IR detector system, biasing of photo diode

When you need an inductor, an inductor makes an excellent inductor.

But the dual microphone thing, electret or mems, sounds like a much better way to go.

It also shifts the phase of the output voltage quite rapidly as the modulation frequency moves across the resonant frequency, or as the temperature dependent resonant frequency drifts across the modulation frequency.

If you want serious filtering, look for multilayer interference filters. they aren't cheap but they can be quite narrowband.

Very nice, sort of like for lock-in detection, I have done that before, works great. Would take the tolerances out of the passive components.

Could also be done with a chirp and auto-correlation, although this does not use the phase, but a beacon reference instead

The docking station is typically in a shed, or close to a building where the GPS signal disappears.

It's moving on grass, and can have bumps etc, so for IR the lope needs to be perhaps 20degrees to avoid loss of signal

Could be done. The guys in charge are worried with change of environment, so what happens when it snows heavily, fog or other surroundings changing effects.

They also need the light to be incident perpendicular to the plane of the filter, otherwise they detune. John

I believe you are looking for a daylight filter circuit.

It's Burr-Brown(I'm old) OPT201.

Some folk are scared of inductors. 10kHz does mean quite a few milli henries.

Of course they might be able to reverse the process and have one receiver on the robot and two emitters on the docking station, a bit like aircraft VOR

A single combo microphone/loudspeaker on the robot would be interesting. Ranging is one obvious use.

Could that tell the direction of the mother ship?

Three or four omni mems microphones could tell the robot which direction to go with full 360 degree coverage.

Maybe even estimate range.

Fancier things, such as narrowish bandpasses and custom wavelengths, tend to be glass and quite a lot more expensive. I usually get those from Omega Optical, but there are European suppliers as well.

For this use, a regular 700-nmm plastic longpass will get rid of most of the daylight, which is what you want. You can also get photodiodes with the filter material included, e.g. the ever-popular BPW34F.

Regular old flat black paint. But a lens on the transmitter will be the biggest win.

It has very low capacitance for its area, it has a lens to increase the detection area, and (crucially for my uses, which are generally at higher frequency) it has very very low series resistance.

The series resistance of the diode contributes Johnson noise that can't be removed by bootstrapping. You don't care too much at 10 kHz, but at higher frequency the 50-300 ohms' worth of Rseries in most diodes will trash the SNR--there's not much use in a 300-pV/sqrt(Hz) TIA if the diode itself contributes way over a nanovolt.

Cheers

Phil Hobbs

Nah, the Johnson noise kills you. It's easier to just calculate or measure the photocurrent from direct sunlight and design around that. You only need enough bias to ensure linear operation at high current, maybe a volt or so.

You will want to put a filter in the second stage to get rid of the nasty high-frequency noise peak. I usually use a two-pole Sallen-Key with equal resistor values, which has predictable gain (1.00) and low component-value sensitivity, and is super simple.

Resist the temptation to do anything floral with the TIA stage, such as LC or *especially* gyrator filtering. A large inductor is a disaster in a TIA, because if it doesn't cause instability, it'll still pick up crap from every VF motor drive on the block, and deposit it right into the summing junction, where you really really don't want it.

You don't need a bootstrap at 10 kHz, but a sufficiently carefully designed DC restore loop can help sometimes. A badly designed one will trash the SNR. At low frequency, the only way to make a quiet current source is to put a large voltage across a large resistor. I usually do that in the emitter circuit of a BJT, to get higher Zout, but you can also do it barefoot.

The thing is, you're going to be dominated by the shot noise of the DC restore current unless its resistor is much larger than the feedback resistor of the TIA. (At high frequency you can use filtering tricks, but not easily at 10 kHz.) Since the DC restore is going to have the same supply headroom as the TIA, it really doesn't help.

If you pick the TIA's feedback resistance so that the IR signal produces

If not, you need to reduce the background with a better filter, or (better) reduce the field of view by using a lens on the receiver, and increase the signal by using one on the transmitter. You win by the square of the angular magnification, which can be a fairly startling number.

Either way, you need to control your detection bandwidth to something reasonable, and remember that LP filter to get rid of the noise peak!

Cheers

Phil Hobbs

Small shielded inductors are cheap, and 10 KHz is not a common switching frequency.

Put the two inductors close together. They will see mostly the same mag fields, so a couple of resistors added somewhere will cancel the pickup.

Or add a third, between them, to drive their bottom ends, again canceling mag field pickup.

Or make each L from a pair, arranged so the pickups cancel.

TV remotes work if you bounce the light off the ceiling in a well-lit room.

But the acoustic approach would be better. Omni MEMS microphones have built-in amps and cost 20 cents.

Yes, it is a form of lock-in detection, adapted to the lack of phase data, other than to know that it changes slowly.

But there is another issue, interference from artificial light, such as streetlamps, at dawn and dusk. Many streetlamps are strongly modulated at a harmonic of the local prime power frequency, 50 Hz or

A classic remedy is correlated double sampling (CDS), where the sample period is exactly one tenth of a second in duration, into which an integral number of power cycles will fit exactly, and so will integrate to zero.

Originally, CDS modulated the transmitter to alternate between on and off, so external light would fall in both odd and even numbered samples, while real data would only fall in one set of samples. Here the TX runs continuously, and it would be a nuisance to synchronize it, so we need a way to do the same in the RX units.

The simplest approach would be to receive at two modulation frequencies, 10 KHz and say 11 KHz, alternating between them. Only the 10 KHz will have data, while both frequencies will have system noise and power-frequency modulated stray light.

There are two identical integrators and a switch. The odd-numbered cycles go into the first integrator and the even-numbered cycles go to the second integrator. These integrator are leaky. One reports their difference as the 10 KHz modulation strength.

There are many schemes like this that could work.

Joe Gwinn

Great recommendations, thanks. I have purchased a bunch of the BPW34F.

So in fact the devil is in the detail. I have not done higher performance optics before, so riding on the learning curve.

The Keysight DMM had/has n powercycles filter, same idea.

Yeah, that is better. Then alternating switching between the two emitters at separate frequencies makes it easier to do detection on the reciever side. Separate envelope detection finds the midpoint when the two lopes are equal. Requires matching of the emitter diodes. Can be done in production or adding a receiver/monitor photodiode at the location of the 2 emitters

Sounds like I should hire you in a reviewer, if I get stuck :-)