ADS1000 Suitable for my application?

[TopGun]

I'm trying to put together a setup to measure Oxygen concentration using a Honeywell sensor that is passive (Gives me a voltage between 10mV and 80mV due to a redox reaction). I need to measure the voltage using an Arduino. Will the ADS1000 be suitable?
Or would I need to pre-amplify the signal with a non inverting Op-Amp?

Datasheet for the ADS1000:https://www.ti.com/lit/ds/symlink/ads1000.pdf] [url]https://www.ti.com/lit/ds/symlink/ads1000.pdf [/url]


Thanks in advance.

[RichC]

The maximum gain from the ads1000 is 8, which combined with a maximum input of 80 mV gives rather poor resolution. That may be adequate, but if you need the full 12 bits then you will need a pre amp.

[Doctorandus_P]

Instead of adding a pre-amp, you can also have a look into IC's that are designed for low level measurement stuff, such as strain gauges or for measuring thermocouples.

HX711 would be a cheap / easy example, but I'm having doubt about long term and temperature stability. It's unlikely you get more then what you pay for.

Do you have a particular reason for being interested in the ADS1000?

According to the datasheet of the ADS1000 there is also an ADS1100 with 16-bit resolution, so even though you still piss away a few bits because the full range don't match with your sensor, you still have four more bits than you would have with the ADS1000.

[TopGun]

>Do you have a particular reason for being interested in the ADS1000?

It's cheap, and apparently pretty temperature stable. Also, it is rail to rail, and single input, which is essential because my device should run on batteries.
But I'm open to any suggestions which you may feel is better suited for my application.

>Instead of adding a pre-amp, you can also have a look into IC's that are designed for low level measurement stuff, such as strain gauges or for measuring thermocouples.
What about the MCP3425? The datasheet states that it is designed for thermocouples and weight scales.
Is the HX711 Accurate? Any experience is welcome. My measurement needs to be pretty accurate to be of any use. The sensor reads the entire range of concentration (0-100%) as voltage from 0-85mV.

[Doctorandus_P]

HX711 is one of the el-ceapo things from Ali.
I consider it fun to play with, but the datasheet is... too short to take the thing seriously. It has just enough info to connect it and how to shift the bits out of it.

There has been an attempt here on Eevblog to use it as a measurement device for shunt resistors. There may be some info on stability and long term drift in that thread. But to keep it simple, I would not consider this chip for a (critical) commecial application.

You have to set yourself goals for resolution, accuracy, drift, speed and other factors and then compare it with datasheets of IC's you can get from a reliable source and decent price.

There simply are too many factors to find the ideal mix.

Also note that the time you spend to find the best chip for your application can be an important factor.
It may be a good Idea to start with a chip that is grossly over specified (and too expensive) for what you need, but is easy to obtain and start with and revise the project later if there is commercial interest in your product.
For one-offs or low production runs, the cost of the components themselves is usually dwarfed by the man-hours needed to design it.

[TopGun]

>It may be a good Idea to start with a chip that is grossly over specified (and too expensive) for what you need, but is easy to obtain and start with and revise the project later if there is commercial interest in your product.
For one-offs or low production runs, the cost of the components themselves is usually dwarfed by the man-hours needed to design it.

I otally agree, which is why I'm better off sticking with something from TI, Analog, Max, etc. Do you know of any good One Chip ADC which can do a low voltage measurement?

[KT88]

Without knowing the exact accuracy requirements, I would start with a 16-Bit sigma-delta / delta-sigma.
The MCP3425 has an internal reference of 2.048V that gives you roughly 30uV/LSB. With some oversampling you could easyly get some more bits of resolution out of the signal chain. This helps at least to see trends a little bit better and would also reduce the noise quite a bit...

Cheers

Andreas

[TopGun]

How does the internal voltage reference affect the resolution?

Sorry if it's a stupid question.

[ledtester]

How does the internal voltage reference affect the resolution?

From the MCP3425 datasheet https://www.microchip.com/wwwproducts/en/MCP3425 ...

The on-board precision 2.048V reference voltage enables a differential input range of ±2.048V.

So the ±2.048V range is divided into 2^16 parts which determines the resolution.

In the case of the ADS1100:



which says that ±V_DD is divided into 2^12 parts making the resolution dependent on V_DD.

[mikerj]

As well as voltage range you also need to consider the output impedance of your sensor.  Is it low enough to drive an ADC input directly, or will it need buffering?  If it needs buffering then you might as well add some gain at the same time.

[TopGun]

So the ±2.048V range is divided into 2^16 parts which determines the resolution.

In the case of the ADS1100:

which says that ±V_DD is divided into 2^12 parts making the resolution dependent on V_DD.

Does this mean that my measurement will be dependent on Vdd if I use the ADS1000, but independent of Vdd if I use the MCP3425 since it has an internal reference?

[TopGun]

As well as voltage range you also need to consider the output impedance of your sensor.  Is it low enough to drive an ADC input directly, or will it need buffering?  If it needs buffering then you might as well add some gain at the same time.

The output of my sensor recommends an output impedance >10KOhms. The ADC says that it has a common mode impedance >10MOhms. That should work, right?

[KT88]

Does this mean that my measurement will be dependent on Vdd if I use the ADS1000, but independent of Vdd if I use the MCP3425 since it has an internal reference?

Yes.
The digital output of an ADC is always the ratio of the input (voltage) and the reference (voltage).
The ADS1000 uses Vdd as reference.
This is especially fine, if you use a source (sensor) with ratiometric (in relation to Vdd) output.
Concerning the input bias current you only need a small capacitor to provide some bypass for short current spikes generated by the SD modulator. The average current will be low enough to be ignored...
In case your sensor has a ratiometric output, the LTC2450 is a low cost 16-Bit SD-ADC that has low power properties...

Cheers

Andreas

[Kleinstein]

The redox cell output should not be ratiometric, but more like an absolute ouput. So the ADC internal reference should not be a problem.

Redox cells and similar sensors may want really high impedance / low bias. So it may be better to have an extra amplifier / buffer.
For the ADC it is not the common mode impedance that matters, but the bias and maybe differential mode impedance.

It depends on the sensor and the accuracy/ noise needs one has.

With some amplification one may also be able to use just the µC internal 10 Bit ADC. It may need some oversampling / averaging for noise suppresion, but may still be easier than an extra ADC.

[TopGun]

The redox cell output should not be ratiometric, but more like an absolute ouput. So the ADC internal reference should not be a problem.

Redox cells and similar sensors may want really high impedance / low bias. So it may be better to have an extra amplifier / buffer.
For the ADC it is not the common mode impedance that matters, but the bias and maybe differential mode impedance.

It depends on the sensor and the accuracy/ noise needs one has.

With some amplification one may also be able to use just the µC internal 10 Bit ADC. It may need some oversampling / averaging for noise suppresion, but may still be easier than an extra ADC.

)

Yes, the redox cell does provide me with an absolute voltage output. The differential impedance of the ADC is >2.4MOhm. The sensor manufacturer (Honeywell says that anything above 10k is sufficient. So am I correct in assuming that it will work?

Also, using the internal ADC if possible will make my life a whole lot easier. I will then just need to choose a simple OpAmp amplifier circuit with one of analog or TI's low noise OpAmps. Maybe something like the OPA377?
(url)Datasheet for the OPA377(url)https://www.ti.com/lit/ds/symlink/opa377-q1.pdf

 

[Kleinstein]

The OPA377 would be good enough for the 10 bit internal ADC. No need to worry about bias with this OP.

Even with a 10 Bit ADC there is likely no need to suppress the 10 mV low limit - just some amplification (about * 30 for a 2.5 V ADC range) may be good enough. The AVR internal ref. is not very good, so maybe plan for an external ref. - at least as an option.
I would also plan to measure the temperature, just in case. The response may be temperature dependent too.

With a kind of higher impedance source and possibly no ideal capacitance (high loss possible), the linearity may suffer a little with directly going to a SD ADC. It would probyble still be OK for the sensor, as sensor linearity is also often not that great.

 

[RoGeorge]

Maybe ADS1118 from TI.  Internal reference, 16 bits ADC, programmable gain, min range is +/-256 mV.

ADS1243 is a 24 bits ADC, programmable gain from 1 to 128, but it needs an external voltage reference.

[TopGun]

The OPA377 would be good enough for the 10 bit internal ADC. No need to worry about bias with this OP.

Even with a 10 Bit ADC there is likely no need to suppress the 10 mV low limit - just some amplification (about * 30 for a 2.5 V ADC range) may be good enough. The AVR internal ref. is not very good, so maybe plan for an external ref. - at least as an option.
I would also plan to measure the temperature, just in case. The response may be temperature dependent too.

With a kind of higher impedance source and possibly no ideal capacitance (high loss possible), the linearity may suffer a little with directly going to a SD ADC. It would probyble still be OK for the sensor, as sensor linearity is also often not that great.

The sensor is internally temperature compensated, so there's no need to worry about that. Also I cannot provide an external reference in my circuit, so I would prefer it to have an accurate internal reference. How bad is the Atmega 328P internal reference? Is it unusable? I only need like +-5mV precision after amplification...

[TopGun]

Maybe ADS1118 from TI.  Internal reference, 16 bits ADC, programmable gain, min range is +/-256 mV.

ADS1243 is a 24 bits ADC, programmable gain from 1 to 128, but it needs an external voltage reference.

I was actually planning to use the ADS1115 if all else fails. I believe that the sensor can directly be connected to this ADC without any pre gain because of the low input range being available, right?

[RoGeorge]

Yes, it should give about 12...13 bits of resolution for 80mV without any additional preamplifier or oversampling.

[Doctorandus_P]

Both the ADS1220 and AD7793 are 24 bit converters with built-in PGA and can be bought cheap soldered on a breakout board from your favorite chinese postage stamp shop.

It might be worth to get a few just to experiment a bit with them.

If the specs of these chips are anywhere near believable, then you will get nanovolt resolution out of them.
But to make such things make any sense you will need to add extremely good shielding and manage thermal differentials.
Here, Marco Reps builds the Open Source 8.5 Digit Voltmeter from CERN



I've also been looking into the HY3131 which is a quite popular DMM frontend with 50000 counts. Dave uses it in his 121GW, Keysight uses it in the EUR600 U1280 and chinese companies use it in EUR50 DMM's such as the HoldPeak HP-770D. and the EUR130 Owon XDM1041 (A benchtop with big TFT display)

Digilent has the HY3131 with some shunt resistors and isolated SMPS modules + data on a breakout board in the redicilous (and annoying) "arduino" form factor.

Annoying thing of the HY3131 is that you can't buy the chip separate.