Integrated Smart Sensor Calibration

1
Integrated Smart Sensor Calibration

• Abstract
• Including at the sensor or sensor interface chip
  a programmable calibration facility, the
  calibration can easily be automated and executed
  for a batch of sensors at a time, thus minimizing
  time and costs
• A calibration method and options for integration
  in the smart sensor concept, in hardware as well
  as software
• An advantage, a step-by-step approach, not need a
  large matrix of calibration data

2

• Introduction
• Integrated Smart Sensors (4 integrated functions)
• Analog sensor readout and signal conditioning
• A/D conversion, to provide a digital output
  signal
• A bus interface, to simplify communication to
  microcontrollers, PCs and other devices
• Calibration of the sensor tranfer curve,
  preferably a digitally programmable calibration
• Calibration of Smart Sensors
• An analog calibration circuit, a limited accuracy
  bu large bandwidth and high processing speed
• A digital calibration facility implementatioin,
  high accuracy but limited processing speed
• To reduce costs, it is important to minimize the
  number of reference measurements, the
  communication and the computation power for each
  sensor to be calibrated.

3

• 1-D Polynomial Calibration Method
• Calibration Principle
• Most methods rely on collecting a complete set of
  measrement data to calculate a correction formula
  or lookup table.
• As for this proposed method, each calibration
  measurement can be used directly to calculate one
  programmable coefficient in a correcion function,
  which can then immediately be used to correct the
  sensor output.
• The next calibration makes use of this corrected
  sensor signal. Each succeeding correction is
  applied in such a way that the previous
  calibrations remain undisturebed.
• Translating, rotating, and bending, more
  calibrations can be done in the same manner to
  further linearize the sensor transfer function.
• Simulation showes, for 3 or more calibration
  steps a good lineariztion is obtained, when the
  first calibration point is at one end of the
  sensor operation range, the second point at the
  other end of the range, and further points
  halfway between two previously selected points.
• Eq. 1-4, more mathematical details.

4

• Implementation analog signal processor for
  polynomial sensor calibration
• Fig.4, the signal flow diagram of a four-step
  polynomial correction. The diagram can become a
  repetitive chain as the dash-boxed part can be
  repeated to increase the order of the polynomial
  correction, which can be implemented in a
  software routine(microcontroller) or in
  hardware(either analog or digital).
• Fig.5, the block diagram of the analog signal
  processor
• Low-cost bipolar process
• All signals represented by differential currents
• Addtion/substraction done by connecting current
  signals, multiplication implemented in Gilbert
  Multiplier
• Current duplicator blocks (IV/VI convertor)
• Multiplying DAC (current-mode, 8bit, cascode
  dividers)
• Bandwidth in the order of 100KHz to 1MHz

5

• Conlusion
• The proposed method doesnt have the disadvantage
  of collecting a matrix of data first and then
  inverting it or use an iterative method to obtain
  the polynomial correction factors.
• Instead, it uses each calibration measurement to
  obtain one specific correction factor, at the
  first to correct the offset, at the second to
  correct gain, at the third to correct 2nd-order
  non-linearity, etc. Each calibration results in
  an additional correction term which is
  constructed in such a way that all previous
  calibrations remain undisturbed.
• This approach also applies for a 2-D calibration
  for sensors that suffer from cross-sensitivity to
  another parameter.
• The error correction is a repetitive procedure,
  which can be implemented in software as a small
  subroutine, or in hardware as a cascade, pipeline
  or a loop of the same building blocks.