We developed a 6-channel SFP transceiver board for a customer using the LMH0366 as re-clocker. Our experience so far...
At lab ambient temperatures (20degC), the devices work as expected for SD and HD data rates. At 3G rates, here is excessive jitter introduced by the re-clocker but it can be fixed by writing the special register values as described in an earlier post...
"1. Register 0x19 bits 4:2 should be changed from 000 to 111.
2. Register 0x0A bit 6 should be changed from 0 to 1.
3. Register 0x1C bit 6 should be changed from 0 to 1."
At low temperatures (we tested at -40degC, our customer reported problems below 0degC), the above statement holds with one exception: HD and 3G data rates do not lock with the special register values. Operating the devices with default settings we achieve lock and writing special register values then improves jitter at 3G rates (SD and HD again seems OK without having to write special register values).
Based on numerous posts on this subject on this forum as well as observation during testing it seems as if the various reclockers' PLLs are interfering with each other. Due to space constraints the components are placed fairly close to each other but we have tried to adhere to all recommended guidelines in the datasheet. In an effort to improve performance (i.e. operate the devices with default settings over all data rates at all temperatures) it would be helpful if you can assist with answers to the following:
1. The register writes: I've found a different post (as far as I can ascertain also referring to the LMH0366) specifying new values for registers 0x0A, 0x1C and 0x13. Are we using the correct registers in our application - being 0x19, 0x0A and 0x1C?
2. What happens to the jitter attenuation if using the special register values? I assume performance to decrease as the loop bandwidth is increased. Can you please confirm?
3. The dev kit (SD3GDAIII) schematic indicates a 4.7uF capacitor being used for the loop filter. This is significantly different from the datasheet recommendation of 56nf. Will an increase in the loop filter capacitor beyond 56nF help?
4. Is there a known coupling mechanism involved when placing a number of LMH0366 devices close to each other - i.e. does the noise enter through the VCC pin, the VEE pins or is the device susceptible to radiated emissions from neighbours (i.e. will a small enclosure over each device help)? We are already using a dedicated 2.5V LDO (LT1963) to supply all devices.
Riaan