With thanks to Stuart G8CYW for all his knowledge and experience on this subject
Separate Tx and Rx heads obviously complicate the building
and alignment of an optical transceiver, so to be able to
use the transmit LED as the receive detector very much
simplifies such a system. The advantage applies all the
more for an entry level transceiver and is what was
originally intended for the Finningley nanowave
transceiver. However, the best I could achieve in terms of
receiver sensitivity was a performance 20 dB down on my
standard SFH-213 based front-end, and as time ran out, the
opto-diode/LED pair arrangement was reluctantly reverted
to. Since Stuart and colleagues up in the North East
successfully use the LED-as-rx approach, all new LEDs that
appear here are still tested for sensitivity and
eventually one was found that was only 10 dB down compared
to the SFH-213 when tested in my standard light tight
receiver test box. That 10 dB reduces to 6 dB when you
take the die area into account and are using a lens ahead
of the front-end.
This current page details the work done on the specific
LED found. For general background reading, Google 'G8CYW
LED as receiver', inspired by this link:
http://een.iust.ac.ir/profs/Sadr/Papers/omd7.4.pdf
Also, the following link contains further interesting
information:
http://makezine.com/projects/make-36-boards/how-to-use-leds-to-detect-light/
There's plenty of other similar articles out there too.
|
The LED in question During an ebay purchase of more10mm LEDs
used in the Finningley optical transceiver, it was
noticed that the type supplied this time were not
as shown in jeledhk's advert. The difference can
be seen opposite. Although the die area on the
type sent (rh side opposite) is smaller and the
heat-sink blade on the main LED body is also
smaller, a sample device did survive being run at
full continuous power for 24 hours, the spec
summary shown below right opposite. The 'full power' rating referred to does need
some interpretation, as all Chinese devices seem
to. Although sold as a 1W device, the maximum
continuous current rating is quoted as 300mA,
which at 2.3 volts is of course only 690 mW. They
are 1W only at the quoted short term peak current
(450 mA). In any case, this is the data for the
original device, and although promised, no data
sheet for the replacement item has materialised to
date. Incidentally, these LEDs first came to my attention via G3XBM, so thank you Roger - the devices have proved reliable enough and the light output is certainly high, plus the inherent lensing reduces the beamwidth to 40 degrees which is about optimum to properly illuminate an A4 size Fresnel lens. |
   |
LED performance as a detector
As an initial test, the photo-diode on a Finningley receiver was removed and directly replaced with the red LED to be tested. Additionally, the 8v reverse bias supply for the photo-diode was swapped for a 0 - 100v variable supply to reverse bias the LED. Thus, the only loading on the live side of the LED remains the same FET gate that the photo-diode experienced. With a noisy (ie, low level) optical signal and 1 kHz amplitude modulation applied, the bias was increased to the point were further improvement in signal to noise ratio stopped, ie reverse breakdown was beginning to occur - in this case, about 64 volts (interestingly, the sensitivity does increase a little beyond this voltage, but the noise increases at a greater rate). The measurements were done in the light-tight test box described here.
A straight comparison of sensitivity between the original SFH-213 and the LED under test is shown below:
| Further measurements and a bit of
thought throw up one good and one bad
characteristic! The good point is that since the
light box is applying an even illumination across
the detector, and the equivalent die area of the LED
is no more than a quarter of that of the
photo-diode, the true difference in sensitivity once
the light is optimally focused onto either device,
should only be about 6 dB. The bad characteristic of the LED is its much greater self capacitance (about 55 pF). This shows up as a deterioration in sensitivity as the modulation frequency is increased at an earlier point than does the photo-diode, and sadly, it occurs immediately above 1 kHz. Capacitance was measured using a 20 Mohm bias feed resistor:  |
 Once reverse breakdown occurs, the graph becomes meaningless, of course, but it's useful just to confirm the onset of breakdown non the less. |
The S/N reduction with increasing frequency response is shown below, together with that of a receiver using an SFH-213 photo-diode front-end:
Outdoor testing
The acid test for the LED-as-receiver configuration is to try it out in the field, I suppose - so here's the result of my first tests down the garden.
A bias battery pack, with some voltage adjustment, was made up using PP3's from Poundland. For the half dozen or so LEDs tried so far, the onset of reverse breakdown has been fairly consistent, being about 64 +-1 volt, so voltage adjustment was made by using a 500k pot across only one of the PP3s.
The output voltage is variable between 60 and 70 volts, PP3s seemingly being 10v these days. Since I don't fit the tilt switch on my Finningley transceivers, I use pin 12 on the 15 way D connector to input the bias supply.
An A4 size Fresnel lens was used to focus the test source onto the detector, refocusing between the LED and photo-diode, depending on the option. The test source was located some 50m down the garden. To hand were two micro-power test sources, one modulated with a single 1 kHz tone and the other with a pair of near 15 kHz tones in warble mode. A lot of effort was spent ensuring that the focusing was optimum for both the LED and photo-diode receivers, resulting in the following:
The bottom two pictures (2 tones at 15 kHz) are so noisy as part of the tone switching that I thought it best repeat the 15 kHz tests with a single tone:
No matter how I adjust the focusing on the two front ends, the sensitivity difference remains within 1 dB of the figures shown in the graph. Though the relatively high LED self capacitance should result in poorer sensitivity at 15 kHz, the sensitivities remain stubbornly similar. I'll re-measure these on another evening, but I think the explanation is that the 1 kHz and 15 kHz test sources, using different red LEDs, were of slightly differing wavelengths and the 15 kHz one was closer to the peak receive wavelength of the detector LED.
| update on
relative sensitivities A few weeks later, the set-up was taken out to test over a 66 km long path between Manton in North Linconshire and High Bradfield, near Sheffield. I joined Richard G0RPH (shown opposite) at Manton, whilst Barry G8AGN and Gordon G0EWN operated from High Bradfield. With Barry using his phlatlight source into an A4 size fresnel lens, there was no problem in identifying his signal, it being a decimal place brighter than anything else on the horizon. His morse ident was audible before I had even got the transceiver tube into its mating section of the light box (right hand box in the picture). Turning his beam away from from us until his signal became noisy but still readable resulted in binoculars being required to see his signal by eye. The signal received by Richard's photo-diode receiver was showing a similar background noise level, confirming that the LED based receiver sensitivity is indeed on a par with that of a conventional SFH-213 photo-diode one. |
 |
One further thing I did do was to swap the Fresnel lens for a 4" circular convex lens. With this set-up, the signal was 10 dB lower which is slightly high given the difference in area of the two lenses.
The small die size result in a beam-width of +- 0.22 degrees when used with an A4 size fresnel lens (of focal length = 350mm).
Narrow optical bandwidth
| As implied above, the
receive optical bandwidth of the red LED is
quite narrow. This is most noticeable when using
my switched red/infra-red mini beacon. When
switched to IR, the beacon has to be placed
immediately in front of the receiver to hear
anything at all. I'll take a measurement at some
point, but it has be at least 60 dB down on the
red output LED. The graph opposite is taken from the link content at the top of this web page and shows how narrow the receive optical bandwidth is for the red LED sample they were talking about. It is also apparent that there is a difference in wavelength peak for the receive and transmit situations. At the moment, I have no test equipment that can produce the same graph for the HK LED. |
 |
Removing the need for a separate bias battery
Whilst it is fine in the short term using a few series connected PP3s to generate the 63v LED bias, long term it is nicer to generate it from the existing 12v supply, and if possible, to locate it on the transceiver pcb. The later requirement runs the risk of pick up of the bias oscillator onto the high impedance receiver input. For that reason, a much higher frequency oscillator frequency was sought and the following arrangement using a 1 MHz ceramic resonator was used. It also takes up minimal board space and the repeatability between resonators is not too bad, particularly if you have a bag full...
| Temporary control pcb: So that headphones can be used, the original agc chip and emitter follower were added to a revised plug-in control pcb. No agc has been added to the mic/tx path. Stopping the SL6270 from 'pumping' seems to be an impossible task, but perhaps it simply needs a more competent designer defining the circuitry. The version opposite is the best I can manage so far. It would be useful to have put the 1 kHz tone oscillator on this board too, but it just didn't happen. |
 |
Update as of July 2017:
The agc chip on the control board has been changed to a standard audio driver IC preceded by a volume control - more output, level controlled and no pumping!. Also, a tone generator consisting of a 4060 divider IC driven from the 1 MHz ceramic oscillator has been added, giving a reasonably stable 980Hz tone.
Final circuits:
 |
 |
 |
 |
| To be continued.... |