Abstract
The measurement of the spectral noise contributed by an
amplifier unit under test (UUT) in regions close to the frequency of the main
carrier signal is necessary to obtain an assessment of the suitability of the
amplifier for applications requiring sensitive measurement of amplitude, phase
and frequency characteristics of a signal. Often it is necessary to employ
expensive noise measuring test equipment for this purpose, but by using the NI
PXI-5421 arbritrary waveform generator with the NI PXI-5122 high performance
digitiser, with large memory capacity installed good performance can be
achieved.
Discussion
To measure the characteristics of a signal to a resolution
of 1 Hz requires the capture of a minimum of 1 second of signal data. When
operating at HF, the avoidance of aliasing errors & artefacts requires setting a
sampling regime that gives a sampling rate at least twice the upper frequency of
interest, and generally higher than that for a margin. In transforming from the
time domain to the frequency domain, it is customary to use a weighting window
to reduce the effects of discontinuity between the start and end of the series.
This introduces unwanted frequency domain artefacts and reduces the frequency
resolution, and the close-in dynamic range. To avoid the artefacts of time
series windowing, the carrier frequency, the sampling rate and the number of
carrier cycles has to be carefully selected to obviate the need for applying
time series weighting. Strict adherence to capturing an integer number of
carrier cycles is required as is maintaining a sampling regime wherein there is
an integer number of samples per carrier cycle.
To reduce the level of noise contribution from the instruments, it is
necessary to maintain coherent signal generation and sampling such that the
instrument-generated quantising noise is minimised and differential clocking
drifts are eliminated.
Signal Processing
After signal capture, the large amount of data has to be managed to
minimise the use of processing resources (memory and CPU time) and to derive the
required frequency spectrum with the desired high resolution. This was achieved
by carrying out initial processing of the captured data block by block to reduce
the amount of data being transformed. Final processing and normalisation of the
data was achieved efficiently to obtain the desired frequency spectrum for
testing against the specification limits (an analysis bandwidth of 7500 Hz with
respect to carrier was readily obtained).
Spectral (ensemble) averaging was applied to arrive at a smoothed
(average) spectrum. During the averaging process, for each data acquisition, the
relative gain and phase stability of the instrumentation (AWG and
Digitiser) were measured and plotted. The following plot is indicative of the
performance of these two units when coherently producing and processing a signal
over a time period of ~6 minutes (1.2 minutes per sample).
SYSTEM DESIGN AND INTEGRATION 
