ADC sample rate

[EYEnet-GROUP]

Hello,

in my project a CY8CKIT-062S2-AI board is mounted to a rotating disk with a sensor on the running surface. The disk spins at 3000 to 5000rpm and therefore the sensor read requires a high sample resolution. The Sensor and an analog amplifier are designed for 600-900kHz. According to the Datasheet of the CY8CKIT-062S2-AI the onboard ADC can process 2Msps. However a simple performance test script results in 41µs read time (24.4ksps) way short of the required 900ksps.

import time
from machine import ADC

adc = ADC("P10_0")

def testFunction():
  time1 = time.ticks_us()
  val = adc.read_uv()
  time2 = time.ticks_us()
  print(time2-time1)

Additionally a Micropython performance test script without the ADC results in 12µs read time.

import time

def testFunction():
  time1 = time.ticks_us()
  time2 = time.ticks_us()
  print(time2-time1)

In this case acquiring the ADC result takes 29µs. To simulate a continuous request the following script counts the number of samples in 1 second and results in 19.227ksps.

import time
from machine import ADC

adc = ADC("P10_0")

def testFunction():
  time1 = time.ticks_us() + 1000000
  i = 0
  while (time.ticks_us() < time1):
    val = adc.read_uv()
    i = i+1
  print(i)

In another test using psoc for arduino the read takes 43µs and the 1s test counts 24.366ksps almost identical to the micropython performance. Is there a setting I am missing to increase the sample rate or a buffer to read from?

[NikhitaR-IFX]

Hi @EYEnet-GROUP,

The timing difference might not be really hardware configuration itself, but the fact that in micropython, every call will add some overhead time due to interpreter execution. Few suggestions I can make to test if performance is improved :

1. Try configuring the sample_ns parameter while creating the ADC object. Details here. By default, we have 1000ns set in our port.
2. Use adc.read_u16() instead of adc.read_uv() because then you get raw values faster and can do optimized operations on application side.
3. You can try use micropython emitters - native and viper which should show improvement in these numbers already. If needed, you can find examples on how to use these emitters here.
4. You could also try to use cache method reference for adc.read_u16() like you would do in python codes which is known to improve the speed as well.

Let us know if any of this helps :)

Thanks and Best Regards,
Nikhita