Hi @donblair,

I came across this post while dealing with the same struggle -- ambient light rejection for turbidity measurements that are open to the environment. The youtube video and the technical note links got me started on a solution, which I'd thought I'd share since this post hasn't had a followup yet.

Luis Orozco of Analog Devices gives the basic setup for a modulated photo-detection setup in Figure 10 of the technical note "Optimizing Precision Photodiode Sensor Circuit Design": a light-source is pulse-width modulated, and a photodiode detecting that light source is connected in photovoltaic mode to a transimpedance amplifier (and a second-stage gain amplifier if needed) and then AC-coupled with a capacitor and filtered. In his very elegant solution the clock that modulates the light source also controls a chopper switch connected to a precision difference amplifier to capture just the positive peak voltage of the AC-coupled signal, which is finally smoothed with a low-pass filter before sampling with an analog-to-digital converter (ADC).

With the parts I had on hand, and a less sophisticated knowledge of electronics than Orozco, I figured that the chopper and difference amp could be replaced with a full-wave rectifier. The performance may not be quite the same but it's hard to beat four plain diodes for price and simplicity. Here's a breadboard photo and a schematic done with "jellybean" components:

So, after the transimpedance amplifier and a second gain stage (on one LM358 op amp), there's a high pass filter at roughly 1.5KHz, a full wave rectifier, and a low-pass filter at roughly 15Hz. The Arduino code at the bottom of the post uses the AVR Timer1 registers to set a (roughly) 8KHz pulse on pin 9, which drives one of two IR LEDs (the other, on pin 12, is directly driven). I connected an Arduino to the breadboarded setup as described in the schematic, and connected the output of the LM358 (MOD in the schematic) to Arduino pin A0 and the demodulated signal (DEMOD) to Arduino pin A1. A basic serial interpreter (below) allows us to run some quick experiments through the Arduino serial monitor window by entering the indicated commands in sequence, hitting the Send button after each command :

1. Reset the Arduino, then turn on just the directly driven LED (command "d") and taking some readings of the output signals ("r"). The signal at pin A1 should read 0 (or rather, it will read at the level of the op amp's offset voltage). This shows that the device suitably rejects unmodulated ambient light.

2. Turn off the directly driven LED ("d"), turn on the modulated LED with a duty cycle of roughly 10% ("p100"), and take some readings ("r"). This shows that the device does a pretty reasonable job of demodulating the photodiode signal. You can do this in an area with a decent amount of ambient light and experiment with covering and uncovering the test setup, and the demodulated signal (A1) shouldn't change.

3. However, we can see some limitations of the system if we then turn on the direct LED again ("d") -- the signal at A1 should drop to 0 (after a few seconds, given the discharge curve of the 10uF capacitor in the low-pass filter). Why does this happen? The direct LED is so bright that it's basically swamping out the signal of the modulated LED, and because there's effectively then no modulated signal leaving the op amp, nothing makes it through the band-pass and rectifier -- which highlights the difference between noise rejection and signal capture.

Electric ghosts seem to love haunting analog circuits on breadboards, so I routed the board layout below and uploaded it to OSHPark for fabrication. Will send a followup comment after assembly and testing.

I'd love to know if you got further on this project -- maybe with a solution a bit less hacky than mine :)

Cheers! Chris

[PLEASE NOTE: I couldn't get the code to format properly, so there will probably be some errors (e.g. "_BV" is being rendered as "BV" in the posted comment). Will try to fix this soon.]

int dc_led_pin 12; char message[8]; int counter, pwm_index, serial_val; bool dc_led_on = false;

void setup() { pinMode(dc_led_pin, OUTPUT); Serial.begin(9600);
DDRB |= (1<<PB1); TCCR1A = BV(COM1A1); TCCR1A |= BV(WGM10) | BV(WGM11); TCCR1B = BV(CS10); pwm_index = 0; OCR1A = pwm_index; }

void loop() { SerialCheck(); delay(100); }

void SerialCheck(){ while (Serial.available() > 0) { char new_byte = Serial.read(); message[counter++] = new_byte; if (new_byte == -1) continue; // if no characters are in the buffer read() returns -1 else if (new_byte == '\n') { if(message[0] == 'p'){ pwm_index = serial_val % 1024 ; OCR1A = pwm_index; Serial.print("setting PWM to: "); Serial.println(pwm_index); } else if(message[0] == 'd'){ dc_led_on = !dc_led_on; Serial.println("toggling dc led"); if(dc_led_on) digitalWrite(dc_led_pin, HIGH); else digitalWrite(dc_led_pin, LOW); } else if(message[0] == 'r'){ Serial.println("reading A0, A1:"); for(int i = 0; i < 500; i++){ Serial.print(analogRead(A0)); delay(1); Serial.print(","); delay(1); Serial.println(analogRead(A1)); delay(1); } } counter = 0; serial_val = 0; memset(message, 0, sizeof(message)); break; // exit the while(1), we're done receiving } else if (new_byte > 47 && new_byte < 58) { serial_val *= 10; // shift left 1 decimal place serial_val += (new_byte - 48); // convert ASCII to integer, add, and shift left 1 decimal place } } }

Hi @iamchriskelly -- sorry for the delayed reply ... and wow -- this is amazing work you're doing! Would love to know how your PCB version worked out.

In case it's still relevant: here is where we'd gotten to with ambient-light-rejection and the turbidity sensor. Potentially noteworthy items, there: - Your previous design is one of our primary references! :) Thanks for the inspiration. - We'd taken a very similar approach to the one you take -- based off the same guide by Orozco. (I'd tried explaining his application note to myself in this research note). - We only used a 'half-wave' rectifier circuit, followed by a low-pass filter -- in contrast with your 'full-wave' rectifier design above. (And I suspect your design might be preferable -- more on that, below.)

From what I can recall, my justification for using a half-wave rectifier (other than its relative simplicity, compared to the 'full-wave' circuit) was just that it was simpler to build on a protoboard. And it seemed to me that, in principle, it would be sufficient as a turbidity instrument: although the negative part of the oscillating signal is always 'zeroed out', the positive part of the oscillating signal still makes it through the low-pass filter; one could then just correlate the resultant amplitude output with turbidity levels in the solution under test.

But in practice -- from what I can quickly gather here -- it might be a better idea to use your full-wave rectifier design -- you're not throwing away half the signal from the system as in the half-rectifier approach, so I think you're going to more sensitive in the low-turbidity regime. And the full-wave rectifier is just a few diodes more -- no problem at all, especially if you're designing a dedicated PCB for the device.

However, we can see some limitations of the system if we then turn on the direct LED again ("d") -- the signal at A1 should drop to 0 (after a few seconds, given the discharge curve of the 10uF capacitor in the low-pass filter). Why does this happen? The direct LED is so bright that it's basically swamping out the signal of the modulated LED, and because there's effectively then no modulated signal leaving the op amp, nothing makes it through the band-pass and rectifier -- which highlights the difference between noise rejection and signal capture.

I also observed this 'swamping' effect -- for example, when attempting to demo the sensor in bright daylight. I guess one could just specify that the circuit needs to be deployed under appropriate conditions -- either in a container that minimizes stray light, or with a light-reducing filter on top of the sensor, or etc. And I suppose that one could try to catch this condition in the firmware, too -- checking whether the sensor has maxed out, and flagging that in that in the output, somehow.

Good luck -- would love to hear any updates / results on your design!

Cheers, D