Talk:Current-feedback operational amplifier

From Wikipedia, the free encyclopedia

[edit] CFB amp feedback network

Is it true that a CFB op-amp has a range of recommended resistor values for the feedback network in order to minimise offset and/or maximise bandwidth? If yes, then is this point worth including in the article? Rohitbd 12:43, August 16, 2005 (UTC)

[edit] Article incorrect?

Actually the Bandwidth of Current Feedback Amplifiers is not larger than that of a VFA. The main difference is that the CFA does not have a constant GBW (gain bandwidth product) but rather a constant bandwidth. However it is not correct (at least the way I understand it) to claim it has a higher bandwidth than a comparable VFA. Steve110 07:20, 17 March 2007 (UTC)

I see some changes were made, but the main premise that CFB are higher bandwidth devices was not changed. This is simply wrong. The Bandwith of a CFA of comparable VOL is the same. The main difference to a VOA is the way these things are looked at. Steve110 22:24, 26 March 2007 (UTC)

The most important point, which the article makes only indirectly, is that the gain-bandwidth of the current-feedback amplifier may be programmed externally, to suit the particular application. The words "not constant gain-bandwidth" suggest this, but do not make it clear.

In a voltage-feedback amplifier, you can scale all of the passives in the feedback network (for some constant k, multiply R and L by k, divide C by k), and the closed-loop transfer function is unchanged. It's only the impedance ratio that matters. (And this makes sense; the opamp's gain is dimensionless, and loop gains must be dimensionless (P/(1 + PC), and 1 is dimensionless, so PC is dimensionless), so the feedback network gain must be dimensionless as well.)

In a CFA, this is not true. Scaling the impedance of the feedback network is roughly equivalent to adjusting the compensation of a VFA. This is useful. For example, most VFB opamps are unity gain stable. This is necessary for many circuits, but wasteful when it's not necessary; if a higher gain is taken, then more closed-loop bandwidth could be obtained by decompensating the amplifier a bit. In the CFA, by choosing the absolute impedance of the feedback network, you can "compensate" the amplifier according to your exact needs. (And this makes sense too; the CFA's gain has dimensions of V/A, so the feedback network gain must have dimensions of A/V, which is an admittance.)

So the VFA might be just as fast as the CFA, but in designing the feedback network for the VFA, you will probably throw that speed away, by making the circuit more stable than it needs to be. In a CFA, this is not true. There's other advantages too; parasitics, like capacitance at the inputs, add extra poles in the transfer function of a VFA, which tend to make you go unstable. The low-impedance input of the CFA does not have this problem. (Contrary to the article's claims, Analog Devices is recommending some of their current feedback amps as photodiode front ends, for DVD players and such. At those speeds, the feedback resistor you can take is limited by bandwidth and stability issues; so if the current-feedback amplifier lets you take more gain without going unstable, then you might win on noise too, since you don't have to take as much voltage gain later. The front-end amplifier's input-referred noise is larger, but it gets multiplied by a smaller constant in the subsequent stages.) 74.61.11.168 (talk) 04:58, 4 February 2008 (UTC)