Electronic – Self-adjusting analog signal translator

Sometimes I have found that having a lot of logic level translators is more trouble than it is worth. Is your device 3.3V tolerant (I'm assuming you're going from 2.5V to 3.3V)? Assuming you are using the Arduino Due and taking a quick look at the Atmel SAM3X8E datasheet, it seems that if VDDIO is 3.3V, V_IH can be as low as .7*VDDIO, or 2.31V and still read a 'high'. Therefore, your 2.5V device should have no problem driving an Arduino input. The question is if your device can handle the 3.3V that the Arduino will put out.

Assuming you don't want to modify the Arduino to simply use 2.5V VDDIO (I bet it might be as simple as changing up a regulator or a resistor), you can actually just use a diode array (for example, a CMEDA-6i, which is a quad diode array that has a convenient forward drop of 1.0V) when going from the Arudino out to the device (you don't need to worry about the other direction). That will limit the voltage the device sees to 2.3V. You can do it even simpler with just resistors (or resistor packs), but you need to look a bit into how much current the inputs on your device actually draw and without knowing what that is, it would be hard to recommend a part.

I don't think that multiplexing raw pH inputs using analog switches is such a cool idea. They are high impedance, and the probes can be destroyed if the muxes latch up due to ESD etc. It won't cost a lot to have a buffer amplifier on each of the 4 inputs, and then hook up a multi-channel A/D converter to their outputs. No need for a mux, and multichannel A/Ds are a dime a dozen.

By multiplexing the input to an amplifier you're also requiring the amplifier to have much higher bandwidth than otherwise necessary, and it might be impossible to design such a circuit while still having acceptable noise.

In your case, you want to take 400 samples/s in total - you're sampling at 400Hz. Your mux would be switching a Ph probe each 2.5ms to the amplifier. Assuming an A/D with 100us acquisition/conversion, the amplifier has to settle to A/Ds resolution within 2.4ms. Let's say we want to settle to 1/2LSB on a 14 bit ADC. It requires at least 10 time constants of the circuit used:

$$n=-\log{1\over{2^{\mathrm{bits}}\cdot \mathrm{LSBs}}}=-\log{1\over{2^{14}\cdot 2}}=10.4$$

This means that the time constant of the amplifier needs to be:

$$\tau={2.4\mathrm{ms}\over10.4}=0.23\mathrm{ms}$$

The amplifier needs to be a 1st order low pass with the cut off frequency greater or equal to:

$$f={1\over\tau}=4.3\mathrm{kHz}$$

This bandwidth is 100x what's needed by the pH probe alone. Your multiplexed amplifier will have ~100x the noise amplitude compared with one that had a 16Hz low-pass response and were be applied to each pH probe directly.

And all this assumes that you are using an ADC that's much faster than the application calls for (10kHz sampling rate for 100us acquisition/conversion), although admittedly that's not a problem with a low resolution ADC.

In closing: Compared to placing the multiplexer after the preamplifiers, the multiplexed amplifier design requies an ADC that can sample 25x faster than 400Hz, and an amplifier that produces 100x the noise. If you can live with those drawbacks, you're OK.