08 April 2014
The truth about phase noise in signal generators
Manufacturers of signal generators set great store by the specification of phase noise, such that, regardless of the application, phase noise is frequently taken as a proxy of the instrument's overall performance.
However, phase noise performance might have little or no impact in some applications. For example, 'close in' phase noise creates problems when the signal generator is being used as a local oscillator, limiting its sensitivity or impairing bit error rate (BER) performance if used as a clock. 'Far out' phase noise, on the other hand, affects wideband communications systems by raising the noise floor and limiting the reception of poor signals.
Choosing a signal generator to match phase noise performance to the requirements of the application can be difficult as manufacturers often characterise phase noise performance at different carrier wave frequencies and at different offsets from the carrier signal. In practice, signal generators are sometimes over specified, entailing unnecessary expense.
Understanding phase noise
Phase noise is the result of small random fluctuations, or uncertainty, in the phase of an electronic signal. It imposes fundamental limitations on the performance of systems by limiting their dynamic range.
In radar and communications systems, compressing dynamic range results in a loss of sensitivity – loss of clarity in radar images and high BER in digital communications links.
Phase noise is commonly referred to as 'single sideband (SSB) phase noise' and expressed in dBc/Hz. It specifies the phase instability of the oscillator measured in the frequency domain and the 1Hz bandwidth allows the noise in other bandwidths to be calculated. It is the most commonly used measurement of phase noise.
What causes phase noise?
Phase noise in a signal generator is a function of the internal oscillator.
Most microwave signal generators used for test and measurement applications feature yttrium iron garnet (YIG) based microwave oscillators, rather than voltage controlled oscillators (VCOs); one reason being their superior phase noise performance.
YIG oscillators have a wide tuning range, which means the multiplication factors can be smaller, resulting in lower phase noise at higher frequencies. The phase noise of YIG based oscillators also drops off more rapidly far out from the carrier than it does in VCOs, so the total integrated phase noise tends to be lower.
A perfect oscillator can be described mathematically by a sinusoidal waveform:
V = cos [?/t]
where ? = frequency in radians, t = time
However, a real oscillator will exhibit an amplitude noise modulation, n(t), and a phase noise modulation, ?n(t).
V = [1 + n(t)] cos [?/t + ?n(t)]
where n(t) and ?n(t) are random processes.
In a good local oscillator (LO), amplitude noise modulation power will be much lower than phase noise modulation power.
Why is phase noise important?
SSB phase noise performance is particularly important in LO duty applications. Here, the LO's phase noise is added to the down converted RF signal during frequency conversion. If LO phase noise is excessive, it masks the information carried on the RF signal (see fig 1). This means the information is lost at the intermediate frequency (IF) stage, resulting in a high BER.
Radar systems, however, require high resolution images of targets, which calls for high spectral purity in an LO or signal generator.
In Doppler radar systems, moving objects reflect the transmitter's signal – the fundamental frequency of which is shifted as a function of the object's velocity – back to the radar receiver. Slow moving objects produce small Doppler shifts, which means they can only be detected if the transmitter signal's phase noise at the reflected frequency is less than the power of the reflected signal (see fig 2).
Because the phase noise of low phase noise oscillators is higher at frequency offsets closer to the carrier frequency, it is even more difficult to detect slow moving objects, such as cars. A radar system that can track both slow and fast moving objects requires a transmitter offering low phase noise over a wide span of offset frequencies (see fig 3).
Low phase noise is also acute in digital wireless communications systems, where narrower radio channel spacings require more selective receivers. To test receiver selectivity, a signal generator must have good spectral purity, otherwise the test system will be measuring the generator, rather than the receiver.
High phase noise may be tolerated in modulation schemes such as QPSK, but leads to symbol errors in more complex schemes, such as 16QAM. In other words, higher order modulation schemes increase the bit rate and bandwidth, but at the expense of increased susceptibility to phase noise.
The right tool for the job
Datasheets will not show which signal generator is best for any given application because manufacturers tend to specify at different frequencies and at different offsets, so a true like for like comparison is not simple.
The typical datasheet specifies performance at a 10kHz offset from the carrier frequency but, for many applications, such a close in offset is almost irrelevant.
A signal generator manufacturer might optimise phase noise performance to provide an impressive specification at this 10kHz offset, whilst compromising performance in the close in and far out regions. The wise user will disregard performance at a 10kHz offset, unless this is important in the application under test. Also beware of synthesiser phase noise being specified at a 100kHz offset, rather than at 10kHz, because phase noise performance is better at the larger offset; and low phase noise figure specified without the carrier frequency or the offset, which is meaningless.
Phase noise in most microwave signal generators increases with carrier frequency because such devices use multiplication to generate higher frequencies and phase noise increases as a function of 20logN, where N is the multiplication factor. So a signal generator that multiplies less than another will probably have lower phase noise at the higher frequency.
Harmonics and spurious signals
All frequency synthesisers generate harmonics and spurious signals; whether they cause a problem depends on the application under test.
Different applications may be sensitive to spurious signals at different amplitudes and frequencies: one synthesiser will often produce a spur that is problematic in one application while a different synthesiser does not. In a different application, the position might be reversed.
No high performance microwave signal generator is consistently better or worse than another at every frequency and amplitude, but devices with poor specifications for spurious signals are simply unusable in certain applications, such as testing radar and electronic counter measures receivers, and electronic warfare systems, where the high level of spurious signals will appear as ghost (false) signals or threats.
Spectral purity is one of the most important considerations when selecting a microwave signal generator; it is also one of the more difficult specifications to fully understand. It is important to realise that headline datasheet numbers for phase noise performance can be misleading, so users must study phase noise plots and consider how they affect the application under test.
Paul Holes is an RF and Microwave field applications engineer with Anritsu (UK).