Wednesday, April 19, 2017

A daylight-tolerant TIA (TransImpedance Amplifier) optical receiver

While the majority of my past experiments with through-the-air free-space optical (FSO) communications were done at night, for obvious reasons, I had also dabbled in optical communications done during broad daylight, with and without clouds.

The challenge of daylight:

Clearly, the use of the cloak of darkness has tremendous advantages in terms of signal-noise ratio and practically-attainable communications distances, but daylight free-space optical communications has some interesting aspects of its own:
  • It's easier to see what you are doing, since it's daylight!
  • Landmarks are often easier to spot, aiding the aiming.
  • Even in broad daylight, it is possible to provide signaling as an aiming aid, such as a mirror reflecting the sun - assuming that it is sunny.
  • The sun is a tremendous source of thermal noise, causing dilution of the desired signals.
  • Great care must be taken when one wields optics during the day:  Pointing at the sun or a very strong specular reflection - even briefly - can destroy electronics or even set fire to various parts of the lens assembly!
As you might expect the biggest limitation to range is the fact that the sun, with its irradiance of around 1kW/m3 (when sunny) can overwhelm practically any other source:  This is why the earliest "wireless" communications methods often used reflected sunlight, notably the Heliograph, where a mirror was "modulated" with telegraph code or the "Photophone", a wireless audio transmitter using reflected light, an invention of Alexander Graham Bell from the 1870s with earlier roots - a device that Bell himself considered to be his greatest invention.
Figure 1:
Optical communications during daylight.  In the center of this contrast-enhanced picture (the red spot) is the light from an optical transmitter using a 30+ watt LED at a distance of 13.25 miles (21.3km).  This image is from my own "modulatedlight" web site.
Click on the image for a larger version.

Means of detection:

While the modulated speech may be produced in any number of ways (vibrating mirror, high-power LED, LASER) some thought must be given on the subject of how to detect it.  While the detector itself need not be spectacularly sensitive due to the nearly overwhelming presence of the thermal noise from the sun, it is worth making it "reasonably" sensitive so that this is not the limiting factor.  An example of an un-sensitive optical receiver (e.g. one that is rather "deaf" and itself is not likely to be sensitive enough even for daylight use) is a simple circuit using a photodiode as depicted below:
Figure 2:
A simple phototransistor-based receiver (top).  This circuit was built by Ron, K7RJ, simply to demonstrate the ability to convey audio a short distance:  It is (intentionally) not optimized in any way and is not at all sensitive.  A similar, but slightly better circuit was found on the Ramsey Electronics LBC6K "Laser Communicator", which was also quite "deaf".  See the article Using Laser Pointers for Free-Space Optical Communications - link that more thoroughly explains this issue.  This image is from my own "modulatedlight" web site and used with the permission of Ron Jones, K7RJ.
Click on the image for a larger version.

The circuit in the top half of Figure 2 (above) depicts one of the simplest-possible optical receivers - and one of the "deafer" options out there.  In this case a biased phototransistor is simply fed into an LM386 audio amplifier and the signal is amplified some 200-fold (about 46dB.)  As noted in the caption, this was a "quick and dirty" circuit to prove a concept and was, by no means optimized nor does it take maximum advantage of the potential performance of a phototransistor.

As it turns out a phototransistor isn't really the ideal device because it is, by itself, intrinsically noisy.  Another, more practical issue is that its active area is typically quite small which means that it won't intercept much light on its own.  Of course, any half-hearted attempt to use any device for the detection of optical signals over even a rather short distance of a few hundred meters would include the use of a lens in front of the detector - no matter its type:  The lens will easily increase the "capture area" by many hundred-fold (even for a small lens!) and will effect noiseless amplification with the added benefit of rejecting light sources that are off-axis.  With the tiny active area of a phototransistor it can be difficult to properly and precisely focus the distant light onto that area and it is likely that unless very good precision in both alignment and focus can be maintained, the "spot of light" being focused onto the phototransistor will be larger than its active area, "wasting" some of its light as it "spill"s over the sides.

One of the biggest problems with a circuit like this is that there will be a level of light at which the phototransistor saturates, and when this happens the voltage across its collector and emitter will go very close to zero and the received signals will disappear, possibly "un-saturating" only briefly during those points in the modulation where the source light happens to go toward zero, resulting in badly distorted sound.  In broad daylight the phototransistor may be hopelessly saturated at all times unless an optical attenuator (e.g. neutral density filter) is used to reduce the total light level and/or more current is forced through it.

Introducing the TransImpedance Amplifier (TIA):

A much better circuit is the TransImpedance Amplifier, a simple circuit that proportionally converts current to voltage.  With this circuit (see Figure 3) one would more likely use a PIN Photodiode, a device akin to a solar cell, in which the output current is pretty much proportional to the light hitting its active area. This is quite unlike the manner in which a phototransistor is typically used where in the former case the impinging light causes a voltage drop across the device.

Figure 3:
A simple transimpedance amplifier.
(Image from Wikipedia)
In this circuit the junction between the inverting (-) and noninverting (+) inputs of the op-amp "wants" to be zero, so as the current from the photodiode increases in the presence of light, its output voltage will increase, sending a portion of that current through feedback resistor "Rf" until the overall voltage is zero.  What this means is that the output voltage, Vout, is equal to the current in the photodiode multiplied by the magnitude of resistance, Rf - except that the voltage will be negative, since this is an inverting amplifier.

As an example, assume that Rf is set to 1 Megohm.  Assuming no leakage and a "perfect" op-amp we can determine that if there is -1 volt output, we must have 1/1000000th of an amp (e.g. 1 microamp) attributed to Ip, the photodiode current.  This sort of circuit is often used as a radiometric detector - that is one in which its output is directly proportional to the amount of light striking the photodiode' surface, weighted by intervening optics and filters and the spectral response of the detector itself.

For more about the Transimpedance Amplifier circuit, visit the Wikipedia page on the subject - link.

This is OK when the photodiode is in complete darkness - or in near-complete darkness, but what about strong light?  We can see from the above example that if we have just 10 microamps - a perfectly reasonable value for a typical photodiode such as the BPW34 in dim-to-normal room lighting - that Vout would be -10 volts.  If this same circuit were taken outside, the diode current could well be many hundreds or thousands of times that amount and this would "slam" the output of the op amp against a rail.

One of the typical means of counteracting this effect is to capacitively couple the photodiode to the op amp so that only changing currents from a modulated signal get coupled to it, blocking the DC, but there is another circuit that is arguably more effective, depicted in Figure 4, below.

Figure 4:
A "Daylight Tolerant" Transimpedance amplifier circuit.
In this circuit the DC from the output is fed back to "servo" the photodiode's "cold" side so that its "hot" side (that connected to the op amp's inverting input) is always maintained at the same potential as the noninverting input, eliminating the DC offset caused by ambient light.  The disadvantage of this method is that it does not lend itself well to reverse-biasing the photodiode to reduce its capacitance, but between the high intrinsic thermal noise levels of daylight and the related photoconductive shunting of the device due to high ambient light, this is largely unimportant.  For the photodiode the common and inexpensive BPW34 may be used along with many other similar devices.
Click on the image for a larger version.

This circuit is, at its base, the same as that depicted in Figure 3, with a few key differences:
  • An "artificial ground" is established using R101 and R102, allowing the use of a single-polarity power supply.  This artificial ground is coupled to the actual ground via C102 and C103 making it low impedance to all but DC and very low AC frequencies.
  • The voltage output from the transimpedance amplifier section (U101a) is feed back via R104 to the "ground" side of the photodiode (D101) to change its "ground".  If there is a high level of ambient light, the voltage at the "bottom" end of D101 (at D101/C107) goes negative with respect to the artificial ground, setting the DC voltage at the non-inverting input of the op amp to zero, cancelling it out.
  • R104 and C106 form a low-pass filter that passes the DC offset voltage to the bottom of D101, but block the audio.  In this way the DC resulting from ambient light that would "slam" the op amp's output to the negative rail is cancelled out, but the AC (audio) signals remain.  The time constant of this R/C network is slow enough to be "invisible" all but the very lowest (subaudible) frequencies, but more than fast enough to track changes in ambient light.
  • By not placing any additional components between the "hot" end of the photodiode and the op amp, the introduction of additional noise from the components (including microphonic responses of the coupling capacitor) is greatly reduced.
In the above circuit the values of R103 and C104 would be chosen for the specific application.  In a circuit that is to be used at very high light levels where high sensitivity is not very important a typical value of R103 would be 100k to 1 Megohm:  Do not use a carbon composition but a carbon film or (better yet) metal film resistor is preferred for reasons of noise contribution.  While tempting to use, a variable resistor at R103 is also not recommended as these can be a significant source of noise.  If multiple gain ranges are used, small DIP switches, push-on jumpers or even high-quality relays - wired to the circuit - could be used to select different feedback resistances, knowing that these devices have the potential of introducing noise as well as additional stray circuit capacitance.  (Such a relay/switch would be wired on the "output" side of the op amp/relay of the feedback resistance(s).)

C104 is used to compensate for photodiode and other circuit capacitance and without it the high frequency components would rise up (e.g. "peak"), possibly resulting in oscillation and general instability.  Typical values for C104 when using a small-ish photodiode like the BPW34 are 2-10pF:  Using too much capacitance will result in unnecessarily rolling off high frequency components, but will not otherwise cause any problems.  A small trimmer capacitor may be used for C104, either "dialed in" for the desired response and left in permanently or optimally adjusted, measured, and then replaced with an equal-value fixed unit.

Again, the reason why the ultimate in high sensitivity is not required on a "Daylight Tolerant" circuit is that during the daytime, the dominant signal will be due to thermal noise from the sun - a signal source strong enough that it will submerge weak signals, anyway:  It need be sensitive enough only to be able to detect the sun noise during daylight hours.

The op amp noted in Figure 4 is the venerable LM833, a reasonably low-noise amplifier and one that is cheap and readily available (and actually works well down to 7 volts - a bit below its "official" voltage rating allowing the above circuit to powered from a single 9 volt battery) but practically any low-noise op amp could be used:  Somewhat better performance may be obtained using special, low-noise op amps, but these would be "overkill" under daylight conditions.

For nighttime use - where better sensitivity was important - a standard "TIA" amplifier that omits the DC feedback loop to cancel out the DC (potentially noise-contributing) components along with higher values of Rf would offer better performance, but for much better low-noise performance (e.g. 10-20dB better ultimate sensitivity) under low-light conditions than is possible with standard components at audio frequencies in a TIA configuration the "Version 3" optical receiver circuit described on the page "Gate Current in a JFET..." - link is recommended, instead.


Additional web pages on related topics:
The above web pages also contain links to other, related pages on similar subjects.


[End]

This page stolen from "ka7oei.blogspot.com".