Hello,
I want to use a ADS1115 analog to digital converter to measure some voltages with a raspberry pi. This works with no problems.
But I need to know how accurate the measurments are. I found this datasheet from Texas Instruments:
https://www.ti.com/lit/ds/symlink/ads11 ... 52FADS1115
Page 7 seems to have some information about the arruracy/measurment errors but unfortunatly I do not understand them. A lot of these informations are for FSR=+-2.048V but what if I set FSR to something else?
Figure 19 seems to have information about the total error. But the y-axis is labels in uV which is not realistic. I asket the TI-customer service about this and I got this answer: "Checking figure 19, the unit looks like an error to me. I believe that should read "mV" instead of "uV"." There is also this older datasheet that has "mV" on the y-axis of figure 19:
https://pdf1.alldatasheet.com/datasheet ... S1115.html
Neighter on page 7 nor in figure 19 I can find comlete information about the accuracy of the ADS1115.
My question in short is: If I measure a voltage in direct (not differential) mode, how big is the maximum possible error?
My goal is to say someting like:
"The voltage of this source is 4V+-0.03V" (values are just examples)
Thank you very much for your help.
Alf
ADS1115 accuracy
Moderators: adafruit_support_bill, adafruit
Please be positive and constructive with your questions and comments.
- sj_remington
- Posts: 1021
- Joined: Mon Jul 27, 2020 4:51 pm
Re: ADS1115 accuracy
The ADS1115 is reasonably accurate but uncalibrated sensor, which means the data sheet provides only representative specifications for a typical example, plus some statistics for a distribution of examples. As you have noticed, the parameters depend on temperature and will drift with time.
Figure 19 (Total Error versus Input Signal) in the ADS1115 data sheet is informative, but only typical and not guaranteed.
If you want to be able to make a statement about accuracy, e.g. for an official report, you need to calibrate the device yourself using a trusted calibrated voltmeter. You will want to measure and report the temperature as well.
Adafruit has a good tutorial on the general approach to calibration.
Figure 19 (Total Error versus Input Signal) in the ADS1115 data sheet is informative, but only typical and not guaranteed.
If you want to be able to make a statement about accuracy, e.g. for an official report, you need to calibrate the device yourself using a trusted calibrated voltmeter. You will want to measure and report the temperature as well.
Adafruit has a good tutorial on the general approach to calibration.
- adafruit_support_bill
- Posts: 88150
- Joined: Sat Feb 07, 2009 10:11 am
Re: ADS1115 accuracy
See the calibration tutorial here: https://learn.adafruit.com/calibrating-sensors
- 4lfb0
- Posts: 3
- Joined: Fri Jun 24, 2022 7:40 am
Re: ADS1115 accuracy
Hi sj,
thanks for the reply.
Ofcourse the accuracy of every sensor is different from the accuracy of another one. And you are absolutely right, the accuracy will change over time.
But is there no maximum error that the manufacturer guarantees for a measurment?
For example for the temperature-sensor TMP117 (https://www.adafruit.com/product/4821) there is a table like this:
±0.1°C (maximum) from –20°C to +50°C
±0.15°C (maximum) from –40°C to +70°C
±0.2°C (maximum) from –40°C to +100°C
±0.25°C (maximum) from –55°C to +125°C
±0.3°C (maximum) from –55°C to +150°C
I would realy like to have an ADC with a table similar to this one. And calibration looks to be quite a lot of work.
Is there a voltage table comparable to the temperature table above for the ADS1115? Is there another I2C ADC with a table like this?
Alf
thanks for the reply.
Ofcourse the accuracy of every sensor is different from the accuracy of another one. And you are absolutely right, the accuracy will change over time.
But is there no maximum error that the manufacturer guarantees for a measurment?
For example for the temperature-sensor TMP117 (https://www.adafruit.com/product/4821) there is a table like this:
±0.1°C (maximum) from –20°C to +50°C
±0.15°C (maximum) from –40°C to +70°C
±0.2°C (maximum) from –40°C to +100°C
±0.25°C (maximum) from –55°C to +125°C
±0.3°C (maximum) from –55°C to +150°C
I would realy like to have an ADC with a table similar to this one. And calibration looks to be quite a lot of work.
Is there a voltage table comparable to the temperature table above for the ADS1115? Is there another I2C ADC with a table like this?
Alf
- adafruit_support_bill
- Posts: 88150
- Joined: Sat Feb 07, 2009 10:11 am
Re: ADS1115 accuracy
The TMP117 is a much simpler device than the ADS1115 so it is simpler to characterize its performance. That said, I always take manufacturer specs with a grain of salt: https://learn.adafruit.com/calibrating- ... aq-1298662
- 4lfb0
- Posts: 3
- Joined: Fri Jun 24, 2022 7:40 am
Re: ADS1115 accuracy
Is there realy no guaranteed limit for the accuracy of the ADS1115? I dont need super high accuracy, I just need to know the accuracy to include it in a scientific paper.
If there is no limit, that would mean that even if the measurment is orders of magnitude bigger or smaler than the real value, it would still be just inaccurate and not wrong.
Is there another I2C-compatible ADC with a guaranteed accuracy?
If there is no limit, that would mean that even if the measurment is orders of magnitude bigger or smaler than the real value, it would still be just inaccurate and not wrong.
Is there another I2C-compatible ADC with a guaranteed accuracy?
- adafruit_support_bill
- Posts: 88150
- Joined: Sat Feb 07, 2009 10:11 am
Re: ADS1115 accuracy
Different manufacturers have different approaches to writing spec sheets. The marketing department likes to show the device in the best possible light. But the engineering department needs to back that up with facts. Sometimes you need to dig deep into the fine print to understand how those statistics were obtained.
TI is actually pretty good about not making unrealistic promises. They list the various error sources and describe how they can contribute to the overall error under certain conditions. But they can't guarantee how any specific chip is going to perform under all conditions. You need to calibrate the device in the system where it will be used.
And here is a spec sheet for the Nuvoton NAU7802 24-bit ADC: https://cdn-learn.adafruit.com/assets/a ... 1653414853
On page 8 they quantify the various error sources in terms of % of full-scale. But the "Test Conditions" column clearly qualifies that as "With calibration".
TI is actually pretty good about not making unrealistic promises. They list the various error sources and describe how they can contribute to the overall error under certain conditions. But they can't guarantee how any specific chip is going to perform under all conditions. You need to calibrate the device in the system where it will be used.
And here is a spec sheet for the Nuvoton NAU7802 24-bit ADC: https://cdn-learn.adafruit.com/assets/a ... 1653414853
On page 8 they quantify the various error sources in terms of % of full-scale. But the "Test Conditions" column clearly qualifies that as "With calibration".
- sj_remington
- Posts: 1021
- Joined: Mon Jul 27, 2020 4:51 pm
Re: ADS1115 accuracy
Yes, of course. But you have to pay for the calibration and certification. That is what is usually understood as a "calibrated instrument".Is there another I2C-compatible ADC with a guaranteed accuracy?
You are getting into professional equipment, and the price goes up by a factor of 10 to 100 or more, depending on your requirements. State those requirements and people can give you better advice.
Or, calibrate the sensor yourself. If you live near a research university, technical college, etc. you might be able to borrow the needed equipment.
- adafruit_support_mike
- Posts: 67485
- Joined: Thu Feb 11, 2010 2:51 pm
Re: ADS1115 accuracy
There is a difference between calibration and certification.
Calibration is simply the act of testing a measuring device against some known quantity. You can do that with any randomly-selected resistor.. the trick is to keep making the same measurement at regular intervals so you can track any changes over time. Those calibration histories allow you to compare measurements made by other devices calibrated against the same reference.
Certification involves tracking the paperwork that connects every device in the calibration process to systems maintained by the international association of standards labs. You get a verifiable paper trail that lets you compare your device to any other certified device on the planet. That includes all the standards for whatever property you're measuring, so you get a reliable estimate of your device's accuracy as part of the certification process. That's the expensive part, and the cost pays to keep that international system running.
There's nothing wrong with using a randomly selected shop reference as long as all the measurements will stay in-house. You don't get a guarantee of accuracy, but "how does a lack of absolute accuracy matter?" is a valid and important design question.
Calibration is simply the act of testing a measuring device against some known quantity. You can do that with any randomly-selected resistor.. the trick is to keep making the same measurement at regular intervals so you can track any changes over time. Those calibration histories allow you to compare measurements made by other devices calibrated against the same reference.
Certification involves tracking the paperwork that connects every device in the calibration process to systems maintained by the international association of standards labs. You get a verifiable paper trail that lets you compare your device to any other certified device on the planet. That includes all the standards for whatever property you're measuring, so you get a reliable estimate of your device's accuracy as part of the certification process. That's the expensive part, and the cost pays to keep that international system running.
There's nothing wrong with using a randomly selected shop reference as long as all the measurements will stay in-house. You don't get a guarantee of accuracy, but "how does a lack of absolute accuracy matter?" is a valid and important design question.
Please be positive and constructive with your questions and comments.