After messing with several different photodiode detector circuits, I
decided to see if I could pass video through an optical-only
("lightbeam") video link.
The answer was: Yes - but this isnot the way to do it over
longer distances - I'll get to this later.
Comment:
Philips has phased out the Luxeon I, III, and V lines in favor
of the lower-power Luxeon Rebel devices. Since I have not
used those other devices, the techniques described here may not
directly apply. For the time being, however, the Luxeon
III devices are still available from various sources.
These newer devices have a "photon density" (related to the maximum allowed
current density of the device in amps/mm2) comparable
to the older ones - but smaller emitters - which means that for
a point-to-point link the brightness at the receive point will
be similar. With these smaller devices, the overall
current consumption will be reduced and with the lower device
capacitance, potentially easier to modulate at high frequencies.
Video-Speed
optical detector:
Figure 1:
Top: Schematic of video-speed optical receiver. Bottom: As-built prototype of this receiver. Click on either image for a larger version.
One of the most vexing things about using solid-state optical
detectors (such as a PIN photodiode) is that you can either make
such a detector very sensitive, or you can make it fast - but not
both at the same time. The main reason is capacitance:
It is the very nature of capacitance to resist a change in voltage -
and high-speed operation clearly implies very fast changes in
voltages over time!
For photodiodes, there are several approaches that one can take to
improve the speed:
Reduce the capacitance by using a smaller photodiode.
The smaller the photodiode, the less area there is, so it has
lower capacitance - but this also means that there is a smaller
area on which the detected light is to be focused and this can
be a complication when using simple optics as it may be awkward
to accurately focus the accumulated light on a piece of silicon
that is smaller than 1 square millimeter.
Reduce the capacitance by using reverse bias. Like any
diode, a silicon photodiode's intrinsic capacitance goes down as
a reverse bias is applied - for precisely the same reason as it
does for Varactor diodes: The width depletion region of
the diode (that area in the silicon where there is nothing to
allow electron flow) increases with higher reverse bias.
While the capacitance can be reduced several fold (compared to
the zero-bias voltage) by doing this, one can only apply so much
voltage before reaching the reverse voltage limits of the diode
itself.
Load down the photodiode to improve the speed speeding up the
R/C constant. Damping down the response with lower
resistance increases the frequency response, but it does so at
the expense of reducing sensitivity because the available output
of the photodiode is reduced. This doesn't take into
account noise contribution from real-world components.
The works of P.C.D. Hobbs go into some detail about extracting
higher frequency response from a relatively high-capacitance
photodiode. In his paper "Photodiode
Front Ends - The Real Story" he details a circuit that takes
several approaches to do this:
The use of a transimpedance amplifier for current-to-voltage
conversion.
Application of reverse bias across the photodiode to reduce
its junction capacitance.
The use of a Cascode input to minimize Miller capacitance -
essentially, capacitive loading effects by the diode's
amplifier.
The use of a bootstrap circuit to minimize the voltage swing
across the photodiode to reduce the "damping" effects of the
photodiode's capacitance.
An adaptation of the circuit is shown at the top of Figure 1 and the circuit works
approximately as follows:
Q1 is wired as a bootstrap circuit. Small changes in
voltage are countered by Q1's current amplification capability
and reduce the voltage swing across D1, the photodiode: If
the voltage isn't allowed to change as much, the capacitance
will have less of an effect on frequency response. This
circuit offers roughly an 8x improvement in frequency response.
Being that the base voltage will necessarily follow the
emitter voltage (which changes as a result of the current
through Q1) Q2, wired as a common-base amplifier and will
respond to this attempted voltage change.
U1 is wired as a transimpedance amplifier, detecting the
changes in Q2's collector current. As with Q1, a
change in current - not voltage- reduces the effects of
capacitance, particularly at higher frequencies.
The loading on the photodiode (D1) is based on Q1 and Q2's
bias current. Note that Q1, the bootstrap circuit,
operates at a much higher current than Q2, the common-base
circuit, but Q2 provides most of the loading on the photodiode,
ultimately determining the frequency response of the circuit.
Reverse bias on D1, the photodiode, is derived from the
voltage drop across the 10k resistor in parallel with C1 plus
the base-collector voltage drop of Q1, and this reverse bias
reduces the junction capacitance of the photodiode.
The upshot of all of this is that the output voltage of U1 will be
generally proportional to the amount of light falling on D1, the
photodiode. U2 and its associated circuity simply provides an
"artificial ground" to avoid the requirement of a bipolar
(dual-voltage) power supply.
Important
note concerning the circuit in Figure 1:
The circuit shown in Figure 1
was designed only to be fast and quick to build and no attempt was made to optimize
sensitivity and I have little doubt that it can be improved in terms
of both bandwidth and sensitivity!
Figure 2:
Top: Schematic of video-speed modulator Bottom: As-built (and operating) video
modulator. Click on either image for a larger version.
Unlike self-amplifying photodetectors (such as photomultipliers or
avalanche photodiodes) the ultimate sensitivity of PIN photodiodes
is limited, in large part, by the noise contribution of the
amplifier connected to the photodetector. In this circuit,
there are a number of noise-contributing components - namely the
transistors Q1 and Q2, but note that the resistors (especially the
4.7k resistor in Q1's emitter circuit) can also contribute
significant noise - so it is recommended that one use metal film
resistors in the Q1/Q2/U1 circuits instead of the more-typical
carbon-film resistors. The resistors associated with U2 are
not noise contributors in this circuit.
Finally, this circuit does not have sufficient gain to provide the 1
volt peak-to-peak video signal necessary to drive a typical video
monitor under most conditions and additional amplification is
usually required. While this could have been done using a
number of different circuits (another LM7171 or two, a UA733-based
amplifier, etc.) I used what was available: The "Channel 1
Vertical Output" from my old Tektronix 465B oscilloscope. I
simply connected the video monitor to the appropriate jack on the
back of my oscilloscope and used the 'scope itself as a
variable-gain video amplifier!
Comment:
Again, if extreme sensitivity is required, the addition of the
extra circuitry shown in the Hobbs circuit (e.g. bootstrapping,
additional resistors) will introduce noise into the system,
fundamentally limiting the absolute maximum
sensitivity of the system. For a more thorough
discussion of a circuit that has far better sensitivity - at the
expense of bandwidth, which makes it generally unsuitable for
video use - see the page Optical Receivers
for low-bandwidth through-the-air communications on this
web site.
Video-Speed
optical modulator:
Initial tests were done using an LED strongly driven by a function
generator and with this setup I could verify that the circuit in Figure 1 did, in fact, have
MHz-range frequency response, albeit with some rolloff. With
these encouraging results, I quickly (in 10 minutes or so) threw
together about the simplest modulator that I could, the result being
shown in Figure 2.
While this modulator works, it
does not work very well as there are several things wrong
with it:
It does not have very good
gain. It would be nice to have a bit stronger
drive signal than that of the 75 ohm video output from the
source. Rbias (the 150 ohm potentiometer) also robs much
of the drive signal.
It does not properly
terminate the video source. Although it wasn't
important in this quick test, this would have been more
important had a longer video cable been used.
This circuit has limited
modulation depth and rather poor linearity. This
circuit will operate with reasonable modulation depth over only
a narrow range of settings of Rbias. At the "optimum"
setting (where reasonable modulation depth, linearity, and
frequency response is obtained) the LED is driven well below its
rated capabilities.
The circuit is not thermally
stable. As the transistor heats up, its beta will
increase, requiring readjustment of Rbias for best results.
The frequency response is
terrible. As evidenced by monitoring the voltage
across the 0.5 ohm (carbon-type - not wirewound) resistor in the
emitter, it was clear that this circuit's frequency response
rolled off significantly by the time one got to the chroma
frequency (3.58 MHz for NTSC) with the amount of rolloff being
about 6dB. The fact that the LED itself has several
thousand pF of capacitance of its own doesn't help, either!
As I mentioned, I threw this circuit together in just a few minutes
just to provide a usable source of video-modulated light and while
there are any number of ways in which this circuit's performance
could have been increased, I won't mention them here. Note
that the driving method described is not capable of fully-driving
the LED at the highest video component frequencies, hence some of
the noted rolloff: A better system would involve a "stiffer"
driver as well as appropriate pre-emphasis designed into the
transmit and receive system to maximize bandwidth at the available
modulation depth - not to mention various methods to achieve
reasonable linearity.
Performance
of the entire system:
Despite its many faults I did succeed in sending video across my
basement (a distance of about 20 feet, or 6 meters) using an LED and
photodiode receiver with plenty of margin to spare. For
transmitting I placed the LED at the focus of a 250x310mm Fresnel
lens to concentrate the beam (but losing at least half of the
optical flux by not using a secondary lens) with the LED simply
sitting atop of CD-ROM Jewel case as shown in Figure 2(the Fresnel lens is
visible in the bottom picture of Figure
3) and using a camcorder playing a pre-taped video segment
as a video source.
The receive circuit (see the bottom picture in Figure 1 and the top picture in
Figure 3) was attached to
some scrap cardboard using double-sided tape and a simple lens
holder was fashioned (also from cardboard) to hold a large
Plano-Convex (PCX) lens such that the photodiode in the receiver was
approximately at its focus. Despite about 10dB of
chroma-frequency rolloff, the pictures received were fairly good, perhaps on
par with a VHS home videotape recorder. In the top picture of Figure 3 one may see, in the
background, the video monitor with the actual transmitted
signal: The white horizontal line on the monitor is from the
pre-flash of the digital camera that I used to photograph the scene.
Figure 3: The video gear
in action. Top: The video receiver is shown being bathed in
red light from the transmitter with the actual video being
received being visible on the monitor in the background. Bottom: The video transmitter. The video
source is from a video camera in the foreground while the
Fresnel lens (being illuminated by the LED) is in the
background, being held up by some lead-acid batteries. Click on either image for a larger version.
As previously implied it takes a bit of careful adjustment to get
this system to work: The setting of Rbias on the transmitter
is critical for best bandwidth and linearity, as is the gain setting
of the video amplifier (my oscilloscope in this case.) I also
noted that at this distance it was very easy to saturate the
photodiode, so the transmitter was purposely mis-aimed to reduce the
signal enough to allow proper operation of the receiver - but this
implies that there was a bit of signal margin in the receiver
itself.
Quick, back-of-the envelope calculations indicate that this system,
using for the receiver the same-sized Fresnel as used on the
transmitter (but optimized) and using a similar-sized Fresnel at the
receiver would be capable of spanning a distance of at least 1 mile
(1.6km) with usable results: Probably much more with more
circuit optimizing - assuming that going through that much air
wasn't going to cause its own problems!
Why
this method is not
good for longer-distance through-the-air video transmitting:
Having said all of this, I would NOT
recommend these circuits for video communications at a distance of
more than several hundred feet under any circumstances. If you
ignore the fact that they are finicky (requiring careful adjustments
of both the transmitter and receiver to work) and that their
frequency response is rather poor (something that could have been
remedied with a little more work) and that the LED could have been
driven to more than four times the output available from this
circuit (theoretically doubling its range) there is the matter of
the effect of atmospheric scintillation on any amplitude-modulated
signal.
Because the video signal was directly amplitude-modulated onto the
LED, critical signals such as horizontal and vertical
synchronization are present as amplitude variations. Also, in
order for a video signal to work properly with any typical monitor
it must be held to fairly close tolerances plus or minus 1dB or
(much!) better - for acceptable viewing: Commercial and even
normal consumer video levels are held to much tighter standards than
that!
Even with medium-sized lenses on both ends it doesn't take
propagation through much of an atmospheric path to experience
scintillation that greatly exceeds 1dB and it is likely that typical
atmospheric conditions along a near-to-ground path length of a mile
or so (1.6km) would probably be pushing the limit. Why?
With the fluctuating light levels, the recovered AM signal will also
be fluctuating wildly: These changing video levels would play
havoc with the TV synchronization signals - not to mention affecting
apparent brightness of different parts of the video picture.
While it is theoretically possible to implement various schemes to
normalize signal levels at the receiver (such as sync-keyed AGC
circuits, etc.) it would still require at least 25dB of signal-noise
ratio of the video signal in order to maintain anything closely
resembling a noise-free picture and it is likely that such circuits
would be problematic in the presence of higher scintillation levels,
anyway.
One way around this problem would simply be to modulate the LED with
a video signal carried on an FM subcarrier: A variation of
this method is used in consumer-grade videotape recorders for the
same reason that we would need to use it: The amplitude
variation of the signal recovered from a videotape would cause
unacceptable artifacts in the playback video. With FM, the
absolute amplitude of the signal is irrelevant - as long as it is
above the noise threshold of the detector. Another advantage
that FM has is that with a signal-noise ratio of better than 10-12dB
(depending on system parameters)
the recovered signal is nearly noise-free. The difficulty with
using FM is that it would require higher-frequency energy to be
modulated, and as we have already seen just getting video bandwidth
modulated - and then demodulated - is a challenge with such simple
circuitry!
One example of this is the system used by the German
Laser ATV experimenter group. This web page is in German: Click HERE for a Google translation to English. In this circuit video was modulated
onto a 20 MHz carrier using an NE564 PLL. Likewise,
demodulation is accomplished also using an NE564: Those
familiar with FM-ATV demodulator circuits (as well as those used in
the early days of analog satellite TV) will immediately recognize
these circuits.
Another example, already mentioned, is the method (often called the
"color under" system) used to record video onto tape as used in
consumer-grade videotape recorders. In these systems, the
luminance (black and white) portion of the signal is modulated onto
an FM carrier in the 2-8 MHz range (the frequency and amount of
deviation depending on the recording system) while the chroma
(color) portion is heterodyne downconverted to something in the
500-900 kHz range (the precise frequency also depending on the
recording system) and both signals are put onto the tape. On
playback, the chroma upconverted and re-united with the demodualated
luminance signal, mostly recreating the original video signal.
It is interesting to note that even though the chroma signal is
subject to significant level variations upon being read from the
videotape (although some techniques are used to stabilize this
level) the human eye is generally insensitive to such variations so comparatively little effort is needed to correct this effect.
It is likely that the "color under" system would eventually be
limited in its range by the amount of scintillatory amplitude
variations that the video monitor could tolerate, although this
effect could be somewhat mitigated by keying an AGC circuit to the
colorburst level of the received signal. The main advantage of
the "color under" system is that it would not require as high a
frequency response as, say, the 20 MHz German system in which the
color is sent along with the rest of the video: Using
photodiodes, reduced frequency response translates directly to
better achievable sensitivity. If photomultiplier tubes are
used, however, the frequency response limitation is not as much a
problem but this comes at the cost of the added complexity,
fragility, and expense of the use of the tube.
Other
related links:
RONJA (Reasonable Optical Near Joint Access) is a system
developed, at least in part, by Twibright Labs
in the Czech republic. This system uses high-brightness
LEDs and reasonably-sized optics that is rated to provide
reliable 10 megabit links at distances of up to 1.4 kilometers
(almost a mile.) Click here for
more technical information on RONJA. Because this is a
purely binary system (on/off) it is immune to the effects of
scintillation - provided that the minimum amplitude of the
scintillatory troughs is above the receiver's threshold.
If you have questions or comments concerning the contents of this
page, or are interested in this circuit, feel free to contact me
using the information at this URL.