Causes of Oscilloscope Errors

   Signal integrity is a function of the oscilloscope and its inputs. Most DSOs have a gain inaccuracy of 1% to 5% at DC. Absolute gain at high frequencies is rarely specified, but the overall Gaussian roll-off of the oscilloscope ensures that transient response is good. The relative gain accuracy of the DSO display is affected by the preamplifier, attenuator, and analog-to-analog converter (ADC), unless an analog oscilloscope with electrostatic deflection or cathode ray tube is used, in which case the accuracy is not affected by the display system. Analog oscilloscopes have a total gain error of 2% to 3% due to errors introduced by the deflection amplifier and cathode ray tube.

　　LCD screens and magnetically deflected cathode ray tubes are driven at television rates, with the composite signal including all logos, menu text, graphics, and waveforms. Thus, the relative accuracy of the waveform to the grid is not affected by the linearity of the display device. The absolute accuracy of the grid lines may be in error, but the signal will be accurately positioned on each grid line. In contrast, the grid of an electrostatically deflected cathode ray tube in an analog oscilloscope is etched on glass, so any nonlinearity in the deflection amplifier or cathode ray tube will increase the overall gain error. Most oscilloscopes have only 8-bit resolution, and a few DSOs offer 10 or 12 bits. For biomedical signals, for example, where there are unknown offset voltages, the greater dynamic range of a 10- or 12-bit system is advantageous. Higher resolution is also required if the complex signal requires vertical amplification to examine fine details. Higher resolution can be achieved by averaging, which is based on the fact that a large amount of inorganic noise is present in the processing, resulting in a typical rms relationship, such that sampling the signal 16 times will improve the signal resolution by a factor of 4.

　　A more reliable approach is to perturb the ADC input with a certain amount of specially weighted noise. During each sampling process, the ADC input is offset a little from the actual input signal value. The offset, direction and distribution of the disturbance are preset to ensure that the average high-resolution data of the disturbance process will not produce spectral distortion. The problem is that the disturbance reduces the peak-to-peak range of the ADC, and its amplitude is equal to the disturbance signal value. Because the degree of improvement in resolution is proportional to the disturbance signal and the number of samples averaged, there is a certain limit to the improvement that can be achieved.