Failing EMC test (ADC channels not stable)

[AR1980]

Hi

We are a company who have used the ESP WROOM 32 (and now ESP WROOM 32D) for a commercial product. Unfortunately we have failed the final EMC test. To be more specific the radiated RF immunity test. Especially in the 80 to 180 MHz range we have fluctuating analog measurements.

Our system has a lot of analog inputs and we are using the ADC in the ESP32. We have external multiplexers, so in the current design we only use one analog input channel on the ESP WROOM 32D and that is ADC1 channel 0.

What we have seen in the EMC test in the external laboratory and afterwards in our own test setup is, that even though we have a fixed voltage on the ADC input, the ADC reading starts fluctuating in the above mentioned frequency range. And we are not talking a few ticks, we are talking rail to rail oscillation, when we hit certain soft spots…..

We have tried filtering it out, but our internal tests show that even a 10 cm wire on one of the other ADC1 channels will make channel 0 fluctuate quite extensively. And we are not able to put a large filter on all these pins as they are used for other high speed communication like UART’s etc….

So my questions is: Have any of you seen anything like this when qualifying your product??? And if yes what have you done to get rid of this???

It is very late in the development stage of out product we find this, so we are in problems here. We have written to the technical support at Espressif, and they have admitted they have an issue here and they say to use coax cable or filter it out with capacitors (but the capacitors is not enough in our application and coax cable all the way to the IO pin is a no-go in a commercial product).

We run the test at a field strength of 10V/m, but we can provoke similar patterns as low as 2-3V/m.

To my best guess it is the internal attenuation/amplification stages of the ADC which starts oscillating, but of course that is my guess, I haven’t designed that ADC in the ESP32……

Currently we will have to look into an external ADC (I2C or an extra micro-controller with a stable ADC like and Atmel AVR or similar), but that takes time and requires a lot of re-qualification.

Any inputs are mostly welcome.

The attached picture show the fluctuations on the analog reading during the immunity test (should be straight lines).....

[WiFive]

You're saying any adc1 pin with a lead even when not configured as rtcio or adc function will still couple the noise into the adc and it will go haywire? What is the green line that looks less affected?

[AR1980]

Correct, even if some of the other ADC1 pins is setup to UART, digital input pin or other pherpherals, then adding a piece of wire to this pin will still influence ADC0 reading.
In regards to the green line. On that picture we are where both using channel 0, channel 3, channel 6 and channel 7. The disturbance was not equal on all channels. Actually we tried making a new version of the HW only using channel 0 (since that was the channel where we saw the lowest disturbance). But not enough, we still see similar issues.....

[WiFive]

Did you try any of the rtc isolate functions and setting the rtc iomux select to not be adc function

[fly135]

Susceptibility to induced noise is increased with impedance. If the circuit feeding the ADC is very high impedance then that may be your problem. I'm using ADC1 Ch6 and didn't see that problem with feeding it a signal from off the board.

John A

[AR1980]

@WiFive in regards to isolating RTC functions:
No, haven't tried that since if we only have something connected to ADC 0 I still have a problem (but you are right that we might be able to get rid of the cross talk from other channels by trying that).

@fly135 in regards to impedance:
We have a voltage follower with a 2 times 1K2 ohm on the output. We have tried to lower this impedance to 2 times 300 ohm without any difference

[Ritesh]

@WiFive in regards to isolating RTC functions:
No, haven't tried that since if we only have something connected to ADC 0 I still have a problem (but you are right that we might be able to get rid of the cross talk from other channels by trying that).

@fly135 in regards to impedance:
We have a voltage follower with a 2 times 1K2 ohm on the output. We have tried to lower this impedance to 2 times 300 ohm without any difference

Hello,

We have started to use channel 0 and channel 3 of ESP32 ADC1 in which we are facing similar kind of problem like getting fluctuation while reading raw value.

Actually, We are using 4-20 mA based sensors over ADC into channel 0 and channel 3. So, We are facing issues like not getting constant or nearest to constant value while reading raw ADC values.

Below are configurations of ADC which we have used.

- ADC Channel 0 and Channel 3 of ADC1
- Resolution - 12 Bit
- Attenuation - 11dB
- Number of Channel - 2
- Result Mode - Polling
- Conversation Mode - Continuous

ESP32 Input Reference Voltage - 1000 mv
VDD_A is 3.3V

So, Did you get any solution to resolve that issue which you are facing? if yes then can you please provide solution for that?

Let me know if need anything else from my side regarding our issue details.

[Ritesh]

Hello Espressif Team,

Any updates as per last few details which I have sent?