Composite Second Order
How Second-Order Beats Accumulate in a Loaded System
Any amplifier, laser, or active device has a transfer function that is not perfectly linear. Expanding the output as a power series of the input, the squared (second-order) term produces beats at f1 + f2 and f1 − f2 for every pair of input carriers, plus second harmonics at 2f1. In a two-tone test these are easy to see, but a real CATV forward path carries 60 to 158 channels across roughly 54 to 1002 MHz. The number of second-order beats that land within a given channel slot grows roughly with the number of carriers, and the composite second order value is the sum of all of those beats referenced to the carrier they sit beside.
Because second-order products depend directly on the carrier frequencies, CSO is unusually sensitive to the channel plan. In a standard (STD) NTSC plan the visual carriers are not harmonically related, so second-order beats fall at predictable offsets of ±0.75 MHz and ±1.25 MHz from each carrier and appear as discrete diagonal lines on a spectrum analyzer. Harmonically related carrier (HRC) and incrementally related carrier (IRC) plans were created precisely to steer these beats onto or very close to the carriers, where the human eye and the demodulator largely mask them. This frequency-plan dependence is the single biggest practical difference between CSO and the noise-like third-order behavior of composite triple beat.
Level Dependence and Cascade Behavior
For a single stage, CSO degrades 2 dB for every 1 dB the per-channel output level is raised, the signature of a second-order mechanism. Across a cascade of identical amplifiers the beats add somewhere between coherently (worst case, 20 log N) and as random power (best case, 10 log N) depending on phase alignment and cable dispersion, which is why long trunk runs budget each amplifier 10 to 12 dB better than the end-of-line spec. Designers trade per-channel drive level against both CSO and noise: pushing level improves CNR but worsens distortion, so the operating point is set where the two effects cross at an acceptable margin.
Governing Relationships
fbeat = f1 + f2 and fbeat = f1 − f2
Single-stage level dependence:
ΔCSO ≈ 2 × ΔPout (dB, for each dB of per-channel level)
Cascade of N identical stages:
CSOcasc = CSO1 − X log10(N), X ≈ 20 (coherent) to 10 (power)
Relation to second-order intercept:
CSO ≈ (IP2 − Pcarrier) − 10 log10(Kbeats) dBc
Where Pout = per-channel output level, IP2 = second-order intercept point, Pcarrier = single carrier level, Kbeats = number of beats in the worst slot. Example: a stage at minus 65 dBc with 20 cascaded amps and coherent addition → minus 65 + 26 = minus 39 dBc worst case.
CSO Versus Other Distortion Metrics
| Metric | Order | Beat location | Level slope | Typical EOL target | Plan sensitivity |
|---|---|---|---|---|---|
| Composite Second Order (CSO) | 2nd | ±0.75, ±1.25 MHz | 2 dB / dB | < −53 dBc | Very high (STD vs HRC) |
| Composite Triple Beat (CTB) | 3rd | On/near carrier (noise-like) | 3 dB / dB | < −53 dBc | Moderate |
| Cross-Modulation (XMOD) | 3rd | Transferred modulation | 2 dB / dB | < −51 dBc | Low |
| Second-Order Intercept (IP2) | 2nd | Extrapolated reference | 1 dB / dB (fund.) | > +50 dBm device | Device property |
| Carrier-to-Noise (CNR) | n/a | Broadband floor | −1 dB / dB | > 43 to 49 dB | None |
Frequently Asked Questions
How does composite second order differ from composite triple beat?
CSO sums second-order products (f1 ± f2); CTB sums third-order products. CSO degrades 2 dB per 1 dB of level while CTB, being third order, degrades 3 dB per 1 dB. CSO beats cluster at discrete ±0.75 and ±1.25 MHz offsets and show as diagonal lines, while CTB is noise-like under the carrier. CSO is far more sensitive to the channel plan, which is why HRC plans were created to hide second-order beats on the carriers.
What CSO level is required for a CATV forward path?
Analog NTSC practice targets better than minus 53 dBc at the subscriber drop, with individual headend and trunk amplifiers specified at minus 60 to minus 65 dBc so the cascade still meets the budget. Distortions add by 10 log(N) to 20 log(N) across N stages, so a 20-amplifier trunk can lose 13 to 26 dB versus a single stage. QAM-only DOCSIS plants tolerate more because digital carriers run 6 to 10 dB below analog video.
How do you measure CSO on a loaded cable system?
Load the plant with the full CW carrier complement, then notch or gate off the carrier under test and measure the residual second-order beat energy in the empty slot at the ±0.75 and ±1.25 MHz offsets with a spectrum analyzer, referenced to the carrier level for a dBc result. SCTE procedures fix the carrier plan, levels, resolution bandwidth, and beat count so readings are comparable between systems.