# 12 bit vs 14/16bit ADC for ph sensor interface

I am making a ph probe interface board for a project. What resolution ADC would you recommend? I am definately going with at least 12 bit, but is it worth it to go with any higher resolution?

What are the implications of using higher res? Does the input need be more stable? Obviously the cost is higher, but am I gaining anything to begin with? Would noise be too high for the extra resolution to be useful?

Are there any ADC's/ic packages that can directly read the small signal voltage from the ph probe(differential) and perform any gain/buffering of the signal as well as perform the digital conversion? Otherwise, I need to design this with opamps. Any recommendations in terms of design for this?

• What's your desired pH resolution? Do you want to be able to resolve 0.01 pH? 0.001 pH? What about the range? 0 to 14 pH? -2 to 16 pH? (Yes, pH can be above 14 and below 0!) – In silico Oct 17 '11 at 2:22
• needed range is 6-9ish(pool water)..would like .001pH if possible, but .01 or .005 could be ok...of course ph sensor will output range from 0-14. – user623879 Oct 17 '11 at 3:14
• If you set up buffering circuits you can cover 6 to 9 in increments of 0.000732 with 12 bits (i.e., (9-6)/4096). But it's a little surprising to need pH to that resolution. I don't think I've ever seen it measured beyond 0.1 unless doing analytic chemistry. – JustJeff Oct 17 '11 at 3:21
• Well that is only the range I expect, however around ph 3-12 should be considered so that it can be calibrated with standard solutions...and I do not want to design the circuit such that a ph outside the range can damage ADC etc.. – user623879 Oct 17 '11 at 4:47

Like JustJeff says, if you design your circuit for a pH range of 6 to 9 you have a resolution of better than 0.001 in 12 bits. If you want the full pH 0 to pH 14 range your resolution will be

$\dfrac{14 pH - 0 pH}{2^{12}} = 0.003pH/LSB$

You can achieve 0.001pH by using a 14-bit ADC.

But it's important to draw a distinction between resolution and precision. Example: a digital fever thermometer which you can buy for less than 10 euro. This gives a reading with 2 decimals, like 36.87°C. But is that also the precision? In other words, is the "7" correct? No it isn't. It's even possible that the accuracy is worse than 0.1°C.
The same goes for a pH meter. You can have a reading with 3 decimal digits, but don't expect this to be the precision, think about the quality of other electronic parts.

edit
Also keep temperature compensation in mind. Temperature variations will cause the pH to change, and also change your probe's reading. Unless you have a highly accurate temperature reading it's no use going for better than 0.01pH precision IMO. Good temperature measurement isn't a sinecure in itself. This site says

"[The] temperature error is very close to 0.003 pH/°C/pH unit away from pH7."

So at pH 10 a temperature variation of 1°C will result in a reading change of 0.01pH.

• Yeah I think 12 bits is likely enough, and I am planning on temperature compensating for this in software. Thanks for your input – user623879 Oct 17 '11 at 8:00
• @user623879: You are also going to have a temperature probe? – endolith Oct 22 '11 at 1:05

As Steven already showed, 12 bits is plenty if the 0-14 pH range is covered linearly. If the signal from the pH sensor is not linear, then you have to find the resolution at the part of the range where it changes most slowly as a function of pH. Unless the output is highly non-linear, a 12 bit A/D is probably good enough and is certainly attractive because you can get that built into a microcontroller.

However, here is another idea. For slow low level signals like this, it can sometimes make sense to use a really high resolution delta-sigma A/D. These are available in 20 bits and more. They trade off speed for resolution and are not that expensive. A typical conversion time might be 15-20 mS, for example, but that should be plenty fast enough for something like a pH signal. The advantage is that this reduces and in some cases eliminates the need for amplification circuitry with its inherent gain, offset, and drift errors. For example, I've done thermocouple measurements with just a little passive filtering in front of a delta-sigma A/D. I understand pH probe signals are high impedance, so you will need at least some buffering. You haven't characterised the pH sensor output, so I can't tell if a high resolution delta-sigma A/D is appropriate in this case.

Note: standard pH calibration solutions are NIST to +/- 0.01 pH. few standard meters read better than 0.02 pH (might output the 0.01pH number but look in the manual) and if you're using a standard pH sensor probe (glass bulb) there isn't a man alive who will have much faith in anything you claim tighter than 0.02pH without a very massive proof detailing your calibration solutions for the entire system (not just 2 or 3 calibration points) as well as your gage R&R in the same report.

That being said, anything closer than 0.01pH is likely measurement error in the probe and wouldn't even be useful for analyzing drifts.

Precision is hard to obtain when you're getting very high resolution. Especially the last digits become non-saying. You could calibrate your pH sensor, but that costs time and you need reference measurements to do that (i.e. get liquids you know are exactly a pH value and make a reference measurement with an already calibrated device).

The only reason I could see a high bit ADC is remotely useful in if you want to see small very drifts in pH without losing your Dynamic Range. This is only useful if the noise floor of your system (from sensor to ADC input) is not exceeding that of your ADC. I have no idea what the noise level of your pH sensor is. Maybe you can refer to the datasheet (if it's mentioned at all).

You could use oversampling to reduce the noisiness of your signal, but then your output sampling rate goes down (or you need to get a faster converter/processing unit!). This only makes life more complicated.

Furthermore, temperature drifts may be a huge factor as stevenvh pointed out, not only from your sensor but also from the electronics. But as I said, if you know temperature is a constant factor in your measurements you can see relative changes more accurately.

About actual ADCs: There are differential ADCs with PGA. PGA stands for Programmble Gain Amplifier. You can amplify the signal by certain factors, like 1x, 8x, 16x or even more. It depends on your pH sensor whether you can easily amplify the region of 6-9pH.