Welcome to the first in a new series of blog posts “Clocks 201” from Silicon Labs (aka “Silicon Technology”) – The Case for Non-Phase Noise – Part 1. Our previous blog series “Clock 101” published 12 articles in about a year and a half. As its name suggests, we intend to cover the topic of clock design in a broader and deeper context than “Clock 101”.
more is less
As you may recall, clock buffers are usually specified in terms of additive jitter. This is because they have no inherent phase noise sources like XOs (crystal oscillators) or VCXOs for phase locked loops. They only include amplifiers and crossovers. Therefore, we generally think that the clock buffer after the XO adds at least some degree of phase noise or jitter compared to using the XO alone.
All things being equal, this is a common situation unless the amplifier has high enough gain to act as a limiting amplifier or LA. As a result, the measurement source + LA phase noise may actually be reduced. I ran into this a few years ago when I was evaluating various clock+buffer combinations. The results don’t always make sense, and at that time, some of the noise I measured wasn’t actually phase noise, so I made this experience specifically titled for this case and shared with the developers.
To understand how this happens, let’s take a look at what we mean by apparent measurement phase noise here.
You may recall that I discussed overmodulation glitches earlier in “Clock 101 #7: The Case for Spurious Phase Noise Part II”. In that article I considered AM (amplitude modulation) and narrowband FM (frequency modulation) or equivalently PM (phase modulation) relatively few carriers. The general idea of comparing AM and FM/PM also applies to AM and FM/PM noise.
One topic I didn’t cover at the time was the phasor or phase vector representation of AM and NB FM, as shown in the diagram below. The carrier vector is represented by thick red arrows, and the modulated LSB (lower sideband) and USB (upper sideband) vector components are represented by thin blue arrows. The modulated vector sum or composite is the thick blue arrow. The modulation frequency is f
The usefulness of the phasor representation is that it shows that the random noise modulation of the carrier can be seen as consisting of two parts, AM and PM. That is, the noise component that causes the carrier amplitude variation is the amplitude modulation component. Likewise, the noise components that cause carrier angle changes are FM or equivalent PM components.
This distinction is emphasized by using scripts L(f) or ?(f) to refer to point noise or “true” phase noise and scripts M(f) or ?(f) to refer to noise. The first paper I know that uses this notation is:
Spectral Density Analysis: Frequency Domain Specification and Measurement of Signal Stability, Donald Halford, John H. Shoaf, ASRisley, National Bureau of Standards, Boulder, CO, presented at the 27th Annual Symposium on Frequency Control, June 12, 1973 – 14th, https://tf.nist.gov/general/pdf/1558.pdf
The noise magnitude of AM+PM will be the RSS or root sum of squares of the individual modulation contributions. Instruments that also deal with these noise components will “see” the RSS noise directly as phase noise. This is known as apparent phase noise, and it is a particular problem with spectrum analyzers, as described below.
A Brief Introduction to Spectrum Analyzers
As I mentioned in clock 101 #7, the spectrum analyzer does not save phase information, so the low modulation AM excitation is similar to the narrowband low modulation FM excitation.
Below is a block diagram of a typical swept spectrum analyzer that illustrates why. It is essentially a calibrated frequency selective peak response voltmeter. In a mixer, the phase difference between the DUT (device under test) and LO (local oscillator) inputs is arbitrary. Spectrum analyzers know nothing about their relative phases, AM and PM cannot be distinguished.
A Brief Introduction to Phase Noise Analyzers
In Contrast, phase noise analyzers have much less effect on AM. The simplified block diagram below gives the basic idea behind the methods commonly used by phase noise analyzers and signal source analyzers. The mixer is usually a double-balanced mixer to suppress homogeneous stage mixing of the product.
Note that unlike a spectrum analyzer, there is a phase-locked loop (PLL) that enforces a specific phase relationship between the DUT and the reference. Further, it can be shown that AM and PM can be distinguished as follows.
Mixer detects PM and suppresses AM if the phase offset is 90 degrees
If the phase offset is 0, the mixer detects AM and suppresses PM
By design, a phase noise analyzer will outperform a spectrum analyzer in immunity to amplitude modulation.
Explore the limits of use
So how exactly does a limiting amplifier or LA help us? The clue is in the behavior of the LA: it removes, or at least minimizes, amplitude changes from the clock signal. Therefore, if a source has both AM and PM noise components, an ideal limiter will remove the AM component noise, leaving only the PM noise (true phase noise). The diagram below gives the basic idea:
Now back to the original work that prompted this article. If a clock source has significant phase noise, including AM and PM noise, then adding a high gain clock buffer or LA after it will remove the AM, resulting in less measured phase noise than expected. The new components apparently produce “subtractive” jitter rather than additive jitter. This is the case with phase noise.
When should AM be considered when making phase noise measurements?
In short, AM is always a potential consideration when making careful jitter and phase noise measurements. Having a limiter on the lab bench is just as important as having a balance bar.
However, in some specific situations, AM can be more problematic than others.
Measure phase noise with a spectrum analyzer or any other instrument that does not adequately suppress AM. Even when using a phase noise analyzer, the limiter is valuable because it suppresses AM beyond the mixer’s rejection capability. This may be necessary when measuring very low phase noise sources.
Measure low frequency, low phase noise sources. You can recall the 20log(N) rule, that is, if a clock’s carrier frequency is divided by a factor of N, then we expect phase noise to be reduced by 20log(N); however, this rule only applies to phase noise. If significant AM noise is present, this component will become larger as we lower the carrier frequency and may affect the measurement results.
Measure sources of known or suspected AM noise. Clock sources with high common-mode noise fall into this category. For example, when testing the power supply rejection of oscillators, we specifically inject power supply ripple. This is why you see a limiting amplifier as shown in Figure 5. PSRR settings in AN491: Power Rejection for Low Jitter Clocks. Check out the highlighted block from the app comment in the image below.
Troubleshooting. Finally, the ability to distinguish between phase noise and excitation, AM noise and excitation is very useful when troubleshooting a system and determining the root cause of performance issues. Additional testing is required to determine the sensitivity of the final receiver to clock impairments including AM noise.
Hope you all enjoyed this issue of the Clock 201 blog post. In the next article, I’ll give some metrics examples and provide some rules of thumb. As always, if you have topic suggestions or time-related questions for this blog, please send them to firstname.lastname@example.org with “Clock 201” in the subject line. Thanks for reading.
The Links: NL2432HC17-02 CM75DY-28H