Trusting Your Senses
The concept of signal integrity as a measure of trustworthiness of data is described using a simple analogy. Contrasts are drawn between integrity and accuracy in regard to the use of the data in a safety related application. The concept of “false plausib
- PDF / 290,559 Bytes
- 14 Pages / 439 x 666 pts Page_size
- 59 Downloads / 222 Views
Trusting Your Senses
G. Hardman, Silicon Sensing Systems Limited
Abstract The concept of signal integrity as a measure of trustworthiness of data is described using a simple analogy. Contrasts are drawn between integrity and accuracy in regard to the use of the data in a safety related application. The concept of “false plausibility” of data is described, with associated failure modes of the analogy examined. An implementation of a yaw rate sensor is described that takes advantage of the use of continuous numeric analysis of the control loop variables to provide very high levels of loop observability in real time. It also considers detectability of failures in times significantly less than the time constant of the sensing system, i.e. fault detection that is not bandwidth limited. Also considered are feedback as a means of verifying the conversion processes within the sensor, and parallel verification of the operation of the sensor software and microcode by using pseudo-random seed data from the sensor.
1
Introduction
In everyday situations, there are often times when an action or reaction relies critically on the ability to sense both accurately and reliably. Decisions that affect safety are based on the trustworthiness of the data. It isn’t acceptable to have a suspicion that something is true, it is essential to be sure beyond a basic level of trust. Take an occurrence that happens millions of times each day. A driver wishing to join a flow of traffic on a major road must confidently and reliably know that it is safe to do so. At junctions fitted with traffic lights, one level of trust is that when the lights are green to join the major road, oncoming traffic is inhibited by a red light. However, it cannot be guaranteed that it is safe to join, as it depends on all other drivers obeying the rules: stop on red. The detection of a green light gives no data as to the safety of joining, as it doesn’t have any information on the rest of the traffic environment. Emergency vehicles may legitimately continue against a red light; an inattentive driver might do so regardless.
285
286
Components and Generic Sensor Technologies
So, what is needed is reliable, trustworthy data. Not necessarily accurate data; there is no need to estimate to within a couple of metres where an oncoming motorcycle is, only whether it is sufficiently far away and travelling sufficiently slowly that it is safe to pull out in front. What this sensing scenario requires can be dubbed “integrity”. That is, a dependence on having data that is sufficiently accurate for the purpose, but with a quality aspect that means that the data can be used unambiguously to initiate or control an event. A high integrity sensor is one which gives an adequate level of performance in terms of the normal parameters, (sensitivity, offset, noise) but in addition provides unambiguous data. A low integrity sensor might be more accurate, (more sensitive, lower noise), but cannot be trustworthy if there are any circumstances where the output is “false plausible”. Dat
Data Loading...
