Hi Nordic team,
I'm investigating a systematic positive measurement error of approximately +8% when using the SAADC on an nRF9151 with its internal 0.6 V reference.
After testing a second board and comparing different SAADC reference configurations, I suspect that the internal reference voltage on one device is significantly below its nominal 600 mV value.
Hardware
We have two boards with identical PCB designs, both using the nRF9151 LACA A1.
Both modules were purchased together in the same order, but the boards were assembled using different processes:
-
Board A (affected): Professionally assembled using an automated SMT assembly process.
-
Board B (working): Assembled in-house using a solder paste stencil, manual component placement, and reflow in an SMD oven.
The modules also have different traceability markings.
SAADC configuration
We're using embassy-nrf with the following configuration:
pub const BATTERY: AdcConfig = AdcConfig {
resolution: Resolution::_12bit,
oversample: Oversample::Over256x,
reference: Reference::Internal,
gain: Gain::Gain1_6,
resistor: Resistor::Bypass,
time: Time::_5US,
};
Measurements and observations
-
On Board A, measurements using the internal 0.6 V reference are consistently approximately 8% higher than the actual input voltage across the measurement range.
-
The error occurs when measuring both
VDD_GPIOand an externalAINinput. -
Testing different SAADC gain settings produces the same proportional error.
-
Switching to
VDD_GPIOas the SAADC reference, using a known supply voltage, reduces the measurement error on Board A to approximately +1%. -
Board B, using the original internal-reference configuration, also measures within approximately +1% of the actual voltage.
-
Testing Board A at different supply voltages between 3.0 V and 3.4 V produces the same approximately +8% error with the internal reference. The error appears independent of supply voltage within this range.
Based on these measurements, we estimate an approximately 7% deviation of the effective internal reference voltage from its nominal 600 mV value.
Datasheet specifications
In the nRF9151 Product Specification v1.1:
-
Section 6.12.7 specifies a nominal internal reference of 0.6 V.
-
Section 6.12.11 specifies a gain error of -3% to +3% for gain settings of 1/6 and 1/4.
I couldn't find a separate specification for the absolute accuracy of the internal 0.6 V reference.
Related Nordic DevZone discussions
I found two discussions that seem relevant:
-
Reference voltage for the ADC module in nRF9160 SiP?
Nordic confirms that the internal reference accuracy is covered by the SAADC gain-error specification, rather than having a separate reference accuracy specification. Although this concerns the nRF9160, I assume the same principle applies to the nRF9151.
-
This discussion reports an approximately 8% SAADC measurement error on an nRF9160 custom board at VDD_GPIO = 1.8 V, which disappears when VDD_GPIO is increased to approximately 2.1 V.
Our issue appears different because the +8% error persists between 3.0 V and 3.4 V, while changing the SAADC reference from internal to VDD_GPIO reduces the error to approximately +1%.
Questions
-
Is an approximately 7% deviation of the effective internal SAADC reference voltage expected, or should this be considered outside specification?
-
Does the specified ±3% SAADC gain error include the internal reference voltage tolerance, as indicated in the related nRF9160 DevZone discussion?
-
Are there any known nRF9151 hardware issues or errata that could explain this deviation?
-
Could differences between automated SMT assembly and manual component placement followed by oven reflow, particularly the reflow temperature profile, influence the accuracy of the internal 0.6 V reference?
I can provide photographs of both modules, including their traceability markings, and additional measurement data if needed.
Thanks in advance for your help!