This entry is a continuation of my current to voltage converter project. In this entry we're going to talk about some of the goals that we'd like this project to meet. Remember that these were not exactly the same goals we had when starting out, but this is a retrospective entry.
Below is the circuit diagram we ended up with at the end of the last entry, and what we will start our discussion this time with.
We need this current to voltage converter to be able to sense optical power over a broad range to make it a generally useful device since we won't know the photo-current.
The feedback resistor Rf determines the transimpedance of the circuit:
Vout = Ipd*Rf, but it also increases the input impedance of the op-amp
L = Rf/(2*pi*GBW) which decreases the resonant frequency of the circuit.
Therefore a higher transimpedance (i.e. voltage output for a given photo-current) means that the bandwidth of the circuit will need to be smaller.
So we have to make a decision: do we want a larger signal or a larger bandwidth. We'll the answer depends on the application, and I as the designer can't make that decision for the user. The result is that there will be multiple gain and bandwidth settings, so we will need some way to switch Rf and Cf between values. We could do this with a continuous control, like a potentiometer, or discrete controls such as a detented potentiometer or a bank of switches.
A continuous control has the advantage of being able to optimize the output voltage to fill the entire range. However, a non-repeatable control invalidates and calibration that might be done since we won't be able to reliably know the gain. Additionally the bandwidth of the detector will also change with the gain setting, meaning that two variables are coupled in a poorly regulated ways. A potential solution is to fix the bandwidth so that all of them would have the same low bandwidth, but that removes a significant amount of flexibility for the user. Using a continuous control to monitor optical powers over long time scales would be a sketchy solution since if the knob ever turned we would be unable to repeat turning it back. A digitally controlled potentiometer could be a viable solution to approximate continuous, repeatable behavior but would require more board space and communication with an external device, and I am not willing to make either concession.
Similar to a digital potentiometer, a detented potentiometer would give relatively repeatable settings and, depending on how many settings are available, could provide a large tuning range in a small package. However, the same problem with bandwidth tuning remains. A selectable capacitor range to complement the resistor would allow one to not have to artificially lower the bandwidth, but additional switches are actually fairly large in comparison to the size of the electronics and would mean increasing the size of the design significantly. In addition, the introduction of a second control means that the user must be more familiar with the intricacies of the device to operate it effectively, which is a non-negligible consideration.
If we decide to use a switch bank, we can change the whole complex impedance value
Cf) with a single switch setting, which means we can choose the optimal
Cf value for each discrete value of
If we decide to add the complexity we can even use a multiplexer switch that takes encoded inputs from a rotary encoder (this is what ThorLabs does in their PDA36A).
But this is a nice solution since we can select from a reasonable number of optimized gain and bandwidth pairs in a way that doesn't waste board space, is repeatable, and can be customized easily by the end user if they want specific settings.
Ok, so we are going to go with the switch bank with three settings with default transimpedances
Rf = (1e3, 1e4, 1e5)
We are also going to want to drive a transmission line with a reasonable termination resistance.
We don't want our fancy transimpedance op-amp to also be driving the output line since it will most likely end up sourcing large currents (for an op-amp) and may heat up, either of which could change critical properties of the device.
So we are going to want a second stage output buffer with maybe some additional gain, and we will want to also implement an RC-filter with the highest bandwidth setting's frequency to prevent amplification of white noise in the second stage.
Finally, we may have a voltage offset added onto the signal to cancel out some of the DC component (we'll talk more about this in the next entry).
Mechanical and System Design
Now let's consider the mechanical properties of the device. We want to be able to install the detector in a optical cage system, examples of which can be seen here or by googling optical cage systems". We would like to mount the detector to the 4 posts in the 16 mm cage system and have a beam splitter pick off a fraction of the light passing through the cage to direct at the detector. We want the detector to be as close to the photo-diode as possible to minimize cable capacitance which would lower the bandwidth, and we want the detector to be swappable without deconstructing the rest of the cage system. Finally the ability to mount filters and lenses to the detector is important so we will want there to be 0.525"-40 (SM05) tapped holes for optics mounting.
To achieve this we are going to mount the photo-diode directly on the back side of the board in a transistor socket so the photo-diode is removable, the case will be designed to hold 16 mm cage rod systems, and the case will have SM05 tapped holes for mounting optics.
The device will accept power from a central power distribution system that provides a low-noise bipolar DC voltage supply, and return the output signal on a coax cable to a differential detector with high CMRR to minimize noise pickup.