Introduction and methodology
Voltage measurements are a cornerstone of electronics design and engineering. It’s pretty simple for the basic engineering needs, where decent multimeter would deliver good measurements. But like any measurement device such multimeter would have to be verified and tested against better reference source to ensure compliance with promised specifications. This is where hierarchy and traceability measurement results comes in. I’d highly recommend amazing documentary The Last Kilogram for the general introduction of metrology ideas.
In this long overdue article I’ll present the reasons, methods and analysis of voltage maintenance system that is used to keep my personal lab electrical measurements with good agreement to international representation of SI Volt. This article is an answer to a complex question “How to do accurate in-house voltage measurements?”. It is a multi-faceted topic with many papers published around it, so please let me know if there is anything missing, lacking or incorrect.
Disclaimer
Redistribution and use of this article, any part of it or any images or files referenced in it, in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
- Redistributions of article must retain the above copyright notice, this list of conditions, link to this page (https://xdevs.com/article/xdevs_volts2025/) and the following disclaimer.
- Redistributions of files in binary form must reproduce the above copyright notice, this list of conditions, link to this page (https://xdevs.com/article/xdevs_volts2025/), and the following disclaimer in the documentation and/or other materials provided with the distribution, for example Readme file.
All information posted here is hosted just for education purposes and provided AS IS. In no event shall the author, xDevs.com site or any other 3rd party be liable for any special, direct, indirect, or consequential damages or any damages whatsoever resulting from loss of use, data or profits, whether in an action of contract, negligence or other tortuous action, arising out of or in connection with the use or performance of information published here.
If you are willing to contribute or add your experience regarding instrument repairs or provide extra information, you can do so following these simple instructions.
Certain commercial equipment, instruments, or materials are identified in this article to foster better understanding. Such identification does not imply recommendation or endorsement by the author or/and xDevs.com, nor does it imply that the materials or equipment identified are necessarily the best available for the purpose. Devices and components were not screened or conditioned prior to use in the electronic assembly, unless explicitly stated otherwise.
Metrology hierarchy and application tiers
Relative to voltage, we need better multimeter or sources to calibrate the meters that we see in the workshop. However, cost of making better measurements does increase a lot as well on every step of the ladder, requiring more time and more expensive delicate measurement experiments with severe maintenance and operation cost. With every next level amount of factors that must be accounted for measurement uncertainty budget are multiplying, transforming total error estimation into difficult statistics and the time-correlated math task.
Image 1: Example calibration hierarchy and conceptual relation between levels for voltage and resistance
There are thousands of meters that provide 1% or even 0.1% accuracy levels, but only a handful of expensive instruments and manufacturers that can offer 0.001% accuracy. And number of system designs capable to accurately provide 0.00001% levels of confidence can be counted with fingers on a single hand. Table below also illustrate the same concept but in different presentation format, highlighting difficulty and resource required for desired accuracy and stability level.
| SI unit | Desired resolution | Possible uncertainty | Implementation cost | Device example |
|---|---|---|---|---|
| DC Voltage | <3½ digits | 1% (10-2) | <$10 USD | Handheld or industrial DMM / Voltmeter, integrated MCU ADC |
| DC Voltage | 4½ digits | 0.1% (10-3) | ~$100 USD | High-end handheld DMM, entry level benchtop DMM |
| DC Voltage | 5½ digits | 0.01% (10-4) | $500 USD | Bench-top DMM, high-performance 24-bit ADC, calibrators |
| DC Voltage | 6½ digits | 0.001% (10-5) | $5,000 USD | High-end benchtop DMMs, high-end calibrators, 32-bit ADC |
| DC Voltage | 7½ digits | 0.0001% (10-6) | $20k+ USD | Best metrology DMM (Keysight 3458A, Fluke 85×8, Datron 1281) |
| DC Voltage | 8½ digits | 0.00001% (10-7) | $80k+ USD | DC reference bank with nanovolt-meter detectors and KVD |
| DC Voltage | Primary Quantum standards, >10 digits | <(10-10) | $0.5M+ USD | JVS , liquid helium cooled or dual-stage cryocooler |
Table 1: Voltage system levels and implementation for desired levels of accuracy
Take a note and think for a second, why fancy expensive commercial 8½-digit DMMs like Keysight 3458A, or Fluke 8508A/8588A and Keithley 2002 are considered as only 7½-digit here. This is not a typo, and number have very obvious reason for it. These instruments can measure and digitize voltage signals with resolution of 8½ digits but uncertainty of such measurement is still in 7½-digit realm at best due to limitations of noise, temperature and temporal stability. It is very important to understand the difference between resolution and uncertainty of the measurement.
Metrology concepts
Measurements of any physical quantity or value can never be exact. One can only know its value with a range of uncertainty. If measurement provides some quantity A, the measurement is written with an uncertainty: A ± ΔA. This expresses that the actual value of A is somewhere between A – ΔA and A + ΔA.
Resolution, precision and accuracy
Let’s go some precision hunting. Some might confuse measurement resolution, precision and accuracy. These concepts are actually different things and can be measured and expressed independently.
Accuracy – difference between the measurement result and true value of the signal, or in other words, closeness of agreement between a measured quantity value and a true quantity value of a measurand.
Precision – closeness of agreement between indications or measured quantity values obtained by replicate measurements on the same or similar objects under same specified conditions. High precision system will provide same results every time the measurement on constant input signal is taken.
Uncertainty – quantity about a measurement’s result inaccuracy due to instrument limits, environment, or operator skill, and is expressed as an range interval with a stated confidence level. Additional parameters can be also attached to uncertainty to provide better information about measurement for statistical analysis.
Resolution – smallest detectable change in measurement result, that cause perceptible change in output value. Resolution in signal-data converters is often provided in bits. Ideal 1-bit ADC will generate value “1” when input analog signal larger than 50% of the range, and value “0” if less. ADC with 8 bits of resolution have 28 = 256 steps (from 00000000 to 11111111), thus able to detect 100% / 256 = 0.390625% change of the range. If range limited between 0.0V and 10.0V input, then such ADC can resolve input analog voltage with 39.0625 mVDC step per 1 bit of digital code output. As result we say this ADC have 8 bits of resolution.
Higher resolution does not provide better accuracy or precision, it only provides smaller value step size. Noise and other system errors reduce actual useful resolution. Together with input signal range resolution provide sensitivity of the ADC.
Imagine yourself at a game, shooting a target with arrows with a goal to evaluate player skills. One hit is one measurement sample. If we take just one sample, we cannot know how accurate player’s skills are. But if we make a series of samples, we can see much more and have quantifiable dataset.
On target case a we have low precision and low accuracy, with shots all over the place. We cannot clearly see what is real value. Need improve both precision and accuracy to get sample hits like on target b. This is high precision with high accuracy. Now if we still use high precision, but our accuracy is low, we will get target image c. That’s why calibration even more important for high-precision equipment, as even if we have high precision, it does not automatically get good accuracy as well without calibration.
There is one more case missing on targets above – with low precision, and high accuracy. Think about it for a minute and missing picture should be easy to imagine.
Practical realization of the voltage unit
Best uncertainty of the voltage measurements is realized by the cryogenic primary voltage standards based on quantum Josephson effect in superconducting junctions cooled to 4.2 K (-269 °C). These complex systems are capable to provide discrete voltage steps without long-term drift effects, capable to achieve uncertainty at room temperature connectors ports better than 10-11. The voltage output generated by Josephson chip is by definition exactly equal nƒ/KJ but in actual measurement setup there are always errors, induced by RF frequency instability, the noise of the null detectors, switch EMF, cable leakage and resistive loss, EMI interference, junction defects, bias currents.
Major error sources that contribute to transfer uncertainty are DUT’s zener reference noise and switch thermal EMF offsets, resulting in uncertainty specification assigned to JVS calibration. Some of these errors can be measured and corrected, some others cannot be predicted, DUT noise or drift stability. Typically, the total uncertainty contribution of a JVS source itself includes multiple readings taken during the calibration method and have a level just within few nanovolts. That is a tiny fraction of own 732B’s typical noise specification 680 nV (SDEV from 10V output regression). It would be possible to achieve at least magnitude better uncertainty for 10 V from JVS transfers if we can get better DC voltage standards using modern technologies. And this challenge is already worked on by people like Beuning Labs LLC. Beuning Labs is a new name in electrical metrology, but it’s not a novice in electrical engineering. They spent many years on development and improving for making a much better 10 V standards closer to the performance of cryogenic quantum JJ-array systems. Some very promising results were presented at the CPEM 2024 and NCSLI 2025 conferences.
It’s key performance capability is 10 times less noise than the very best selected existing DC voltage standard products currently on the market. New standard is reusing the ideas implemented in famous time standards such as cesium/rubidium atomic clocks to applying them to enhance performance in both short-term and long-term domains.
The new standard, named Z10 is built in a convenient 19” 4U enclosure with a traditional internal battery back-up, rugged construction and user friendly layout. Outputs are available from both high-quality low thermal EMF 5-way binding posts and with rugged multi-pole circular connector for integration in the systems. Standard does not require any expensive cryogenics or external equipment to operate.
Beuning Labs had live demo of the two measurement setups: first with Keysight 34420A measuring differential voltage between two performance-selected commercial 10V zener standards and second setup replicated same nanovoltmeter reading differential voltage between two new Z10 standards. Even with noisy and quick setup at the conference floor this demo was able to showcase record low noise levels, at least number of times better than pair of golden 732B’s.
Such standards would be a good refresher to somewhat stale commercial DC voltage standards market which have not seen an improvement in performance for last 40 years. I am sure we will hear more about this Beuning Lab’s Z10 standards later on and see more real-world performance results once people start getting these in their labs. These characterized standards can be then used to calibrate and adjust secondary voltage standards, which can be a traditional room-temperature high-performance solid-state zener devices, in shape of commercial units like Fluke 732A/B/C and alike.
Current commercial standards have typical stability at long terms in the order of few µV/V/year if carefully maintained and left continuously powered. Calibration and industry labs which using these standards must send them to primary labs for calibration measurement in every one or two years to maintain low uncertainties at few µV/V. This is required because of drift that solid-state zener sources have a predictable part (depending on a quality of the device) and a random one (induced stress from temperature, pressure, humidity variations, shipping, handling abuse, etc.). Although it is possible to make a good estimation and correction to predictable drift component, the random instability has an unpredictable behavior and no mathematical models can account for that. As a result total uncertainty of the zener DC voltage reference would still increase over time, even if kept in safe vault underground, like the famous IPK.
Over last ten years here at xDevs.com lot of experiments were completed with our own custom designs in search of the improved stability zener DC voltage reference. Most of these utilize best reference on the market – Analog Devices (former Linear Technology) LTZ1000 and LTZ1000A Super-Zener. Some publications about these designs and results are available here, here and here. These designs demonstrated short-term stability well within under 1 ppm over period of days/weeks, with errors mostly induced by environment and measurement setups, rather than reference itself.
One such prototype xDevs.com 792X FX LTZ1000A reference, assembled in Fluke 792A Power Pack. This standard provides +10 VDC nominal, with estimated annual drift less than -3 ppm. Key feature is high output drive capability with minimal loading errors, due to use of true 4-wire Kelvin connection output. Unlike Fluke 732B, this unit is able to sink or source currents up to 25 mADC with minimal error to the output voltage. Since 2019 we also started testing and building prototypes with newer Analog Devices ADR1000 and ADR1001 zener ICs which are able to demonstrate significantly better short-term and 1/f noise performance. I’ve presented results of that research at CPEM 2024 in Denver with published paper in IEEE proceedings.
But how one would verify performance of high-quality laboratory zener reference without the ultimate Josephson Junction Voltage Standard? With a bit of systematic approach, it is possible to do relatively good measurements even without JVS next door. None of the commercial digital multimeter is able to directly measure output of high quality zener standard’s drift to best uncertainty because internally all these DMMs are limited by their own internal zener IC. Sure, one can cherry pick golden 8½-digit DMM, put it into super-precise temperature chamber to maintain perfect +23.00 °C temperature at constant humidity and perform whole lot of filtering and optimizing to have a chance of great +10 V measurement uncertainty. Sounds expensive and time consuming, but many beginner metrology enthusiasts are often falling into this expensive trap, buying more and more expensive muiltimeters until bank account runs dry.
However, a much smarter way is to skip a problem of measuring tiny changes of few µV on top of large 10 V signal and instead focus on doing differential measurement. This method is not free and do require additional hardware, such as multiple high-quality nominally equal 10 V standards. This method was explained by W.G. Eicke and J.M.Cameron in technical note NBS 430 Designs for Surveillance of the Volt Maintained by a Small Group of Saturated Standard Cells.
This publication was released in October 1967, almost 60 years ago and way before any 8½-digit DMM existed. But these methods are still very useful to apply today. This series-opposition differential method compares two nearly equal DC voltage standards by connecting them in series with opposite polarity, so that their large output voltages almost fully cancel. Instead of measuring tiny changes on top of ~10 V, we can now focus and optimize to measure only a small residual voltage. This is the key uncertainty win, as absolute accuracy of the voltmeter largely cancel out and only linearity, noise, and offset stability matters. Hence variations of this NBS 430 technique are widely adopted for any sub-ppm and sub-0.1-ppm DC voltage calibration work.
Implementation of series-opposition voltage array system at xDevs.com
DC Voltage standard stability verification to levels below a few µV/V and complicated analysis of results is not an easy challenge. Different solutions and approaches were tested in practice by xDevs.com members since 2016. One of very good solutions was acquisition of the specialized low thermal scanner, such as Dataproof 160 or 320 and measuring each DC Voltage reference output individually by some high-end long scale multimeter, such as Keysight 3458A or Datron 1281. But this method is prone to errors and would depend on noise and temperature stability of DMM. Gain errors can be cancelled out by using calibrated DC standard with known drift as a reference, but other issues remain in this method.
Better approach is to perform the comparison of voltage differential in series opposition between two different DC standards. This significantly reduced requirements for DVM performance and even lower-end 6½-digit instruments such as HP 34401A used in lowest 100mV range can be used now to obtain sub-ppm uncertainty. But best instrument for this task is often sensitive nanovoltmeter which can easily resolve even fractions of microvolt between loosely matched 10 V outputs of various standards. 1 µV of 10V equals 0.1 ppm resolution, and nanovoltmeter like Keithley 2182A or Keysight 34420A have own noise floor around 0.03 µV. Given internal zener reference noise in commercial DC standards at level around ~1 µV such comparison method is well suited for detecting even minute changes in output voltage EMF with best uncertainty. This procedure is also covered with diagrams in detail here.
This method is also what Fluke recommends performing calibration of their own 732B/C standards. In fact, even in Josephson Voltage standard system such as NIST SRI 6000 or Supracon JVS the calibration of unknown source or DC standard is done in very same manner. JVS array +10 V output configured as quantum-accurate noiseless level and connected in opposite to the compared zener, with commercial nanovoltmeter like HP 34420A measuring difference in nanovolts. This provides a direct link to intrinsic realization of the voltage, limited only by quality of the scanner/switching system, parasitic EMFs and nanovoltmeter detector short-term noise. Any gain and linearity errors of the nanovoltmeter can be easily calibrated and cancelled out with programmable JVS array output as well.
Block diagram outlines the key system components and overall design of voltage monitoring equipment. Nanovoltmeter Keithley 2182 (and later Agilent 34420A) were used to measure the small differential voltages between any of the two zeners in the array.
Design was inspired by systems and publications from recognized National Measurement Laboratories and their work used to determine long-term stability of voltage standards and artifacts. Some of the notable publications are listed here below:
- Tang, Y. , Kupferman, S. and Salazar, M. (2005), An Evaluation of Two Methods of Comparing Josephson Voltage Standards of Two Laboratories, IEEE Transactions on Instrumentation and Measurement
- T.J. Witt, Maintenance and dissemination of voltage standards by Zener-diode-based instruments
- Bilateral Comparison of 1.018 V and 10 V Standards between the INRIM (Italy) and the BIPM, November to December 2023
- Andrius Bartašiunas, Rimantas Miškinis, Dmitrij Smirnov, Emilis Urba, Stability and predictability of Zener reference voltage standards: long-term experience by the Lithuanian National Metrology Institute
- André Bülau, Daniela Walter, André Zimmermann, A 10 V Transfer Standard Based on Low-Noise Solid-State Zener Voltage Reference ADR1000
- P Kranjec, D Fefer, A Jeglič, A group self-controlled solid-state voltage reference source, November 1998
My measurement process for each datapoint on the chart following this automated algorithm:
- The nominal 10 V electromotive force (EMF) outputs of each standard were connected to 16-channel scanner, one at a time.
- Each output terminal of the measured standard was connected in series opposition to the verified good standard for a duration of the sampling interval.
- The EMF differences between the unknown standard and the known good standard are measured using a digital nanovoltmeter
- Up to 24 samples were taken consecutively and stored into the local log-file database for further analysis
- Absolute voltages for each channel are calculated using differential values from known standard(s) calibrated and traceable to NMI labs.
Nanovoltmeter operated on its most sensitive 10 mV or 1 mV range and processed in real-time with a custom data acquisition software developed in-house by Igor O. Each individual data point for each standard represents the mean of up to 48 measurements (24 in positive and 24 in negative polarity) and zero drift correction measurements. Once data point for particular standard is collected, channels are switched to next standard and whole process repeated. The whole scan of all 15 channels (one channel used for zero short, used to perform quick test measurements) takes about 90 minutes. All standards, scanner, nanovoltmeter and logging Raspberry Pi 3B microcomputer are placed on a shelf in normal environment with typical residential airconditioning/heating system.
Result of continuous operation of such system (in 24/7/365 monitoring state, with few exceptions in few periods) can be summarized on the chart below.
It’s a bit busy chart, so let’s unpack it with a series of simpler charts. First let’s dig into labels used for each of the measured zener standard.
| Name of device | Nominal output | Manufacturer, part-number | Power | Zener design |
|---|---|---|---|---|
| 732A-404 – standard | +10 V | Fluke, Model 732A, S/N 404 | AC power+battery | Motorola SZA263 REFAMP |
| 732A-319 – standard | +10 V | Fluke, Model 732A, S/N 319 | AC power+battery | Motorola SZA263 REFAMP |
| 792X(FX) – standard | +10 V | xDevs.com, 792X S/N 102 | AC power+battery | FX module with LTC LTZ1000A |
| 732B2 | +10 V | Fluke, Model 732B | AC power+battery | LTC LTFLU-1ACH |
| 732Bx | +10 V | Fluke, Model 732B, N/A | AC power+battery | LTC LTFLU-1ACH |
| FX1 | +10 V | xDevs.com, FX S/N 101 | DC power ±12.6 V | FX module with LTC LTZ1000A |
| Quad LTZ1000A | ||||
| Single ADR1000 QVR | ||||
| Dual ADR1000 QVR | ||||
| Quad ADR1000 QVR | ||||
| Quad ADR1000-10V QVR | ||||
| Fluke 732C |
QVR module combine up to four zener IC cells on a single 6-layer PCB. This project is described in this article in better detail. Some of the RAW data and analytics related to QVR project is also available via this GitHub repository page. Outputs of each zener cell is averaged with passive resistor network into combined low noise 6.6 V DC level, which is filtered and fed to zero-drift chopper amplifier with discrete bipolar trasistor output stage for current source/sink capability. Zener cells powered by low-noise LT3045 linear regulator, except heaters. Heaters were powered separately from direct DC input (usually +12.5 V … +13 V). Output amplifier also powered directly from bipolar DC inputs (+12.5 V … +13 V and -12.5 V … -13 V).
Now with this information in mind, let’s take a look on top three zeners that are used as anchor standards and good known “keepers” of the voltage in my lab. Results of 10.5 million samples to make 29012 data points with nanovoltmeter since 12 November 2022 until 25 December 2025 can be rendered on the following chart.
Calibration results summary
Calibration results and performance tracking
This reference was calibrated by Taiwan’s National Measurements Institute in August 2019 . NIST SRI6000 PJVS system was used as a primary traceable reference during that key calibration.
Calibration report for xDevs.com 792× 10V DC Voltage Reference, August 8 2019, Taiwan NMI
Calibration report for xDevs.com 792× 10V DC Voltage Reference, March 3 2020, TMI Lab, FL, USA
Accredited external calibration history so far:
| xDevs.com 792X | Calibration Date | Assigned value | Uncertainty | Shift | Shift/year | Days |
| TW NMI calibration (NIST PJVS Quantum system) | 08/08/2019 | 9.9999838 VDC | 0.03 µV/V | -1.82 µV/V | -1.16 µV/V | +573 |
| TMI calibration | 03/03/2020 | 9.9999751 VDC | 0.39 µV/V | -2.69 µV/V | -1.26 µV/V | +208 |
Here we have all list of traceable calibrations to other references of similar grade, and estimate annual stability of finished 792X project.
| Calibration date | Method | Measurement result | Uncertainty | Conditions | Accredited | Δ to predicted |
|---|---|---|---|---|---|---|
| January 2018 | Direct 3458A+3458B, K2002 measurement | +10.0000134 V | ±2.20 µV/V | Battery power | No | No data |
| March 2018 | Direct 3458B | +10.0000066 V | ±2.30 µV/V | Battery power | No | No data |
| May 2018 | Direct 3458A | +9.99998983 V | ±2.30 µV/V | Battery power | No | No data |
| October 2018 | Direct 3458B | +9.9999885 V | ±2.40 µV/V | Battery power | No | No data |
| December 2018 | Direct 3458A, corrected | +9.9999862 V | ±2.90 µV/V | Battery power | No | No data |
| March 2019 | Direct 3458s,2002 | +9.9999793 V | ±1.80 µV/V | AC power | No | No data |
| June 2019 | K155 vs F732B SI | +9.9999779 V | ±0.60 µV/V | Battery power | No | No data |
| 8 August 2019 | PJVS SRI6000 transfer @ CMS ITRI | +9.99998382 V | ±0.03 µV/V | Battery power | NMI | Zero reference |
| March 2020 | HP34420A vs F732B | +9.9999728 V | ±0.40 µV/V | Battery power | Yes | No data |
| October 2020 | 3458A transfer from 732B | +9.9999716 V | ±0.44 µV/V | Battery power | No | No data |
| November 2020 | JJ-referenced D4910 array | +9.9999698 V | ±0.30 µV/V | Battery power | No | 1st reference |
| September 2021 | NLab VBank system | +9.9999800 V | ±0.70 µV/V | AC Power | No | No data |
| June 2022 | D4910 TBank system | +9.9999795 V | ±0.80 µV/V | Battery Power | No | No data |
| July 2022 | D4910 TBank | +9.9999806 V | ±1.00 µV/V | Battery Power | No | No data |
| November 2022 | JJ-referenced D4910 XBank | +9.99997612 V | ±0.30 µV/V | Line Power | No | 2nd reference, +0.42 µV/V |
| December 2022 | X-Bank array + nVM | +9.99997557 V | ±0.32 µV/V | Line power | No | 0.04 µV/V |
| January 2023 | X-Bank array + nVM | +9.99997414 V | ±0.35 µV/V | Line power | No | 0.16 µV/V |
| February 2023 | X-Bank array + nVM | +9.99997320 V | ±0.38 µV/V | Line power | No | 0.24 µV/V |
| March 2023 | X-Bank array + nVM | +9.99997353 V | ±0.41 µV/V | Line power | No | 0.18 µV/V |
| April 2023 | X-Bank array + nVM | +9.99997500 V | ±0.44 µV/V | Line power | No | 0.02 µV/V |
| May 2023 | X-Bank array + nVM | +9.99997514 V | ±0.47 µV/V | Line power | No | -0.01 µV/V |
| June 2023 | X-Bank array + nVM | +9.99997482 V | ±0.50 µV/V | Line power | No | 0.00 µV/V |
| July 2023 | X-Bank array + nVM | +9.99997481 V | ±0.53 µV/V | Line power | No | -0.01 µV/V |
| August 2023 | X-Bank array + nVM | +9.99997453 V | ±0.56 µV/V | Line power | No | -0.01 µV/V |
| September 2023 | X-Bank array + nVM | +9.99997369 V | ±0.59 µV/V | Line power | No | 0.05 µV/V |
| October 2023 | X-Bank array + nVM | +9.99997373 V | ±0.62 µV/V | Line power | No | 0.03 µV/V |
| November 2023 | X-Bank array + nVM | +9.99997355 V | ±0.65 µV/V | Line power | No | 0.03 µV/V |
| June 2024 | X-Bank array + nVM | +9.9999762 V | ±0.8 µV/V | Line power | No | 0.? µV/V |
| 25 October 2024 | JJ-referenced D4910 XBank | +9.99997031 V | ±0.3 µV/V | Line power | No | 3rd reference, 0.? µV/V |
| 11 November 2024 | X-Bank array + nVM | +9.99997038 V | ±0.35 µV/V | Line power | No | 0.? µV/V |
| 1 December 2024 | X-Bank array + nVM | +9.99997068 V | ±0.35 µV/V | Line power | No | 0.? µV/V |
| 5 March 2025 | X-Bank array + nVM | +9.99997007 V | ±0.4 µV/V | Line power | No | 0.? µV/V |
| 20 June 2025 | X-Bank array + nVM | +9.99997198 V | ±0.4 µV/V | Line power | No | 0.? µV/V |
| 21 October 2025 | X-Bank array + nVM | +9.99997024 V | ±0.5 µV/V | Line power | No | 0.? µV/V |
Drift of this custom LTZ1000A-based referenced in 2266 days between key points on 8 August 2019 and 21 October 2025 is -1.358 µV/V. This translates into voltage drop at rate -0.22 µV/V/year, which is pretty good for DIY single-device zener reference that is being constantly powered.
Such long-term drift study also highlights importance of periodic calibrations with very good uncertainty, ideally directly referenced to quantum standards such as Josephson Voltage System with minimal amount of intermediate transfer standards or transportation. Table and chart in this section represent only condensed key summary of all the measurements and long-term data captured by xDevs.com members and in-house automation software. We have gigabytes of data points captured on this and other standards including commercial units like Fluke 732A/B, Datron 491x and others. Due to magnitude of sources used and complexity of such large dataset analysis with logistic challenges this data it is not available to public at this time.
One also can notice important evolution of our in-house voltage measurement and monitoring capability improvements over the years time span, which is also reflected in improved uncertainty of the points. Current setup is built around array of various zener standard, DataProof 160A low-thermal scanner , Keithley 2182A nanovoltmeter and in-house software implementing modified algorithm measuring matrix opposite polarity of each zener cell, heavily inspired by NIST/NBS Technical Note 430. All measurements are done in both polarities to cancel thermal EMF errors. Use of dedicated nanovoltmeter on most sensitive range instead of long-scale DMM such as HP3458A helps to remove the contribution of the meter’s tempco and noise from zener voltage result.
Conclusion and future progress
Hopefully this article provides overall insight into the background experiment that I’ve been running for last decade to improve and learn about precision DC voltage standards and methods to reduce uncertainty of keeping “lab volt” in agreement to internationally adopted SI representation.
Discussion about this article and related stuff is welcome in comment section or at our own IRC chat server: xdevs.com (port 4808, channel: #xDevs.com). Or if you like to contact me directly via traditional mail please find contact information here.
Projects like this are born from passion and a desire to share how things work. Education is the foundation of a healthy society - especially important in today's volatile world. xDevs began as a personal project notepad in Kherson, Ukraine back in 2008 and has grown with support of passionate readers just like you. There are no (and never will be) any ads, sponsors or shareholders behind xDevs.com, just a commitment to inspire and help learning. If you are in a position to help others like us, please consider supporting xDevs.com’s home-country Ukraine in its defense of freedom to speak, freedom to live in peace and freedom to choose their way. You can use official site to support Ukraine – United24 or Help99. Every cent counts.
Modified: Jan. 1, 2026, 5 a.m.