PCB for a Kelvin (4-Wire) Milliohm Meter

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

fmilburn — Thu, 09 Jul 2020 23:24:34 GMT

That looks great! I am without a label printer currently as the old one broke years ago - did not know there was transparent tape. Nice that it works over the LED. I worked on the new PCB a bit last night but did not get very far. Still planning to get something done in the next week or so....

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

shabaz — Sat, 27 Jun 2020 00:43:10 GMT

Meter mod:

The meter has two white connectors. One connector is for the meter supply voltage to power it up. The other connector is for the measurement input to the meter.

This modification is required on that measurement related connector.

Using a pair of pliers or cutters, carefully pull off the white socket as shown in the photo below, which is the meter measurement input connector.

It will slide off and leave the two pins remaining.

You can desolder the pins now quite easily, without needing to remove the PCB from the meter enclosure.

Next, with a knife/scalpel, cut the track indicated with a purple circle in the photo below.

[View:/resized-image/__size/620x465/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/5280.contentimage_5F00_188968.jpg:620:465]

Solder two wires to act as the meter input in lieu of the previous connector. The positive connection will be the corner pin, same as it was with the connector.

However the negative connection is to be soldered to the resistor shown in the photo below. By doing this, the input to the meter module becomes a differential input.

[View:/resized-image/__size/620x827/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/3146.contentimage_5F00_188969.jpg:620:827]

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

shabaz — Fri, 26 Jun 2020 23:53:16 GMT

It was quicker to just photograph the mods:

[View:/resized-image/__size/956x664/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/4213.contentimage_5F00_188967.jpg:956:664]

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

shabaz — Fri, 26 Jun 2020 23:37:51 GMT

Hi Frank,

Good news, tonight I've built the V1.3 board, with a few minor component value changes, and it's working pretty much spot-on, with no adjustment. I still need to do more tests, but I measured a resistance wire which I calculated to be 608 mohm based on its dimensions, and it measured 606 mohm with the meter.

This time around I deliberately did not use any high accuracy component, all resistors were 1% thick film. I had to adjust a few values due to the components I had, but nothing was measured and calibrated out, so that I've got a very typical board if someone else were to solder with standard components too.

Also, it measures all the way down to 0.0000 ohms now (I tested with a gold-plated zero-ohm link, it measured 0.0001 ohm) with a slight circuit mod to the display module. I'll write up the mods in the next day or two. Currently I am only using the 0-4 ohm range, because I didn't have a trimmer pot to install on the PCB for the 0-40 ohm range. However since it's working well without any trimming, I might just stick in a fixed resistor for now so I can use that range too.

And the instability is fixed.. I think it was being caused by the long wires, being inductances, causing the constant current source to oscillate. Putting a resistor (10k is fine) between the two high side input connections, and another between the two low side inputs, made the issue completely disappear. I've not seen it once after doing the mod, yet if I remove the resistors, I can make the problem appear every few readings with the length of wiring I happened to have. So, in summary, I think the design is reliably providing measurements, and at least if the room temperature doesn't change much then it looks like no accurate parts are required, beyond 1% normal thick film, and for those who want to, they can use the trimmer to calibrate with a known resistance. More tests would have to be done to confirm the effects of temperature change, but since a significant part of that is automatically compensated, I'm hoping it's a non-issue for reasonable lab temperature differences experienced over time.

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

andbro — Fri, 12 Jun 2020 13:27:30 GMT

Hi,

Does this project is aborted?

Where we can find the latest schematic and Gerber files?

Thank

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

clem57 — Fri, 23 Aug 2019 10:49:52 GMT

Thanks [mention:cff82ed12b714001a1ba5205f2cb553c:e9ed411860ed4f2ba0265705b8793d05] for a well document process.[View:/resized-image/__size/16x16/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/contentimage_5F00_930.png:16:16] Maybe you could run a poll and see how many folks have interest. If there is enough, have folks sign on and share the costs of the board if you think appropriate.[View:/resized-image/__size/16x16/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/contentimage_5F00_3.png:16:16]

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

fmilburn — Thu, 17 Jan 2019 05:06:38 GMT

I have done a number of tests in the last day or so and have not made much headway. In order to easily change out resistance, capacitance, and diodes in front of the instrument amplifier flying leads were made to a breadboard as seen in the photo below (no added passives in the photo):

[View:/resized-image/__size/620x430/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/3051.contentimage_5F00_188960.jpg:620:430]

That made it easy to try out different passives but adds unknown capacitance which gave different response to what I was getting before. I never found a stable combination. Even so, here are a couple of new observations:

Addition of resistance of 10K in front of diodes / capacitors perceptively slowed down the equilibrium readings. Both 10K and 1K gave poorer stability results than what was seen with no resistance although I did not do enough tests to quantify. I did not test resistors after the capacitors / diodes.
The breadboard must be adding unwanted capacitance. The Schottky diode placement that worked for me the other day did not work today.
There was an interesting behavior in one configuration where the instability would occur if the probe were quickly opened and closed but did not occur if the probe was held open a while before closing (presumably allowing the capacitor to charge or discharge). This probably occurred before and wasn't noticed.

I am stuck for the moment but need to return to my LoRa project for a day or two so I can complete it before the deadline. The schematic has been modified but I will hold out posting it until working values can be determined.

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

fmilburn — Tue, 15 Jan 2019 05:18:26 GMT

Summary: Good news - it looks like everything could be working now - long post below but I wanted to document it.

I have probed around the inputs to the instrument amp with the oscilloscope and the difference when an instability occurs is now clearer. But first a quick summary:

Using a voltage divider to set VREF a few millivolts above ground appears to solve the problem with very low resistances, but it does not solve the instability problem. The circuit used in the oscilloscope captures does not have this modification as it was made on a different board.
Using IN5819 Schottky diodes as shown in the sketch below solves many of the instability problems but there is still one left when the positive probe is closed last and the sense side of the probe is kept in contact (i.e. the positive current side closes last).
An IN4148 diode connected as shown on the far left in the sketch below does not solve the instability problem, but doesn't seem to hurt it either. It was confirmed that the diode keeps the current flowing at 10 mA.

[View:/resized-image/__size/620x815/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/7774.contentimage_5F00_188949.jpg:620:815]

The probes in the following oscilloscope screen shots were connected as follows:

channel 1 (Yellow) connected to inverting / - input of the instrument amp
channel 2 (Blue) connected to non-inverting / + input of the instrument amp

WORKING

The following is with the project working properly and measuring ~6 milliohms. No trigger....

[View:/resized-image/__size/480x234/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/5621.contentimage_5F00_188950.bmp:480:234]

The legend at the bottom tells us that the major voltage divisions are 20 mV for both probes and that the major time divisions are 5 us. The number on the far right is the trigger setting. There is no voltage offset - they are superimposed to show the very small difference in signal and the noise.

In the following screenshot the positive probe is closed last and the apparatus works properly. There is a voltage offset of 200 mV in the oscilloscope between the two probes and channel 1 (inverting) is triggering. Vertical and horizontal scales have been changed form the last capture. The non-inverting input touches first, bounces and then settles in an exponential fashion without undershoot. The inverting probe touches later, bounces for a longer period and after a while finally connects and follows an exponential rise without overshoot.

[View:/resized-image/__size/480x234/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/6201.contentimage_5F00_188951.bmp:480:234]

Here is another one where the positive probe is closed last and the apparatus works. This time the spikes go in opposite direction.

[View:/resized-image/__size/480x234/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/3157.contentimage_5F00_188952.bmp:480:234]

If the positive probe is closed last, and the current side is kept firmly in contact the following occurs and everything works:

[View:/resized-image/__size/480x234/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/4628.contentimage_5F00_188953.bmp:480:234]

NOT WORKING

And this is what it looks like when it fails. The wire being tested is held tightly against the sensing side of the positive probe while it is closed. Both inverting and non-inverting inputs rise together quickly and fall together in an exponential fashion.

[View:/resized-image/__size/480x234/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/7271.contentimage_5F00_188954.bmp:480:234]

Below is a similar event where the voltage offset is removed.

[View:/resized-image/__size/480x234/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/8741.contentimage_5F00_188955.bmp:480:234]

The curves are pretty much right on top of each other. In approximately 30 trials it always gave this shape. Since the wire under test is being held tightly against the sensing jaw there is no bounce to speak of.

The failure only occurs with small resistance under test. So there is minimal voltage differential between the two inputs to the instrument amp. The Schottky diodes do something to upset this when the negative probe is closed last or as in the last working test above the current side is in continuous contact. Also, bounce seems to help out.

I hypothesized that Jon is right about capacitance and decided to try different small values of capacitance to the inverting side of the circuit. My thinking was that this would decouple the inputs. It now looks like this:

[View:/resized-image/__size/620x815/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/6607.contentimage_5F00_188956.jpg:620:815]

It works! My range of through hole capacitors is small but it works with 0.47 uF but not with 0.1 uF or smaller. I have not tested it over a broad range yet - just with the roughly 6 milliohm so more experimentation is needed. The shape of the oscilloscope trace has not changed that much to my eye.

Of course the datasheet admonishes me not to do this and I have some concerns.... Per Jon's suggestion above have a go at loading the inputs with additional capacitance and see how far you need to go before the amp oscillates to see how much margin there might be.

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

shabaz — Tue, 08 Jan 2019 18:38:17 GMT

Hi Frank!

I had the chance to explore it today, and I think that instability thing may be related to the inputs going out of range. Basically, when the Kelvin clips are in the open position, mains pickup gets amplified, and you'll see 60Hz on the output (50Hz for me). It disappears once the clips are in the closed position, either with a resistor connected or no resistor connected. However, at some stage I managed to get it into a situation where although the clips were in the closed position with no resistor connected, I was expecting to see about 4.6V on the output (since the input is effectively a very high impedance and out of range), but I didn't. Instead, the output was a 50Hz square wave!

Leaving everything the way it was, I touched one of the input pins at the circuit board end (to not disturb the clips), and immediately the 50Hz signal disappeared and the output went back to the high 4.6V as expected.

I have not been able to replicate it again, but I'll keep trying.

I don't understand why this happened, but perhaps the inputs need some limiting, e.g. a low-voltage TVS. I'm going to look to see what I've got, and add it across the inputs.

Also the datasheet refers to the input being protected against fast signals above supply, but apparently not for slow signals (e.g. mains pickup perhaps).

So, although I don't understand it all yet, I figure it's worth a try experimenting with limiting the inputs, to see what happens.

[View:/resized-image/__size/373x422/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/5861.contentimage_5F00_188942.png:373:422][View:/resized-image/__size/413x375/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/5488.contentimage_5F00_188943.png:413:375]

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

fmilburn — Sun, 06 Jan 2019 05:10:53 GMT

Thanks to those who continue to show interest in the project and contribute! It became a joint project some time ago :-)

[mention:b0bc65b9ecdc4307bd967592f00e340a:e9ed411860ed4f2ba0265705b8793d05]

[mention:4f57fc9d538949ad9eb336ddb9469bb8:e9ed411860ed4f2ba0265705b8793d05]

[mention:d214a0a0f5594ee19515b2a3782e7070:e9ed411860ed4f2ba0265705b8793d05]

[mention:f80b53cee57c44bc9d7c577d07d7c791:e9ed411860ed4f2ba0265705b8793d05]+

Find below Version 1.1 which I propose to revise to 2.0 later this week after review and comments.

[View:/resized-image/__size/620x479/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/5238.contentimage_5F00_188935.jpg:620:479]

A summary of changes in bold based largely on the list from Shabaz above is given below:

1. The trimmer pot component outline seems slightly offset to the actual Bourns pot physical outline - I am using the Bourns PV36W footprint from the KiCad library. But the trimmer I am using is a Vishay which fits the outline well. If this is not a big issue I will leave it as is.

2. Change R9 to be 10 ohm (reduces the volt drop) - done

3. Change C4 to be 1uF (to add a filter) - done

4. Add 100nF decoupling near U4 - done and also added a 10 uF footprint since the datasheet showed it

5. Might be better to space out the pins for the TO-92 package, for the transistors. But no real need, so if there isn't space, this suggestion should be ignored. For the LM334, the spacing is fine - will look at doing this when the PCB is revised

6. It wasn't clear where to obtain power for the panel meter. The connection marked +5V was used, since the panel meter prefers that voltage or higher. But the +5V connection is before the power switch, so if I had used a power switch, then the panel meter would have stayed lit (I didn't solder a power switch) - done, added a power connection for the panel meter after the switch. It is also shown in the Miscellaneous box.

7. Maybe have an option to use a DC power jack instead of USB (i.e. have both footprints perhaps, or just pins for soldering flying leads). For my version, I didn't solder a USB connector, but used a DC connector soldered on flying leads - done. Note that the outer barrel is connected to ground and the pin is power. Let me know if polarity is wrong. It won't make any difference if it is panel mounted with flying leads of course. Also see footprint below and let me know if placement on the PCB other than at the edge where the USB connection is located is important.

[View:/resized-image/__size/620x247/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/4743.contentimage_5F00_188936.jpg:620:247]

8. Shield of USB connector should be connected to 0V perhaps. Would make a nice place to clip a scope or meter 0V connection. - You will have to educate the mechanical engineer ;-) What pin should be connected to for 0V? I have an idea, but how would this be used with a scope or meter Vs. GND?

9. Maybe no need for jumper JP2 and JP3, and alow a SPDT switch to be wired to a new connector at R24 and R25 and ground. - done, but have a look at where I placed it just after LCURRENT. Also, my symbols may be confusing but this is meant to be a DPDT with the amplifier on the panel. I will make it such that it could also be SPDT or Jumper on the PCB.

11. The labels that are on the PCB should also be on the schematic, because it was confusing understanding which ones related to which part of the schematic, e.g. Cal+ and Cal- are not marked on the schematic; they are marked as JP1 on the schematic - done

12. Change the kelvin connection labels to Hcur, Hsense, Lcur, Lsense or similar perhaps, maybe on underside of the PCB - done, will need to remember to add to the PCB

13. Add a 0-ohm resistor in the connection between the op-amp and the comparator, so that it can be desoldered and isolated for testing (i.e. still have the meter connected, but not the comparator circuit) - done

14. Maybe no need for the power LED, since the panel meter will be lit. - will leave it for now unless we run out of room on the PCB. It would be useful if the user elected to use a multimeter instead of a panel meter.

15. Possibly change U4 to LMV7271MF (it is pin-compatible, the footprint can remain the same). I have not tested this yet, I've left U4 currently unpopulated. - done

16. Add 0-ohm resistors in front of amplifier. These might be populated later per Section 4.4.5 in the datasheet if they help reduce the intermittent instability.

17. Possibly enlarge the PCB yet still fit within the Bopla enclosure. Move mounting holes to fit enclosure.

18. Question for Shabaz: Any additional recommendations on spacing other than TO-92 footprint? Were pad size and spacing otherwise OK for other components? What about placement along the edge of the PCB?

Other desired areas for improvement and ideas for this version are:

1. Resolve intermittent instability which drives the output to a low reading

2. The meter reads approximately 1 milliohm when the Kelvin probes are shorted together. There was discussion about improving very low resistance readings in the post by Shabaz and thoughts from Jon. If there are proposals for this, please sketch and I will try to test before finalizing the version 2 schematic. This would be an improvement on the specification which is +/- 1 milliohm down to 1 milliohm.

3. Auto-ranging and other features will not be pursued for this revision

I am sure I left things out and further suggestions are welcome.

Frank

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

shabaz — Sun, 30 Dec 2018 01:19:12 GMT

I started building this tonight : )

I've soldered all the resistors and pots, and then trimmed to the desired resistance.

Hopefully tomorrow I'll get the rest all soldered up, and try powering it up before putting it in the case!

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

makethingstoday — Mon, 17 Dec 2018 03:10:35 GMT

Hey Dude, congrats on first place! Well deserved for an awesome project and write-up!

Any plans to make the PCB files available so we can make one of these for ourselves?

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

shabaz — Wed, 14 Nov 2018 15:43:03 GMT

Hi Frank,

Thanks for the PCB! It looks even better in real life - it is very compact.

It's like it is almost designed for the enclosure. Everything will fit, but it will be cosy! : )

It is almost a palm-sized box - very small at 85x81x40mm.

[View:/resized-image/__size/620x349/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-05d9f427-be3c-4420-b5ef-614b91f3e913/7103.contentimage_5F00_188930.jpg:620:349]

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

genebren — Wed, 07 Nov 2018 00:50:49 GMT

Frank,

If you do decide to update you board and you want to use SMD transistors, I have a bunch of 2N3906/2N3904 devices in SOT23 packages that I would gladly send you ( I bought partial reels of each type on Ebay years ago and I used starting using them recently to build tiny H-bridges).

Gene

RE: PCB for a Kelvin (4-Wire) Milliohm Meter

jc2048 — Wed, 07 Nov 2018 00:07:36 GMT

Just a small observation. You've used an older bipolar op amp as the comparator (the LMV321 is a low-voltage variant of the LM324 style parts). The input is a differential pair of PNP transistors with a current source sitting above them, so the input range doesn't extend up to the positive rail - the maximum (according to the datasheet) is about a volt short, though in practice you'll see it get a bit closer than that. [That almost-a-volt comes from the Vbe of the transistor - perhaps 0.5V at the lowish current the pair run at - plus the drop that's necessary for the source to run properly.] You might say 'why not use NPN transistors?', but then it can't get down to ground on the input and they used to think operating down to ground was more useful.

Anyway, if you have the regulator in circuit, you probably won't get the comparator operating properly above 4V.

There's a further complication with using old op amps as comparators and that is that sometimes they have protection components to stop the two inputs moving too far apart [the reason (I think) is that one way to get the input bias down is to use super-beta transistors on the inputs which have a very high gain but, because you get the gain from having a very narrow base region, have low breakdown voltages]. Unfortunately, those structures were rarely shown on equivalent circuits and seldom hinted at in the figures in the tables. I seem to remember the 324 devices are ok as comparators, but it's not certain the LMV321 would be.

Another observation is that the instrumentation amplifier won't operate all that well with its output close to ground (particularly not once you get within a few mV of GND). Marketing departments are very fond of 'rail-to-rail', but if you think how you would design an input or output to function right up to both rails you'll quickly see that it's a very difficult challenge. If your measuring device doesn't need to be ground referenced, one way to improve things would be to lift the Vref above ground (the meter negative would go to the Vref rather than GND). Effectively, on the output side, that gives you a small negative rail and the output is then operating in an area where it performs reasonably well. [This isn't an area in which I have any real expertise, so I can't tell you the best way to do that, though I could come up with a few possibilities (it doesn't need great precision because everything is relative to the Vref). Probably best to ask it as a question, if you wanted to try it, and have someone who knows what they're talking about answer.] You'd need to modify the comparator a bit to work to the same reference voltage rather than ground.

Forgot to say - nice project and a professional looking board. There's nothing wrong with mixing conventional and SMD, but you might consider using SMD transistors (SOT23 packages are easy to solder).