Characterizing the MAX6341 Voltage Reference

 

The quality of the voltage reference is key to a precision multimeter so it was worth spending some time to run some simple tests on the MAX6341 to verify that it will be fit for the task of building a precision multimeter.

        MAX6341 Key Features

·         Ultra-Low, 1ppm/°C Max Tempco 

·         Very Low, 1.5µVp-p Noise (0.1Hz to 10Hz) (MAX6325) 

·         ±0.02% Initial Accuracy (MAX6350) 

·         ±15mA Output Source and Sink Current 

·         Low, 18mW Power Consumption (MAX6325) 

·         Industry-Standard Pinout 

·         Optional Noise Reduction and Voltage Trim 

·         Excellent Transient Response 

·         8-Pin SO Package Available 

·         Low, 30ppm/1000h Long-Term Stability 

·         Stable for All Capacitive Loads 

Pictured below is the schematic that was followed for this test. There is no reason to stray from this design in the final product as the chip manufacturers generally know how to get the best out of their products.

Initial Tolerance and Drift Test

            To perform this test, I relied on the accuracy and data logging capabilities of the Keysight 34461A. 

The circuit was built up using the same TDK Lambda boost converter that will be used in the final product. This ensures that any peculiarities that may arise from using the regulator are present in these tests.

Below are the results of this test that was performed over an hour.

As this test was more a test of the initial tolerance and drift over time, as opposed to over a temperature range, the results were extremely impressive.

This IC has a pin for voltage trimming but in reality, it doesn’t appear necessary. Over the course of an hour the voltage drift was ± 0.00002 or ±0.02%.  Which exactly matches Maxim's stated the initial accuracy as 0.02%. Any physical trimming that would be performed is likely to incur more errors due to the inaccuracy of analog potentiometers or resistors used.

It’s not strictly essential that this voltage is the same as what the manufacturer states. What’s more important is that it doesn’t drift over time and temperature. It is easy to compensate ,in software, for any slight difference in voltage in the final product.

Noise

A quick measurement was taken of the references output noise on an oscilloscope. The yellow wave form is the voltage reference and at first glance the results look alarming! However, the green wave form is the output of the boost converter and once a simple subtraction math function is performed (the purple wave), a much more acceptable level appears.

This math function is showing a Peak to Peak value of 60mV. Effort will be put into the final design to get this figure as low as possible but for now at least there’s a baseline figure to work from.