An audio interface device for speech-bandwidth
optical communications systems


What it does:

After doing several medium-distance experiments in the field using optical voice links it became clear that it would be nice to have a single device that performed multiple functions:
 The audio interface device:

There are several circuits contained within this device and, for the most part, they operate independently.  Four schematics are shown:
 The audio amplifier, audio recorder interface and other interconnections:
Figure 1:  The audio amplifier and interconnect.
larger version.
Audio amplifier and interconnection

Referring to Figure 1, the speaker amplifier is based on the LM386-4:  It should be noted that ONLY the "-4" version of the LM386 is rated for safe operation above 12 volts, so be aware of this when obtaining the part.  (Another "12 volt safe" equivalent is the LM386D, which is second-sourced by at least one other manufacturer.)

The audio amplifier drives a good-quality speaker and a 4-ohm unit is recommended as this allows a higher audio power to be delivered to the speaker than an 8 or 16 ohm speaker.  An important design feature is the selection of C404, a 0.1uF capacitor and R403, the 10k volume control.  These resistor/capacitor values limit the low-frequency response of the speaker amplifier (down by 3dB at 150 Hz) because of the fact that a small speaker cannot efficiently reproduce low frequency energy, so there's no real reason to amplify them!  Were this simple highpass filtering not done, the speaker amplifier could more-easily be driven into distortion trying to amplify 100/120 Hz hum from light sources - or even clip on low-frequency voice components.

Switch S1 is a 6-position, 2-pole rotary switch used not only to select the audio amplifier amongst the audio sources, but the second half (the "B" section) is used to enable/disable the tone output of the audible S-meter.  Were this not disabled, bleedthrough of the tone would probably occur when the input gain and volume controls were set to high levels.

To facilitate recording for archival or later analysis J4, a stereo jack, is provided.  R407 adjusts the audio level of the receive signal source down to a "safe" level into the digital audio recorder, while R402 similarly adjusts the audio level from the modulator or another audio source, allowing one to use a stereo recorder to record both the original, unprocessed receive audio, but also the accompanying audio transmitted back to the far end.  Note that it will likely be necessary to experiment with the specific audio recorder that one is using to determine a "safe" level.  Once this level is found, setting R402 and R407 for a level at least 6 dB below the peak (e.g. half the voltage) will allow a reasonable margin to prevent the likelihood of clipping under any normal circumstances.

It should be noted that the digital audio recorder records the receiver audio after it has been amplified, but before any filtering or processing.  The reason for this is that, at a later date, one may simply play the recorded audio back through this unit to hear the "raw" audio before any processing (such as "de-scintillation") because any processing of the audio done prior to recording may mask some properties of the recorded audio that may prove useful in later analysis - namely the analysis of scintillation.  It is vitally important that the audio recording device have a high enough sampling rate to avoid aliasing problems with the incidental audio and the 4 kHz pilot carrier.  (Note:  One can simply play the recording back through the unit to "re-process" it as it was originally heard in the field if desired.

Finally, Figure 1 also details the power supply.  S2 is the on/off switch with R405 and D401 providing a "power-on" indicator - which is also useful for locating the device in the dark and providing spatial orientation for the user - especially if the LED is placed near the volume control.  C411 should be located right next to U401, the LM386-4 to provide the best power supply filtering and amplifier stability.  R406, a 10 ohm resistor, and C410, a 220uF capacitor, isolates the power supply of the audio amplifier from the rest of the circuit, further improving stability:  Without these components the slight amount of audio that appears on the power supply could result in feedback at high gain levels.  U402 and C409 provide a quiet, stable 5 volt "mid-supply" reference used as both a virtual ground in the audio circuitry and a voltage reference for the clip warning indicator.

When wiring the audio amplifier section (U401, C411 and other components) it is best to locate these away from the rest of the circuit, making only a single ground and V+ connection to the rest of the circuit to prevent the likelihood of ground loops.  All of the components to the right of R406 should be located close to each other and away from the audio amplifier - especially U402 and its associated capacitors.

A note about the wiring of J1, the TX audio input and the presence of C401 and R401 in the ground lead of the jack.  This was done to break up a possible DC ground loop between the modulator's power supply and that of this circuit.  Without these components it is possible, under some conditions, for the modulator's full supply current to appear on the TX audio ground lead, possibly resulting in damage to the digital audio recorder or leading to audio feedback problems associated with ground loops.  It is recommended that, if this unit and the modulator are not operated from the same power source, that the ground (negative) leads be tied together externally.

Note that there are actually two "headphone" (or external speaker) jacks:  J3 and J5.  J3 is a "disconnect" type of jack, causing the speaker to mute when something is plugged into it whereas J5 does not disconnect the external speaker.  J5 was added later as it was noted that while using headphones improved the ability to copy weak, noisy signal, it also muted the speaker making it impossible for anyone else present to hear the same audio.  It is worth noting that if both the built-in speaker and headphones (plugged into J5) are used it is likely that the volume in the headphones will be very high so one should choose headphones that have a volume control built into the cord to allow the speaker volume to be high enough to be useful, yet prevent the headphone-wearer from being deafened!

Adjustments:

As mentioned above it will likely be necessary to determine the input level at which the audio recorder begins to clip.  For the receive audio a 1kHz tone is inputted and the audio level is adjusted just to the point where the "Clip warning" indicator just starts to come on.  With the audio recorder connected to provide loading, R407 is adjusted so that the signal going to the audio recorder is at no more than half the voltage required to drive the recorder into clipping:  If your recorder does not have a VU meter or a means of determining clipping it is recommended that one export the audio file to an audio editing program such as "Audacity" and look at the amplitude of the recorded signal.

Similarly, for the transmit audio monitoring the modulator is set to maximum output and the monitor port of the modulator is connected to the "TX Audio In" jack, J1, while R402 is adjusted for a signal that is no more than half that required to drive the recorder into clipping.

Operation:

Other than the volume control (the operation of which is obvious) there is S1, the 6-position selector switch, of which only 5 positions are currently used:
  1. In the first position, one can hear the audio from the transmitter.  This can be useful to monitor what is being sent to the far end, but it should go without saying that feedback would likely result if an open microphone is nearby!  This is the audio that is being sent to one of the stereo channels of the audio recorder.
  2. The next position is the audible S-meter.  As described below, this generates a tone that is proportional to the power (in dB) of the received signal.
  3. The next position is unused.  It may be used to "mute" the speaker without adjusting the volume.  (Actually, I couldn't figure out what to do with it...)
  4. The next position is the "raw" audio input which monitors the audio output by the optical receiver with no filtering.  This is the audio that is being sent to the other channel of the stereo audio recorder.
  5. The next position is notch-filtered "raw" audio - the same as above, but with the 4 kHz pilot carrier removed.  No scintillation compensation is applied.
  6. The final position is the "scintillation compensated" output audio, as described below, with the 4 kHz pilot carrier removed.
Component notes:
The Input buffer/amplifier and audible S-meter:
Figure 2:  Input buffer amplifier and audible S-meter.
Click on the image for a larger version.
Input buffer amp and audible S-meter

Input buffer amplifier and gain adjust.

Figure 2 shows the input buffer amplifier (U3C) used to amplify the signal to a nominal level as well as to provide a low-impedance audio source for the other circuits, audio amplifier, and record audio output.  This circuit includes some RFI protection (C101) and is capable of up to 40dB of gain.  The reason for having variable gain is due to the fact that the signal levels from an optical receiver depend on how much light is available from the distant transmitter, plus the depth of its modulation.  Because this signal level could easily vary over 10's of dB depending on circumstances, R103 allows adjustment of the gain to boost this signal to the highest "safe" level.  This not only keeps this signal level in a "safe" range for the audio recording device - well above the noise floor - but it, along with the Clip warning indicator (see below) prevents any circuits from being overdriven, something that would likely result in distortion.



Audible S-meter

One of the difficulties in achieving precise optical alignment is being able to judge the amplitude of the recovered audio.  When the optical path is first established one typically uses a bright light to provide an initial "eyeball" reference for the location of the distant transmitter so that the users know where to look!  At some point it will be necessary to "talk in" the alignment of the distant transmitter - something that is typically done over a radio or telephone link.  This process can be be quite tedious as it is very difficult to convey to the other end exactly what is being seen and when. 

Assuming you gotten things "close" and the "rough" alignment of the transmitter and receiver has been completed, there is also the task of the fine-tuning of the link.  Often, a modulated tone is used and adjustments are made to obtain the loudest tone - but this has several difficulties:
One way to minimize some of these problems is through the use of a tone that varies in frequency according to the signal strength.  By design, both my PWM and Linear modulators have built-in tone generators, capable of produce a variety of audio frequencies - including a very precise, fixed 1 kHz tone, modulated at 100% power and it is this tone that may be used to provide an amplitude reference for peaking the received signal.  Note that this unit is intended to be used with an Amplitude-Modulated lightwave communications system and cannot work with an FM-type system:  This scheme relies entirely on the fact that the 1 kHz tone being received from the distant end will have a loudness that is proportional to strength of the optical signal being received and this would not necessarily be true with a 1 kHz tone modulated atop a frequency-modulated carrier!  Having said that, it is likely that a purely subcarrier system would be designed to have available a means to both amplitude modulate and detect an amplitude modulated test signal such as this.

Note that using an amplitude-modulated tone for peaking is really the only practical way to discern the strongest optical signal.  I've seen suggestions made for somehow monitoring the current (or voltage) from the photodetector itself, but it should be remembered that not only does this parameter vary over many 10's of dB - a fact that makes it extremely difficult to detect when at low levels - but it is easily swamped by normal variations in the photodector's operations, such as temperature or even ambient light - plus there can be quite a bit of DC bias on the photodiode in the first place.

Instead of trying to relay the "loudness" of the received 1kHz tone via a radio or telephone link, this circuit produces a tone that has a frequency that is roughly proportional to the power of the received tone, in dB.  By using the pitch to indicate signal strength rather than loudness it is much easier for most people to peak the signal as one simply adjusts for the highest note - unless one suffers from severe "Tone Deafness."  In actual testing and use it has been found that variations of less than one dB are easily detectable despite the fact that the circuit has well over 40dB of usable range. 

U3D is a bandpass filter centered on 1 kHz.  Because the Q is fairly low, tuning isn't very critical and using standard 5% parts, it should be centered fairly close to 1kHz - but R118 may be tweaked slightly if it is off-frequency.  This filter is vitally important as its purpose is to remove extraneous noise, such as most of the broadband white noise from the receiver as well as the majority of 100/120Hz energy (and most of its harmonics) from lighting, leaving intact the 1 kHz tone from the distant transmitter.  As noted below, use good-quality and stable capacitors for C103 and C104 - see the "component notes" below.

U3A is a simple logarithmic amplifier that increases the dynamic range of the audible meter from 15-20dB to well over 40dB by having an output voltage that increases more-or-less linearly for each doubling of the input voltage.  By doing this, the voltage changes relatively little over a very wide range of signal levels, but even fairly minimal changes in input levels can still be detected.  Note that the output of this very simple logarithmic amplifier will increase (and thus, the tone) with lower temperature because of the change of the forward voltage drop across the diodes, but since we are using this only for relative readings this shouldn't be a noticeable problem to the user.

The output of the log amp, a sort of rounded square wave, is slightly filtered by C106 and R107 and then amplified by U3B to several volts peak-to-peak (under high-signal conditions) and following this is a simple rectifier and filter consisting of D103, D104, C108 and R110, the output of which is a voltage that is roughly proportional to the level of the 1 kHz tone in decibels.  A sample of this voltage is buffered by U1C and is made available on the front panel, allowing one to use a voltmeter to check the relative signal level in addition to observing the tone frequency if this is desired.

R110 is used not only to set the time constant of the voltage filter, but it also scales the voltage downwards for input to U4, a 4046, which is used as a VCO (Voltage Controlled Oscillator) - a necessary step owing to the fact that U4 is powered from the +5 volt supply and the voltage across C108 could exceed 5 volts.  On U4, C109 sets the general operating frequency range, and for this reason a somewhat temperature-stable capacitor should be used (about anything other than a disk ceramic capacitor would be fine) while R11 sets the maximum frequency and R112 set the minimum frequency:  Without R112, the output of U4 may cease under no-signal conditions when it tries to faithfully produces a "zero hertz" output in response to a zero volts input.  As shown, the frequency range is from about 100Hz (no signal) to 2.5 kHz (maximum signal.)

Controlled from the rotary switch, Q101 is used to enable/disable U4's oscillator:  It was noted in early testing that if the input gain and amplifier volume control were turned way up, the oscillator of U4 could be heard in the background from cross-coupling of wiring even when it wasn't selected:  Disabling the oscillator when the audible S-meter was not being used cured that problem.  R114, C111 and C112 are used to reduce the level of the output tone, block DC, and filter some of the harmonics from it.

Adjustment:

The only required adjustment is that of R110.  To do this, first apply a voltage to the "VCO Enable" input to turn on the oscillator.  Next, set R110, "Tone Range Adjust" to mid-rotation and R103, the input gain control, to minimum and then apply a 1kHz tone to the main audio input.  Increase the level of the 1kHz tone until the highest-pitch tone is observed at the "Tone Out" point:  You may need to increase R103 to achieve the highest-pitch tone.  Now, adjust R110 to obtain the highest pitch possible - and then adjust it down slightly (by a few musical notes) to allow for a bit of extra "headroom" in the drift of the logarithmic amplifier's diodes with temperature.

Component notes:
 Operational notes:

It is best to start off by adjusting the input gain control (R103) so that under no-signal conditions, the tone pitch just starts to increase, being keyed by noise as evidenced by a randomly wavering tone pitch:  In this way even the slightest presence of a 1 kHz tone from the distant end will begin to register as an increase in the pitch of the tone.  During alignment it is very important that one makes sure that the audible S-meter isn't being "pegged" (at the highest frequency) - something that is easily checked by occasionally adjusting R103 to reduce gain and also by noting that the pitch of the tone becomes constant rather than waver due to atmospheric disturbances which chould be clearly audible on any paths over a kilometer or so.

As you might suspect, scintillation shows up as a randomly varying tone pitch, but even so, it is still easy to determine the best signal - in spite of the constantly-changing tone!

Using the audible S-meter:

Here is a clip demonstrating the use of the audible S-meter when aligning an optical link using ordinary, cheap laser pointers:

Recording from September 3, 2007 - For more info, see the "Revisiting the 107 optical mile path" web page:
In the first few seconds of the recording one can hear brief "hits" of the 1 kHz tone from the modulated laser in the right channel while in the left channel there are momentary changes in the pitch of the tone from the audible S-meter as the laser briefly sweeps past the receiver.

Once the signal has been acquired, note that although the pitch is varying it's still easy to tell what the "average" pitch is.  For the person at the other end, aiming the transmitter, hearing these "hits" allows rapid correlation of the movement made with a peak in signal - even if it is brief - potentially allowing a repetition of the motion that caused the signal to be seen.  In this manner one may be able to quickly "dial in" the aiming of the transmitting device and then adjust for the highest tone pitch, indicating the maximum signal and optimal aiming.  It should be noted that this is best done using good, old-fashioned analog communications (e.g. amateur radio FM simplex or via a repeater) rather than a digital communications link such as a cell phone or a digital radio system due to the unavoidable delay caused by these latter systems and this delay can be confusing to the person trying to aim the transmitting device until one becomes accustomed to it.

"Breaking" the S-meter's low-signal threshold:

As noted above, when using the audible S-meter it is important that when it's set up that it will trigger on the optical receiver's noise even when there is no signal.  To demonstrate, note the first fiew seconds of the above recording:  When the laser isn't there one hears a randomly varying tone from the audible S-meter from its keying on the receiver's noise.

This is important because the S-meter has a low "low signal threshold" that must be exceeded before one will ever get a reading and once this level is crossed, the tone will start to increase when even more signal (such as the 1 kHz tone from the modulated light source at the other end or even noise) is detected.  If one didn't first exceed this threshold on noise the S-meter would be much less-sensitive.

This is done under no-signal conditions (only noise from the optical receiver) by increasing the gain of the input amplifier by adjusting R103 so that the tone's pitch rises from the "no signal" frequency (around 100 Hz or so) to a randomly-varying tone of several hundred Hz.

Avoid "pegging" the S-meter on strong signals during alignment:

Conversely, when signals are very strong it's possible to "peg" the S-meter.  When this happens the pitch will be high (around 2 kHz) but not be varying much and in that case, adjust R103 downwards to lower the pitch to a midrange.  If the S-meter is pegged, finding the best peak may be difficult as the pitch won't seem to change much when the receiver's aiming is adjusted.  When in doubt, simply reaadjust R103 for a "mid-range" pitch.


Comments on in-the-field use:

The "audible S-meter" has already been used successfully many times in field conditions:
- On a 15 mile path using 3-watt LEDs and Fresnel lenses:  The first time that this system was used was in aligning the optical gear on our "standard" test shot across the valley.  On this occasion, the far end's LED was very easy to see with the naked eye and "rough" alignment was done, via voice feedback on the radio, only to the point of just being able to see the beam from the far end.  At that point, the near end used that weak beam to peak the receiver using the audible S-meter.  Then, the tone from the audible S-meter was then transmitted to the far end via radio and they used the tone as feedback for precise aiming:  The entire process was quick and painless!

- On a 15 mile path using a standard red Laser pointer:  Previously, peaking the very narrow beam of a laser pointer was a very tedious and frustrating processes:  The delay between the far-end observer seeing the laser and being able to tell the person trying to aim it made the aiming processes only slightly less frustrating than something completely useless.  In this case, the already-aligned receiver (the alignment having been done during the LED testing) was used for receiving and a 1 kHz tone was modulated onto the laser.  The far end simply swept the beam back and forth, listening to the S-meter's tone from the far end via radio.  Because even the briefest "hits" could be heard - even if they are off-point and not readily visible to the naked eye - it was possible to align the laser pointer precisely on peak, even with the laser pointer being mounted on a cheap photographic tripod.  Even so, it was difficult to the make minute adjustments to the tripod without "over-shooting" and knocking it completely off-point.

- On a 107+ mile path using 3-watt LEDs and Fresnel lenses:  During this test, the air was extremely hazy due to wildfires in California, but views through an 8" telescope revealed that the a vehicle's headlights could just be seen through the haze.  Using the telescope and voice feedback, the far-end's LED transmitter was approximately peaked.  While it was modulated with a 1 kHz tone, the near end's receiver was then swept until a deflection on the audible S-meter was noted.  At this point the S-meter's tone was then transmitted back to the far end and used to peak up the transmitter.  Final touch-up was done on both ends by each end alternately using the other's 1 kHz tone for receiver peaking, as the beamwidth of the receiver is narrower than that of the transmitter and is the more-critical adjustment.

- On a 107+ mile path using laser pointers in two directions:  Seeing conditions were good and the laser pointers were adjusted using the pointing devices described on the "Using Laser Pointers..." page mounted to standard tripods.  The use of this pointing device make the pointing of laser pointers much easier than it had been!

- On a 173+ mile path using 3-watt LEDs and Fresnel lenses:  While initial alignment was accomplished by sighting the far end using a telescope, full end-to-end alignment was completed using the audible S-meters at each end.  Despite the extremely weak signals, the system still worked nicely, able to detect the presence of the alignment tone from the far end before it became readily audible to the "naked ear."

- On a 22+ mile daylight path using very high power and Fresnel lenses:  Despite the high noise level contribution of daylight, the system worked flawlessly and allowed precise aiming of both ends of the optical path.

Clip warning indicator:
Figure 3:  Clip warning indicator.
Click on the image for a larger version.
Signal clip warning indicator

The purpose of the "clip warning" circuit in Figure 3 is to provide an indication that the peak audio levels are approaching a level that may result in clipping (distortion) in the audio recorder in addition to (possibly) exceeding the useful range of the other circuits.  If this condition is noted, signal levels may be reduced by readjustment of the input gain control, R103.

The input audio signal is full-wave rectified by U5C and U5D to allow easy detection of both positive and negative audio peaks while U5B is configured as a comparator with a slight amount of hysteresis, the "indicator" threshold being set with R208.  Because U5B inverts, positive-going pulses are produced when the input level exceeds the threshold level, and this pulse charges C202, a capacitor used to stretch out the pulses to allow longer-duration illumination of the clip indicator LED.  The capacitor-filtered voltage is buffered by U5A and this voltage is used to turn on Q201 and, in turn, D204, the "clip warning" LED.

Adjustment:

R103 is set to minimum gain, and a 1kHz sine wave is applied to the main audio input.  Using an oscilloscope, the main amplifier/buffer audio output (pin 8 of U3C) is monitored while the audio input level and/or setting of R103 is adjusted for a sine wave with a 1.5 volt peak-to-peak amplitude:  R208 is then adjusted so that the clip LED just illuminates.  This adjustment should yield a clip indication at a level that is safely below clipping of other portions of the audio chain.

Note:  If a digital audio recorder is used, experience has shown that the "clip" light's threshold should be set to illuminate at 10-15dB below the clipping threshold of the digital audio recorder.  In actual use, it is easy to forget to check the settings of the input gain control, allowing the audio level to be too high.  In most cases, an audio recorder with 16 bit A/D resolution has adequate dynamic range and suitable signal-noise ratio (even the inexpensive ones usually have at least 70dB S/N) so that audio with peaks 10-20 dB below the clipping level will reproduce adequately.

Component notes:
Scintillation Compensator:
Figure 4:  Block diagram of the scintillation compensation system.
Click on the image for a larger version.
Block diagram of the scintillation compensator system

One difficulty encountered with optical through-the-air communications is that of scintillation.  This effect manifests itself as "twinkling" in distant light sources such as stars or distant streetlights.  As the through-the-air length of the path increases, an optical signal is increasingly affected by this phenomenon, resulting in often rapid and extreme, random variations in signal amplitude.

There are several ways to minimize scintillation:
In this case, because we are using AM to modulate the light sources, there is another method available to us to combat the effect of scintillation and that is to track and compensate for changes in the amplitude of the signal.  A practical way to do this is to transmit an amplitude reference along with the audio to be used to restore the signal at the receive end using a keyed AGC system.  It should go without saying that this system can only recover a signal that still exists:  It can do nothing for those portions of the signal that have disappeared into the noise floor as its purpose is to level out the amplitude of the received signal to improve intelligibility.

As mentioned before, both the PWM and Linear modulators include built-in tone generators and one of these is a 4 kHz "pilot" tone, modulated at 25% of full output (12dB down) in order to serve as an amplitude reference.  A frequency of 4kHz was chosen because it was above the speech range, but still within the frequency response passband of even a fairly low-bandwidth optical receiver.

Figure 4 is a block diagram showing the operation of the system:
Figure 5:  Schematic of the scintillation compensator.
Click on the image for a larger version.
Schematic diagram of the scintillation compensator system
Figure 5 shows the schematic of the scintillation compensator.

U1A is a MFBF (Multiple Feedback Bandpass Filter) circuit tuned to the pilot carrier, 4 kHz.  A notch output is obtained by summing the input signal (the junction of R301 and C301) and adding it with the out-of-phase bandpass signal from the output of U1A using amplifier, U1D.  The result is that at the notch frequency, the two signals cancel each other out yielding audio being output by U1D that has been filtered of the 4 kHz pilot tone.

The 4 kHz bandpass output of U1A is routed through a simple highpass filter (C312/R312) and into an active 2-pole highpass filter consisting of U1B.  This amplifies the signal somewhat and it is then passed through another simple highpass network consisting of C315 and R317.  The result of this filtering is that signals at 1 kHz are attenuated by more than 40dB while signals around 4 kHz are passed easily.  This asymmetrical bandpass response removes those frequencies at which the majority of voice energy resides, that is those below around 2.5 kHz and were that energy still present, the filter would track the audio content in addition to the pilot carrier level.  This filtered 4 kHz output is then fed into the rectifier input of U2A, an NE571 compandor IC, with C308 being used to set the AGC time constant.

The output of U1D - the notch-filtered audio - goes into the NE571, which is wired as a compander, with a "gain cell" connected across the feedback path of the '571's built-in amplifier.  The result of this is that as the amplitude of the pilot carrier goes down, so does the output from the rectifier, which increases gain of the amplifier.  Likewise, as the pilot carrier amplitude increases, the gain of the amplifier is commensurately reduced.  Because of the design of the NE571, the rectifier and gain cell track within a few dB over more than 40dB of dynamic range resulting in an audio output that is in lock-step with the amplitude of the pilot carrier, and in this way, the scintillation is removed.

Limitations of this method

It should be immediately pointed out that while this method will effectively combat scintillation effects of a received signal, it cannot possibly recover a signal that has already been lost to the noise.  What it does do is to reduce the annoying effects of scintillation that can cause individual syllables or even words to be lost as the signal momentarily drops in amplitude, and it reduces distortion to a degree in that it largely removes the random amplitude variations that are imposed on the audio waveforms that particularly affect lower frequencies.  If there is sufficient signal margin, the lost audio will simply be brought up as the signal fades, improving intelligibility. but if the signal is already fairly weak, instead of audio, one will simply hear bits of noise - but it seems that this makes for a more intelligible audio source than without the compensator, as the human brain seems to be able to deal better with bursts of noise where syllables should be rather than silence.

When designing a pilot-based, keyed AGC system such as this, one important design point must be to take into account two important factors:
It was through empirical testing that both of these factors were determined.  In past testing, two primary tests were done:
As expected, the use of the Laser represented, by far, the worst-case scenario.  It was noted that significant (30 dB) amplitude variations within a period of 30 milliseconds were common with many smaller amplitude variations occurring in well under 10 milliseconds.  The LED, on the other hand, had much slower fading, typically 10dB of change occurring in about 60 milliseconds.  An example of the differences between coherent and noncoherent light sources may be heard here. 

Using the Laser as the worst-case example, the design goal of this circuit was to be able to compensate for the variations experienced in that test.  Through the use of a recorded audio clip I was able to evaluate the operation of the compensator on an actual clip that had been recorded in the field and to this end the time constant of the AGC circuit was set at approximately 1 millisecond.  The AGC's ability to operate at this rate is irrelevant unless the pilot carrier's detection bandwidth is suitably wide as well, which means that the filter must not be so narrow that the pilot tone's passing through it would not be able to track the rate-of-change.  As can been seen from the circuit in Figure 5, U1A acts as a 4 kHz bandpass filter, the "Q" of which is purposely low:  If the Q were fairly high, it would be able to reject off-frequency energy better, but it's response to variations in the amplitude of the pilot carrier would be slowed, not to mention that the filter itself would tend to delay the pilot carrier as it changed amplitude, causing the AGC response to slightly lag the change in audio.

To further improve rejection of off-frequency energy, a 4-pole highpass filter was employed, following the 4 kHz bandpass filte as noted.  The use of a highpass filter was chosen because it could effectively reject the frequencies at which the vast majority of the energy was present, namely below 2 kHz, but it wouldn't have too much of an effect in slowing the response of the variations of the 4 kHz pilot carrier.  Circuit simulations indicate that the group delay of this entire filter at 4 kHz is on the order of 250 microseconds with an attenuation of greater than 10 dB at 2.5 kHz and more than 40dB at 1kHz.

Adjustment:

The only adjustment that is really required is that R303 be adjusted for the best rejection of the 4 kHz pilot tone:  At least 20dB of rejection should easily be obtainable, but well over 40dB could be managed if either R305 or R306 were made slightly variable:  Typically, one would temporarily parallel a 1 megohm trimmer potentiometer across R305 or R306 (one may need to try both, as it could be either R305 or R306 that might need to be adjusted) and alternately adjust R303 and the trimmer to obtain the best notch.  One would then (carefully) disconnect the 1 Meg trimmer potentiometer, measure its value, and then replace the trimmer with a fixed resistor of a standard value closest to that of the trimmer.

In order to maintain notch stability, C302 and C303 should be temperature-stable units, preferably polystyrene, silver mica, possibly C0G or NPO ceramic, or even mylar - but never with ceramic disk capacitors of the "X", "Y" or "Z" type (e.g. X7R, Y5P, Z5U, etc.)

The component values shown are appropriate for a 4 kHz pilot tone that is 12dB below the peak audio.  If the pilot tone is of a lower amplitude it may be necessary increase the value of R308 in order to reduce the amount of signal appearing at the VGA (Variable Gain Amplifier) to prevent distortion.  Alternatively, one could modify the values of the U1B highpass filter to amplify the 4 kHz tone to compensate.

Operation of the scintillation compensator:
Figure 6:  The audio interface unit, outside and inside.
Outside view of the audio interface
The "guts" of the audio interface unit

When changing operational modes, always turn down the volume control - especially if using headphones!

As you might expect, in the absence of a pilot tone the gain of this circuit will immediately go to maximum and if the user is wearing headphones this could result in painfully-loud audio!  While this will likely result in the appearance of just a lot of noise or other audio it could also cause feedback to occur - particularly if the volume control and/or input gain control is set very high and the audio input jack is unterminated!

If signals are extremely weak it may be useful to switch to the "Scintillation Compensation" mode, even if a pilot carrier isn't being transmitted.  Without a pilot tone present the AGC will increase the gain considerably, providing even more gain than is available even if R103 (the input gain control) is adjusted for maximum.  If this is done keep in mind that owing to the extreme amount of audio gain (as much as 80dB) that feedback may result, especially if the volume control setting is near maximum.  It should also be remembered that the amount of signal being sent to the audio recorder is dependent upon the setting of R103, but is not in any way affected by the operation of the scintillation compensator - and, if you are using the AGC of the compensator to increase gain, it should be remembered that decreasing the setting of R103 too much may degrade the recording.  In other words, if you are running the gain "wide open" using the scintillation compensator, it is best to keep the gain high using R103 (but avoiding feedback) but the volume control set to a lower level.

Comments:
 Component notes:
 Observations using the scintillation compensator:

Initial testing of the scintillation compensator was done with a previously-recorded audio clip that consisted of music and speech transmitted along with the pilot tone via a Laser using a large-aperture emitter and this same clip was used to test the design of the scintillation compensator and was the basis for the empirical determination of the needed time constants in the AGC system.

The audio clip below consists of the following segments, demonstrating the operation of the scintillation compensator.  This audio clip contains exactly the same audio played twice - first without the scintillation compensator, and then with the compensation active.  In each case, the 4 kHz pilot tone was removed by the notch filter.
Remember:  Both portions of the above clip were transmitted using the laser/telescope combination:  Ignore what I said in the voice announcement!

Comments about the audio clip:

In experimentation it  has always surprised me that despite severe scintillation, the effects on speech intelligibility was less than I would have expected.  This makes sense owing to the rather redundant nature of speech and the ability of the brain to "fill in the gaps."  In the case of the Laser scintillation, the periods during which the audio was badly attenuated were brief enough that typically, only a syllable or two might have been lost, but in most cases, enough of the original speech remained to be able to fill in the blanks.  Nevertheless, listening to such audio can cause "ear fatigue" and usually requires that the audio gain be turned up quite high - often enough that the peaks of the audio are extremely loud and causing clipping of the audio amplifier or, if you are using headphones, risk hearing damage.  Having de-scintillated audio can greatly reduce the peak-to-average ratio and mitigate several of these factors.

Upon listening to the above clip there are several things that you might have noticed:  The background noise is mostly 120Hz (plus harmonics) from urban lighting - an inevitable result of the fact that the optical path spanned a metropolitan area.  Had this test been done in an area free of artificial light sources the noise floor would not only have been free of hum, but would have been a number of dB lower, thus resulting in a better overall receive system signal/noise ratio.  In the "un-compensated" clip, the background hum stays constant - as you'd expect - but in the "compensated" clip, the level of the background noise fluctuates wildly as the AGC tracks the pilot carrier.

Another interesting thing is that the "un-compensated" clip has what sounds like clipping-related distortion, but this is much diminished in the "scintillation-compensated" clip.  As it turns out this is, in fact, distortion caused by the original scintillation when the rate of change of amplitude is a significant portion of the period of the lower-frequency audio components.  In particular, the bass notes and certain speech components are distorted as their waveforms are too "slow" to ride atop the amplitude envelope caused by the scintillation.  After compensation, much of this distortion is corrected, as can be heard from the clip.

Transmit Audio Null:

This circuit takes a sample of the transmit audio (that used to provide "transmit audio" to the recording device) and uses it to help remove some of the transmit audio intercepted by the receiver.

When the transmitter and receivers are fairly close to each other, there is inevitably some interception of the transmit beam's signal via the receiver - due to Rayleigh scattering, dust in the air, and other things.  Because the optical transceivers may be run full-duplex (that is, transmitting and receiving at the same time) some of the transmitted audio can make its way back to the speaker - where it is picked up again by the transmitter's microphone, possibly resulting in feedback.

One obvious method to deal with this problem is with the use of headphones, which reduces the amount of "receive" audio being picked up by the microphone.  This is also useful when the "other" side is also using a speaker/microphone combination and the audio that you are transmitting is coming back across the link - also (possibly) causing feedback!

It isn't always practical for everyone to wear headphones - especially if you have more people than headphones/jacks - so it is often desirable to run an "open speaker."  While little can be done to prevent feedback from your audio coming across the link from a "open mic/speaker" at the other end, you can do something about your own transmitted audio being picked up by your receiver.  The answer to this problem is similar to that solved by a Telephone Hybrid - that is, a device that can amplify one source of audio while rejecting another source of audio appearing on the same signal path.

Since we already have available a "copy" of our own transmit audio, we can use it to cancel out some of the transmit audio that appears on our receiver.  To do this, we amplify our own transmit audio with a phase opposite of that bit of transmit audio that is appearing on the receiver, and then add the two together.  If properly done, the transmit audio that is appearing on our receiver will be canceled out, leaving intact the audio being received from the far end.

This circuit has two controls:  An "A/B" phase switch, and a "Null" control.  Because the phase of the incoming signal is unknown - depending on the electronics of the modulator and receiver - the circuit produces two "copies" of the transmit audio:  One "in-phase" of the source audio to our transmitter, and the other 180 degrees out of phase:  A simple switch allows us to select one or the other to determine which is the appropriate phase to achieve cancellation.  The "Null" control allows differing amounts of our transmit audio to be inserted into the mix:  Too much or too little audio will not result in optimal cancellation.

Limitations:

With such a simple circuit it is not likely that "complete" nulling of the backscatter audio will occur over a wide frequency range andthe main factor that limits the performance of this circuit is the fact that the optical receiver, by its nature, does not have a flat phase and amplitude response over the entire audio spectrum.  What this means is that some portions of the audio frequency range will be nulled out better than others:  For example, if one adjusts it for optimal nulling of, say, a 1 kHz tone, there will be a significantly reduced amount of nulling of frequencies that are much lower and higher.  Overall, one can expect to achieve over 10dB of nulling - particularly if the circuit is adjusted for the optimal null of the tone frequency at which feedback is most likely to occur.

To achieve better nulling, it would be better to feed a copy of the transmit audio that had the same phase and amplitude characteristics as that of the receiver, and the easiest way to do this would be to construct a "sampling" receiver - that is, an optical receiver identical to that being used to receive signals that was, instead, coupled only to the transmit beam.

This circuit cannot null out another artifact of the transmit beam being intercepted by the receiver:  Noise.  Inevitably, additional noise will be added by the transmitting LED, scattering medium, etc. and because this is random - and wasn't present in the original transmit audio - it cannot be removed.

The schematic of this portion of the circuit has yet to be added.


Additional comments on construction:

After the pictures in figure 6 were taken, several modifications were made:



Return to the KA7OEI Optical communications Index page.

If you have questions or comments concerning the contents of this page, feel free to contact me using the information at this URL.
Keywords:  Lightbeam communications, light beam, lightbeam, laser beam, modulated light, optical communications, through-the-air optical communications, FSO communications, Free-Space Optical communications, LED communications, laser communications, LED, laser, light-emitting diode, lens, fresnel, fresnel lens, photodiode, photomultiplier, PMT, phototransistor, laser tube, laser diode, high power LED, luxeon, cree, phlatlight, lumileds, modulator, detector
This page and contents copyright 2007-2015 by Clint Turner, KA7OEI.  Updated 20150819
Page count since 201005: