Optical transistors are inherently slower than optical diodes. The earliest and slowest was the 4N35 optocoupler, which had a rise and fall time of 5 microseconds at 100 Ohm, but it was still quite common, and its bandwidth could not exceed 10 kHz, which could be used in applications such as EEG or pulse-width modulation motor control. For example, the original MIDI specification in 1983 recommended the use of optocouplers such as PC-900 or 6N138, so that the transmission speed of digital data could reach tens of kps. In order to achieve the fastest speed, optical transistors need to be appropriately biased and provide the appropriate load, for example, the 4N28 can operate up to 50kHz with optimal bias, but only 4kHz is typical.
Designing the circuits of optical transistor-type optocouplers requires sufficient margin to withstand large fluctuations in the parameters of common electronic components. Some parameters fluctuate destructively, for example, if the transfer function of the optocoupler in the feedback circuit of the DC-DC converter changes and parasitic oscillations, or if the optocoupler has an unexpected delay that causes a short circuit in one of the pins of the H-bridge, the supplier's data sheet only lists the worst-case values for the key parameters. Actual components often make things worse than the worst-case scenario in the datasheet in unpredictable waysBob Pease found that some 4N28 current converters could vary from 15% to over 100%, and the datasheet only listed their minimum values of 10%. The beta of a transistor bipolar transistor may vary from 300 to 3000, so the bandwidth changes to 10:1.
Optocouplers that use a field effect transistor (FET) as a sensor are rare, such as vactrol, and can be used as an analog potentiometer for remote control if the FET spanning voltage does not exceed several hundred mV. Optical FETs do not need to inject a charge into the output circuit when they are turned on, and are especially suitable for sampling and holding circuits.