RGMII voltage levels

[abquke]

Are the voltage levels of a RGMII device the same as its DVdd supply? That's the guess I have that makes the most sense...

A more sticky question might be regarding mixed voltages between two devices. Can they be mixed (1.8 and 2.5 say) as long as the lower voltage device's Vhigh > V_high-min of the higher voltage device?

[NiHaoMike]

A more sticky question might be regarding mixed voltages between two devices. Can they be mixed (1.8 and 2.5 say) as long as the lower voltage device's Vhigh > V_high-min of the higher voltage device?

If the lower voltage device outputs a high that's a comfortable margin above the threshold at which the higher voltage device calls high, yes. The other way around will need measures to attenuate the voltage if it's not tolerant of the higher voltage. The trickiest to deal with are bidirectional lines.

[ajb]

Generally the IO voltage will be the same as Vdd unless otherwise specified.  Some devices (among PHYs and switch ASICS anyway) will have separate VddIO pins to power the IO buffers so that the IO voltage can be selected separately from the operating voltage(s).  You could think of these parts as having level-shifters built in, conveniently preconfigured with the correct IO directions. 

[abquke]

I'd read a TI App Note on voltage translation (SCEA065B) where they showed off a SN74AXC8T245 translating a RGMII bus. I got all excited only to find that the SN74AXC8T245 didn't seem to be available. I asked about a near-replacement on the TI E2E forums on the important parameters and the response was more or less "Yeah, we don't guarantee anything for that application".