ADC performance, achieving 16 bits (or very close to it)

[DrM]

In numerous postings it has been said that the limit for the Teensy ADC is substantially less than 16 and less than the performance metrics stated in the processor data sheets. However, measurements using solderless breadboards and jumper wires, are not an appropriate way to assess the performance of the ADC of the Teensy processor.

In fact the Teensy ADC really is capable of performance very close to 16 bits, if you use it correctly. In the following post, I describe a design that achieves noise levels within the last bit of the 16 bit ADC on a Teensy 3.1

https://forum.pjrc.com/threads/53173-USB-DAQ-with-the-Teensyhttp://

NXP seems to concur, and forwarded to me an application note on how to increase ADC accuracy in general, and for the ADC's in this family of processors. This can be found at

https://www.nxp.com/docs/en/application-note/AN5250.pdf

All of these recommendations are satisfied in the board that I described in the link above.

Note there that good layout is important, as well as careful part selection, circuit design and analysis.

 

[BJB]

Am not certain if the A5250 reference makes your point. It states the 'obvious'. Gain and offset errors are the (relatively) easy stuff. INL, DNL, reference stability, Z-mismatch, digital noise and most other stuff are not so easy in either hardware or software. Agree that some portions of the 'noise' can be removed via statistical stuff and averages done in code, and that other stuff can be done using nice analog front-end. A few years back, did a PCB (custom 5-layer) that got 13.8 bits of legit ADC data on a T3.2, but this was under a narrow range of inputs and an extremely limited range of environmental conditions, and was not connected to a computer, and was run off of a battery, and was most careful to provide shielding from the eye of Sauron.

Perhaps you are a wizard of signal conditioning and PCB layout and can do what many of us plebeians cannot. Dunno, but to quote Paul Stoffregen, "I think it'd be really interesting to see more details.."

 

[DrM]

One of us (not me) taught the instrumentation electronics course for some decades, and now has instruments around the outer planets. We collaborated on the design and part selections. Notably, he rejected most of my first choices for parts. I did the layout and routing and assembled the boards.

Did you notice the guidance on ground and signal traces? We actually went a little further than that. We also have a very quiet -5V supply that was particularly sensitive to layout.