pH Amp circuit

[TeraHz]

Hi all. I'd like to get your opinions on the attached circuit.
The idea is to read the voltage signal from a pH probe which generates from about -450mV to +450mV (~50mV per pH unit, 7 pH being 0V). I will be using a MCP3424 ADC which has "input range of  ± 2.048V differentially". So all I'm trying to do is amplify the  ±450mV to ±2V (about 4.5 gain). I have 5V available so I'm using a MAX861 charge pump to generate -5V for the opamp. The datasheet says "even though the MAX860/MAX861’s output is not actively regulated, it is fairly insensitive to load-current changes" so I'm hoping to be able to skip the negative voltage regulator and still get a stable output from the opamp. Is that possible/wise?

Anyway, thanks in advance.

[Mechatrommer]

your opamp seems fine, but how actually do you wire the power supply? (i dont have max861 datasheet)

[TeraHz]

Thanks. I don't have the full circuit here but the PSU +5V signal is coming form a 5V regulator and GND is common with the regulator.

MAX861: http://datasheets.maximintegrated.com/en/ds/MAX860-MAX861.pdf

[SeanB]

Normally most pH probes use a CA3140 as the opamp, and use a 10-100n film capacitor across the 390k feedback resistor to reduce hum pickup. The probes have a response time in the tens of seconds range, and need to feed into a very high impedance input. The capacitors on the input need to be very low leakage types, and preferably they should be wired clear of the board along with pin3 of the opamp to reduce leakage current. Otherwise you will have to guard band them with a ring connected to pin 2.

[TeraHz]

I'm planning to use a TL082 which also has a very high input impedance (something like TOhm). Will add the cap as well.

I have no idea what the guard band thing is  . I was thinking of using regular ceramic caps (1210 or one smaller if I can't find the right value). No idea how are they in terms of leakage.

[SeanB]

Anything but ceramic........... TLO82 has way too much drift and noise at DC for this use, the CA3140 is the default unit, preferably in the TO100 can as well. Guard band is a thin trace that encloses all the PCB trace that is connected to the input, it includes the resistor and the input side of the capacitors, and is there to reduce PC board leakage current to the input. Connected to pin 2 as it will be the same potential. This is one board where you will have to clean it after assembly very well to get all flux residues off. The probes normally are modelled as a voltage source in series with a very high value resistor, typically over 10Mohm. They are able to source only nanoamp currents to drive the input.

[CarlG]

With a single-ended output, you have no use for the ADCs differential input capability. Note that min input voltage for MCP342x is Vss - 0.4V. So you have to rethink this a bit.

Switched-cap inverters are generally noisy, so the least you should do is add ferrite beads on the input and output of the MAX861 (and add more C to the OPA since the MAX861 decaps gets isolated from the OPA supplies). However, that will only take care of the high frequency noise. You may have to filter out the fundamental f_S and harmonics too (with LC-filters). Some inverters may also have pulse-skip modes that results in low frequency (<<f_S) noise.

It might also be wise to add some (better) RFI filtering and ESD protection. A start may be a series ferrite bead and 2N4391's to the supplies (diode coupled).

[TeraHz]

So it is never that easy

SeanB, thanks for the explanation. I'll not use the TO082. While I was looking at the CA3140, I couldn't find the TO99 package anywhere so it will have to be the SOIC-8. Another option that I saw is the LMC6081 which has 10fA input bias current.  Would it be better to get rid of the input resistor/caps altogether and wire the BNC input pin directly to the OPAMP?

CarlG, bummer, I thought the input on the channel can handle negative voltage.  Noted on filters. Will it be better to just put a 3.3 and -3.3 linear regulators to cleanup the signal for the opamp?

So it looks like after cleaning up the voltages I'll have to add another amp to shift the signal into 0-2V range so that my ADC can read it...

Given the 'it is never that easy' statement above, I doubt this will work, but can I do something like this instead to bias the input and get away with the negative rail/charge pump part altogether:

[SeanB]

RC on input will reduce input noise from RF sources. You cannot apply bias to the sensor, it will be destroyed in short order. Remember as well that the reference electrode ( the lead that is connected to ground) also has a voltage developed on it, so ideally you should not have any connection to ground in your circuit other than the sensor itself. Most of the circuits using pH sensors are isolated from the rest of the unit, either after the ADC or by being battery powered, for just this reason. As well you need, for best accuracy, to read the temperature of the solution by the sensor, and use a lookup table for the solution you are using ( or the major ionic components in it)  to compensate for the effects of temperature.

Sadly concentration has an effect as well on pH, and needs correction as well, along with choosing an appropriate electrolyte medium inside the glass vessel and the compensation electrode as well. then you can go on about different electrodes like Antimony, used for process controols, and get really complicated.

[TeraHz]

So basically I'm going to need a second opamp to add the bias to the signal to bring it over 0V for the ADC to read it.

I'm going to use this circuit (three of them, actually) on a raspberry pi to monitor the pH and ORP (different input range) in my aquarium. I already have a working circuit that involves a voltage inverter, two voltage regulators (one for 5V, another for -5V), and two opamps (for shift and amplify to 0-5V) and it is good enough. But I'm trying to reduce the component count to make it simpler to build. I don't need super accuracy, not even full 0-14pH range.  Just up to the second decimal (x.xx) and will be mostly in the range of 4-9pH, so the entire range of the pH is not that important. The common ground hasn't been a problem so far. Temperature compensation and probe calibration will be done in software as well.

Ok, I'm going to combine the ideas and add a second opamp and come back with a new circuit.

Thanks so far!

[charlespax]

The idea is to read the voltage signal from a pH probe which generates from about -450mV to +450mV (~50mV per pH unit, 7 pH being 0V). I will be using a MCP3424 ADC which has "input range of  ± 2.048V differentially". So all I'm trying to do is amplify the  ±450mV to ±2V (about 4.5 gain).

Have you considered connecting the probe directly to the MCP3424 and using the internal 4x Programmable Gain Amplifier (PGA)? That would simplify your circuit a great deal. I'm using the MCP3424 with PGA for my data logging thermometer and it works great. I haven't worked with pH sensors yet thought, so I know if a direct connection would work the way I think it would.

[TheAmmoniacal]

I tried building a pH meter not too long ago, seemed to work in theory, never in practice..

Just converts the +- 400 mV swing to a +- 2.4 V at a 2.5 V offset (with temperature compensation on the probe). Just planned to use an arduino as the ADC at first, with all the rest done in code.

Never got it working, ditched the project. Post your schematic if you get a working pH meter!

[ElektroQuark]

I'm waiting for some time to make one based on this TI application note: AN-1852

[Kleinstein]

The MCP3424 has internal gain - so no need for extra amplification. The ADC also has plenty of resolution, way more that useful for PH.

If you can use a virtual ground (GND of Probe set to something like 2.5 V) there is no need for a negative supply and thus noisy converter. Just have a low Bias RR op as buffer and use die differential input of the ADC.
There are plentey of 5 V Rail to Rail CMOS OPs with bias in the pA range. So even simple ones like MCP6002 would be OK.

The filter at the OP input is needed for RF suppresion and ESD protection. Depending on the OP extra low leakage diodes mit be a good idea. The diodes get less cirtical if bootstrapping is used to keep the voltage at essentially 0 - so need to use a JFET as a low leakage diode, just als BAV199 or similar should work.

If the probe need to be connected to GND,  the negative supply is needed - in addition there should be a protection at the input in case no supply is present.

[vindoline]

I'm planning to use a TL082 which also has a very high input impedance (something like TOhm). Will add the cap as well.

I have no idea what the guard band thing is  . I was thinking of using regular ceramic caps (1210 or one smaller if I can't find the right value). No idea how are they in terms of leakage.

pH meters are non-trivial circuits. pH probes are very high impedance sensors, you can't draw any current from them. Generally "electrometer" grade FETs or opamps are used as the input section. I believe that the TL082 will have too much input offset current for this application. Good luck.

[dom0]

The idea is to read the voltage signal from a pH probe which generates from about -450mV to +450mV (~50mV per pH unit, 7 pH being 0V). I will be using a MCP3424 ADC which has "input range of  ± 2.048V differentially". So all I'm trying to do is amplify the  ±450mV to ±2V (about 4.5 gain).

Have you considered connecting the probe directly to the MCP3424 and using the internal 4x Programmable Gain Amplifier (PGA)? That would simplify your circuit a great deal. I'm using the MCP3424 with PGA for my data logging thermometer and it works great. I haven't worked with pH sensors yet thought, so I know if a direct connection would work the way I think it would.

The MCP3424 has an input resistance in the order of only a few M?. Way to low for pH probes.

As for the PGA, like in many ADCs it won't be some kind of operational amplifier or even normal linear amplifier. Often it's not an amplifier at all, it's changed biasing, reference current switching and similar techniques.

[Zero999]

There are a range of op-amps which should work. The CA3130 is certainly. You could probably even try the TS272 or MCP602.

Going from the TI application  note, I don't see why the reference electrode can't be biased at half the supply.

I wonder if you can buy a pH probe with the buffer amplifier built-in? This would seem to be the best design: put the amplifier as near to the transducer as possible. That's what's done in other high impedance transducers such as electrect microphones.

[RonZ]

I designed and built a pH meter about thirty years ago. I used a CA3140 as the input Opamp it worked well with a combination electrode but I had to use the following trick. The Opamp +  input pin was not soldered to the PC or any other component but bent up at ninety degrees from its original position and then connected with a short solid wire directly to the BNC input pin the Opamp was wired as an x 1 buffer. The input Opamp was defluxed  and cleaned with analytical grade Iso-propanol to reduce surface leakage. The measurement was done using a ICL7106 chip using a reference consisting of a LM334Z feeding a LM336 2.5v active reference trimmed to 2.500 volts stability is extremely important. However it would not work with a dual electrode pH setup. As previous posters have said a pH meter is a non-trivial circuit because the pH electrode is an extremely high output impedance because it is an electrolytic cell using porous glass as the electrode separator the output impedance is greater the than tetra ohms from memory. Also I used a Teflon insulated BNC connector to connect to the pH electrode. Once you have it built you will need to calibrate it using at least 3 pH buffer standards (ph3, 7 and 10).

[charlespax]

The idea is to read the voltage signal from a pH probe which generates from about -450mV to +450mV (~50mV per pH unit, 7 pH being 0V). I will be using a MCP3424 ADC which has "input range of  ± 2.048V differentially". So all I'm trying to do is amplify the  ±450mV to ±2V (about 4.5 gain).

Have you considered connecting the probe directly to the MCP3424 and using the internal 4x Programmable Gain Amplifier (PGA)? That would simplify your circuit a great deal. I'm using the MCP3424 with PGA for my data logging thermometer and it works great. I haven't worked with pH sensors yet thought, so I know if a direct connection would work the way I think it would.

The MCP3424 has an input resistance in the order of only a few M?. Way to low for pH probes.

As for the PGA, like in many ADCs it won't be some kind of operational amplifier or even normal linear amplifier. Often it's not an amplifier at all, it's changed biasing, reference current switching and similar techniques.

Thanks. That's really good to know. I'll make sure to have a proper opamp stage before the ADC. I plan on using https://www.sparkyswidgets.com/product/miniph/ as a reference design. Then I'll drop that in place of one of the thermocouple inputs of my data logger.

[Gyro]

The LMC662 is (I think) the lowest input current op-amp available that is cheap and readily available. Input current is in the region of 3fA. Input offset voltage isn't as low as the more expensive (smd only) LMP7721 but its TC Vos is very low.

[charlespax]

The bottom of this page isn't rendering properly for me, but other pages on the forum are. Is it just me?

[Zero999]

The bottom of this page isn't rendering properly for me, but other pages on the forum are. Is it just me?

I'm having the same problem. I had to click the reply button to see the bottom half of this page.

[dom0]

Me too. Everything after RonZ post is cut out. I can read replies in "Post reply" and per Mail.

[charlespax]

The bottom of this page isn't rendering properly for me, but other pages on the forum are. Is it just me?

I'm having the same problem. I had to click the reply button to see the bottom half of this page.

Yeah, I had to click reply also. I sent a message to the admin.

[Brumby]

There seems to be something funny going on with this page at the end of the post from RonZ @ 01:23:03 today.

This is a test post - to see what happens.