Last March 2016, I made my first voltage reference based on LTC6655 BHMS8 and recorded the first stability results. I haven't the time to make PCB and breadboard with low noise LDO and several decoupling capacitors was used. For the first 8 days, using Keithley 2002 DMM, I measured voltage fluctuations in the range of 124 ppm (min 2.500100V, max 2.500410V).
The result was not so great, but expected because any voltage reference needs hours for stabilization in the long term. The value is specified in the datasheet under long term drift name. In my case, expected long term drift was 60 ppm after the first 1000h. Usually this characteristic has logarithmic nature and with the time is decreasing.
Next measurements were made at September 2016 with better results. After approximately 4000 hours, during period of 2 weeks measurement, I got 52 ppm stability : min 2.500320V, max 2.500450V.
The last measurements, made in April 2017 were much better. The result was 5.2 ppm stability: min 2.500255V, max 2.500268V. This result was more with regard of the room's temperature changes. According the datasheet, the expected temperature coefficient has typical value of 1 ppm/°C and maximum value of 2 ppm/°C. The recorded temperature difference in the room was 7°C (17°C-24°C). This is less then 1 ppm/°C. All measurements were performed automatically for about 70 hours during period of 9 days. If I compare them with the last September results, they are within 25 and 70 ppm range for about 5000 hours period.
All measurements for the last year were made with the same DMM. It was calibrated in March 2016 and March 2017.
I will keep running the voltage reference for the next 12 mounts. I want to collect voltage drift data after stabilization to figure out the expected DMM accuracy using this low priced voltage reference.
