Group Actions
Engagement
  • Author Author: shabaz
  • Date Created: Date Created
  • Views 9043 views
  • Likes 12 likes
  • Comments 32 comments
Related
Recommended

Building Frank's Milliohm Meter

shabaz

Introduction

This short blog post documents the specific component values and any mods done to the revision 1 PCB from    Working Prototype of a Kelvin (4-Wire) Milliohm Meter  Project14  Test Instrumentation entry!

It is a project intended to create a 4-wire measurement meter for 0-4 ohm and 0-40 ohm ranges, with sub-milliohm granularity.

 

It is a physically small test instrument, but works well! In very limited tests, the typical discrepancy between a calibrated commercial meter and this project was ballpark 0.1% although from that one can't promise that will definitely always be the case.

Lego blocks shown for size comparison:

image

 

I recorded a 10-min video with some basic test results:

 

 

Schematic

I followed the schematic (Rev 1.0) in Frank's blog post here:

Working Prototype of a Kelvin (4-Wire) Milliohm Meter

Frank kindly sent me a spare PCB to construct it on.

It is better to wait for a Rev 2 board, but here's the values/mods on my PCB currently, I used these values for the tests:

image

 

Enclosure and Assembly

The Bopla enclosureBopla enclosure has two end panels, these were cut/filed. The body of the case needed no modifications, except for trimming some tabs slightly, which support the end panels, but are unnecessary. They were trimmed by scoring and snapping off with pliers (it is soft ABS plastic), so that the panel meter and banana sockets could fit:

image

The 24A banana sockets24A banana sockets (just slight overkill) had to go on the rear panel, there's no space on the front panel : )

image

For the DPDT latching push switchDPDT latching push switch a couple of pieces of 22x10x3mm plastic were used, epoxied, to hold the switch in the correct position behind the front panel. The panel meter is from aliexpress.

I didn't screw or glue in the PCB for now:

image

View from the other end, you can see the switch attachment to the front panel more clearly:

image

 

 

Tests

I've not done much testing, but as can be seen in the video, I measured some resistances and compared with a Keithley DMM 6500. I used some resistance wire that I'd purchased for a different project, and crimped some ferrules on the ends (this is all experimental..):

image

 

Here are the results:

Test # Value according to DMM 6500 Value measured by Milliohm Meter project
1 8.6-8.7 milliohm 8.7 milliohm
2 15.5-15.6 milliohm 15.4 milliohm
3 125.5 milliohm 127.5 milliohm
4 319.9 milliohm 320.4 milliohm
5 3.319 ohm 3.323 ohm

 

The largest discrepancy between the two meters is a 1.6% in test #3, but the difference is around 0.1% ballpark in all other tests, so it could well be that I didn't clip the connectors on the same way for test#3. There is a small difference in location each time the clips are connected and disconnected - it won't be completely identical, although I could have reduced that by not moving the test clips, and swapping the banana plug end each time. Anyway, I only intended to do quick sanity tests, not anything more accurate for now.

 

Also, I've not tested at a value lower than 8.7 milliohm.. there may well be a lower limit of a few milliohms, probably not of practical importance once it is in the (say) 0-3 milliohm ballpark perhaps, but I will check with a lower resistance than 8.7 milliohm at some point.

There is some occasional instability that needs investigating (mentioned in a discussion comment here: https://www.element14.com/community/people/fmilburn/blog/2018/10/18/pcb-for-a-kelvin-4-wire-milliohm-meter#start=25

Note: there was some instability, but this is the solution; add two 10k resistors:

image

Meter Improvement (Conversion to Differential Input)

See the comments below  PCB for a Kelvin (4-Wire) Milliohm Meter for some information about how the meter can be improved, to squeeze a bit more accuracy out of it. The information is reproduced here to make it easier to follow: 

image

image

Bigger is Better?

After completing this build, I decided to start building another one, based on an interim board from Frank (it is not a version 2.0, more a version 1.5 of sorts).

This new one is broadly the same as the one described in this blog post, but with a slightly newer version of PCB. I put it in a bigger case, because I wish to eventually make this one battery powered. It is incomplete, I still need to add the battery power circuitry, and I need to drill the holes for the banana sockets.

image

A view inside the unit:

image

 

Summary

This is just a rev 1.0 project, but it is promising how accurate and stable it is, considering the low cost. It functions well.

It was a lot of fun to construct and try it out this Xmas! The design has a lot of flexibility, and is easy to solder and experiment with. I'm looking forward to seeing the design evolve - it will only improve I think!

I still can't get over how small it looks next to the other test instruments though : )

There is a rev 2 discussion here:  New Improved DIY Milliohm Meter V2.0 

image

Parents

  • That's very neat and compact. The picture of it next to the Keithley box is lovely - perhaps you should have entered it in the Question of Scale competition!

     

    How does it get on measuring the DCR of inductors? Is it stable?

     

    I'm also wondering whether you'd get a sensible value for RDS(on) for a power MOSFET. Don't see why not, though the manufacturers do their tests at a much higher current (the MOSFET I've used in the load is tested at 27.5A!) and it does vary a bit with current. It would be a very easy experiment if you wanted to try it (perhaps bias the gate with a 9V battery so that there are no complications in that area).

  • in reply to jc2048

    Hi Jon,

     

    I tried a MOSFET, and got decent results I think.. so it's another nice use of the milliohm meter!

    For the test, an IRFZ44 was used, and a voltage from 4 to 10 was applied to the gate, while the milliohmmeter was connected to the Drain-Source.

    image

    According to the ratings on the datasheet, the on resistance should be less than 28 milliohm at 10V. My results are in the chart below - I measured a value of 18.2 milliohm at 10V, so it meets the spec!

    image

  • in reply to fmilburn

    Thanks for the feedback! 

     

    For now I will keep it simple.  I like the idea of a Pi - it might even be a Pi Zero W that keeps the enclosure small but if done that will be a later revision.

     

    Regarding the loss of accuracy down in the milliohm region, on my current build I read 0.99 milliohm with the inputs shorted.  Looking back at the version I constructed on perfboard the reading was 1.01 milliohm.  So, not quite as low as what Shabaz is seeing but still +/- milliohm which was the original design spec.

     

    The thing I would most like to resolve in the next version is the occasional instability.  I can add pads for the series resistors that Shabaz highlights from the datasheet above (well spotted) on the next version if that works.  In section 4.4.6 of the datasheet they warn of input capacitance - I realize the leads can contribute but do you think there is anything on the board that should be changed?

     

    I have lost enthusiasm for the second range because I don't like the (admittedly simple but prone to possible mental error) change of scale.  This would be especially prone to mental error with auto ranging.  The only way to fix that I can see is with a microcontroller or Pi and ADC.  But it doesn't hurt to leave it in and it does not have to be populated.

     

    Regarding the schematic and board revisions, I plan to post a comment later today back on the post that has the Version 1 schematic.  In that comment I will include a revised schematic that captures planned updates.  Please feel free to critique.  After receiving comments I plan to capture them and revise the PCB within a week and make an order.  The PCB will be designed to fit the enclosure that Shabaz is using, including mounting holes, etc.   If you are interested in building it, please contact me by personal message and I will send you the PCB.  If you need any of the components on the board let me know as I will be making another order soon and will include them.  This is just a hobby for me and I appreciate the help - sharing parts is of minimal cost, so hopefully it is partial repayment for your time and interest.

  • in reply to fmilburn

    Frank,

     

    Projects have a way of taking on a life of their own and growing without bounds, so I think it is wise to keep it simple.  Having said that, if you were to want to add some features a simpler microcontroller (MSP432, Atmel, ATtiny or ATmega) would be a good choice.  If you need help on this sort of extension, I would be glad to help out on something like that.

     

    This has been a great project and you have accomplished a lot.  Well done!

     

    Gene

  • in reply to genebren

    Thanks Gene!  I almost certainly will need help :-)

  • in reply to fmilburn

    Hi Frank,

     

    I've tried placing an extra capacitance on the inputs, close to the meter, but it seems to not cause instability : ( I've been unable to reproduce the symptom in this way. I tried 100pF, and also 1nF. With that capacitance plus the test leads, it was still possible to successfully measure resistances. I've not seen the instability today : (

    Regarding the dual ranges, personally I think that's currently an awesome feature of your project, since typical handheld multimeter 2-wire resistance measurements still are not great at 4 ohm.

  • in reply to shabaz

    The instability is a bit frustrating.  There are long periods where I don't see it and I did not see it yesterday or today.  But when I do see it, it has always gone away by simply removing a Kelvin probe and replacing it.  I don't remember ever seeing it above say 10 milliohms.  Need to start keeping a record of when it occurs.

Comment
  • in reply to shabaz

    The instability is a bit frustrating.  There are long periods where I don't see it and I did not see it yesterday or today.  But when I do see it, it has always gone away by simply removing a Kelvin probe and replacing it.  I don't remember ever seeing it above say 10 milliohms.  Need to start keeping a record of when it occurs.

Children
  • in reply to fmilburn

    Hi Frank,

    Hehe yeah it's an annoying thing in the system.

    I'm going to read the entire amp datasheet today, I've not done that so far, and look with a 'scope, in case anything can be spotted. It was surprising that adding more capacitance at the input didn't cause instability, because the document had called that out. Maybe it's something like needing a capacitor across the gain resistor, if there is some high-frequency noise that is being amplified or something (just speculating).

  • in reply to shabaz

    Perhaps have a look at what the LM334 does when you apply the clips. It is a step change for the part and it might ring quite badly as it scrabbles to get the current down to and then aligned with the set level.

     

    The circuit in the datasheet for adding an external transistor shows this

     

    image

     

    so they don't expect it to be perfectly stable servo-ing the transistor without modifying the compensation a bit.

     

    If you have a small amount of lead inductance too [in the test clip leads], then it all becomes quite complicated.

     

    Mind you, I don't see why that should cause the amplifier problems it couldn't recover from if it all stays within the common-mode range of the inputs, so it's probably not the answer you're looking for.

  • in reply to jc2048

    Hi Jon,

     

    Thanks for this, I'd missed seeing this in the datasheet! I'll check around this area too.

  • in reply to jc2048

    I tried several things this afternoon.  First I examined whether the instability is correlated with the resistance being measured during 100 readings:

    • Resistance = 1 ohm:  No instability over more than 100 readings
    • Resistance = 24.7 milliohm:  No instability over more than 100 readings
    • Resistance = 13.2 milliohm:  Instability on average every 9 readings but varying from 2 to 28 readings during 100 readings

    This at least partially explains why sometimes I see instability and sometimes I don't.

     

    I probed around U3 (instrument amp) but I am not experienced at this and can offer no insight on the problem yet.

     

    I probed around U1 (10 mA current source) and see a tremendous amount of bounce when the probe is connected.  I can't say but if anything, when there is reduced bounce it seems more likely to be unstable but it may be my imagination. For example, here is a stable capture on pin 2 of U1 when the probes are first placed.

    image

    Here is an unstable one on pin 2.

    image

    If this seems like a useful thing to explore further I can make more captures.

  • in reply to fmilburn

    Another thing to try. With the Kelvin probes, the two sides are deliberately separated and only connect when the clip is closed. So, when you apply the clip, you may be connecting to the current source first and then the amplifier, or vice versa. So how about trying that in a deliberate manner, with the low-value test resistor, and see if it ties up with the instability in any way?

     

    Further thing to try - connect a diode (1N4148 or 1N914) from R9 to the emitter of the transistor so that the current source always has a load and there is always a path for the current to flow. When your test resistance connects, the diode will turn off and the current will take the path through the test resistance instead. The change at the start will then be less abrupt and easier for the source to deal with.