 50 Hz.
• The digitization noise of the ADC can be approximated to be around 1.5nV/√Hz. A bandwidth of 50Hz will make it close to 10nV rms. This is way smaller then the smallest voltage the ADC will have to measure, and that is 0.2V*2 = 400mV.
• The instrumentation amp bias current has a maximum value of 250pA and is negligible compared to 10μA which is the recommended level of excitation for the diode.
• R109,R110, C117 and C118 form a double pole low pass filter with a cutoff at 200Hz(calculated earlier). The bandwidth of the instrumentation amplifier varies from 400KHz to 1 KHz for a gain of 1 to 1000 respectively. The double pole low filter thus removes most of the noise and thus the unnecessary part of the spectrum. Add a section here describing the low pass filter at the output of the sense. What frequency? etc. What is the BW of the instr. amp?

RTD Current drive

• For the RTD current drive, we set the jumper as follows
  • P101---> Connect it with a resistance of a 100Ω. This leaves the gain resistor value to be around 99.2Ω, resulting in a gain of 504.
• This results in a transimpedance of -5.04x107 V/A for the drive circuit, where the voltage is being specified at the DAC output and the current is through the sensor.
• Output is sourced by a sine wave( at around a 100Hz with sidebands of 10Hz) produced in firmware carried out to the DAC.
• Feedback network transforms this into a varying current across the sensor, which can be modeled by a sine wave, exactly like the one carried out to the DAC
• The primary source of noise is the input referred noise produced by the op-amp U101, however in the sidebands of the sine wave (a total sideband of 20Hz at around a 100Hz) (what bandwidth is the sideband?) this noise will be no more then 13.2nV rms as shown earlier. The output produced at U101 will never be less then 420μV and so will be at least 60dB higher then the noise.
• We don't need to worry about the THS4131 bias current as an analysis performed at the input nodes of the amplifier tells us that the current will be divided among R101 and R102, producing opposing voltages, which will in turn cancel each other out.

RTD Voltage sense

• For diode voltage sense, we set the jumper as follows: *P102---> Connected with a resistance of 49.9Ω. This leaves the gain resistor to be at a value of 49.9Ω resulting in a gain of 1001 for U103.
• Sensor voltage will range from 20μV-63μV, and will take the form of a sine wave.
• The readout noise will be dominated by the first stage instrumentation amp, which has a noise of around 0.28μV peak to peak between 0.1Hz and 10Hz for a gain of 1000. However for the RTD, 1/f noise will not be a big problem as instead of a DC signal, an AC signal will be used. For the AC signal the noise encountered in its sidebands will be 44.7 nV peak to peak (for a sine wave at a 100Hz with a total of 20Hz in sidebands) (for what bandwidth??). The least voltage read off across the RTD is going to be 20μV, again safely over the noise by at least 50dB.
• The digitization noise of the ADC can be approximated to be around 1nV/√Hz. A bandwidth of 50Hz will make it close to 10nV rms. The least signal to be measured, 20mV(Gain factor of U103 amplifies the voltage across the sensor ), is again above the noise.
• The instrumentation amp bias current has a maximum value of 250pA and cannot be ignored in the RTD case. The least current flowing through the sensor can be around 2nA. 250 pA is the worst situation possible, the typical bias current will be around 40pA. To take care of the worst situation high impedances have been connected to ground and the input terminals of the op-amp to provide a path for the bias current. Need to be quantitative about if this fixes the problem or not..

Heater Circuit

The purpose of the circuit is to inject precise amounts of heat into the cryogenic environment. Therefore a switching circuitry is required which can handle large amounts of current.

MAX4660 SPDT switch (single pole/double throw)

The MAX4660 is a medium voltage CMOS analog switch with a low on-resistance of 25Ω max specifically designed to handle large switch currents. With a switch capability of up to 200mA peak current and 150mA continuous current (MAX4660), this device parts can switch loads as low as 50Ω. It can replace reed relays with a million times the speed and a virtually unlimited number of lifetime cycles. Normal power consumption is only 3mW, whether the switch is on or off. These parts are TTL/CMOS compatible and will switch any voltage within their power-supply range.

The switch offers a typical on-resistance of 25Ω; however there are variations of this quantity according to the following graph. It is important to note that this graph only relates to a single supply operated switch. A double supply lowers the resistance even further but it is not worth the extra step of getting another supply.

OPA551 (High power operational amplifier)

Wide supply range: ±4V to ±30V

• High output current: 200mA Continuous
• Low noise: 14nV/√Hz
• Fully protected:
  • Thermal Shutdown
  • Output Current-Limited
• Thermal shutdown indicator
• Wide output swing: 2V From Rail
• Fast slew rate: 15V/μs
• Wide bandwidth: 3MHz

Heater circuit mechanism:

• Shutdown mode
  • '!heaterSwitchFpga' is pulled low and the heater is connected to ground at both of its terminals
  • DAC input to the op-amp is 0 volts.

• Operational mode
  • '!heaterSwitchFpga' is pulled high and the heater is now connected to the op-amp at one of its terminals.
  • Voltage provided by the DAC is pulled up to 22.6V. Assuming the R-on resistance of the switch to be at 25Ω and the heater resistor to be 200Ω, this leaves us with a current of around 100mA. Power dissipation in each resistance is approximately,
    • R104 --> 10mW
    • resistance of MAX4660 --> 250mW
    • resistance of the heater --> 2W
  • In case of a thermal shutdown of the op-amp (OPA551 has a thermal shutdown mechanism through which it shuts down at temperatures around 160degreeC) the FPGA will be notified through the ‘flagHeaterShutdownFpga’ going high. In that case, either the event can be used to revert to shutdown mode described above or be logged and analyzed to prevent a thermal shutdown of the op-amp in future.

---

This topic: CryoElectronics > DigitalFMux > TemperatureReadoutSystem Topic revision: r17 - 2009-08-25 - GraemeSmecher

---