Phase noise aspect

A quite confusing topic is phase noise performance. None of the available D-ATV sites on the web shows any information about phase noise needs. So... what impact will phase noise have on digital modulated signals?
Well, this depends on several parameters. Phase noise will distort the constellation diagram and has negative effect on the correct decision for a given point in the constellation. It shall be clear that higher order constellations with lots of closely spaced decision point are more susceptable to phase noise than a simple constellation as QPSK.  QPSK has a big phase margin (45 degrees) before a symbol unintentionally falls into the next quadrant. Therefore, QPSK is again quite robust. The single sided phase noise spectrum of an oscillator is usually defined as the noise level normalized in a 1Hz bandwidth versus frequency offset from the carrier.
Phase noise levels become higher at closer frequency offsets of the carrier. In practical situations we can get some control on the phase noise performance of a VCO (Voltage controlled oscillator) at closer distances from the carrier if we embed the VCO into a phase locked loop. A free running VCO has worse stability and therefore some stabilizing loop is needed to get a steady output signal. The phase locked loop is a control system which compares the actual phase of the VCO output with the phase of a very stable reference signal, for most cases an external crystal oscillator. The loop itself has some transfer function. The transfer function of the phase locked loop is typically characterized as a high pass characteristic. This means that the phase locked loop will suppress phase deviations of the VCO at lower frequencies up to some cross-over point. The amount of suppression is determined by the loop gain, which decreases slowly at higher frequencies up to the cross over point (loop bandwidth). Above this cross over point (which is actually the corner frequency of the high pass characteristic of the closed loop) the loop cannot control the phase deviations of the free running VCO anymore. We can therefore conclude the following: within the loop bandwidth of the PLL, the free running VCO phase noise level will be mainly determined by the control loop. Outside the loop bandwidth, the loop will loose control and therefore the phase noise performance is fully determined by the actual phase noise level of the free running (uncontrolled) VCO itself. In practice, the PLL control bandwidth is chosen such that a desired overall phase noise mask is achieved.
Because noise is concentrated at closer distances from the carrier, negative effects of phase noise become more visible at lower symbolrates. It is similar as the difference between wideband FM and narrowband FM systems which are typically used for FM-ATV and narrowband amateur radio. So what phase noise levels are typical for DVB-S systems? Well, now we come at a mystifying area. If we look at specifications of DVB-S receivers then we first see that most of them have a limited lower symbolrate specification, which is typically around 2-3Msymbols. This is due to practical phase noise issues. A typical DVB-S receiver setup for satellite reception starts with the LNB which is located at the dish. The 11-12 GHz received signals
from the satellite are downconverted to a lower IF range of 950-2150 MHz. This requires some Local Oscillator signal. The LO is realized with a DRO (Dielectric Resonator Oscillator). A well designed DRO can achieve quite low phase noise levels. The typical DRO's which are used in the low cost commercial LNB's for satellite reception are mostly limited to levels of approx. -75 dBc @ 10kHz offset from the carrier. Below are the typical phase noise requirements given for a typical digital satellite reception LNB:

Typical LNB phase noise requirements for DVB-Satellite reception:

-65 dBc at 1kHz
-75 dBc at 10 kHz
-85 dBc at 100 kHz

Given these values we might suggest that for our D-ATV systems these values are good enough and we don't have to spend too much designtime in order to achieve this phase noise performance. However, this would be a premature conclusion. The values above are in fact the typical noise values for satellite television reception which mostly run at high symbolrates. As stated before, higher symbolates have less problems with worse phasenoise. Also, at receiver side the carrier and symboltracking loops have some loopbandwidths. Therefore also the receiver timing recovery loopbandwidths define if the link is more or less susceptible to phase noise. Unfortunately in most cases we do not have any insight in loopbandwidths of receiver carrier/symbol/timing recovery loops and therefore this will be a uncertain factor.
We should design the transmitter such that this doesn't become the weak part of the total link. Some searching on the internet revealed the recommended transmission parameters for DVB broadcasters and operators of DVB multiplexes using the Eutelsat Hot Bird system.


Figure 1. Earth Station Phase Noise Mask

As this figure points out, the phase noise levels for the transmission system are tighter than the typical LNB phase noise requirements. Simulations of several phase noise masks show that the phase noise mask of the typical LNB introduces significant jitter on the constellation points. The Eutelsat spec simulates very sharp constellation points as expected.

This page is under construction and will be expanded soon.