short range (<cm) full duplex IR data exhange?

[max_torque]

I've been looking at using infrared data exchange to effectively "Hide" an interface for a remote device, as the TX and RX diodes can be hidden behind a visible light blocking filter.  There are lots of off-the-shelf half duplex IR transcievers, some with built in emitters etc, but they are all reasonably slow, and always seem to "mirror" any TX back onto the RX output, i guess to avoid reflected IR from causing issues, and that means they are restricted to half duplex coms.  I'd really like full duplex at up to say 500kbps or something like that. This looks like it will need two different modulation frequencies and RX detectors that are a narrow pass at those different frequencies to be able to discriminate the TX's apart.

How hard is this likely to be?  Modulating the TX diode is trivial, and i can't see why a basic driver couldn't do that at around 2 Mhz or more?
The RX path is harder, requiring a band pass filter (a tuned/resonant filter makes some sense here i think) into a precision rectifier and then a comparitor to chop the bits out of the modulated stream.

The total transmission distance ca be very short, i'm currently thinking about a magnetically attached "probe" that just snaps into the correct place over the hidden TX and RX diode "window"

Anyone done something similar?   

[jbb]

Well, there were the assorted IRDA standards.  They got nibbled at by Bluetooth for phone accessories and then eaten alive by WiFi.  But they did show that quite high data rates can be achieved.  In a nice example of "what was old is new again," people are now talking about LiFi (Light + WiFi) for ultra high speed LAN work.

I guess you could do some sort of on/off coding with an IR LED at 2 Mbit/s.  You would need some modulation to remove the DC and low frequency components of the data - maybe Manchester Encoding?  (Note: the receiver will experience DC drift and 50 / 60 / 100 / 120 Hz lighting flicker will absolutely be present somewhere.)

On the receive side, a photodiode => TransImpedance Amplifier (TIA) => band pass filter => comparator might work to recover the data stream.  You'd need something to decode it.

[max_torque]

The diodes should be fairly well shielded from most ambient light when coms are operational, as the "probe" from the master will cover the diodes on the slave unit. I considered using a hall effect switch to sense the magnets on the probe and only try to establish coms when that switch is shut, and hence the system will ignore any spurious inputs when the ldr is uncovered?

[pwlps]

How hard is this likely to be?  Modulating the TX diode is trivial, and i can't see why a basic driver couldn't do that at around 2 Mhz or more?
The RX path is harder, requiring a band pass filter (a tuned/resonant filter makes some sense here i think) into a precision rectifier and then a comparitor to chop the bits out of the modulated stream.

For RX the standard method is a PLL-based tone detection which is much more sensitive than any circuit based on band-pass filter (rejection of out-of-band signals is excellent with a PLL), and it isn't too difficult to make.  I tried it once in a somewhat similar project, using  the tone decoder chip LM567, it was quite simple to set up and worked well, however the chip is limited to 500kHz or so.   I don't know if equivalent chips exist for higher freqs.  I also played with another version using a 4046 PLL, then to make a tone decoder I needed to add a two D-flip-flop freq divider/quadrature generator and a second mixer (like in the LM567), this was a bit harder to set up but eventually worked too. There is also a new version of 4046, the 74HCT9046A, which goes to higher frequencies, but I didn't try it.

[jmelson]

I did an IR receiver using the QSE159 from Fairchild.  Part # QSE159 on Digi-Key.  I just used a bank of IR LEDs for the transmitter.
It worked very well, seems to be totally immune to visible light, and simple to use.  In my particular application, it was NOT a 40 KHz on-off modulated carrier, so I needed a sensor that did not depend on modulation.

Jon

[jbb]

If you use a CRC check (sorry, reduncant acronym) then random environmental noise should get thrown out. However you might have too many interrupts to deal with...

A separate presence sensor is more to go wrong, but could be useful for simply disabling the IR path. Especially if your device is battery powered (receiver might be quite hungry).

[soldar]

The total transmission distance ca be very short, i'm currently thinking about a magnetically attached "probe" that just snaps into the correct place over the hidden TX and RX diode "window"

If you are thinking of the two communicating units being pretty much touching then you could separate the two channels so you have one IR channel going one way and another IR channel going the other way.

[JVR]

Why not just use RF?

[max_torque]

RF is out because of the specific environment this device is used in (sorry, i can't be too specific....)

I could arrange for the RX and TX channels to have two completely independent and separate optical paths, but it would be easier for them to share the same path, as that will remove any alignment issues and make for a smaller probe.  I'm going to have a play around with that LM567, and see what data rate i can robustly get within it's maximum 500KHz demodulation frequency.

[eliocor]

to reduce crosstalk (full duplex) just use two different wavelenght transmitters (850nm and 940nm). Put on the receivers a comparable IR filter.

[langwadt]

why do you need real full duplex? 500kbps isn't particularly fast, just do 1mbps and toggle between rx and tx

[pwlps]

RF is out because of the specific environment this device is used in (sorry, i can't be too specific....)

I could arrange for the RX and TX channels to have two completely independent and separate optical paths, but it would be easier for them to share the same path, as that will remove any alignment issues and make for a smaller probe.  I'm going to have a play around with that LM567, and see what data rate i can robustly get within it's maximum 500KHz demodulation frequency.

If you need higher freqs you might also have a look at this app note from TI:
http://www.ti.com/lit/an/slaa618/slaa618.pdf

[mikeselectricstuff]

High speed IRDA works well but  all of the nice transceiver devices are obsolete now.4M and 16Mbit - Digikey only has stock left of this one
https://www.digikey.co.uk/product-detail/en/rohm-semiconductor/RPM973-H11E2A/RPM973-H11E2ACT-ND/3769159

Forget full-duplex - witha free-air link you will always get reflections -  if necessary ping-pong at a higher rate to get the same result

[JVR]

RF is out because of the specific environment this device is used in (sorry, i can't be too specific....)

I know the feeling. However, design is making the best compromises. Is having a 500Kbit link worth technician time and design/manufacturing issues? Or will a 100Kbit link do the same thing with off the shelf components, albeit a little slower.

Last resort would probably be inductive or capacitive comms, but both also have other issues you will need to solve for.

[mikeselectricstuff]

If you're basically going through a wall of a case then inductive may be a good choice, and doesn't need a window.

Unshielded drum style inductors placed on-axis a few mm apart couple very well - 500k might get tricky but 100k should be no problem.

[max_torque]

Forget full-duplex - witha free-air link you will always get reflections -  if necessary ping-pong at a higher rate to get the same result

This is why i as looking at using a different TX and RX frequency in order to fully discriminate against cross talk   

[langwadt]

Forget full-duplex - witha free-air link you will always get reflections -  if necessary ping-pong at a higher rate to get the same result

This is why i as looking at using a different TX and RX frequency in order to fully discriminate against cross talk   

for light that would have to be different colors of light

[ogden]

Forget full-duplex - witha free-air link you will always get reflections -  if necessary ping-pong at a higher rate to get the same result

This is why i as looking at using a different TX and RX frequency in order to fully discriminate against cross talk   

Very bad idea. You can discriminate crosstalk only by discriminating optical paths. Half duplex IRDA can reach TX/RX speeds you are looking for, so take it as it is. Consider that those who made IRDA knew what they are doing - they made it half-duplex for a very good reason.

[eliocor]

If you want to use the same optical path, and mantain a full duplex communication, your only solution is using different IR light wavelengths: 850nm and 940nm, followed by optical filters on the receivers. 
For your tests, start with known good one filters (expensive!): 
https://www.edmundoptics.com/f/traditional-coated-700-1650nm-bandpass-interference-filters/13917/ 
once done, you can test with cheaper (but not always available and of [maybe] shady quality) ones, like:
https://www.aliexpress.com/item/940-Nm-Infrared-Narrowband-Filter-NBP940-Half-width-40-Nm-Bandwidth-8-8-0-55mm-0/32982732595.html 
https://www.aliexpress.com/item/High-quality-IR-coated-film-filter-band-pass-850nm-laser-transmissivity-92-black-glass-filter-infrared/32864007904.html 
 
If the bandwidth of your filters is too narrow, you have to take care about emitter wavelength shifts do to temperature changes

[pwlps]

If you want to use the same optical path, and mantain a full duplex communication, your only solution is using different IR light wavelengths: 850nm and 940nm, followed by optical filters on the receivers. 

The "only solution" is a bit exaggerated and discouraging for the OP's project.  The FSK demodulation bandwidth is controlled by the loop filter time constant so that there is a of course a tradeoff between the frequency selectivity and the maximum bitrate, in other words sacrifying the bitrate you can have as many FSK channels as you want in a single IR wavelength channel. 

[langwadt]

If you want to use the same optical path, and mantain a full duplex communication, your only solution is using different IR light wavelengths: 850nm and 940nm, followed by optical filters on the receivers. 

The "only solution" is a bit exaggerated and discouraging for the OP's project.  The FSK demodulation bandwidth is controlled by the loop filter time constant so that there is a of course a tradeoff between the frequency selectivity and the maximum bitrate, in other words sacrifying the bitrate you can have as many FSK channels as you want in a single IR wavelength channel.

doesn't help if a device's receiver is swamped out by its own transmitter