Logging temperature meter (ltm) v1 – reducing ADC noise

Logging temperature meter (ltm) is a ESP8266 based temperature measurement and logging device.

Above is the prototype.

The ESP8266 incorporates a single 10bit SA ADC with a single input.

Thermistor bias supply

The first trial used the on board 3.3V regulator for bias current for the thermistor. It turns out that is most unsuitable as the voltage varies quite a bit with WiFi operation. A TL431 precision programmable reference shunt regulator from the Vin line (nominally 5V) was used to derive a more stable bias source of ~2.5V.

Series resistance calculation

Solution is similar to that of a zener shunt regulator.

\(Rs_{min}=\frac{Vs_{max}-Vz_{min}}{Il_{min}+Iz_{max}}=\frac{5.5-2.4}{0+0.1}=31\) \(Rs_{max}=\frac{Vs_{min}-Vz_{max}}{Il_{max}+Iz_{min}}=\frac{4.5-2.5}{0.001+0.0025}=570\)

Select 560Ω. A smaller value will be needed for low resistance thermistors.

So, for the test case of a 10k3950B thermistor, a series resistance of nominally 3.3kΩ from the regulated supply to the thermistor to ground, and the junction of thermistor and 3.3kΩ is tied to the ESP8266 dev board pin A0 (which had a voltage divider to the ESP8266 ADC pin).

WiFi ADC noise contribution

Even then, there was still ADC noise associated with the WiFi operation. WiFi output power was dropped from the default 20.5dBm to 0dBm which reduced that component of noise to a lower level.

Note there was an issue listed, WiFi.setOutputPower setting ignored after some time, it open but inactive.

Summary: new NTC bias source, WiFi 0dBm

Above, the frequency distribution of 500 observations using the revised power supply arrangement and low WiFi power. Standard deviation is 0.45° which is acceptable, down from some 6°.

As mentioned, doubt has been case on whether setting low WiFi power is persistent.

Summary: new NTC bias regulator, WiFi OFF

Above, the frequency distribution of 500 observations using the revised power supply arrangement and low WiFi power. Standard deviation is 0.090° which is acceptable, down from some 6° with no measured, down from 0.45° for low power WiFi.

The findings here also affect rfpm2 and similar code changes have been incororated into that project.

A work in progress…