Optical Communications at 481THz

Updated 17.7.12


My interest in communicating at optical frequencies (infra-red and visible) goes back to the mid 1960s when a school friend (Francis Wood) and I did some experiments using a small torch bulb and an OC71 transistor (with its paint scratched off). The torch bulb was modulated with a small audio amplifier and the OC71 used as a photo transistor at the receiving end. Despite no real optics we managed to communicate across the street one way and were very pleased.

Since then technology has moved on rapidly. Today there is a growing group of enthusiasts working around 300-500THz with quite advanced equipment and achieving results that are staggering. In the UK the DX record is now well over 100km. In other countries the distances covered have exceeded 250km using cloud bounce techniques and high powered lasers or LED arrays with excellent optics.

My experiments so far (most recent posts first)

On this page I intend to record my own experiments nearly 50 years later than those first tests. Like my experiments on VLF, LF and 500kHz this will grow as I build more gear, refine it and improve it. As a keen QRP man I expect my approach will be "keep it simple" if at all possible. So, it is unlikely my kit will be the biggest and the best. Rather, I expect to share my modest attempts with relatively simple equipment and optics.


First 2-way Optical QSO (10.5km)

This evening, thanks to Bernie G4HJW, I managed my first 2-way optical QSO over a path of around 10.5km.  For the AM baseband QSO, Bernie used his Finningley transceiver with 100mm optics and I used my simple AM TX  with my modified KA7OEI receiver with unbiased BPW34 connected directly to the FET gate.  For some reason, the RX in my dual tube transceiver (the one on the larger tripod) was about 3-4dB less sensitive than the separate RX, so in the end I used this to copy Bernie at a solid 58-9 with scintillation. There is lots more work to do to make a better system bit this first 2-way optical contact was very satisfying.


Bernie G4HJW was 20dB S/N in 0.67Hz bandwidth earlier this evening at my local test site 8.63km NLOS from his QTH "over the horizon". Bernie was using his Phlatlight beacon with 1.082kHz tone. The sky was clear and starlit, so this was clear air scattering not cloudbounce.  About 15 minutes after I first copied him it seemed to get more hazy and I was struggling to find his signal at all.  Best results appear to be with a clear sky and aiming at the horizon. 

On 3 attempts copying G4HJW's optical signal it has been (1) 30dB S/N, (2) no copy and (3) today 20dB S/N in 0.67Hz bandwidth over an 8.63km NLOS path. I am still using the BPW34 detector in 100mm optics.


Successful NLOS test over 8.63km!

Bernie G4HJW set his high power optical beacon running a 1.082kHz tone this evening pointing in the Burwell direction on a non line-of-sight path from his home QTH at Little Wilberham. He then came over to Burwell to compare his RX and my own from a test site on the edge of the village clear of the village lights. Distance was 8.63km.

Both on Bernie's RX kit and my own we got solid copy by ear of the beacon at a distance of around 9.8km NLOS using 100mm optics. Once found, the signal was solid.  I then set up the optics on the tripod and fed the signal into Spectran where the signal was 30dB over noise in a 0.17Hz bandwidth.

There was hardly a cloud in the sky, so this was largely by scatter from the mist/dust in the air. Bernie says signals have been stronger on other nights. Best reception was with the RX aimed just above the horizon, maybe 5-10 degrees at most above.

This was very gratifying as it is the furtherest NLOS signal I have copied and it verified my RX kit is now working at good sensitivity.

When I got back home I did try to copy the beacon by ear out of the bedroom window but the tree cover and lights made copy not possible. However I will try again later in the week with Spectran running and narrow bandwidth.

A recording of the signal is attached at the bottom of this page. This is best played back using Spectran with a bandwidth of around 0.17Hz.  A photo of Bernie's beacon is linked here.


This evening I tried a 7.6km non line-of-sight (NLOS) path between Burwell and Newmarket Heath with the 0.7W input red LED beacon running QRSS3 on an 820Hz subcarrier (at home) and modified KA7OEI receiver at the RX end, both with 100mm optics. My PC was running Spectran.

The best I can say is reception was very doubtful:  the screen capture above shows a signal at the correct subcarrier frequency in the right direction, but too weak to identify CW characters. The recent 3.6km NLOS test produced good 10dB S/N signals in 0.67Hz bandwidth - much easier.

With a largely clear sky, few clouds and a slight haze, it was not ideal conditions for cloudbounce testing. This test will be repeated when cloud cover is better and visibility is clearer. QRSS30 would give me another 10dB so this may be worth a go if QRSS3 doesn't work on a second try. There are several variables that make weak signal forward scatter detection difficult: (a) what vertical angle to use at both ends, especially with few clouds (b) amount and height of cloud cover, (c) horizontal direction of aim.  With QRSS3/30 modes it is a case of small aiming increments and wait to see if a signal appears on the screen. It is quite hard work.

The result is disappointing, but there will be plenty of other occasions to repeat this test and do others. All good  fun.


Success with non line-of-sight "cloudbounce" test (3.6km) today

Armed with my simpler QRSS3/CW beacon (see schematic on the left) I did a very successful non line-of-sight (NLOS) cloudbounce test this evening using my 1W red LED in 100mm optics (run at 340mA). TX was my "G3XBM" message in QRSS3 (3 second dots CW) at 820Hz subcarrier.

With the beacon aiming out through the double glazed shack window at nearby Burwell windmill (as an aiming point) I set off for a road at Landwade which was 3.6km away "over the hill" and on a NLOS path from here. At Landwade I set up the 100mm optics and my variation of the KA7OEI head feeding into my laptop running Spectran. Immediately I got a good signal from the beacon 3.6km away. Signal was around 10dB S/N in 0.67Hz bandwidth. The signal was neither visible as a red glow nor audible in the earpiece despite listening quite hard and panning around for best signal.

This was my first proper NLOS test and it is extremely encouraging. I did try to elevate the RX to higher points in the sky but best reception was with the optics aiming at the lights of Burwell village in the distance i.e. as low as was possible in elevation. At the TX end I was aiming to just clear the slight rise in ground to the east of me near Burwell windmill.

Weather conditions were light patchy low cloud with pretty decent visibility. I did notice QSB as cloud cover varied.

I'm really lucky finding this test path as I can put the TX beacon on the bedroom shack windowsill and fire towards the windmill. In daytime I would be able to align the RX better as I was having to guess the best direction with only Burwell church visible. I had to tweek the alignment to what I thought was the right direction. I did not spend a lot of time trying to peak the signal and better copy is possible.  In all honestly I did not expect this test to be successful.


This evening I did some further tests with treebounce and what I'd loosely call skyscatter. The sky was cloudless, but a little hazy.

Kit: TX (250mA LED, 100mm optics in one room of house), RX in another room KA7OEI based head (doors closed, no light leakage) with 100mm
optics. PC running Spectran positioned to minimise light pollution to RX head.

First test was 10wpm CW off the tree (0.3km path length total) with strong signals received 35dB S/N in 5.4Hz bandwidth. Signal v.clear in the earpiece too. I could just make out the red light glow in the distant tree branches. Aiming critical.

Second test: TX and RX elevated to aim at roughly same patch of clear sky. QRSS60 signal sent from TX. Signals detectable in Spectran in 0.17Hz bandwidth, weakly, but definitely there. It is less clear on the capture than on the real screen. For this second test I made no great attempt to optimise the RX aim, just aimed at what I thought was roughly the same patch of sky. Now I can't be sure whether the signal is purely from scattering off mist/dust particles or what, but I think it is unlikely signals are coming off other objects as I am aiming quite high into the sky (about 45 degrees up) clearing nearby stuff.

Screen shots of both the 10wpm treebounce (above) and QRSS60 skyscatter signals (left) on Spectran are attached.

These tests suggest that with very slow QRSS I may be able to get a NLOS signal to G6ALB in the next village 3km away. I wasn't expecting the QRSS skyscatter test to work at all, so was surprised. When the weather improves I'll go out /P with the PC and RX and see if I can detect the "forward" scattered signal at much greater range (1-3km). With proper cloudbounce it should be better I would think.


Today I did my first test beyond the end of my street and in daylight.

My 1.082kHz beacon and 100mm optics were set up pointing out through my double glazed bedroom window and aimed it at a local feature called the Devil's Dyke which is 1.6km (1mile) exactly from home. This is the furtherest line-of-sight (LOS) path from home. Then walked along the dyke to roughly the point where I was aiming and started looking with my handheld 100mm optics receiver. Much to my joy and surprise, I heard the beacon before I spotted it by eye. 

The beacon could be  copied over a stretch of about 50m along the dyke. S/N I'd guess at around 20dB (by ear) in speech bandwidth in daylight. Next time I'll take the laptop and measure S/N with Spectran. It's a bit of a trog up there at night in the dark though as it is some way from the roadside carpark.

I simply had a cascode KA7OEI optical head feeding a feedback biased common emitter stage into a crystal earpiece, although the recovered audio was a bit
low in the wind. I am using a BPW34 detector with some reverse bias but the diode anode is connected directly to the FET gate. Down the street that seemed to give best results, although I have not done extensive tests.

Now 1.6km may not sound like much, but this is my best distance so far.  So, some progress in the right direction.


Last night I made my first recordings of my optical beacon in action. In one room of the house the 100mm optics beacon was set running QRSS3 and aiming at a tree 150m away across a field. This was a deciduous tree with just bare branches. In another room of the house, optically well isolated from the TX, I set up the 100mm optics KE7OEI head feeding an earpiece and my PC running Spectran. The signal scattering off the tree (total path length 0.3km) was 30dB over noise in a 2.7Hz bandwidth. This suggests (I think) that other non line-of-sight paths with a total distance of up to 1km might be possible by reflection off trees, buildings etc. I have still to try a true line-of-sight path at any distance with my current system but it will certainly be good for many kilometres.

Yesterday my 1W ultra-bright LEDs arrived from Hong Kong. When fitted to the beacon these will be VERY bright and will extend the range possible on 481THz yet further. All good fun! 


Today I completed my first keyed optical beacon transmitter (with 100mm optics) which I'll be using for tests at greater range than hitherto. The 481THz (red light) beacon sends my callsign in either 10wpm CW, QRSS3, QRSS30 or QRSS60 as well as a continuous carrier or a 30sec on/off sequence. The TX uses a sub-carrier of 1.082kHz or 8.659kHz, the frequency being derived from an HF crystal divided down in a 4060. The message comes from a K1EL K-ID2 programmed PIC. Initial tests this afternoon at dusk allowed me to copy the 1.082kHz 10wpm CW sub-carrier signal by reflection off a wall across the street at decent strength.

I now need to find some longer (local still) optical paths to test the beacon TX with my 100mm KA7OEI receiver. It would also be fun to try some non line-of-sight tests locally , possibly including cloud-bounce. I think this may be possible with QRSS30 or QRSS60.


Last evening I tried some optical tree-bounce: aiming my optical beacon out of the house window at a tree 150m away I then walked down the street to see if I could detect the signal bouncing back from the tree. Considering the 300m path length and that the light was simply that reflected/scattered off the bare tree branches the signal was actually quite good with clear copy of the 1kHz tone.


I've now bought some NE5534 low noise op-amps so can implement the full G8CYW optical head circuit which many users have successfully adopted. Stuart describes this as "exquisitely sensitive". I have BPW34 PIN diodes but think the sensitivity should be similar to that with the SFH2030 diodes used in Stuart's design.

In the first instance I need to remeasure the noise floor with the optical head in total darkness to see if the changes indeed make it better than the current discrete component circuit. My first quantitative noise measurements suggest far higher noise floor than I expected.

Assuming this gives improved performance over my 0.4km local "up the road" test range, my next step is to attempt a longer path, so I am looking for a path of around 2-3km to test with my baseband beacon. If this works with decent S/N then I'll retry looking for the GB3CAM optical beacon. At 32km this is a good test of system performance. At the test site at Nine Mile Hill the traffic noise was high (optically and audibly) so I also need to increase the signal level in the earpiece used to align the RX.

There is also soon to be a second optical beacon near Cambridge and this should be a little closer and an easier signal to find. However, I need to do tests before this is installed at Dry Drayton.

Incidentally, I managed to overcome the parallax issue in the optics alignment so now a distant street light is spot on in the cross-hairs of the spotting scope.


A failure

Flushed with my local successes (0.4km) last night, I decided to have a go at looking for the GB3CAM optical beacon 32km away from a local high spot where I know others have copied it sending its 1/15kHz FSK sub-carrier data at 481THz (red light). I set up in the daylight so I knew where to aim approximately and had the PC ready with SAQ receiver (0-22kHz SDR) and Spectran all ready. After dark I started to seriously hunt for the elusive signal, but without success.

Failure can be put down to several possible causes:
  1. Difficulty in aiming. Although I knew where to aim, I am uncertain that my spotting scope is correctly aligned on the 100mm optical head and that there is no parallax error. Looking for a very weak signal is therefore hard to start with.
  2. Inadequate sensitivity in the optical head. Although I could hear the 50Hz buzz from every street light and house light locally, maybe there is another 6-12dB to be had in basic sensitivity in the optical head.
  3. Inadequate volume on the recovered baseband audio. I was monitoring the "live" feed with just a crystal earpiece attached to the head and, although this is fine at home and down a quiet street, it was too low a level to hear weak signals over the traffic noise from the main road which adjoined the field gateway at the RX location.
  4. Path loss. According to G4HJW, the beacon keeper, the signal varies from 5-30dB over noise at this location, assuming one is aligned correctly. Although the weather was cloudy and good visibility there was drizzle that came in as I was testing.
  5. QRM from car headlights. As cars came along the main road near the test site, I got quite bad buzz from their headlights.
  6. The beacon keeper had turned off the beacon! It happens.
So, not to be beaten, I will do things to address all these problems and try again quite soon. I think I need a better local test beacon that I can use to (a) check aim, check RX sensitivity and adequacy of the recovered audio.


Today, in daylight again, I did a repeat of my recent tests with the baseband pre-focused red LED TX (20mA), looking for the signal
0.2km along the road. This time I also took my PC and looked for the signal with Spectran and the SAQ SDR receiver. The audio tone was around 40-50dB over noise on both software receiving systems.  At this range I would have expected far better S/N. The noise floor looks high. When I blocked off the lens the noise floor went UP about 15dB. Why I wonder? GM4SLV has suggested it was the heat (IR) from my hand and he may be right.

The picture shows the test set-up. The dashes in the Spectran trace are a result of me interrupting the beam with my hand.

If you click on the image you can see the pre-focused red LED pointing out of the house window and the receiving kit in the bottom foreground.

The image on the right shows the signal as received using Spectran. The signal is well above the noise floor, but the noise floor is well above where it should be! My field test to go searching from GB3CAM will have to wait until the system is improved.

Even with just the stand-alone 20mA TX diode (no additional lenses) and this RX system, the range for voice comms should be be over 10km I think based on 6dB additional loss for doubling distance.

UPDATE AT 1730: For the first time I tried the system in the dark at night. I managed to get around 0.4km from the LED beacon at home (the limit I could find walking along the street and into a field) where the audible signal in the crystal earpiece was STRONG. This is very promising. Interference from sodium street lights was an issue although the RX optics have sufficiently narrow beam width to avoid these.


I now have a 100mm lens set-up in some grey drainpipe with 4x sighting scope properly attached on a tripod together with a "simplified version 3" KA7OEI head (of sorts) with a BPW34 detector. In my version the cascode stage is just followed by a single common emitter stage driving a crystal earpiece. there is also an emitter follower so I can connect to my PC (to use with Spectran/SDR) or a transverter.

A 5mm 10 degree red LED using just its own built-in lens was shone out of the stairway window aiming down the street in daylight with 1kHz modulation. It was taking around 20mA. I then walked with the kit to the far end of the street and the signal was solid (S9) in the RX earpiece at a distance of about 0.25km.

The system has been locally tested using SM6LKM's SAQ receiver on the PC (as a 1kHz direct conversion receiver) and using Spectran. All looks very promising in terms of S/N and sensitivity.  Next week I hope to go looking for GB3CAM from a site 32km east of the beacon at a location called Nine Mile Hill. See Bernie Wright's page about the optical beacon at http://www.earf.co.uk/light_beacon.htm.


In the last couple of weeks I have managed to test various head systems using BPW34 PIN diodes and high brightness LEDs as emitters and detectors. Results suggest that a Toshiba extreme brightness (10000mcd) red LED can be used both on TX and RX with good sensitivity allowing a "single optics" transceiver to be built. Range with 100mm lenses should be around 5km at least. On a recent test, with a less bright LED, 100mm lens at the TX end (with the TX firing through a double glazed window) the signal could be easily detected at the far end of the street (200-250m) in daylight. 

The photo shows the "Heath Robinson" TX optics positioned on the stairs of my home! Note the 4x gun sight which was obtained via eBay. I think this will be an essential part of the alignment procedure, as will good stable tripods at each end.

Next stage of the experiment is to complete one complete transceiver unit. This will probably be a RX/TX head together with a transverter built into the same unit. This will allow FM, CW and SSB operation on subcarrier frequencies up to around 30kHz. Hopefully this will be compatible with other local 481THz operators. It should also allow me to look for the GB3CAM beacon on 481THz at RAF Wyton in Cambridgeshire.


Today I carried out my first optical communications tests since 1966. I built a small "baseband" (i.e. not on a sub-carrier based) optical transmitter producing a tone at around 800Hz feeding a standard low cost high brightness red LED at around 10mA current. The LED has a small built-in lens which produces a beam of around 20 degrees.

For the receiver I built the first stage of the optical head described in the RadCom articles (March-May 2011) using an identical  red LED (reverse biased) as the detector. I also tried a BPW34 detector, but first impressions were that it was not that different. This was followed by a couple common emitter transistor stages using my ubiquitous 2N3904s feeding a crystal earpiece. With the TX "beacon" running from a 9V battery and aiming out of my garage I walked across the road with the receiver and a 4 inch magnifying glass. Across the road, at about 25m range, this produced a quite respectable signal as long as the magnifying glass was focussed onto the RX LED. This was all a bit "Heath-Robinson" as I had to hold everything in my hand and move things around to get it spot on. There was quite a bit of interference from the street lights nearby.

I have no idea how much "antenna" gain there is in the built-in lens on the TX LED (a few dB?) or with the magnifying glass on the RX but with this set-up correctly aligned I would think 100m range is certainly possible.  This is just the beginning of a series of tests, but I am already happy that the optical head is sensitive and that good, well aligned optics will be essential to get decent distances. More TX power is easily available by using a power LED.

Next stages are:

  • A better beacon TX capable of operating at higher power on both baseband and subcarrier frequencies.
  • Putting the optical head into a screened enclosure, even if a temporary one.
  • Starting to think about optics. Using the same LED on both RX and TX will save on optics as just one set is needed at each end of the link.


UKNanowaves Yahoo Group

Modulated Light

Optical (through the air) Communications KA7OEI

GB3CAM Light Beacon

Roger Lapthorn,
Feb 26, 2012, 2:06 PM