23 May 2018
My most recent little hack has been a linear fiber-optic transmitter. This is actaully a pair of circuits, one that converts a continuous input voltage into a proportional optical signal, and a matching circuit that converts the optical signal into a continuous output voltage. when I used to do EMC testing, I frequently used fiber optic transmitters and receivers to get signals from inside the EMC chamber to the outside where I could see them on a scope. In this situation, running copper cables might have both interfered with the measurements, and may not have been expecially good for my health as some of the testing was done at very high RF levels (100V/m+). In more everyday circumstances, fiber optic signal bridges can be useful when trying to transmit signals in noisy environments. They are also usefulr when bridging signals across differences in local grounds - and I am not even talking high voltage here - even a few hundred millivolt ground delta between the item being monitored, and the monitoring equipment can cause all kinds of measurement problems.
Most linear fiber optic transmitter/receivers I have seen rely on either pulse width modulation or pulse frequency modulation. While either of these techniques has a huge advantage in that they are relatively oblivious to the exact characteristics of the communications 'channel (the transmitting LED/LASER, optical fiber, and the receiving photodiode), it can be difficult to get a lot of bandwidth out of them unless some pretty sophisticated supporting electronics are used. These techniques are mainly useful when you need stable gain and offset - but this comes at the price of eating a lot of channel bandwidth. While communications-grade fiber optic components offer gigagits/second of bandwith, getting this performance out of the supporting electronics can be difficult. It is even more difficult when you are using inexpensive components based on LEDs and phototransistors designed for plastic fiber optic cable, which may only support maximum pulse rates in the hundreds of kHz or low MHz range.
My first cut at the linear transmitter/receiver shown below achieves a only a fairly low bandwith (-3dB point ~ 10kHz), which is mostly limited by the transmit receive circuitry. As the LED and photodiode pair used in the channel support Mbps data rates, I am sure that there is a lto of room for improvement here.
The transmitter is shown in the picture below - uses some simple op-amp circuitry to convert an input voltage into a proportional current used to drive the fiber optic transmitter (blue component on the right side of the board). A DIPswitch allows far adjustment of input span and offset so that the device can accomodate signals of 0-5V, 0-10V, +/-5V, and +/-10V. While the light output of LEDs is fairly linear with respect to drive current, this only holds over so wide a range, so a key feature of this circuit is to bias up the LED current so that even at the bottom of the input range it is always 10% of of the full-scale.
The receiver (shown below) is basically the inverse of the transmitter, except that instead of driving a current into an LED, the receiver's photodioe (the black component on the left side) feeds an I/V converter, and the switches are used to set coarse gain and optionally bias the signal to get a bipolar (+/-) output. Because the fiber optic cable has an attenuation based on length, and becaue the transmitter LED and receiver photodiode are not matched parirs, potentiometers are needed to make scale adjustments once the system is assembled.
Below you can see a picture of the system assembled on my bench with plastic fiber-optic cable, There is only maybe 6 or 8 feet of fiber optic cable connecting the units.
The scope-shots below show the system's response to 1kHz square and sine waves, with yellow being the input signal and the blue being the output signal. There is about a 20us delay between input & output.
And in the scope-shot below, you can see the -3dB point behavior. Note that the response also is somewhat non-linear. Guess we aren't going to be using this circuit in any audiophile applicatons :(.
Finally, one of the major drawbacks of analog signal trasmission via fiber optics is not only that the signal strength diminished with the fiber length, but it is also dependent on how you form the fiber - as in bending it. As an example, I pinched the fiber into a tight radius and you can see the signal drop significiantly compared to the -3dB exmaple shown above.
Based on the results above, I may experiment with this a bit more to see if I can get some better bandwidth. I am also kicking around a PWM-based system, as well as a digital one using an ADC and DAC to go analog-to-digital-to-digital, but that's kind of cheating and it might just end up being Yet Another Microcontroller project - you know the kind, a microcontroller, two analog parts, and a pile of code..