I am working with a team to explore some new concepts for integrated optical waveguides that may significantly improve bend loss and coupling into and out of the waveguides. We are looking most seriously at optical waveguides for high-speed signal interconnects in circuit boards.
My question is for those working in this field. What is the typical acceptable bend loss, or insertion loss, that we we should be shooting for? What performance goals would attract attention to our technology? Any thoughts or experience you have on this subject would be very helpful!
Thank you!
The bending loss strongly depends on the bending radius and the waveguide size. For a conventional silicon strip waveguide with a core size of 500 nm x 220 nm surrounded by glass and the bending radius of 1 micron, the bending loss is ~2 dB per bend at telecommunication wavelength (1550 nm). But the loss drops rapidly as you increase the bending radius.
The insertion loss for this kind of waveguides is ~3 dB/cm. Another important figure of merit for dense photonic integration is cross talk. Light confinement is also important for nonlinear optics and quantum optics applications.
The bending loss strongly depends on the bending radius and the waveguide size. For a conventional silicon strip waveguide with a core size of 500 nm x 220 nm surrounded by glass and the bending radius of 1 micron, the bending loss is ~2 dB per bend at telecommunication wavelength (1550 nm). But the loss drops rapidly as you increase the bending radius.
The insertion loss for this kind of waveguides is ~3 dB/cm. Another important figure of merit for dense photonic integration is cross talk. Light confinement is also important for nonlinear optics and quantum optics applications.
All your losses depend on the selectivity of your photo-detector at the end of the optical circuit. For a given sensitivity, and input optical power, then you can deduce how much you can loose per element (curvature, crossing, add-drops, intersection...) to ensure there is still enough signal at the end.
At a second time, consider the crosstalks... if any, since it increases the noise level, and then impose a higher signal for the photodetection.
I agree with both answers posted: for accurate loss targets, you want to look at your loss-budget for the system, which means how much power does the detector need to hit some error rate (maybe than 1dBm?) and how much power does your transmitter output (maybe 10-20 dBm?).
Typical fiber-to-chip coupling losses should be
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Optics