M
Member
Most RoF link datasheets quote an output P1dB and everyone blames the modulator. In a passive intensity-modulated link the compression you measure at the PD side is usually the photodiode, and it moves with photocurrent while the MZM's does not. Here is the 20-minute test that separates them, and the numbers I expect on a 3.5 GHz link.
Why the two compress for different reasons
- MZM: at quadrature the transfer function is sinusoidal, so the fundamental compresses as the drive approaches roughly 0.4-0.5 Vpi. For Vpi = 3.5 V into 50 ohm, Ppi is about 21 dBm, and 1 dB compression of the optical modulation lands near 12-14 dBm of RF input. It is set by Vpi and bias angle, not by optical power.
- PD: compression comes from carrier sweep-out and the load. Output RF power at photocurrent Iph into RL is Iph^2*RL/2. At 8 mA that is about +5 dBm, and most high-linearity PDs start bending 3-4 dB above that. It is set by photocurrent, not by the modulator.
The two-current test
1. Bias at quadrature, drive one tone well below compression (say -10 dBm at 3.5 GHz), and attenuate until the PD current reads 2 mA. Sweep up until the output drops 1 dB and record the input power.
2. Raise the optical power until Iph reads 8 mA. Repeat the sweep.
3. If P1dB falls by roughly 4-6 dB when you quadruple the photocurrent, you are PD-limited. If it holds within 1 dB, you are MZM-limited.
4. Cross-check with bias angle: detune 10 degrees off quadrature. An MZM-limited link loses 2-3 dB of P1dB and grows even-order almost immediately; a PD-limited link barely notices.
What actually helps
- PD-limited: lower photocurrent and recover the gain electrically, or move to a PD with a higher saturation current. Chasing +3 dB of optical power to cover a 2 dB gain shortfall is exactly how people end up with -8 dB of SFDR.
- MZM-limited: reduce Vpi, or bias slightly below quadrature if you can accept the even-order growth. Backing off 3 dB from compression has bought us 6-8 dB of SFDR at 3.5 GHz.
Two things that look like compression but are not
- The post-PD amplifier's own P1dB. If the output compresses but photocurrent and the module's RF output both look clean, terminate into a known load and repeat before blaming anything optical.
- Bias-tee DC saturation: the bias tee load can shift the PD bias point as photocurrent rises, so the compression you see is actually a bias-angle change. Watch the PD DC bias while you sweep.
One field note: on a multi-km feeder the same module measured 2 dB better in P1dB on the bench than up the mast, and it was the splice loss changing the photocurrent, not the module. Log photocurrent with every compression sweep or the comparison is meaningless.
Question for the board: what P1dB do you measure at 8 mA photocurrent on a 1550 nm link, and does your datasheet state the photocurrent it was taken at? Half the ones I read do not.
Disclosure: our RoF modules come from F-tone Networks (f-tone.com / fiberda.com). The numbers above are from our own qualification sweeps, not from a datasheet.
Why the two compress for different reasons
- MZM: at quadrature the transfer function is sinusoidal, so the fundamental compresses as the drive approaches roughly 0.4-0.5 Vpi. For Vpi = 3.5 V into 50 ohm, Ppi is about 21 dBm, and 1 dB compression of the optical modulation lands near 12-14 dBm of RF input. It is set by Vpi and bias angle, not by optical power.
- PD: compression comes from carrier sweep-out and the load. Output RF power at photocurrent Iph into RL is Iph^2*RL/2. At 8 mA that is about +5 dBm, and most high-linearity PDs start bending 3-4 dB above that. It is set by photocurrent, not by the modulator.
The two-current test
1. Bias at quadrature, drive one tone well below compression (say -10 dBm at 3.5 GHz), and attenuate until the PD current reads 2 mA. Sweep up until the output drops 1 dB and record the input power.
2. Raise the optical power until Iph reads 8 mA. Repeat the sweep.
3. If P1dB falls by roughly 4-6 dB when you quadruple the photocurrent, you are PD-limited. If it holds within 1 dB, you are MZM-limited.
4. Cross-check with bias angle: detune 10 degrees off quadrature. An MZM-limited link loses 2-3 dB of P1dB and grows even-order almost immediately; a PD-limited link barely notices.
What actually helps
- PD-limited: lower photocurrent and recover the gain electrically, or move to a PD with a higher saturation current. Chasing +3 dB of optical power to cover a 2 dB gain shortfall is exactly how people end up with -8 dB of SFDR.
- MZM-limited: reduce Vpi, or bias slightly below quadrature if you can accept the even-order growth. Backing off 3 dB from compression has bought us 6-8 dB of SFDR at 3.5 GHz.
Two things that look like compression but are not
- The post-PD amplifier's own P1dB. If the output compresses but photocurrent and the module's RF output both look clean, terminate into a known load and repeat before blaming anything optical.
- Bias-tee DC saturation: the bias tee load can shift the PD bias point as photocurrent rises, so the compression you see is actually a bias-angle change. Watch the PD DC bias while you sweep.
One field note: on a multi-km feeder the same module measured 2 dB better in P1dB on the bench than up the mast, and it was the splice loss changing the photocurrent, not the module. Log photocurrent with every compression sweep or the comparison is meaningless.
Question for the board: what P1dB do you measure at 8 mA photocurrent on a 1550 nm link, and does your datasheet state the photocurrent it was taken at? Half the ones I read do not.
Disclosure: our RoF modules come from F-tone Networks (f-tone.com / fiberda.com). The numbers above are from our own qualification sweeps, not from a datasheet.
