how to measure 0-5V analog sensors with Kinetis 3v3?
You must divide them down. Vdda within 100mV of Vdd, VrefH <= Vdda.
You can find the ADC electrical specifications in Datasheet.
The input voltage should between VREFL and VREFH.
I agree Earl Goodrich II , you must divide them down.
Simplistic answers like "divided down" can be very misleading to anyone new to using A/D's. Slap a resistor divider on the front the A/D and call it a day is a sure way to get a system with very poor performance.
Yes the answer is "divide down". How to divide 5 by 1.5 with a low enough impedance to drive the A/D correctly and not a low enough impedance to impact the thing being measured is not necessarily a simple task.
Take a look at how the TI INA159 works as one possible way to accomplish this goal:
Amplifiers and Linear - Difference Amplifier - INA159 - TI.com
To get more detailed answers you must supply a more detailed question about what your system is really doing.
Absolutely -- nothing about analog system design is EVER 'trivial'. But a 10 word question only gets a five word answer. Sure, a full design requires consideration of source and destination impedance, but also noise filtering and protection, bandwidth (and anti-aliasing?), and other details in support of particular 'accuracy' and 'throughput' goals. Is the source ratiometric or absolute, and how does that impact the conversion reference source? Can a differential signal treatment help avoid some complications like unknown ground offset? What is the optimal 'common mode range' for interface electronics? What is their impact to distortion and accuracy? Where is the tradeoff for implementation cost/complexity versus engineering cost/complexity? The list goes on...
Thanks for the answers
but, well, I'll be having less accuracy with "divide down", and this is not good for my application:
Reading, in real time, pressure sensors in different chambers. This sensors heve output 0-5V...
Another option might be to use a sensor that speaks I2C or SPI. You would need a 5V power supply for the sensor and then use a digital isolator, something like the SI8660, to convert the 5v data to 3v3.
This sort of thing is a problem I always run into. The answer depends on what is important to you.
Big question is, are you willing to Calibrate your channels or not? If calibration is required than all you need is a circuit that is stable and linear. A divider consisting of a dual op-amp and two 1% resistors might be fine in this case., because any scaling error will be taken care of by calibration.
Note there are two types of Calibration, manual and auto. In autocalibration you use a multiplexer to measure zero and a voltage close to full scale. And then to math to adjust your sensor readings. (is was very common in industrial control)
Without calibration, then the answer is, it depends. But typically look for 'Voltage Divider (TCR Matched)' parts, matched resistor pairs. Or just high precision resistors. Accuracy costs money, how accurate do you need?
Also note, the weak point is always the ADC's voltage reference. The second weak point is the ADC's offset. As 12 bit ADC might have a 1-2bit offset. A 18 Bit audio ADC might be very stable, very linear, stable, low noise, and crummy offset and span. It's been awhile but I used to see numbers around 1%. But if you have calibration or auto-calibration, then no problem.
The aforementioned INA159 is a 'pretty darn good' divider, with a max gain error of only 0.024% and 100uV offset. And TI has 'related' parts, even some with a front-end MUX like the PGA280. But of course the 'conversion reference' you put into Kinetis is another source of 'gain error', and actually matching that kind of tolerance is 'difficult'. IF you have the option, I would personally recommend gaining direct bridge-transducer access and going fully ratiometric differential. That is, drive the Bridge and the Kinetis Vref with the same 'reasonably stable' voltage, and feed the differential pair directly into the Kinetis differential inputs. This removes all those sources of gain/reference error.
Retrieving data ...