Content area
Full Text
Hello Bob: I'm designing wideband photodiode amplifiers and using Jerald Graeme's excellent book (Photodiode Amplifiers-Op Amp Solutions) as a reference. In the case of a composite transimpedance amplifier (TIA) discussed in Chapter 6, do you know whether the phase compensation requirement that dictates a value of CF for stability/gain peaking (i.e., the formula on page 58) changes? (Refer to my column "What's All This Transimpedance Amplifier Stuff, Anyhow?" (www.electronicdesign.com, ED Online 4346) Be sure to write down these questions: What BW (min and max) do you need? What noise do you need in that BW? What is your Z source (R, C)? What is the minimum and maximum signal size? Data? These questions can help you define your circuit and amplifier needs. These are usually defined by real circuits and not by a formula. If one circuit doesn't solve your problem, you may need another circuit. No book makes it easy! /rap) I've designed and built TIAs before, but not high frequency. I'm getting ready to design a photodetector for an analog application to measure pulses (amplitudes) as small as 10 µA and as fast as 50-ns pulse width (10-ns rise/fall times). (If you have signals as fast as that, you are definitely interested in op amps with low V noise, and the I noise probably won't be so important, unless you get silly. You have to get low V noise/(Z^sub S^) at the high frequency of interest. And you haven't mentioned your CS, so I can't guess what your Z^sub S^ is. It had better be a small C^sub S^... /rap) Right now I have a spreadsheet (People who use spreadsheets expect some kind of quasi-linear problem. These TIA problems force you to change your whole circuit, so it's not very linear. /rap) where the fixed parameters like the op-amp parameters and diode parameters are entered. Then R^sub F^, damping, and signal current (pulse amplitude) can be played with. What gets calculated is the C^sub F^ (compensation), bandwidth, noise (diode shot, amp current, Johnson, amp voltage, and SNR). The tricky part is the amplifier voltage noise, which experiences noise gain that dominates wideband designs. That is why the datasheet for TI's OPA656 FET amp recommends using the OPA846 (replaced OPA686) and OPA847 (replaced OPA687) bipolar amps, which have lower voltage noise and input capacitance. Besides the TIA topology, I entered in the composite topology and bootstrap but didn't see a real advantage to those. The composite reduces noise bandwidth (and signal bandwidth) with a second op amp in the loop with the TIA, so it probably has some advantages over a separate filter after the TIA. (I have never been enthusiastic about that approach-not a winner. /rap) The bootstrap just made the pole from the compensation capacitance (which is a sum of a few capacitances) dominate versus the second-order pole from the limited open-loop gain and feedback zero. This second-order pole is the bandwidth given by the standard TIA bandwidth formula with 45° of phase margin. But if the R^sub F^, C^sub F^ pole is less than that, it dominates, and bandwidth corresponds to 1/(2*Π*R^sub F^*C^sub F^).