NUCLEO-H723ZG BOARD ADC NOISE

[josegalicia]

Hi, I am planning on using STM32H723ZGT6 chip on one of my boards. I got the NUCLEO-H723ZG to measure the 12-bit ADC noise. From the datasheet for STM32H723ZGT6, the max ADC noise should be 5 times the LSB or 4.02mV. (1 LSB = 3.3/4096) 

I am getting 22.5641mV of noise. Which is more than 5 times the max ADC noise from the datasheet. Below is an image of the data. I didn't use any averaging. I used a 1.5V battery as my ADC input. 

I am wondering if this is expected ? I want to have 2 or 3 bits of noise on my board. Based on this, it seems I would be getting 5 bits of noise? Thank you!

[Andrew Neil]

The Nucleo boards aren't optimised for analogue ...

[ahsrabrifat]

The STM32 Nucleo board's power supply may be noisy due to USB or switching noise from the onboard LDO/Switching Regulator. Add capacitors (10uF + 100nF) near VREF+ and ADC inputs.

[josegalicia]

Hi Andrew, I was wondering if there were any reference designs or internal configuration of the chip for best ADC noise performance. I did my my design with capacitors (10uF + 100nF) near VREF+ and ADC inputs as mentioned below by ahsrabrifat. I also followed Application note AN2834 for best results and my results are the same as NUCLEO-H723ZG Board, namely 5 bits of ADC noise. If there are any reference designs to best improve my ADC noise it would be very helpful!

[MasterT]

The best option is to do oversampling. 

What is sampling rate required? How you drive inputs of the adc?

[josegalicia]

Hi Master T, thank you for the response. We don't want to do oversampling so we are not slowed down. I am running the ADC clock frequency at the max, namely 75MHz.

Is there anything else you recommend? 

[MasterT]

I get lower noise in the analog domain ( adc + dac ) on my nucleo-H743zi2 if I completely switched off  Ethernet. 

I find out, that even I don't use E-net anywhere in the project, no do any config - assuming all hardware circuits should stay in sleep mode : it's not a case with nucleo-144 E-net equipped.  LAN8742A-CZ-TR is always running and generates huge power spikes. 

To solve issue desolder L1. Also remove jumpers listed in table 13.

https://www.st.com/resource/en/user_manual/um2407-stm32h7-nucleo144-boards-mb1364-stmicroelectronics.pdf 

Another things to keep under control  is clock. Has to be crystal - precise, not use any HSI. In case ST-Link provided, reconfigure to 25 /3 MHz, or it's generated by HSI in the ST-Link itself.

[Carl_G]

Are you using the internal VREF or an external VREF? Maybe plot the VREF along side your sample data.