Saturday, March 23, 2019

Calibration tools part 2 : DIY


This post is about what calibration tools can be made by yourself without buying expensive ones. This does not mean that they are cheap either. I already made several tools, some are in progress and some I have a plan to build in the near future.

Kicad projects of already done tools can be found in the following link:

https://github.com/opendcm/opendcm/tree/master/tools/calibration


1. Temperature


Before to start any calibration, it is essential to know which are the environment  conditions under which measurements will be performed. The most important condition is the temperature. The best tool is platinum resistance thermometer (PRT). Why exactly platinum? Because it has very linear characteristics, very good long term stability and available in tight tolerances. In the Internet there are plenty information about PRT and comparison with thermistor (NTC) and thermocouples. PRT sensors can be purchased for few euros for a thin film based sensors with B class tolerance (~0.4°C for 23°C) or A class tolerance (~0.2°C for 23°C). Their wires are extremely fragile and I do not recommend to solder them directly to the sensor wires. I broke two sensors trying to solder them. Third time I soldered the sensor's wires into a small proto PCB and after that I soldered the wires from the measurement system.

Ceramic cylindrical thin film temperature sensor
Another flat thin film temperature sensor attached to PCB 



If you already have DMM which can measure directly PRT sensor, you only have to attach a small PCB with 4 banana plugs to the DMM and to start to measure ambient temperature with the sensor.

If you do not have exactly that DMM function, the second option is to use IC that converts resistance value to temperature. A cheap MAX31865 PCB can be used for this purpose. Be aware that the accuracy can be several times worse than the PRT accuracy if the default value for reference resistor is used because of the self-heating. The bias voltage is 2V and using 400Ohms reference resistor will result in 5mA current through the RPT. This is 2.5mW power dissipation that will cause 1.25C self-heating for my DM-509 sensor which has 0.5C for mW self -heating value. The self-heating can be reduced with higher value of the resistor, but you have to sacrifice the resolution. Alternative option is the ADS1220 IC in ratio measurement mode with stable reference resistor, but I was not able to find such an open source project. So it is already in my future sub-projects.
(Update 7 August 2019 : Initial version of PCB for ADS1220 is ready and working. Check in the github repository for more information.)

Currently I'm using a thin film platinum sensor with 100 Ohm at °C class 1/3 DIN made by LabFacility which is measured by Advantest R6581T DMM with the built-in RTD reading capability in 4 wire mode. This gives me about 0.22 °C accuracy for 23°C which includes accuracy of the sensor, self-heating and the 100 Ohms range of the DMM without calibration. Using two point calibration at 0 °C using ice bath and at 29.76°C using Gallium element the accuracy can be increased. A very informative post can be found here.



A nice to have tool is temperature controlled box which can be used for measuring temperature drift. I planned to make such a box with thermoelectric cooler (TEC).

Most of the DMM calibration manuals describe in which points the calibration must be performed. In a case where the information is missing, the minimal amount of points for calibration are two: 0 value and maximal value of the range.

2. Zero calibration tool


The calibration of the zero value can be done with the simplest DIY tool: just short the DMM inputs with pure copper wire (to avoid the error from the thermal EMF) or make a small PCB with 4 holes for 4mm banana plugs. The holes must be shortly connected with the PCB tracks.
A special attention must be taken for used materials in the following ranges:
- Calibration of millivolt range.
- Calibration of  1V/10V for 6/7/8 digits DMM.
Connecting two different materials (for example: beryllium copper in the binding posts and nickel in the banana plug) can produce 10uV if the temperature difference of 1°C exists between them. So in this case you have to wait until a thermal equilibrium happens between the binding posts and the zero calibration tool. Sometimes this can take several minutes and to avoid a delay it is better to use at least ENIG plated PCB and gold platted banana plug made from gold plated copper or  brass.

I made some designs and the final ones can be found in the blog's github page. I ordered an ENIG PCB from OSHPark and purchased 4mm banana plugs model SA 404 made by Milti-Contact/Stäubli (gold over brass). A M3 screws made by brass are used for connection between the PCB and the plugs.

This short PCB tool can be used for noise measurement of the DMM as well.

3. Full scale calibration tools


The calibration of the maximum value is far more complicated if commercial calibrators are not used. Generally three approaches can be used:
- a calibrated reference for each range. This can be costly if high accuracy is required.
- transferring a single high accuracy reference to lower ranges.
- non-calibrated short time stable references for each range connected in parallel to the DMM under test and calibrated DMM. The calibrated DMM must have better accuracy than the DMM under test.

The reference (voltage, current or resistance) should have high stability for long term period, low noise (less than the minimum resolution), low drift due to temperature changes (Temperature Coefficient or TC) and low hysteresis (reference value changes due to temperature cycling). The initial tolerance of the reference is not so important if the reference is calibrated. Most of the DMM accept 10% tolerance of the reference value. This reduces the schematics complexity for adjusting to exact full scale value (10V/1KOhms/1A/etc). For example :

  • If you want to calibrate the 10V range, your voltage reference should be between 9V and 11V.
  • If you want to calibrate the 100KOhms range, your resistance reference value should be between 90KOhms and 110KOhms.

The reference value accuracy should be better than the stated in the instrument under test specification. Usually the 4:1 Test Accuracy Ratio (TAR) is applied: the accuracy of the reference should be 4 times better then the accuracy of the instrument's range. If ISO-17025 should be followed, Test Uncertainty Ratio or TUR should be used instead TAR.
Exact numbers depend on DMM specification for the specific range under calibration. For example:
- If the 10V range of the DMM has 40ppm per year accuracy (35ppm reading + 5 ppm range: typical for 6 digits DMM in the 18-28°C temperature range), your 10V reference should have less than 10 ppm accuracy at the moment of the calibration.
- If 10K range of the DMM has 110ppm per year accuracy (typical for 6 digits DMM), your reference resistor should have less than 27.5ppm accuracy at the moment of the calibration.
- The noise of the reference should be less than the absolute value of the last digit of the DMM range. E.g. less than 10uV for 10V range of 6 digits DMM.
- If the calibration takes long time (30-60 min) and can not be done in temperature controlled room, the temperature coefficient of the reference also should be less than temperature coefficient of the DMM range stated in the specification. E.g. the voltage reference temperature coefficient should be less than 6ppm/°C which is the typical value of the 6 digits DMM in the 10V range. As much as the reference's TC is lesser than the DMM TC, the calibration will be more accurate.

Here is a list with the best voltage references which can be made by yourself.

The first one is a single 10V voltage reference based on LTZ1000(A) IC invented by Linear Technologies (last year acquired by Analog Devices). It can be used for calibration for the most bench top DMM. With this reference, the 10V range can be calibrated directly and using resistance divider the 1V/0.1V ranges can be calibrated as well. The LTZ1000 voltage output is ~7.15V, the long term stability is ~2uV for √kHr (~0.3ppm) with 0.05ppm/°C drift and 1.2uVp-p noise. Depending on the resistors used in 7 to 10V amplifier, 2-5 ppm per year stability and 0.1-1ppm/°C TC can be reached. Such voltage reference can be used for calibration of DMM with up to 8 digits DMM.

There is a very informative forum thread in eevblog.com site about this IC. 3 LTZ1000 designs can be found in https://github.com/pepaslabs repository and two more can be found from the eevblog member TiN in the xdevs.com site.

I tried to make a voltage reference two years ago, but used components and layout caused a very high TC. I kept it running for almost 1 year and after that I moved the used LTZ1000 to a new PCB. This new PCB was based on design made by Dr. Frank (who is an eevblog member) with small modifications. This time I used Vishay RNC90Z for the important resistors and Vishay 300144 for 7 to 10V converter which had about 0.6 ppm/°C TC.

The 10V voltage reference was calibrated in September 2018 with 2.2 ppm expanded uncertainty by the Bulgarian Institute of Metrology (BIM). The measured value was 10.004907V. The next calibration will be made this June 2019. I tried to calibrate it last month, but I got a call from BIM that it was unusually hot and when I got it back I realized that the LTC1052 op-amp, which amplified the LTZ1000 7V output, was burned. I added a small PCB with current buffer and over-current protection and now I'm waiting for stabilization. The output voltage already drifted up with about 500uV.

Inside 240x120x100 mm ABS case I was able to assembly:
- the LTZ1000 board
- small PCB with current buffer and over current protection
- 12V 6Ah SLA battery (tested successfully for ~8 days battery operational)
- PCB for battery charger and under voltage detector
- 220V/15V transformer with main switch and fuse.

The new box with binding posts, LED indicators, power switch and fuse.

From left to the right : binding posts, LTZ1000 PCB, battery, SLA charger and the power supply

LTZ1000 PCB inside the box
LTZ1000 PCB

Front Panel



At front : the battery charger, at rear : 220V transformer for the power supply

Third voltage reference is on the way. It will use more expensive resistors and low noise current buffered output with over-current protection. I'm estimating to reach less than 5 ppm year accuracy for the 10V output and with good insulation to have less than 0.1 ppm/°C TC. A Kicad project for LTZ1000 is available in my project's github repository together with SLA BQ24450 based charger.

The second best voltage reference after LTZ1000 is based on LM399 IC. It has slight worse parameters than LTZ1000: ~8ppm/√kHr long term stability with typical 0.3ppm/°C drift and typical 7uV noise. It is much cheaper to built compared to LTZ1000. It can be used for calibration of up to 6 digits DMM. If multiple LM399 are used, the noise can be decreased.

After LTZ1000 and LM399 which are ovenized buried zeners, the best non-ovenized zeners I was able to find are shown in the next table:


Model
Nominal VoltageLong Term Stabilitymax TCNoise
Vppm after first 1Khppm/°CuVp-p
REF102CP10V52.55
LT1021B7754
AD588BQ/KQ10V151.56
LT1021B101556
LT1021B51553
LTC6655BHLS852020.6
MAX6350C_A53013

Based on their specification, they can be used for DMM calibration with up to 6 digits.

The second tool, which is in progress, is a short time stable multiple voltage references based on LTC6655-2.5V. It can be used for calibration of DMM using a second already calibrated DMM. It will have 100V/10V/1V/0.1V outputs and alternatively can be used for 20V/5V/2V/0.2V ranges calibration with LTC6650-5V. The first prototype PCB had wrong footprint for the LTC6655 and the second modular PCBs (separate analog, power and control to switch between voltages) is in progress.

The first version of the low noise multi voltage reference using LTC6655.


For the resistance reference I used 10KOhms resistor from Vishay VHZ series (Y507710K), stored in small aluminium cast case. It is already calibrated in the Bulgarian Institute of Metrology and I got 0.41ppm expanded accuracy. I already adjusted my Advantest R6581T in temperature controlled room in the company where I work. The difference between the resistor reference value and the measured by the DMM before the adjustment was 5.2 ppm (17 ppm year spec for the low power specification).






In the next table I collected information for resistors which can be used as reference. They have less than 35 ppm year accuracy for shelf life and less than 5ppm/°C.


Manifacturer
Series
Model
Order code
TCR, ppm/C
Case
Shelf life
TypMax
Vishay PG, Metrology grade
VHZVH102ZY50770.20.6Sealed2 ppm / 6 years
VH102JZY07840.20.6Sealed2 ppm / 6 years
VHZ555Y16350.21.8Sealed2 ppm / 6 years
VHZ555JY07830.21.8Sealed2 ppm / 6 years
Vishay PGVHPVHP203Y60720.22Sealed2 ppm / 6 years
VHP203JY60730.22Sealed2 ppm / 6 years
Vishay PGHZVHP202ZY60710.22Sealed2 ppm / 6 years
Vishay PGVHP100VHP100Y00780.33N/ASealed2 ppm / 6 years
Metrology gradeVHP101Y40780.33N/ASealed2 ppm / 6 years
VHP102Y50780.33N/ASealed2 ppm / 6 years
Metrology gradeVHP103Y60780.33N/ASealed2 ppm / 6 years
Vishay PGVHVH102KY078712.5Sealed5 ppm / 1 year
VH102KY078812.5Sealed5 ppm / 1 year
VH102KY578712.5Sealed5 ppm / 1 year
VHS102Y007722.5Sealed5 ppm / 1 year
VHS102Y008722.5Sealed5 ppm / 1 year
CaddockUSFUSF240N/A2Aluminum oxide ceramic20 ppm / 1 year
Vishay PGZZ201LY14540.20.6Molded epoxy25 ppm / 1 year
Z201Y14530.20.6Molded epoxy25 ppm / 1 year
Z204Y14410.20.6Molded epoxy25 ppm / 1 year
Z205Y14430.20.6Molded epoxy25 ppm / 1 year
Z206Y14470.20.6Molded epoxy25 ppm / 1 year
Vishay PGSS10xK/LY006212.5Molded epoxy25 ppm / 1 year
S10xK/LY078612.5Molded epoxy25 ppm / 1 year
S10xK/LY010112.5Molded epoxy25 ppm / 1 year
S10xK/LY010212.5Molded epoxy25 ppm / 1 year
S10xK/LY010312.5Molded epoxy25 ppm / 1 year
S10xK/LY410112.5Molded epoxy25 ppm / 1 year
S10xK/LY410212.5Molded epoxy25 ppm / 1 year
S10xC/J/DY000722.5Molded epoxy25 ppm / 1 year
S10xC/J/DY078522.5Molded epoxy25 ppm / 1 year
S10xC/J/DY001122.5Molded epoxy25 ppm / 1 year
S10xC/J/DY001222.5Molded epoxy25 ppm / 1 year
S10xC/J/DY001322.5Molded epoxy25 ppm / 1 year
Alpha ElectronicsMAZN/A1Molded epoxy25 ppm / 1 year
YN/A2.5Molded epoxy25 ppm / 1 year
Vishay PGRNC90RNC90ZY1189N/A2Molded epoxy25 ppm / 1 year
RNC90SY1506 N/A2Molded epoxy25 ppm / 1 year
RNC90YY0089N/A5Molded epoxy25 ppm / 1 year
RNC90TY1508N/A5Molded epoxy25 ppm / 1 year
S555Y0088N/A5Molded epoxy25 ppm / 1 year
Z555Y1288N/A5Molded epoxy25 ppm / 1 year
CaddockUSFUSF340N/A5Aluminum oxide ceramic20 ppm / 1 year
General ResistanceEconistor8G16 & 8G2435Molded epoxy35 ppm / 1 year

I bought several Vishay resistors from Z201, VH102K, VHP101 series and made small adapter in order to calibrate them directly to a DMM using WAGO connector:


Adapter for measuring resistors.

4. Transferring tools

Transferring tools are used to calibrated one or more references (voltage or resistance) from already calibrated single reference. For example : if you have calibrated 10V reference in a lab, you can calibrate your 100V or 1V reference at home using resistor divider.

The gold standard for transferring voltage references is Kelvin-Varley voltage divider (KVD) with null-meter. It is possible to be made at home, but if you want good accuracy, the price is extremely high. If you need an only fixed ratio like 1:10, it is cheaper to build a Hamon divider.

DIY Hamon divider


For transferring resistor references are used resistive bridges : Wheatstone bridge, Direct Current Comparator (DCC) Resistance Bridge. Again they are expensive instruments and the alternative is to build a low noise current source and to use DCV ratio mode of the DMM.

5. Calibration services

I'm sorry, but currently I can not provide such of service because:
- I do not have enough free time.
- There is no interest: in the last 2+ years, only one reader of this blog contact me for such service.
- PayPal changed their policy for refund and now the fees are much higher.
After spending almost one year for calibration research I was not able to find affordable calibration service in Europe for hobby project like mine. I asked several companies for 6 digits DMM calibration (without adjustment) and they offered a price starting from 200 Euro without the shipping. In USA the cost is ~100USD, but shipping is more expensive.
The calibration services for single reference value from National Metrology Institutes give very low uncertainty (few ppm or sub-ppm) for price range from 100 Euro to 500 Euro per single reference depending on the country and the method of calibration.
Thus I decided to provide a kind of calibration service at a low price and a reasonable uncertainty. It will be available only in Europe. In contrast with the ordinary services where the device under test is sent to calibration laboratory, I'm thinking to send the calibrated references to a client who wants to self-calibrate and adjust their equipment. I'm not sure how many people will like this idea, so if somebody is interested, please write a comment with your e-mail address in this page to connect with me.
I'm thinking to provide zero calibration tool, high accuracy thermometer, 3 sets of resistors (100R, 1K, 10K, 100K) with different level of accuracy and TCR, capable to calibrate 4-8 digits DMM, low noise multi-voltage reference for transferring accuracy (0.1V/1V/10V/100V or 0.2V/2V/20V) and 10V based LTZ1000 and 10KOhms based VH102Z resistor calibrated by Bulgarian Institute of Metrology. The client should deposit money via PayPal equal to the cost of the references. After that, the references will be sent by mail. After receiving them, the client will have 2 weeks for calibration and/or adjustments of their equipment. The return shipment is paid by the client. The deposit will be refunded after receiving the references. Total price will be the calibration service cost plus shipping cost.
My goal is to provide voltage and resistor references for under 50 Euro (excluding the shipping) capable to calibrate:
- 4 resistance ranges and 3/4 voltage ranges.
- own resistance/voltage references against references calibrated by National Metrology Lab.
In the table below initial estimated price for this service can be found:


Price in Euro
Calibration toolsDepositService costEst. shipping
Zero251Included with other tools
RTD Thermometer501Included with other tools
Z201 100R/1K/10K, S102 100K60105-10
VH102K 100R/1K/10K, VHP101 100K80155-10
VHP101 100R/1K/10K/100K120205-10
0.1/1/10/100V or 0.2/2/20V, low noise, short term stable Vrefs1502010
VH102Z 10K calibrated by National Metrology Lab2502510
LTZ1000 10V calibrated by National Metrology Lab5002530-40

The references except the 10V and 10KOhms will be calibrated against freshly calibrated Advantest R6581T DMM before the shipping and after receiving back. The 10V LTZ1000 and 10K VH102Z references will be calibrated once or two times per year. The uncertainty for references will be based on expanded uncertainty given by Bulgarian Institute of Metrology for 10V and 10K references and the 24h accuracy from Advantest R6581 specification.

It will take me some time to calibrate and to make temperature characterization of all references that I have and plan to build, so probably I can start this service  at end of the next year if enough people are interested in and if I receive metrological traceability from the Bulgarian Institute of Metrology. Currently I only have a measurement protocols that do not guarantee traceability.


3 comments:

  1. Hm, LTZ1000 must have airdraft baffle on top and bottom side, made from foam. Also LTZ chip should be soldered flush to board, without gap to PCB. Shorter the kovar pins are - the better. Do NOT use any kind of clips/connectors/jacks between LTZ and your output. Must be minimum junctions possible, ideally directly soldered shielded copper wire to LTZ pin and output connector, or your TC will be negatively impacted.

    Also before you jump into calibration service and taking customer's money, check your R6581 all ranges linearity. You can find nasty surpises there (as I did in my 6581T). 6581T is NOT a metrology transfer unit, and cannot be used as such without full characterization.

    ReplyDelete
  2. Hello Ilya,
    Thanks for the tips!
    This is a test version and I do not pay too much attention in the details. But for the next version I have to re-read the monster eevblog thread about LTZ1000.
    About characterization : you are absolutely right about R6581T. It is not metrological grade DMM and characterization is required. I hope to make linearity test this year against JAVS in the BIM. They have Supracon JAVS bought before 2-3 years, but still last year it was not officially used for calibration service. This is the reason I wrote that I can provide such service at the end of the next year. I have to prepare everything : temperature controlled chambers, several standard calibrations with BIM, certifications with metrological traceability and this will take time. May be even 2 years will be not enough.
    Kind Regards,
    Krasi

    ReplyDelete