Lecture
All faults that, for one reason or another, occur in a PC or affect its operation are caused by errors, which can be classified into the following main types:
- errors in programs;
- erroneous operator actions;
- errors in information storage and transmission devices;
- errors in equipment:
Detection of errors in programs consists of the detection system identifying violations of formalized actions on the part of the program, which lead to the appearance of errors in computations. Such violations include, for example, accessing invalid or forbidden addresses, the appearance of invalid operation codes, and so on, i.e., everything that can somehow be formalized and provided for in the detection system as a way of checking these formal requirements. Obviously, such protection is only capable of detecting elementary errors in the program, since it is difficult to create a sufficiently simple system for detecting errors in the logic of solving a task.
Errors of this kind are more easily detected by the programmers or operators themselves, who run the program in accordance with the instructions, than by the machine.
Erroneous operator actions are difficult to predict. An operator may launch the wrong program, not to mention other, more «minor» errors — pressing the wrong button, transferring control to the wrong place, and so on. The whole difficulty of the issue lies in the fact that the cause of operator errors is not only, and not so much, inattention, but rather increased fatigue during work and the operator's internal state.
Research in recent years clearly shows the need for special attention to the problem of increasing the reliability of the human factor in control systems of varying complexity and purpose. The effectiveness of human-machine systems drops sharply when the operator's ability to cope with the duties assigned to them decreases. The human operator's ability to perform their functions promptly and accurately over a given period of time is influenced by many factors, among which the psychophysical characteristics that determine their state are perhaps the most significant. Therefore, the possibility of eliminating operator errors is related both to creating optimal working conditions and to formalizing the operator's actions in a way that allows criteria for evaluating these actions to be introduced. However, determining which part of the operator's activity can be formalized for error detection remains a problem that has not yet been fully resolved.
Errors in data to be written to memory and stored are eliminated by an error-correction scheme before writing, or by restoring the information in memory after receiving error signals. For this purpose, the original information is retained for a certain time to allow correction of the received data that was distorted as a result of an error. In some machines, information is stored with redundant bits, which facilitates the task of correcting it. There are various codes used in the storage devices of computing machines.
Errors during information transmission over communication channels are similar to errors in storage devices. These errors are corrected during transmission (using special correcting codes), or the information is restored in memory (usually by the method of retransmitting data received with an error).
When errors occur in the machine's logic circuits, a restart is performed if the error is a single occurrence. If the error is repeated or persistent, repair or reconfiguration of the system is performed (excluding the faulty unit while preserving further system operability).
If errors occur in the monitoring circuits themselves, the operator must choose the further mode of operation. However, if it is necessary to continue computations, they must remember that during this period the machine will be left without protection.
Faults in monitoring circuits can be of two types: the appearance of a false error signal and the absence of a signal when an error occurs in the monitored circuit.
Faults of the first type are detected in two ways: by stopping the device when an error occurs, after which the state of the monitored circuit is analyzed and a conclusion is drawn, or by running a special test program that diagnoses the signals of the monitoring circuit during its operation.
Faults in monitoring circuits of the second type pose a greater danger than the appearance of false error signals. Therefore, monitoring circuits are either periodically checked using a test program, or, if a particular monitoring circuit cannot be covered by such a check, they are duplicated.
Faults in power supply, cooling or mechanical devices of the machine can cause erroneous results to appear, similar to faults in logic circuits. In this case, the machine must be stopped and the fault that has arisen eliminated. Faults in power supply and cooling systems are detected by means of sensors and monitoring instruments. Faults in the machine's mechanical devices are harder to establish, so the main guarantee of their operability is timely preventive maintenance and keeping these devices in good technical condition.
Before troubleshooting and fault elimination, a number of actions must be performed to localize the source of the error.
Problems during POST
During the power-on self-test, errors most often appear due to incorrect hardware configuration. When a POST error occurs, check the following:
Hardware problems after booting
Sometimes problems occur after the system has booted, without any change to the hardware or software. To eliminate such errors, perform a number of actions.
Software problems
Software (especially the newest) can be the cause of errors. Most often this happens due to incompatibility between the software and hardware.
Adapter problems
Adapter problems most often arise due to incorrect installation or resource allocation (interrupts, direct memory access channel, and input/output addresses). In addition, do not forget to install the very latest version of the driver known to the operating system for this adapter.
Comments