Calibrating my bench equipment - which is most trustworthy?

Discussion in 'Test & Measurement Forum' started by cspwcspw, Jul 13, 2018.

  1. cspwcspw

    Thread Starter Member

    Nov 8, 2016
    34
    2
    Hi

    I have a low-end cheap setup, mostly stuff from Banggood, with two cheap multimeters, a bench PSU kit that I just put together that I needs calibration, zener diodes, resistors (although no low-tolerance ones that I am aware of), some regulators like 7805, 7812, op-amps, laptops with supposedly 5v USB supplies, some of those cheap voltage measuring units (with LED displays) and some Arduino and ESP32 microcontrollers.

    I don't have expensive (for me) Fluke multimeters or a high end oscilloscopes that get sent for factory calibration once a year. (Let's assume I don't have access to "reliable" stuff either).

    What is likely to be my best approach to getting a "gold standard" reference that I can believe? Do I start at the 3.3v pin on an Arduino and treat it as gospel? Or does a zener diode likely give me a better reference than a linear regulator? Are multimeters generally better at either measuring current or voltage, (internally of course they are the same but we hope they used low-tolerance circuitry), and is it generally better to try to measure near the middle of a range rather than at extremes? (Maybe volts rather than millivolts, or KV).

    Are there any algorithms for "successive refinement" of calibration in the case that your resistors are at best 10% accurate, etc.

    Thanks
    Peter
     
  2. dl324

    AAC Fanatic!

    Mar 30, 2015
    6,943
    1,606
    The usual way is to use an accurate voltage reference. I've seen some 10.000V references, but you can't use that to calibrate all scales of a measuring device.
    No. If you're referring to a voltage regulator output on the Arduino board, it could be about 3.0-3.6V and still be withing the "3.3V" spec.
    Zeners typically come in 5 and 10% tolerance. Voltage regulators are typically 5%. An issue with Zeners is that none of them have ideal characteristics, so zener voltage will vary with forward current and temperature.
     
  3. ScottWang

    Moderator

    Aug 23, 2012
    6,252
    970
  4. Kermit2

    AAC Fanatic!

    Feb 5, 2010
    4,166
    1,125
  5. KeepItSimpleStupid

    Distinguished Member

    Mar 4, 2014
    2,772
    492
    Here http://www.nutsvolts.com/magazine/article/build_a_01_accurate_voltage_reference a short blurb. Searching will get you more.

    Many multimeters will use a thick film hybrid resistor divider that has 10 M end to end, BUT you can overload one of them and mess up ay voltages higher and equal to the blown one.

    A low-end model multi-meter may have a 200 mV low range and everything else (higher voltage ranges) is based on a resistive divider.
    Measuring calibration on the lowest and highest range gives you the best overall health.

    As for current, most multi-meters measure the voltage across fixed resistors depending on range. My Fluke 77 has a 300 mA or below range and a 10A range. For the 10A range you have to use a different jack. <300 mA is fused, The 10A range is not. It's also not meant for continuous measurements.
     
  6. cspwcspw

    Thread Starter Member

    Nov 8, 2016
    34
    2
    Thanks for these tips. I have ordered a voltage reference module, and found the article referred to by KeepItSimple to be a good starting point for possibilities.
    Peter
     
  7. MrChips

    Moderator

    Oct 2, 2009
    16,724
    5,127
    How does one calibrate one's own test equipment without having to go to qualified test and calibration labs?
    There are so many parameters to calibrate relevant to the electronics field:
    • voltage
    • current
    • resistance
    • capacitance
    • inductance
    • time
    • frequency
    • temperature

    Where doe one begin?
    Do you need NIST traceable calibration? Let assume that for most of us, NIST traceable is not a requirement.

    To begin, you can purchase a high quality instrument such as a DMM and use that as your standard. Of course, that will require some substantial capital outlay. Or you can buy a high quality calibration standard for the particular parameter, again that cost money.

    For the DIY, let us look at one common parameter, voltage. For starters, zener diodes are a poor choice because the I-V characteristic curve will show that the zener voltage is highly dependent on the diode current.

    Search for the best cost effective precision voltage reference. In our lab, we have 50-year old Weston Cells which are based on mercury. After so many years, the voltage is still accurate to better than 0.1% at 1.018V.

    [​IMG]

    Your best bet will be to find a semiconductor device designed specifically as a high precision voltage reference.
    Go to any of the major manufacturers or suppliers and select your desired output voltage and precision.

    Here are two examples:

    Texas Instruments REF6050 5V precision voltage reference, ±0.05% initial accuracy, US$7.44.
    MAXIM MAX6250, 5V precision voltage reference, ±0.02%, US$3.82.

    Take it from there one step at a time, i.e. one parameter at a time.
     
    atferrari and jpanhalt like this.
  8. MrAl

    AAC Fanatic!

    Jun 17, 2014
    5,120
    1,109
    Hello there,

    I have only one traceable standard, a temperature meter, and for voltage i use an IC chip solution that cost just around 5 dollars.

    The chip is a 4.096 volt voltage reference chip. If i remember right it was made by Maxim.
    A 500 dollar (USD) bench meter measures it at 4.0960v and a 300 dollar bench meter measures almost the same.

    I needed a good reference around that voltage (4 volts) because i wanted to accurately measure the charge termination voltage of Li-ion cells using different charge methods and i did not want to have to depend on the accuracy of any particular meter.

    This is of course only one voltage level, but you can get other levels using some tricks.
    For example, to get exactly half of that you can use two resistors in a voltage divider and that will give 2.0480v. The way to be sure you are getting half the voltage is to swap the upper and lower resistors of the voltage divider and see that the voltage is the same for both topologies. If the voltage changes when you swap, then the two resistors are not exactly the same value and so the voltage comes out different. When they are the same, the voltage will read the same no matter what resistor is on top and what is on bottom.

    The voltage references do age, so check the data sheet to find out information on that too. The high accuracy ones are pretty good though.
     
  9. DickCappels

    Moderator

    Aug 21, 2008
    4,863
    1,470
    If you are real careful, mainly in choosing the right op amp, you can use those matched resistors as the feedback resistors in a noninverting amplifier. Using your 4.096 reference as the noninverting input gets you an 8.192 volt reference. Using precision metal film resistors in the range of 10k to 100k from the same batch might help they stay matched over minor temperature changes.

    This technique can be extended until you tire of matching resistors.

    PWM is also a good way to make a divider with a wide range of ratios but needs a little circuitry (some fast analog switches) and needs to be calibrated.
     
    atferrari likes this.
  10. MrAl

    AAC Fanatic!

    Jun 17, 2014
    5,120
    1,109
    Hi,

    Yes good idea i'll have to keep that in mind. The non inverting amp will get me to 2x the voltage with the same resistor set.

    I guess we could look into getting more voltages too with the matched resistors and it doesnt hurt too much to match one more to get three matched and then one more to get four, etc. :)
    I guess a large batch of resistors of the same value would be nice.

    I dont know if you remember the "old" days when we would use carbon resistors and use a "V" triangle file to cut into the side and then into the carbon and thus tweek the resistor value upwards just a little. Sometimes i miss those days :)
     
  11. crutschow

    Expert

    Mar 14, 2008
    19,518
    5,443
    You can use your ohmmeter to match resistors (to the precision of the ohmmeter).
    For matching purposes of equal value resistors, the absolute accuracy of the ohmmeter doesn't matter, only its display resolution.
    With that you can make voltage dividers or op amps with voltage gain accuracy equal to the display resolution.

    For example, with a common 3½ digit multimeter, you should be able to match 1k or 10k or 100k resistors with 0.1% precision. These can then be used for 0.1% accurate voltages from a voltage reference or as 0.1% gain elements in a precision op amp (an op amp minimizes errors due to loading).

    But for this to work only the same value resistors can be used in the divider or for the op amp gain.
    More than one resistor can be put in series, of course, to get the desired total resistance.

    Even if you don't have enough resistors to get a 0.1% match, you can still calculate the divider output voltage or op amp gain to 0.1% from the measured resistance values.

    That should be good enough for most hobby requirements.
     
  12. MrAl

    AAC Fanatic!

    Jun 17, 2014
    5,120
    1,109
    Hi,

    I guess i should have mentioned that the reference i use is 0.02 percent, but because of measurements of that reference with expensive equipment i know it is almost dead on. I would not want to loose that preciseness so i use a meter that reads with five digits. So i get a reading of 4.0960 volts. I'd want to keep that accuracy using that DC volt meter. So really i am after a tolerance that allows up to 4.0968 volts or down to 4.0952 volts, but nothing worse, and it would really be better than that using the DC meter because i am starting with a reference that i believe is better than 0.01 percent.
    As you can see at least for me high accuracy is a necessity.
     
  13. atferrari

    AAC Fanatic!

    Jan 6, 2004
    2,968
    1,106
    Hola @MrChips

    Not intending to derail this thread, my question: what you use to verify the above? Whatever it is, isn't it your actual standard instead of the Weston cell?
     
  14. crutschow

    Expert

    Mar 14, 2008
    19,518
    5,443
    Okay.
    But I don't think the TS or most hobbyists need that high degree of accuracy, which requires at least a good 5-digit voltmeter.
     
  15. MrAl

    AAC Fanatic!

    Jun 17, 2014
    5,120
    1,109
    Hi,

    Yes i agree. Most of the time 0.1 percent is good enough. I just dont feel comfortable starting out with a 0.01 percent tolerance and then end up degrading that to 0.1 percent when getting the 0.01 is just a little bit more work.

    So yes i fully agree, and i know even 0.5 percent and even 1 percent is often good enough.
    For example when i measure the charge termination voltage of an Li-ion cell i like to get that reading with better than 1 percent because the spec i think is 1 percent. So i like to be working with equipment that can do at least 0.5 percent or even 0.1 percent. That way ican be reasonable sure i am getting a reading that is better than the spec.

    Some of the people i have worked with in the past were so particular that they wanted a judgement reading from a digital meter that was BETTER than the resolution of the meter. We might ask how that is possible. The technique is to get at least another 1/2 digit by watching the last digit bounce up and down. As it goes from say 1.101 to 1.102 and back again, we would judge that to mean the reading is 1.1015 but it's harder to judge 1.1014 or 1.1016 so it was usually acceptable to use that 1.1015 if the reading did bounce between two numbers.
    That was for a defense aircraft company. Various militaries were very demanding.
     
  16. MrChips

    Moderator

    Oct 2, 2009
    16,724
    5,127
    Good question. The E.M.F. written on one of the cells is 1.0189, dated 1-3-47.
    Fluke 83 meter reads 1.018V
    B+K 391A reads 1.0183V

    That suggests that they agree to within ±0.1%.
     
  17. DickCappels

    Moderator

    Aug 21, 2008
    4,863
    1,470
    How sensitive is the standard cell to temperature?

    I have found that a company named Apex Microtechnology offers a 10V ± 1mv reference (0.01%) with temperature coefficients of 1 to 2 ppm/°C -there is also a 0.6 ppm/°C version but it seems rare. The price is about US$85.00. Still considering it.

    https://www.apexanalog.com/resources/products/vre310ds.pdf
     
  18. MrChips

    Moderator

    Oct 2, 2009
    16,724
    5,127
    The original design was a saturated cadmium cell producing a 1.018638 V reference and had the advantage of having a lower temperature coefficient than the previously used Clark cell.[1]

    One of the great advantages of the Weston normal cell is its small change of electromotive force with change of temperature. At any temperature t between 0 °C and40 °C, Et/V = E20/V − 0.0000406 (t/°C − 20) − 0.00000095 (t/°C − 20)2 + 0.00000001 (t/°C − 20)3. This temperature formula was adopted by the London conference of 1908[2]

    The temperature coefficient can be reduced by shifting to an unsaturated design, the predominant type today. However, an unsaturated cell's output decreases by some 80 microvolts per year, which is compensated by periodic calibration against a saturated cell.

    Reference:
    https://en.wikipedia.org/wiki/Weston_cell


    The un-saturated type of cell is more common, as it is portable. The electrode material and mercurous sulfate are held firmly in place with porous plugs, so the cell won't be damaged if inverted. The temperature coefficient can be ignored in most applications, but is given by:

    Et = E20 - 0.000005(t-20)

    Reference:
    http://conradhoffman.com/stdcell.htm
     
  19. DickCappels

    Moderator

    Aug 21, 2008
    4,863
    1,470
    I read that as 5 ppm/°C which is better than all but very few bandgap references can give today, though with modern bangap references, if you drop it on the floor it has little is any measurable effect on performance :)
     
  20. MrAl

    AAC Fanatic!

    Jun 17, 2014
    5,120
    1,109
    Hi,

    You might want to try Maxim. 0.02 percent is pretty good for around 5 dollars (USD).
     
Loading...