I think it's current in. http://www.analog.com/static/imported-f ... SM2164.pdfyes gain cell is non inverting, looks like voltage input, not current input.
The 30K is the input V-I converter. The 30K wrapped around from the output back to the 2164 input alters the gain control slope (I think) since it's providing NFB current into the input summing node. The sign of the feedback is negative due to the output I-V. I just figured out that the same trick can be performed with the THAT2180 but in a slightly different way because it's non-inverting overall.
To invert the gain control response with the 2180 assume that there's the typical 20K Rin and a Cin to the VCA. The output I-V also has the typical 20K Rfb and a small Cc. Since the input stage of the THAT VCA is inverting (internally by the log converter) summing the overall input to the I-V's inverting node (using a third 20K brought from the left-hand side of Rin/Cin to the I-V inverting in) allows the VCA to subtract the input from the output. In other words, when the VCA Ec is at 0V (and the VCA gain unity) the overall output is 0 because the VCA's output current is being subtracted from the input current. (Assuming resistor match, precise unity gain etc.) As the VCA goes into attenuation, the overall output increases. Instead of a normal Ec vs "G" response it appears that you get 1/G.
Why this is useful in the context of the phaser I'm not sure yet, but a 1/G Ec response may linearize the effect.
EDIT: Wouldn't it be 1-G?
For G=0 the output is 1. For G=1 the output is 0.
As to the series RC network at the input, it does add a pole but I think it's there for 2164 stability (per the datasheet). I think the C that's doing all the work is the large-ish Cfb around the output. So I bet the dominate (single) pole is Cfb.
I think the SSM2164 topology could be adapted to the THAT2180/2162. The cell I show (and have demoed) is more of the classic approach but I'm not sure that when transformed into the other topology would necessarily sound any different. It would appear to save an op amp buffer though and reduce the requirement for BIFETS.