The Quiet Metric That Dictates Link Budgets
When engineers argue about receiver performance, they usually fixate on sensitivity figures and noise temperatures. Fair enough. But the single number that tells you whether a demodulator will actually lock onto a signal is the carrier-to-noise ratio (C/N). Unlike SNR—which gets measured after despreading, decoding, and a pile of processing gain—C/N is a raw, pre-detector measurement. It’s the fundamental balance between the power of your carrier and the noise floor inside the occupied bandwidth. No clever baseband trickery can fix a C/N that’s fallen off a cliff.
I’ve spent years chasing down underperforming satellite downlinks and microwave hops where the link budget looked pristine on a spreadsheet. Then you put a spectrum analyzer on the LNB output and the C/N is 2 dB below the modem’s acquisition threshold. The carrier is there, but it’s drowning in noise. The culprit is rarely the modem. It’s almost always something physical: a connector that’s been quietly corroding for a decade, a dish that shifted half a degree in a storm, or a local oscillator with a phase noise profile that belongs in a horror movie.
C/N vs. SNR: The Distinction That Catches People Out
Datasheets love to blur the line between C/N and SNR, but they’re not the same animal. C/N is measured in the transmission bandwidth before any despreading or decoding. SNR is what you get after the receiver has applied its processing gain. In a spread-spectrum system, the difference can be enormous. A direct-sequence signal might sit 20 dB below the noise floor and still deliver a perfectly usable SNR after correlation. If you mistake the C/N requirement for an SNR spec, you’ll either overbuild the link by 30 dB or—worse—assume the modem can dig deeper than it actually can.
When I’m on site, I measure C/N with a spectrum analyzer set to the occupied bandwidth of the carrier. For a 10 Msps QPSK signal, that’s about 10 MHz. I integrate the carrier power across that bandwidth, then measure the noise power in the same bandwidth with the carrier switched off. The ratio is your C/N. If the modem needs 8 dB and you’re seeing 7.5, the link will drop packets. No amount of begging or configuration tweaks will change that.
Phase Noise: The Silent C/N Killer
Most link budgets assume the noise floor is flat, white, and thermal. Real-world oscillators don’t read textbooks. Phase noise from a cheap local oscillator spreads carrier energy into adjacent frequencies, effectively raising the noise floor inside your receiver’s bandwidth. I’ve seen systems where the thermal C/N was a comfortable 12 dB, but the effective C/N—what the demodulator actually sees—was closer to 9 dB because of a noisy LO in the block upconverter.
This hits narrowband carriers especially hard. A 100 ksps signal has a tiny bandwidth, so even modest close-in phase noise can eat a significant chunk of it. Wideband carriers are more forgiving because the phase noise skirt occupies a smaller fraction of the total bandwidth. When a link underperforms and the thermal C/N looks fine, I go straight for the phase noise measurement. A good spectrum analyzer with a phase noise option will reveal the problem in seconds.
Measuring C/N in the Real World
Field measurements are where theory meets rusty bolts and bird droppings. You can’t just point a spectrum analyzer at the IF output and read a number off the screen. The resolution bandwidth needs to be narrow enough to see the noise floor clearly, but not so narrow that the sweep takes forever. I usually set it to about 1% of the symbol rate and use the marker noise function, which normalizes the reading to a 1 Hz bandwidth. Then I add 10log(symbol rate) to get the noise power in the occupied bandwidth.
One trap: measuring carrier power at the peak of the spectrum and assuming the noise floor is flat. In a filtered system, the carrier has a shaped spectrum, and the peak power isn’t the total power. You need to integrate the power across the full occupied bandwidth. A modern analyzer with a channel power function does this automatically. I’ve seen 2 dB discrepancies between peak-and-assume and true integrated measurements. That’s the difference between a link that works and one that flaps every time a cloud passes.
C/N Requirements Across Modulation Schemes
Different modulation and coding combinations demand wildly different C/N thresholds. A BPSK signal with rate 1/2 FEC might lock at 2 dB C/N. An 8PSK signal with rate 3/4 trellis-coded modulation might need 12 dB. That 10 dB gap translates to a factor of 10 in transmit power, or a dish that’s more than three times the diameter. Choosing the right modulation for the available C/N isn’t just good practice—it’s the whole game.
I’ve watched operators chase throughput by jumping from QPSK to 16APSK without touching the RF chain. The link works fine under clear skies, then crumbles the moment rain starts. A 16APSK carrier might give you 50% more bits per second, but it also demands 6 dB more C/N. If your margin is only 3 dB, you’ve just traded reliability for a speed bump that disappears in bad weather. I always recommend logging C/N over at least 24 hours before signing off on a modulation change. See what happens at 3 a.m. when the dew settles on the feed horn.
Rain Fade and the Margin You Actually Need
At Ku-band and above, rain attenuation is the dominant fade mechanism. It’s a double punch: the rain absorbs and scatters the carrier, reducing received power, while the warmer sky raises the system noise temperature. Both effects hammer C/N. A link budget that only considers free-space loss and clear-sky conditions is a fantasy document.
I design for a minimum of 4 dB rain fade margin at the target availability. For 99.5% availability in a temperate climate, the ITU-R rain model might predict 6 dB of fade at Ku-band. That means your clear-sky C/N needs to be at least 10 dB above the modem’s threshold. If you’re measuring 8 dB on a sunny afternoon, you’ll be down during a moderate shower. The fix is straightforward: a bigger dish, a more powerful amplifier, or a more rugged modulation scheme. Pick one, because the rain won’t negotiate.
Interference: When the Noise Floor Isn’t Just Thermal
Not all noise comes from the thermal agitation of electrons. Adjacent satellite interference and cross-polarization leakage can raise the effective noise floor without showing up as a distinct spike on the spectrum analyzer. Your C/N measurement looks fine, but the C/(N+I)—the ratio that actually matters—is worse. This is a chronic headache on crowded satellite arcs where everyone is squeezing into tighter frequency slots.
To spot interference, I do a simple carrier-off test. Switch off your own carrier and watch the spectrum. If the noise floor drops, something else was contributing. The difference between the noise floor with your carrier on and off gives you the interference power. Subtract that from your C/N and you’ve got C/(N+I). If that number is below the modem’s threshold, you’ll see errors even though the C/N looked healthy. I’ve resolved more than one “mystery” link problem this way.
Antenna G/T and Its Direct Grip on C/N
The receive antenna’s gain-to-noise-temperature ratio (G/T) is the single most important parameter for a satellite downlink. A 1 dB improvement in G/T gives you exactly 1 dB more C/N. That’s why earth station operators obsess over low-noise amplifiers and reflector alignment. A dish that’s off by half a degree can lose 0.5 dB of gain, and that comes straight out of your C/N budget.
I once spent a frustrating week on a link that was 2 dB below spec. The LNB noise figure was fine. The antenna gain measured correctly on the test range. The culprit was the feed horn alignment—it was offset by 3 mm, which reduced illumination efficiency and let in extra spillover noise. After realignment, the C/N jumped by 1.8 dB. No amount of modem configuration changes would have recovered that. The problem was literally mechanical.
Practical Ways to Squeeze Out More C/N
Improving C/N comes down to two levers: more carrier power or less noise power. On the transmit side, keep the high-power amplifier in its linear region. Compression creates spectral regrowth, which throws power outside your occupied bandwidth and can cause adjacent channel interference. On the receive side, minimize cable losses before the first amplifier stage. Every 0.1 dB of loss there is a direct 0.1 dB hit to C/N. Use the shortest, fattest, lowest-loss cable you can afford.
Filtering helps, but it’s a trade-off. A bandpass filter before the LNB can reject out-of-band interference that would otherwise saturate the amplifier and raise the noise floor. But the filter’s insertion loss directly reduces C/N. I prefer waveguide filters for their low loss and high power handling, though they’re bulky and expensive. For lower frequencies, a well-made cavity filter is a practical compromise. Just don’t slap a cheap filter in line and wonder why your C/N got worse.
FAQ
What is the difference between C/N and Eb/No?
Eb/No is the energy per bit to noise power spectral density ratio. It relates to C/N by the formula Eb/No = C/N + 10log(bandwidth/bit rate). Eb/No is handy for comparing different modulation and coding schemes because it normalizes for data rate. C/N is the raw RF measurement you can grab with a spectrum analyzer. Both matter, but C/N is what you can actually measure in the field without a modem.
How does antenna size affect C/N?
Antenna gain scales with the square of the diameter. Double the dish size and you get 6 dB more gain, which directly adds 6 dB to C/N, assuming the noise temperature stays the same. Larger antennas also have narrower beamwidths, which can reduce interference from adjacent satellites and improve the effective C/(N+I).
Can I improve C/N by using a better LNB?
Yes, but only if the LNB’s noise figure is the dominant noise source. The system noise temperature is the sum of the antenna noise temperature and the LNB noise temperature. If the antenna is already seeing 100 K of ground noise, dropping the LNB noise temperature from 50 K to 25 K only improves C/N by about 0.8 dB. The improvement is much more dramatic when the antenna noise temperature is low, like in a cold-sky condition.