LVPECL with VCC of 3.3V on a Cyclone3

The quick answer: No.

The long answer:

The Cyclone 3 handbook says LVPECL requires VCCIO of 2.5 V, and is only available on Clock Inputs of Cyclone 3 devices (Table 6-4 on page 6-13 of the handbook)

That being said, quartus doesn't know what your real VCCIO is connected to, so you can say it's a 2.5V bank, but have it all hooked up to 3.3V, but the timing, and threshold levels wouldn't be guaranteed in this state, so it would be use at your own risk. However, if you state one item as 2.5V in the bank, (For the LVPECL), you would have to declare all IO's in that bank as 2.5V IO's, so you risk invalidating more than just the LVPECL thresholds and timing, but all the other IO's that are in that bank.

If you need LVPECL outputs, you need to have an external LVPECL driver anyway. Since the Cyclone 3 doesn't support LVPECL drivers.

Pete

What dod you mean, LVPECL in- or output?

LVPECL output can be emulated by respective resistive dividers, LVPECL as input can be interfaced with a LVDS receiver and level shifting resistors. But no differential I/O standard is specified to work with VCCIO of 3.3 V.

The problem is, that Cyclone III doesn't provide a differential input standard for VCCIO of 1.8V. You are able to divide down a single ended PECL output to met the SSTL-18 reference voltage, but the signal swing is below the specification, and I fear, the level will be also affected by temperatur variations. Without a true differential input, it's difficult to achieve reliable operation.

Setting the comparators VCCO to minimal speicified 2.7V may bring you nearer to an acceptable solution. Personally, I would prefer external LVDS receivers.