40m QIL Cryo_Lab CTN SUS_Lab TCS_Lab OMC_Lab CRIME_Lab FEA ENG_Labs OptContFac Mariner WBEEShop
  PSL  Not logged in ELOG logo
Entry  Sun Aug 27 16:15:07 2017, awade, Craig, Notes, TempCtrl, Using AD590s for vacuum can sensing 
    Reply  Wed Aug 30 16:50:36 2017, rana, Notes, TempCtrl, Using AD590s for vacuum can sensing 
Message ID: 1882     Entry time: Sun Aug 27 16:15:07 2017     Reply to this: 1887
Author: awade, Craig 
Type: Notes 
Category: TempCtrl 
Subject: Using AD590s for vacuum can sensing 

So the plan was to use AD590 temp -> current sensors with a transimpedance amp to sense the temperature of the vacuum can. These AD590s run on 4-30 V and have a response of 1µA/K referenced to zero with 40 pA/sqrt(Hz) noise.  This means that at room temperature of 293 K we will have a 293µA offset.  If we plan to heat the can to about 40 C (initially) then we should set the max range of the temperature to voltage conversion to rail at maybe 50 C.  For a plain tranimpedance amplifier, that Kira and Kevin were planning to build for the 40m temperature control project, this means that the range of operation is small compared to the offset for the CTN application. 

The acromag cards are configured to have ±10 V over 2^15 bits. The least significant bit (LSB) is therefore 20V/2^15 bit = 0.61 mV/bit.  For an AD590 hooked up to a tranimpedance amplifier one would select a 30.96 kΩ transresistance gain to achieve 10 V @ 50 C.  This make the total gain from temperature to volgate G_{temp to V} = 1µA/K * 30.96 kV/A = 30.96 mV/K.  The minimum possible resolution is therefore LSB/G_{temp to V} = 19.7 mK/bit.  Not great if we are going for 1mK temperature stabilization. 

We can potentially halve the resolution to 9.9 mK/bit by setting the acromag range to 0-10 V. But its still a factor of 10 off what we nominally want.  At the moment we are seeing 100 kHz shifts over a matter of 1 min when the door is left open or I use the heat gun in the same room.  

From previous post estimating differential temperature shift (PSL:1874) we estimated ∆T based on the BN shift in frequency and the expansion coefficient of of Fused Silica.  So for the 100 kHz shift that we are seeing   \Delta T = \Delta \nu\lambda/c\alpha = 0.64 \textrm{ mK} (using alpha 5.5e-7 [1/K]) over a minute. At the moment we have absolutely no idea what the can-temperature to cavity-differential-temperature conversion is.  For a passive system an absolute temperature variation of 1 mK puts an upper bound on differential change of 1 mK or less.  But if our minimum achievable resolution is 10 mK (because of the huge AD590 current offset), then we are going to be limited in our stabilization. The differential variations from one side of the can to the other are another issue once we actually start sensing from a single sensor and heating.  

We should think about nulling this large current offset of the AD590 to zero the output at about 16-18 C. That way we go from a ∆T/T = 0.1 to something closer to 1: i.e. we would be using the full ADC range. Not sure how to construct such a large current nulling element in these transresistance amplifiers.  Another option is to have a second amplification/adding stage where a null voltage can be added in a second op amp circuit.  We need to think about the various noise trade offs: the 40pA/sqrt(Hz) of the AD590, noise of the op amp in the transimpedance stage and the voltage gain stage.  A liso model should be constructed to find the optimal ballance of gain between the stages to minimize noise.


We were wondering if maybe, in the future, we should wrap the can in THICK copper foil, then heaters, then foam, then aluminum foil would be a better way to go: this would increase the thermal conductivity about the can by a factor of 5-10 (from the new copper wrapping). It would still be slow to propagate inwards into the stainless steal but at least it will move around the perimeter of the can faster.


ELOG V3.1.3-