Modulation Choice for a 28 MHz Microwave Channel
A network operator must upgrade a licensed 7 GHz microwave link to carry 80 Mbit/s of net payload inside a 28 MHz channel. The hop is 22 km with 1.2 m dishes, and the radio supports adaptive modulation and coding with a channel roll-off of 0.25 and FEC rates from 1/2 to 7/8. Under clear sky the receive level is -60 dBm; the link is designed for a 25 dB multipath fade depth and the receiver noise figure is 4 dB. The vendor quotes, for BER 1e-6 including implementation margin: 64-QAM rate 3/4 needs about 26 dB C/N, 16-QAM rate 3/4 about 19 dB, and QPSK rate 1/2 about 7 dB. The transmitter uses a 30 dBm P1dB output amplifier with a single fixed output back-off setting chosen for the highest modulation. The customer requires the link to "stay up" at the 25 dB fade depth and wants to know what payload that guarantee actually represents.
1. Determine the symbol rate and identify which modulation/coding states can deliver 80 Mbit/s net in this channel.
Read the worked answer
Symbol rate. With roll-off alpha = 0.25, occupied bandwidth = Rs x 1.25, so Rs = 28 / 1.25 = 22.4 Mbaud.
Bits per symbol needed. 80 Mbit/s net over 22.4 Mbaud requires (80 / 22.4) / FEC rate bits per symbol. With rate 3/4 that is 4.76, so 4 bits (16-QAM) is not enough:
- 16-QAM, rate 3/4: 22.4 x 4 x 0.75 = 67.2 Mbit/s - short.
- 16-QAM, rate 7/8: 22.4 x 4 x 0.875 = 78.4 Mbit/s - still short, and 7/8 is the best rate offered.
- 64-QAM, rate 3/4: 22.4 x 6 x 0.75 = 100.8 Mbit/s - meets the requirement with 20 Mbit/s of headroom.
- 64-QAM, rate 7/8: 117.6 Mbit/s - more capacity but needs a higher C/N.
So 80 Mbit/s is only achievable in 64-QAM; the lowest compliant state is 64-QAM rate 3/4, and 16-QAM cannot reach the target even at the strongest coding rate. This matters for the availability discussion: the required payload sits in the top modulation state, so any fade forces a throughput reduction, not just a margin reduction.
2. Compute the clear-sky and faded C/N and state what payload can actually be guaranteed at the 25 dB fade depth.
Read the worked answer
Noise floor. N = -174 + 10log(28 x 10^6) + 4 = -174 + 74.5 + 4 = -95.5 dBm.
Clear sky. C/N = -60 - (-95.5) = 35.5 dB. Against 64-QAM rate 3/4 at 26 dB, that is 9.5 dB of headroom, so 100 Mbit/s (and the required 80 Mbit/s) is available until the fade exceeds about 9.5 dB.
At the 25 dB design fade. C/N = 35.5 - 25 = 10.5 dB. At that level the only states that close are QPSK rate 1/2 (7 dB required, 3.5 dB margin) and marginally QPSK rate 3/4. QPSK rate 1/2 carries 22.4 x 2 x 0.5 = 22.4 Mbit/s.
Answer to the customer. The link will stay up at a 25 dB fade, but the guaranteed payload at that depth is roughly 22 Mbit/s, not 80. 80 Mbit/s is guaranteed only down to about a 9.5 dB fade, so the real question is what outage probability the customer will accept for the full rate, which depends on the site's fade statistics rather than on the radio alone.
Options if 80 Mbit/s must survive deeper fades. Larger dishes (1.8 m instead of 1.2 m gives about 3.5 dB), a shorter hop, a second parallel channel, or a 2+0 XPIC/CCDP arrangement on two polarizations to halve the bits-per-symbol demand. Each should be traded against the licence conditions for channel and polarization use.
3. The factory sets one fixed output back-off for all modulation states. Explain the consequence and what you would specify instead.
Read the worked answer
Why back-off exists. A 64-QAM signal needs an error-vector magnitude around -30 dB, i.e. about 3% RMS. Driving a 30 dBm P1dB amplifier near compression generates AM-AM and AM-PM distortion of the same order as the constellation spacing, so roughly 6 dB of output back-off is needed: the transmitter runs at about 24 dBm.
Consequence of one fixed setting. If the same 6 dB back-off is applied when ACM drops to QPSK, the transmitter still runs at 24 dBm although QPSK would tolerate only 2-3 dB of back-off. That throws away 3-4 dB of transmit power precisely in the fade states where every dB of link budget matters. The ACM thresholds then sit 3-4 dB lower than the data sheet predicts, the hop drops to a lower state earlier, and the effective fade margin is silently reduced.
What to specify. Per-state back-off, switched with the modulation and coding state, with EVM and adjacent-channel leakage ratio verified at each state (not just the top one). Check three consequences: the licence's maximum EIRP and any power-flux constraint must still be met at the low-modulation states; the far-end AGC and receiver compression must tolerate the higher level; and the adjacent-channel performance in QPSK must be re-measured, since higher output power with a wider relative constellation is not automatically cleaner. Also ask the vendor to state the C/N thresholds measured at the antenna port with the production back-off table, not with the laboratory optimum.