New Phase Noise Measurement Techniques and Ultra-Low Noise SAW Oscillators
By Guillaume De Giovanni, President, Noise eXtended Technologies & Michel CHOMIKI, R&D Director, TEMEX

The use of the ultra-low noise oscillator is a demanding new application in the global military and instrumentation markets. To improve the phase noise performance of these oscillators, it is mandatory to improve the measurement test set and techniques. This article deals with two aspects of designing ultra-low noise oscillators: using an improved test set and techniques such as cross-correlation to measure ultimate phase noise, and ultra-low noise oscillators based on SAW (Surface Acoustic Wave) technology.

Introduction
Phase noise performance of ultra-low noise (ULN) oscillators has always been driving the specifications of phase noise analyzers. Once adequate for advanced designs, a noise floor of -178dBc/Hz is not enough anymore. Today, the cross-correlation process is the only technique that allows close to thermal noise floor measurements, around -195dBc/Hz.
More than 40 years after the first laboratory studies, SAW (Surface Acoustic Wave) technology is emerging as the favored approach for ultra-low noise reference for many next-generation radar systems under development today.

Thanks to its dual demodulator architecture, Noise eXtended Technologies’ (formerly Europtest) Dual Core Noise Test System (DCNTS) system can cancel most of its internal noise, as well as the noise of the reference sources. This article will demonstrate the impressive noise performance of the DCNTS.

Today’s SAW Oscillators
SAW oscillators, such as those designed and manufactured by TEMEX, are commonly used for instrumentation and military applications including airborne radars. The center frequency of SAW oscillators is in the range of 300-600 MHz, with a noise floor of less than -180dBc/Hz. By comparison with multiplied crystal oscillators achieving the same final frequency, the noise floor of SAW oscillators is improved by at least 12 dB, while close-in noise is quite similar.

Phase Noise Test Sets
The DCNTS is the only automated full dual channel test system on the market. The DCNTS sets the new reference standard in the market after the well-known HP3048A phase noise test system became obsolete. The DCNTS integrates two complete phase noise test sets, allowing the user to unleash the magic of cross-correlation.

The DCNTS uses two channels, which are directly derived from the PN9000 phase noise test system, as shown in Figure 1.

Both single channel paths are powered by a common power supply and lead to a dual channel digitizer instead of two digitizers. This does not impact the performance of the DCNTS. The DCNTS relies on both hardware and software to measure extremely low phase or amplitude noise. Phase noise test systems are integrated systems, with software a key element to successful measurement of low phase noise.

When Software Rescues Hardware
While most of us dream of a noise-free amplifier, a zero dB conversion loss phase detector, the real world tells us otherwise quite often. Certainly we cannot find these perfect pieces of hardware, but we can almost overcome this situation thanks to math.
Digital signal processing (DSP) used in the below equation integrates a cross-spectrum function in order to keep the common spectral distribution present in the two independent channels, A and B, of the DCNTS.

where X and Y are the complex FFT results of the phase or amplitude signals defined below:

This complex multiplication operation is done for m iterations and all the complex cross-spectrums are averaged, cancelling the non-coherent part, while revealing the coherent component linked to the only common points: the power splitter input and everything before it.
The obtained improvement performance can be seen in the measurement equation below:

If we relate in dB the cross-correlation gain will be

The handy Table 1 will help us estimate the number of averages (cross-correlations) and make us realize its basic drawback: measurement time. Fortunately, the DCNTS has optimized settings to reduce those measurement times.

As we can suppose, this gain would increase as m grows, but there will be some hardware limits eventually, which stop the improvement. This limitation is particularly true in added phase noise (also called residual phase noise) measurements but not so much the case in absolute phase noise, as described in this article.

Ultra Low Noise Saw Oscillators
Phase noise measurement
TEMEX proposes a new line of SAW oscillators featuring ULN. Typical curves at 320 MHz are displayed in Figure 2. With the DCNTS, it is now possible to extract the exact phase noise of each oscillator from the measurement of a set of three similar oscillators.

The result is very impressive; the noise floor is better than -180 dBc/Hz, while the close-in phase noise is comparable to the best of UHF crystal oscillators multiplied up to the same carrier frequency, as demonstrated in the next paragraph.

SAW versus Crystal Oscillators
The frequency range of high performance crystal oscillators is from 5-150 MHz. The frequency range of SAW oscillators is from 300 - 600 MHz. In order to compare phase noise performances, one must be at a same carrier frequency and take into account the degradation factor of 20.log10(N) when multiplying the carrier frequency by N.

A phase noise comparison is proposed in Figure 3 between a TEMEX crystal oscillator at 100MHz (also measured on DCNTS) and the ULN SAW oscillator at 320 MHz. The degradation factor is 10.1 dB (20.log10(3.2)).

In Figure 4, the improvement in phase noise performance is evident. For offsets below 1 kHz, SAW and crystal phase noises are similar. For offsets above 1 kHz, phase noise is improved with SAW, and the improvement reaches 12 dB for the noise floor. In a radar application, that means an improvement in the detection of high-speed stealth targets for both range and detection sensitivity improvements.

Leeson’s Model
Let’s recall the equation:

The variables are defined as follows:
G: compressed power gain of the loop amplifier
F: noise factor of the loop amplifier
k: Boltzmann’s constant
T: temperature (K)
P0: carrier power (Watts) at the output of the loop amplifier
F0: carrier frequency
QL: loaded Q of the resonator in the feedback loop
aR and aE: flicker noise constants for the resonator and loop amplifier, respectively

All the variables may be measured separately and then the phase noise may be simulated using Leeson’s model.

Reference Table 2 for a list of the values of the variables in the case of the SAW oscillator at 320 MHz. Figure 5 presents both DCNTS measurement and Leeson’s model simulation of the SAW oscillator at 320 MHz.

To our knowledge, it is the first time that the measurement confirms so accurately the relevance of the Leeson’s model for high performance SAW oscillators, and therefore the pertinence of the design of the SAW oscillator.

Acknowledgment
Noise XT (formerly Europtest) would like to thank Enrico Rubiola (Femto-ST, France), who provided a great analysis of the cross-correlation “magic” and who inspired us many times in our quest of the lowest phase noise we could get.

TEMEX would like to thank Noise XT (formerly Europtest) for their patience and helpful collaboration in the measurements of ULN crystal and SAW oscillators with the new DCNTS.

References
[1] Rubiola, E. The Magic of Cross-Spectrum Measurement from DC to Optics, invited, Proc. 22 European Frequency and Time Forum (EFTF) art. no.186, Toulouse, France, 23-25 April 2008.
[2] Rubiola, E. The Effect of AM Noise on Correlation PM Noise Measurements, and 1/f Noise in RF and Microwave Amplifiers, TimeNav’07 Conference, Geneva, Switzerland, 28 May - 1 June 2007.
[3] Rubiola, E. V. Giordano, Correlation-based Phase Noise Measurements, Rev. Sci. Instrum. vol.71 no.8: p. 3085-3091, August 2000.

Noise eXtended Technologies

Email this article to a friend!