Group Actions
Engagement
  • Author Author: shabaz
  • Date Created: Date Created
  • Views 6929 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

  • in reply to jc2048

    I tried the experiment with the Kelvin probes.  Using the positive probe, it was possible to get the instability whether the current source was kept connected, or the amplifier side was kept connected.  Data is limited, but the instability was more likely when contact was continuous with the amplifier.  When it is kept in contact with the amplifier then the reading is ~ 1 mA when the jaws are open.  It stays there when closed and unstable.  The resolution on my oscilloscope is quite low in that range and there is a lot of noise - I have not been able to make a single shot capture.

     

    Have not tried the diode idea yet.

  • 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.

  • 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 jc2048

    Hi Jon,

     

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

  • 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.