Hi Tom, the dB channel loss in this case is talking about the loss associated with the electrical interface from the ASIC switch to the optical engine.

Again we see a battle to get the optical engine as close to the switch as possible. The amount of loss is determined by distance and the impedance or resistance associated with the electrical traces and connections.

POET has the ability to get very close because heat is not as big an issue with POET's Optical engine due to the athermal waveguides.  The dielectric stack layers offset the thermo optic effects of heat which causes wavelength drift. POET initially provided the metric when they began working on the mux demux 5 years ago:

Low rate of change of the filter response with temperature of minus 0.014 nanometre per degree

Interestingly AYER LABS tunes their micro-rings to 125’C to fix the wavelength to the highest operating temperature of the monolithic silicon. So they use heaters to keep the wavelength within range so that the signal can be read.

POET does not need to do this because their AWG mux demux and associated waveguides are athermal. The waveguides associated with the dielectric stack are not thermo optic dependent. Suresh recently spoke of how clean the RF signal is with a wide open eye diagram at low power and how the waveguides are agnostic to frequency. No wire bonding. Connections are the closest possible  through bumps at the surface where the electrical traces are for ease of access and design flexibility.

Wish I could spend more time on this but we are heading out for a daytrip….freedom!

One more quick note:The applications associated with the DML lasers is where POET really begins to show its muscle as latency is removed and energy efficiency and the reduction of devices ie re-timers and so on become a very big cost saver with further footprint reduction to allowing this new breed of optical engine to get upclose and personal with the Switch and AI ASIC.