Stanford Research Systems SR785, Руководство по эксплуатации и программированию

Страница: 146 / 546

2-66 Signal Inputs
SR785 Dynamic Signal Analyzer
Most noise sources, including the amplifier input noise, are Gaussian in nature. The
amount of noise is proportional to the square root of the bandwidth in which the noise is
measured. A noise measurement made using normal units would yield different
amplitudes depending upon the linewidth of the FFT span. By using PSD units, the
measured noise is independent of linewidth and span.
The input noise of the SR785 inputs is about 10 nVrms/
√
Hz. The input noise of the
analog to digital (A/D) converter is about 300 nVrms/
√
Hz (referenced to a full scale of
1 Vpk). What does this mean in terms of the FFT noise floor?
For input ranges below -30 dBV (input gain greater than 30 dB), the measured noise
floor is determined by the signal inputs. This is because the signal gain (>30) times the
input noise (10 nVrms/
√
Hz) exceeds the input noise of the A/D converter.
For input ranges above -30 dBV (input gain less than 30 dB), the measured noise floor is
determined by the A/D converter. There is not enough gain to amplify the input noise to
a level greater than the A/D converter itself. The noise floor in this case is 300
nVrms/
√
Hz relative to 1 Vpk or -130 dBpk/
√
Hz relative to full scale.
All of this assumes that the signal input is driven from a low impedance (and noiseless)
source. Remember, resistors have a Johnson noise equal to 0.13x
√
R nVrms/
√
Hz. Even a
50
Ω
resistor has almost 1 nVrms/
√
Hz of noise! A signal source impedance of 6 k
Ω
has a
Johnson noise equal to the SR785’s input noise. To determine the overall noise of
multiple noise sources, take the square root of the sum of the squares of the individual
noise figures. For example, a 6 k
Ω
source impedance has 10 nVrms/
√
Hz of noise. Added
to the input noise, the overall noise is (10
2
+ 10
2
)
1/2
or 14 nVrms/
√
Hz.
Input Impedance
The input impedance of the SR785 is 1 M
Ω
. If a higher input impedance is desired, then
the SR550 remote preamplifier must be used. The SR550 has an input impedance of
100 M
Ω
and is AC coupled above 1 Hz.
Anti-aliasing Filter
After all of the signal amplification, there is an anti-aliasing filter. This filter is required
by the signal digitization process. According to the Nyquist criterion, signals must be
sampled at a frequency at least twice the highest signal frequency. In this case, the
highest signal frequency is 102.4 kHz and the sampling frequency is 262 kHz so things
are ok. However, no signals above 131 kHz can be allowed to reach the A/D converter.
These signals would violate the Nyquist criterion and be undersampled. The result of this
undersampling is to make these higher frequency signals appear as lower frequencies in
the digital data stream. Thus a signal at 175 kHz would appear below 102.4 kHz in the
digital data stream and be detected in the spectrum.
To avoid this undersampling, the analog signal is filtered to remove any signals above
160 kHz (when sampling at 262 kHz, signals above 160 kHz will appear below 102
kHz). This filter has a flat pass band from DC to 104 kHz so as not to affect
measurements in the operating range of the analyzer. The filter rolls off from 104 kHz to
160 kHz and achieves an attenuation above 160 kHz of at least 90 dB. Amplitude
variations and phase shifts due to this filter are calibrated out at the factory and do not
affect frequency domain measurements.