Choice of turbidity sensor options

To help illuminate variation in light underwater I'd like to attach a turbidity sensor to the light logger I'm going to be redeploying when allowed.

The sensor I'd planned on using (here) uses the onboard ADC to get voltages, which a formula converts to NTU. The issue is that this gives a very low precision (other people judge it to be 7.25NTU per ADC count. For reference values I've taken in clean-turbid bay water range up to 22 NTU). Playing with it at home that seems to concur with what I found.

A couple of studies (here and here) have used light to frequency sensors like the TSL235 to make a turbidity sensor with <1 NTU precision. I've got a few TSL237sensors already. The basic code seems simple. The engineering to make the enclosure is harder, particularly under the current circumstances where I can't access a 3D printer or cuvettes.

At this point I'm not looking for something where I know the NTU value. Until complete lockdown ends I won't have access to proper turbidity metres or calibration standards. I'm looking for a sensor that has a relatively high precision so that I know when the turbidity is going up or down.

So, the question is whether I can do something with the premade sensor to increase the precision of the ADC (would one of these gain amplifiers help?) Or should I continue on with making my own?

Thanks for any thoughts.

Is your goal to create a sensor that has the same performance as commercial turbidity sensors but at a fraction of the price?

As much as that would be nice I don't think it's realistic, I'd just like something that would reliably tell me whether the water is turbid enough to affect the light coming to the light sensor, and an index of how turbid it is.

If I could get <1 NTU resolution that would be good enough. Even the portable turbidity meters have a resolution of 0.02 NTU.

Why not buy a commercial sensor?

Cost basically. It's hard to get a hard price on them, but if Alibaba is any guide you're looking at $400+ for a commercial sensor. The portable turbidity meter I mention above is 1500AUD.

It looks like you have this sensor

which is a simple light source and phototransistor receiver.

The mechanical one piece design prevents you from increasing the sensitivity by simply increasing the path length. You could however disconnect the light source and shine a laser pointer through the side of something transparent like a fish tank with the sensor on the far side so you have an effective path length of many centimeters.

A photo transistor sensor isn't appropriate. Transistor gain is not well characterized.

To measure light photo-diodes are used. A particular photodiode's response is much more
stable as it only depends on the photons arriving, not other parameters like gain.

I would suggest a good photodiode in a reverse-bias configuration with a transimpedance
amp after it is needed to make any kind of stable measurement. Reverse bias puts
a photodiode into the most linear / well behaved configuration as every photo-generated
electron-hole pair gets counted as the electric field sweeps them apart to the terminals.

As you are looking to look for a small dip in light w.r.t. to clear water conditions the stability
of the emitter and sensor are both important for solid readings. The emitter should be
driven from a constant current source circuit.

mikb55:
It looks like you have this sensor
DFR Turbidity Sensor Review & Thoughts - Codrey Electronics
which is a simple light source and phototransistor receiver.

The mechanical one piece design prevents you from increasing the sensitivity by simply increasing the path length. You could however disconnect the light source and shine a laser pointer through the side of something transparent like a fish tank with the sensor on the far side so you have an effective path length of many centimeters.

Thanks. I can't really increase the path length in the current set up, it needs to be quite small.

MarkT:
As you are looking to look for a small dip in light w.r.t. to clear water conditions the stability
of the emitter and sensor are both important for solid readings. The emitter should be
driven from a constant current source circuit.

Thanks, if I were to build my own I'd be using a light-to-frequency sensor like the TLS237, which seems to have a good response?

Re: the constant current source circuit, I don't know how to do that, but thanks for the recommendation, I'll learn.

In the interim I've got an ADS1115 16-Bit ADC which increases the response from the ADC to 16bit (15 in my case). Regarding what you were saying about photo transistors, am I trying to make a silk purse out of a sows ear by using this to get a higher resolution response from the sensor?

The TLS237 seems very reasonable - the datasheet explains it uses a photo diode and then current
to frequency converter, they know what they're doing!

Constant current for an LED is available as a chip, there are various constant-current LED drivers
around.

Without constant current you may find temperature changes have noticably effect on output,
especially if the supply voltage to the current limiting resistor isn't much greater than the LED
forward voltage. LED forward voltage varies with temperature, and this may affect the current
more than you'd want.

Calibrating the reading in clean water would need to be done at the same temperature as the
test sample to eliminate variation - having a constant current driver would bypass this precaution.

Thanks for the tips. I had a try with what I've got here, and I can't do the mechanical precision required, but it's something I'll definitely try when I get access to a 3D printer, NTU standards and ultrapure water.

I did some tests with the 16-Bit ADC added to the original sensor. Response is reasonable, hard to say without standards (adding 0.05ml of milk to a 1L container of water is about as good as I can get). It is a lot more responsive than it was with the addition.

Thanks for the help.