Zero Ohm DIY 4-wire standard

[VintageNut]

This thread is in response to a PM asking for me to provide pictures of the inside of my Ohm Labs balanced tetrajunction.

Sadly, the Ohm Labs device case is epoxy sealed at the screws. To open the case I would have to remove the epoxy and thus void the warranty.

What I did for a DIY project is to try my hand at making a tetrajunction. Attached is an excerpt of US patent 5867018 showing various configurations of physical manifestations of a balanced tetrajunction.

The final device is mostly figure 2(d).

Details to follow in separate posts in this thread.

 

[VintageNut]

The main portion of the device is a solid copper disk. Diameter is about 2" or 50.8mm. The thickness is about 3/8" or 5 mm.

Four holes are drilled and tapped. The three perimeter holes are 120 degrees apart. The fourth hole is in the center of the disk.

The screws are solid copper tattoo machine screws. 

[quarks]

thanks for sharing, to bad the Ohmlabs Standard cannot be opened

BTW this would fit into to the zlymex posts
https://www.eevblog.com/forum/metrology/teardown-standard-resistors/msg997453/#msg997453

[VintageNut]

The device was measured using a Keithley 2450 to force 1A while a Keithley 147 was used to measure voltage.

The measurements obtained are attached. It must be noted that to connect this device and measure it is an endeavor that requires hours. The instruments must warm up and stabilize. The microvolts offsets created by connecting the device and cables requires quite a long time to settle. The microvolts are everywhere in the current path and are temperature dependent. Touching anything creates a thermal imbalance that must come to ambient equilibbrium.

[doktor pyta]

I see You have a lock-in amplifier. Have You tried to provide AC excitation current (eg. current source chopped by two MOSFFET) and measuring nanovolts using lock-in ? This should provide quick (few minutes) information about the geometrical symmetry. Do You think it is a bad idea?

[quarks]

As I do not have a Nanovoltmeter, I have just made a quick meassurement with my DIY Zero Ohm Standard
and tried the following setup:

Calibrator  10,00000 A (both AC and DC)
DMM 200mV range (zeroed before meassure)
around 0,0005 mV DC
around 0,0095 mV AC/100Hz (not really a stable reading)

I meassured several times to confirm it
If I did get it right my calculation gives 50 nOhm with DC  and about 950 nOhm with AC current

[VintageNut]

I see You have a lock-in amplifier. Have You tried to provide AC excitation current (eg. current source chopped by two MOSFFET) and measuring nanovolts using lock-in ? This should provide quick (few minutes) information about the geometrical symmetry. Do You think it is a bad idea?

I have never thought to try something like this. It will require some research and thought to construct the setup. If you have a proposed schematic and instrument connections that would help.

The lockin amplifier was used to perform some demodulation of phase modulated distortion. Eventually I found that a software defined radio Windows app was more robust than the lockin for my requirement at the time. 

[VintageNut]

As I do not have a Nanovoltmeter, I have just made a quick meassurement with my DIY Zero Ohm Standard
and tried the following setup:

Calibrator  10,00000 A (both AC and DC)
DMM 200mV range (zeroed before meassure)
around 0,0005 mV DC
around 0,0095 mV AC/100Hz (not really a stable reading)

I meassured several times to confirm it
If I did get it right my calculation gives 50 nOhm with DC  and about 950 nOhm with AC current

The problem with using a DMM is that you are using the very bottom of the range. I suspect that the absolute uncertainty of the 200mV measure range of your DMM is larger than your measurement. You cannot rely on a 0.0005 mV reading of a 200.0000 mV range. You want your measurement to be a significant percentage of the measure range. This is why a very old school nanovoltmeter is ideal for nano-ohms measurement.

The KE147 uncertainty is 2% of full scale. For the most sensitive 30nV range, that uncertainty is 0.6nV. The noise is specified at less than 0.6nV RMS.

There is a KE147 on eBay right now for USD$40. That is very cheap. Most likely it will require some repair. I own multiples for future spare parts if the chopper wears out.
http://www.ebay.com/itm/Keithley-Nanovolt-Null-Detector-model-147-/262615923728?hash=item3d2520fc10:g:ZnYAAOSwvzRX0bCY

[quarks]

I only did this as a quick and dirty test with what I have and was not ment to measure absolut values, only to get an idea how close to zero it might be

[VintageNut]

I only did this as a quick and dirty test with what I have and was not ment to measure absolut values, only to get an idea how close to zero it might be

What exactly is the range uncertainty of your DMM for the 200mV range?

The KE2002 has a 200mV range uncertainty of 9ppm. That is 1.8 uV. At 10A of drive current that makes your uncertainty 180 nano-ohms. The uncertainty of the measurement is 3X what you are measuring at 50 nano-ohms.

If your DMM has a greater than 9ppm range uncertainty then your measurement uncertainty is greater than 180 nano-ohms.

[HighVoltage]

Very interesting thread and too bad you can not open one up.

Where die you buy this Ohm-Labs standard? I might buy one for a reasonable price and open it up.
Are there other companies that make a similar zero ohm standard?

[quarks]

My DMM is much better than K2002, but as I only checked the relative change after zeroing it, it does not matter.

[VintageNut]

Very interesting thread and too bad you can not open one up.

Where die you buy this Ohm-Labs standard? I might buy one for a reasonable price and open it up.
Are there other companies that make a similar zero ohm standard?

I bought it directly from Ohm Labs. They build them one-at-a-time to order. There is a waiting time which was not a problem for me.

As far as I know, only Ohm Labs makes these. They are used for 4-terminal junctions of resistor transfer standard arrays. They decided to place one into a box and sell it as a zero ohm standard.

[3roomlab]

this reminds me of this thread on DIY plug. but compare to fluke plug, how many micro-ohm is fluke's?
https://www.eevblog.com/forum/testgear/diy-34401a-calibration/

[VintageNut]

That is a good question. I do not think that you can measure one of these with a DMM because these are used to calibrate zero of a DMM.

On my DMM7510, the Keithley Cal Short measures about 1/2 mico-ohm which is going to be way less than the uncertainty of the measurement. 1/2 micro ohm is 1/2 ppm of the 1 ohm range.

I will put this on my list to measure with the KE147 nanovoltmeter and a SMU. It will require an enclosure to shield electric and magnetic interference as well as thermal interference from the very un-regulated lab temperature. Its on my list to acquire a cast iron pot for measuring low ohms devices.

[3roomlab]

That is a good question. I do not think that you can measure one of these with a DMM because these are used to calibrate zero of a DMM.

On my DMM7510, the Keithley Cal Short measures about 1/2 mico-ohm which is going to be way less than the uncertainty of the measurement. 1/2 micro ohm is 1/2 ppm of the 1 ohm range.

I will put this on my list to measure with the KE147 nanovoltmeter and a SMU. It will require an enclosure to shield electric and magnetic interference as well as thermal interference from the very un-regulated lab temperature. Its on my list to acquire a cast iron pot for measuring low ohms devices.

EMF definitely is a itch in my brain somewhere.
i had been curiously browsing taobao about mu-metal sheets. i am guessing, the higher the flux handling the better the shield? but iron POT = WIN !
https://item.taobao.com/item.htm?spm=a1z0d.6639537.1997196601.67.vHJ5X8&id=539901820269

[lukier]

I wonder how well it would work. Machining mu-metal (bending these sheets, hammering, riveting etc) would disturb the crystalline structure thus you would have to perform the annealing again, see here:

https://en.wikipedia.org/wiki/Mu-metal

Heat treatment in a hydrogen atmosphere is not something I would DIY at home

[acbern]

It does not really make much sense to shield the zero-ohms-standard in mu-metal if the cabling and meter does no have the same shielding.... and they dont, while having the majority of magnetic flux exposure simply because of mechanical dimensions.
sounds a little like audio-voodoo

[VintageNut]

It does not really make much sense to shield the zero-ohms-standard in mu-metal if the cabling and meter does no have the same shielding.... and they dont, while having the majority of magnetic flux exposure simply because of mechanical dimensions.
sounds a little like audio-voodoo

The front-end copper pieces of the KE147 are inside an aluminum enclosure that is also Mu metal shielded.

The first time I used the KE147, the meter needle oscillated. The instrument was sitting on top of the SMU forcing current. Not a good idea. Creating distance between the SMU and the KE147 was the solution.

Magnetic fields hitting the the inside of the instrument will cause beat frequencies.

The KE147 cable is shielded. Doubtful is there is any Mu metal in the cable shielding.

[3roomlab]

It does not really make much sense to shield the zero-ohms-standard in mu-metal if the cabling and meter does no have the same shielding.... and they dont, while having the majority of magnetic flux exposure simply because of mechanical dimensions.
sounds a little like audio-voodoo

i definitely have no way of measuring its induced effects properly, and from what i can see here (the materials presented in the video), even the materials are so varied. i assume only in very specific conditions will it be a very bad problem.

[Vgkid]

I haven't watched the video yet. I recall that steel is one of the most cost effective shielding materials for low frequency shielding. Not as good as copper, but thickness does play into it.

[ap]

MU metal is a solution to shield low frequency magnetic fields where other solutions are not possible (e.g. for audio transformers). The preferred way to do this when cables are involved (and also within the zero ohms standard) is to use twisted cables. The twisting causes noise voltages to be induced in the cable pair with opposite polarity, cancelling each other out. In an amplifier PCB, for obvious reasons, MU-metal is the choice if the complete circuitry should be shielded, as the layout can hardly be equivalent to a fully out-cancelling design (but many nV-meters still dont do that, e.g. the 34420 iirc). Mu-metal shields for cables do exist, but are pretty exotic. It can enhance the effect of cable twisting, especially when high current conductors cause strong magnetic fields to be attenuated.
The zero ohms resistor, btw, as mentioned in the patent, solves the problem to generate zero ohms in applications where the 4 connected signals are physically located in certain positions (e.g. in the SR1010). If one just wants to make a zero ohms resistor with 4 (low emf, if need be) banana posts in a case, an equivalent solution would be to connect a wire to each of the posts, twist them together as pairs, and connect them at one point in an appropriate manner. That way, no voltage can be developped within the sense wires by current in the force wires. No need for a triangular piece of metal or so as in the patent.

[Kleinstein]

For strong magnetic fields steel is good, as it has a saturation, but it is not good at low fields as it stays magnetized. So than at least use electric steel like that used for transformers.

For low field Mu-metal  is the right material. If subject to mechanical stress / strain than some amorphous or nanocrystaline materials can be better, as Mu Metal is sensitive to deformation.

For the zero Ohms shielding is tricky as the magnetic material can also increase the coupling for certain combinations. So it makes some sense if the current and voltage contacts are fixed - but can cause more trouble than good if you want the contacts interchangeable.
The non interchangeable zeros are much easier and can get lower resistance values, as they don't rely only on symmetry but can use aspect ratios too.

To distinguish resistance, induction voltage / inductive coupling and thermal effects one might have to look at the time dependence. So a lockin amplifier might be Ok, but with care to the phase.

[Assafl]

How do you balance the round tetrajunction (how do you decide where to file the piece)?

I ask because of all the shapes, the tetrahedral (2b) seems to me to be the easiest to work out the balance (in theory - mainly since a tetrahedral is a superposition of six Wheatstone bridges). 

[VintageNut]

How do you balance the round tetrajunction (how do you decide where to file the piece)?

I ask because of all the shapes, the tetrahedral (2b) seems to me to be the easiest to work out the balance (in theory - mainly since a tetrahedral is a superposition of six Wheatstone bridges).

Very good question. I do not have an answer. I did not attempt to balance my DIY version. I only wanted to characterize it as built. The intended purpose is to have something better than the PCB plug-in banana "short" that is used to calibrate DMMs to zero ohms. The actual resistance of that "short" is something on the order of 2 micro-ohms (from memory).

If someone wants to explores this filing of the geometric shape, I will assist as I can.

[3roomlab]

How do you balance the round tetrajunction (how do you decide where to file the piece)?

I ask because of all the shapes, the tetrahedral (2b) seems to me to be the easiest to work out the balance (in theory - mainly since a tetrahedral is a superposition of six Wheatstone bridges).

what you said gave me an interesting idea. the "balance" then turns out to be a physical puzzle to be solved, by which all 4 points must be connected with the same "resistance" no matter which point it intend to "go" to, and it can only be done by emphasizing/enforcing the "diagonal" length. then the perimeter length have to be "bent" but still carry the same length, see pic. do i make sense? now then, all 6 connections/paths have the same "resistance" and "distance". i hope this idea is not 1 of the patents anywhere?

ignore the scale in mm, then expand the diagonal/bends to achieve the 0.75 inch 4mm banana port?
BUT, now that this is all in equilateral "distance/resistance" WRT banana port wise, the insides of the DMM connection wires are not, or are they ? oh dear !

[VintageNut]

I do not follow what you are explaining. Are you using the triangle shaped device as an example?

There are three unique configurations to force current and measure voltage.

Call them configurations 1,2 and 3. You characterize each configuration. Then you sort the resistances from low to high. The highest resistance is the worst one and is the limiting factor. You cannot make the highest resistance lower. You can only make the lower resistance configurations higher. You adjust the lower two configurations to be the same as the highest one.

This is my educated guess for the triangle shaped device. For the round device, I am not certain if filing one of the 120 degree arcs of the circle makes the resistance lower or higher for that arc.

The Ohm Labs device in my office/lab is specified at no more than 50 nano-ohms. The measured value is 4 nano-ohms. So, when you are adjusting, you are changing the value by a very tiny and very difficult to measure amount.

[3roomlab]

im visualizing it as point to point (the 4 points are imaginery ABCD spots). i think it this way, is because i dont understand the triangular theory behind it. i assume in my own way the 4 points to be connected need to see each other as zero ohms (or shortest path?)  in which ever path they need to take, and each point has only 3 ways to reach the other 3 points. in my mind it is as if, all the 4 points need to exist in the same spot (= travel zero distance to reach each other), which is impossible.

this reminds me of ohms per square. odd corner to even corner resistance = diagonal tip to tip resistance?

[VintageNut]

I think of the device this way.

The 4 points are labeled 1,2,3,4.

Let's start by forcing current through points 1-2 and measuring voltage across 3-4.

If you draw a straight line from 1 to 2 and another straight line from 3 to 4 you will see that the lines are perpendicular to each other and the lines do not intersect. This is how very little voltage is induced in line segment 3-4 when a current is forced through 1-2.

The other unique combinations are 1-3, 2-4 and 1-4, 2-3.

Perform the exercise of drawing the current path and you will see that the measured voltage is in an area where no electron should be flowing.

[3roomlab]

hmmm yes i think i get it

in this pic, again i see the "ohm/square" inside, and each corner connected in a uniform way, and the other pair not "seeing" the induced voltage.

in my case, i think i have flatten the 4 point triangle into 6 wires, so the wire frame seem to have a "bypass" for 3-4 not being able to see induced 1-2 voltage (or rather sitting in the middle with "no" voltage). being a wireframe, does it subject itself to being sensitive to the perpendicular induced current? since 3-4 are equal path/distance and "perpendicular" in arrangement, a wireframe should be similar like a solid copper piece? (but with much higher resistance)

[VintageNut]

The device is excited at the vertices. Since the two line segments share no copper, the only method to induce a voltage into line segment 3-4 by forcing a current into line segment 1-2 is via magnet flux lines around line segment 1-2 cutting line segment 3-4. The beauty of this geometric arrangement is that there is only one tiny flux line of 1-2 that cuts through 3-4 and the flux line is not in a direction that will induce a voltage between the points 1-2. Rather, the flux line will cause a voltage to appear across the thickness of line segment 3-4. Quite a clever device indeed.

I am not sure what your drawn ohm-square represents.

If are using segments 1-2 and 3-4, cut away all of the material except the two line segments. Then start adding material nearby to these two line segments. Mostly likely, the only electrons flowing are very near to these two line segments. So, the material very far away from these line segments probably is not used for this configuration.

I do not think that you can analyze this device all-together. You have to analyze it as three distinctly different  devices, each of which must be evaluated by itself.

[3roomlab]

ohms/square? im confusing myself terribly about how to calculate the resistance 

so unlike the tetrahedral, the flatten version forces 2 points to appear to be in the mid point of the other 2 if current passes.
although worlds apart, i think it may have some advantage over the normal short plug.

how does the keithley short plug achieve 0.5u-ohm? that trace is small, could it be 5000u-ohm? a 220mil x 1060mil x 1 oz @ 35C = 2500 u-ohm according to saturn PCB calculator.

[Assafl]

In the tetrahedral scenario what you basically have is a Wheatstone bridge that has 4 resistors, and a resistor across the Up/Down (UD) and another resistor across left/right (LR). The shape matters little.

As usual, if all bridge resistors (ignore the top/down TD and left/right LR resistors) are equal - and you apply a voltage across the top/down there will not be a voltage across the LR resistor since the bridge is balanced.

Measure all options 12, 13, 23... Figure out which has the lowest -imbalance. The Top/Down resistor is the culprit for the imbalance in the other options. Fix it, redo the testing.

I don't believe that mOhms to nanoOhms can be done by filing the junction mOhms yes. uOhms - Ehhh. nOhms? No way...

Maybe they drill fine holes, tap them, and insert a screw into the tetrajunction. By adjusting the screw "depth" the resistivity will change ever so slightly. The smaller the screw, the finer the resolution... Then goop it as a fancy "do not touch" warning... 

Or parallel a few gigohm resistors in parallel to your thick tetrajunction arms.
(that would be amazing - a 0ohm in parallel with a glass embedded mega to multi GOhm resistor - talk about wasting money...).

Hmmmm - this process should also work for your tetrajunction.

So let's theorize a process:

1. Figure out which pair has the highest imbalance (say force 1-2 measure 3-4 results in the highest NV reading).

2. So your highest imbalance is either 1-3 2-3 1-4 or 2-4. (the imbalance is always in the bridge - not the resistors across TD or LR).

3. You can deduce the pair by the polarity. So if the 3 is positive (to 4) the offending "resistors" are either 1-3 or 2-4. Feed 1-3 (and measure 2-4) and Feed 2-4 (and measure 1-3). Whomever has the lowest imbalance - assuming feeding 1-3 results in a lower voltage across 2-4 - means resistor 1-3 needs to be lowered somewhat. Solder a resistor across 1-3. Repeat.

(Note: I am a bit on the fence about how to relate the imbalance nano volts to gigohm resistors, though.... Should be simple wheatstone bridge analysis, but having a superposition of 6 of these makes my head spin. The symmetry reeks of a simpler way to solve the equations...)

4. Perhaps one doesn't need a trick. Solder a resistor across 1-3. Measure 1-3 and 2-4 again. Decrease the value of the resistor until measurements 1-3 and 2-4 balance out. Repeat the entire exercise. Do this until all 4 balance out.

Note: The resistors need to be soldered where the wires connect to the tetrajunction. Not to the tetrajunction itself, though, as heating it will change its properties. A solder lug under the screw perhaps?

[Assafl]

ohms/square? im confusing myself terribly about how to calculate the resistance 

so unlike the tetrahedral, the flatten version forces 2 points to appear to be in the mid point of the other 2 if current passes.
although worlds apart, i think it may have some advantage over the normal short plug.

how does the keithley short plug achieve 0.5u-ohm? that trace is small, could it be 5000u-ohm? a 220mil x 1060mil x 1 oz @ 35C = 2500 u-ohm according to saturn PCB calculator.

By the nature of the symmetry all the shapes in the diagrams are the same "shorted bridge" as the tetrahedral. If they were not - the voltages would not zero out. However, the mechanics of achieving the symmetry change between the different geometrical shapes.

I like it. It can make a pleasant party quiz entry: What has a resistance for 2 wire measurements - but a real zero for 4 wire measurements? A superconductor? Nope - it is zero for both. Yup - a tetrajunction!

[VintageNut]

In the tetrahedral scenario what you basically have is a Wheatstone bridge that has 4 resistors, and a resistor across the Up/Down (UD) and another resistor across left/right (LR). The shape matters little.

As usual, if all bridge resistors (ignore the top/down TD and left/right LR resistors) are equal - and you apply a voltage across the top/down there will not be a voltage across the LR resistor since the bridge is balanced.

Measure all options 12, 13, 23... Figure out which has the lowest -imbalance. The Top/Down resistor is the culprit for the imbalance in the other options. Fix it, redo the testing.

I don't believe that mOhms to nanoOhms can be done by filing the junction mOhms yes. uOhms - Ehhh. nOhms? No way...

Maybe they drill fine holes, tap them, and insert a screw into the tetrajunction. By adjusting the screw "depth" the resistivity will change ever so slightly. The smaller the screw, the finer the resolution... Then goop it as a fancy "do not touch" warning... 

Or parallel a few gigohm resistors in parallel to your thick tetrajunction arms.
(that would be amazing - a 0ohm in parallel with a glass embedded mega to multi GOhm resistor - talk about wasting money...).

Hmmmm - this process should also work for your tetrajunction.

So let's theorize a process:

1. Figure out which pair has the highest imbalance (say force 1-2 measure 3-4 results in the highest NV reading).

2. So your highest imbalance is either 1-3 2-3 1-4 or 2-4. (the imbalance is always in the bridge - not the resistors across TD or LR).

3. You can deduce the pair by the polarity. So if the 3 is positive (to 4) the offending "resistors" are either 1-3 or 2-4. Feed 1-3 (and measure 2-4) and Feed 2-4 (and measure 1-3). Whomever has the lowest imbalance - assuming feeding 1-3 results in a lower voltage across 2-4 - means resistor 1-3 needs to be lowered somewhat. Solder a resistor across 1-3. Repeat.

(Note: I am a bit on the fence about how to relate the imbalance nano volts to gigohm resistors, though.... Should be simple wheatstone bridge analysis, but having a superposition of 6 of these makes my head spin. The symmetry reeks of a simpler way to solve the equations...)

4. Perhaps one doesn't need a trick. Solder a resistor across 1-3. Measure 1-3 and 2-4 again. Decrease the value of the resistor until measurements 1-3 and 2-4 balance out. Repeat the entire exercise. Do this until all 4 balance out.

Note: The resistors need to be soldered where the wires connect to the tetrajunction. Not to the tetrajunction itself, though, as heating it will change its properties. A solder lug under the screw perhaps?

In my opinion you cannot look the device as a resistance bridge. You have to look at as a collection junctions. Not semiconductior junctions but conductor junctions. The resistance of the junction is not zero but the voltage across the junction is nearly zero when current normal to the junction is forced.

These things are used to connect arrays of resistors together without adding significant resistance to the individual resistors.

I spoke with the owner of the company and yes a file is used to modify the contour of the device to change the behavior of one of the configurations. I discussed the device and how it is manufactured and tested when I was pacing the order. It is tested at the factory by forcing 5A and measuring the nanovots drop.My 4 nano-ohm device develops 20 nV when 5A is forced through it. They manually reverse the power supply wiring to achieve an offset-compensated ohms measurement.

[Assafl]

Is there a lack of equivalency between all the forms of tetrajunctions in the diagrams as far as DC network analysis is concerned? If so why would that be?

To me they are all a blob with 4 contacts that happen to be perfectly symmetrical (perhaps with the exception of 2a).

A tetrahedral clearly looks like a shorted bridge. The zero ohms stems not from any attribute of the copper, but from the perfectness of the symmetry.

I wonder how much filing they need to do to get to 20nV...

[VintageNut]

A network is a flat 2-D arrangement of elements that each have a well-known V-I behavior. When a current flows through an element, a corresponding voltage is developed across that element according to its V-I behavior.

In the tetrajunction, the current flows through element 1-2 while the voltage is measured across element 3-4.

Its more like a transadmittance amplifier than a resistor network. Actually three different transadmittance amplifiers in a 4-terminal package.

[Assafl]

I think it it transimpedance if your ration is voltage output to current input.

The same can be said of a current driven Wheatstone bridge. Now you could argue that most Wheatstone bridges are voltage driven - true - but one could use Thevenin's theorem and substitute the current source with a voltage and resistor...

Be it lumped or not - all Thevenin requires is that it is a linear circuit. 

[VintageNut]

Sorry for incorrect term. You are correct. Transimpedance.

For voltage drive, the manufacturer of the commercial tetrajunction may be using a power supply and relying on the current limit for a constant current.

Since I own some SMUs, I prefer to use a SMU.

Here is the difficulty that I see trying to model this device using a bridge, etc. The device has 4-terminals while it has 6 distinctly unique current paths. If the terminals are labeled 1, 2, 3, 4, the unique current paths are

1-2
1-3
1-4
2-3
3-4
2-4

So, you could model as a bridge and add two resistors where the voltage source and galvanometer would normally go. That would probably be a good starting point.

As for finding the offending place to file material away by reasoning, that is giving me a headache. If I were going to try to modify my DIY device to balance it, I would construct more than one and experiment to find a reproducible procedure to balance the device.

This device is extremely difficult to measure. The error sources of induced AC and DC microvolts  are over 100X larger than the voltage being measured. The thermal drifts caused by touching the device and the wires to connect it require hours to settle out. Not seconds. Not minutes. Hours.

So, You decide to file some material off. You disconnect the circuit, take the device in hand and file some material. You reconnect the device and try to measure it. Thermal equilibrium is hours away and it will probably be the next work day before the device is stable enough to measure.

You cannot use any modern digital DMM style equipment to measure this voltage. Modern nanovoltmeters cannot. The only modern commercial instrument that can measure this voltage is the obsolete Keithley 1801 married to a 2001 or 2002 DMM.

The clever thing to do would be to construct a mounting apparatus that would allow filing the device without touching it. The mounting apparatus would have to be enclosed in an electrically and magnetically shielded enclosure that is able to be easily opened and closed for repeated adjustment of the device. Another low ohms project!

[Conrad Hoffman]

Long ago the late Bob Pease described some sort of resistive paper that could be cut out and used to analyze this sort of problem. Voltages could be applied, then the paper probed with a conventional meter. I suppose one could also stamp the shape out of resistive foil, manganin or something, and probe it. Can you tell math isn't my best subject?

[VintageNut]

Long ago the late Bob Pease described some sort of resistive paper that could be cut out and used to analyze this sort of problem. Voltages could be applied, then the paper probed with a conventional meter. I suppose one could also stamp the shape out of resistive foil, manganin or something, and probe it. Can you tell math isn't my best subject?

Interesting. It is worth considering.

[guenthert]

ohms/square? im confusing myself terribly about how to calculate the resistance 

so unlike the tetrahedral, the flatten version forces 2 points to appear to be in the mid point of the other 2 if current passes.
although worlds apart, i think it may have some advantage over the normal short plug.

how does the keithley short plug achieve 0.5u-ohm? that trace is small, could it be 5000u-ohm? a 220mil x 1060mil x 1 oz @ 35C = 2500 u-ohm according to saturn PCB calculator.

By the nature of the symmetry all the shapes in the diagrams are the same "shorted bridge" as the tetrahedral. If they were not - the voltages would not zero out. However, the mechanics of achieving the symmetry change between the different geometrical shapes.

I like it. It can make a pleasant party quiz entry: What has a resistance for 2 wire measurements - but a real zero for 4 wire measurements? A superconductor? Nope - it is zero for both. Yup - a tetrajunction!

You guys are pulling my leg, aren't you?  *Again* (https://www.eevblog.com/forum/metrology/teardown-standard-resistors/msg999854/#msg999854) 

If you want a four terminal device with current going into one pair and no voltage across the other pair, there would be a much simpler solution, no?  What in the world is the use case of this (other then fooling the uninitiated?).  Hey, it's not even April 1st. 

(edited due to erroneous quoting)

[3roomlab]

https://www.eevblog.com/forum/metrology/teardown-standard-resistors/msg911075/#msg911075

oh gawd ! zlymex's tetrajunction !
and post #73 by quarks. "TRANS-RESISTANCE"

[VintageNut]

The ESI document states that the junction they use are less than 100 nano-ohms. Zlymex shared his DIY version that is something less than 50 nano ohms. The commercial version that I have from Ohm Labs is no greater than 4 nano ohms. My first attempt at the circular version with no adjustments yielded 20 nano-ohms or less in any orientation.

Having a geometry that is adjustable seems to make a big difference if you are trying to get as close to zero as possible.

[Assafl]

You guys are pulling my leg, aren't you?  *Again* (https://www.eevblog.com/forum/metrology/teardown-standard-resistors/msg999854/#msg999854) 

If you want a four terminal device with current going into one pair and no voltage across the other pair, there would be a much simpler solution, no?  What in the world is the use case of this (other then fooling the uninitiated?).  Hey, it's not even April 1st. 

(edited due to erroneous quoting)

Hmm. Quite true. If you want zero volts there are other ways to get zero volts... And my guess is if the bias current into the D/A was indeed zero, those would indeed represent a zero ohm resistor.

My guess as to why this is a "better" approach is that since the bias current is not zero, how you take the entire current from the kelvin force connection, allow the bias currents as they would apply to the sense terminal, and yet maintain a true zero ohm resistor? (so that errors due to the bias current are indeed fully nulled). In this case it will be a perfect zero ohm resistor.

I say "better" in quotes as it will be large and heavy, and have large thermal EMFs if not hooked perfectly and allowed to stabilize for hours. It is also very tedious to calibrate. And it needs longer cables than a 4w short. So one has to be careful when using.

Just as to how to much of a difference does Fluke's 4w short approximate the same thing (putting a hairline short between the shorted force and shorted sense terminals) - I don't know - I always wondered if Fluke's positioning of the hairline short was random or intentional (about a 2/3 to 3/4 of the way down).

For me the beauty is of a mathematical nature and lies in that fact that a geometrical exercise can find a virtual zero ohm resistor that doesn't really exist. I really dig that - even if it happens to serve a none existing issue. 

[Assafl]

As for finding the offending place to file material away by reasoning, that is giving me a headache. If I were going to try to modify my DIY device to balance it, I would construct more than one and experiment to find a reproducible procedure to balance the device.

Find the culprit "most to blame" leg of the bridge. I don't know if the process I postulated above will work but I see no reason it won't.

But first try to see if ultra high resistors can balance it (on the 4 wire zero side - not the power supply or nano voltmeter side).

Then try to balance it by using exceptionally high value resistors screwed to the binding posts of the short. Don't touch the short.

Once you are able to balance it that way - the question becomes how much drilling is a 100M vs 1G worth. Oh - and where to drill/ scratch/scuff....

This device is extremely difficult to measure. The error sources of induced AC and DC microvolts  are over 100X larger than the voltage being measured. The thermal drifts caused by touching the device and the wires to connect it require hours to settle out. Not seconds. Not minutes. Hours.

Maybe hook it to a bunch of reed relays so you could rotate the bridge without touching it...

The clever thing to do would be to construct a mounting apparatus that would allow filing the device without touching it. The mounting apparatus would have to be enclosed in an electrically and magnetically shielded enclosure that is able to be easily opened and closed for repeated adjustment of the device. Another low ohms project!

In a box with binding posts....

I applaud you trying to make this work. I neither have the equipment, nor the patience to do something like this.

Who needs Smith charts when one can have so much fun with DC?

[VintageNut]

Are you suggesting placing a resistor across the current force pair or the voltage measure pair? I do not see how you could reduce the apparent resistance because the device is not a resistor. It is more like anti-resistor. Where the voltage is measured, no current is flowing. The current is flowing in a segment normal to where the voltage is measured.  If you try to place anything in parallel, you will probably induce microvolt errors.

i definitely like the reed relay idea. I have a KE7001 and a matrix card that could be used for this purpose. What I don't have is the cast iron pot and the design of a mechanism to hold the tetrajunction.

After constructing a DIY version, it makes me appreciate the commercially produced device. At $400 it was a bargain to have something mounted in a box with high quality binding posts that was 5X better than my DIY device. They do not stock these. They make them one at a time to order. I had to wait something like a month before they shipped it.

[Conrad Hoffman]

You can download a free and surprisingly capable FEA program called FEMM. I usually use it for magnetics but it will also do current flow problems. I've never done it before, so may have some horrible error, but here's the result setting the upper hole at 1V and the lower left hole at 0V. You can see the other two holes do come close to equal potentials.

[Assafl]

Such a beautiful FEA plot. Just like network DC analysis would have us predict. An infinite array of copper resistors.

No skin effect (to the horror of the audiophiles). No Leyden jars...

Can you imagine what this plot would look like at Ghz frequencies?

[3roomlab]

You can download a free and surprisingly capable FEA program called FEMM.

wow this is a fun simulator
the slightly crooked X junction did show a skew in potential diff

*update
question for conrad : in the 2nd pic, is there a way to cross the horizontal segment over the vertical without then forming a new node? (so it mimics pic3).
it appears the program does not do multiple planes but its all fine,

but i guess you peeps can see what i am trying to mimic (a shorting plug with 6 legs of equal traces trying to be a tetra-thingy)
if i can get the FEMM to make the centre float off and not connect?
does my flatten tetra thingy make any sense?

[VintageNut]

This is extremely cool. Nice work! My guess is that no more than microvolts will develop across the driven pair of contacts. The resistance there is probably micro-ohms. That would be extremely useful information as to what exactly is the resistance between vertices of the device. Another exercise!

Something I had not thought of that your FEA points out is that while the potential diff may be 0, the value of each point is somewhere between the two driven vertices. Very interesting info!

You can download a free and surprisingly capable FEA program called FEMM. I usually use it for magnetics but it will also do current flow problems. I've never done it before, so may have some horrible error, but here's the result setting the upper hole at 1V and the lower left hole at 0V. You can see the other two holes do come close to equal potentials.

[Assafl]

What about the sense terminals? Don't they take some (tiny) bias current where the sense terminals are?

[VintageNut]

The sense portion of a 4-wire device consists of a low (hopefully) impedance source and a high impedance voltmeter.

For the tetrajunction, the device sense terminal are some micro-ohms or less and the voltmeter is 1 megohm or greater.

How can adding a parallel resistance help this in any way?

The tetrajunction is not a resistor. It is a 4-terminal device that changes current on two terminals into voltage on the other two terminals. You have to look at it as a trans-resistance amplifier with a gain of something on the order of 10^-9. The only way to improve it is to change the gain. That is the question. How to adjust the gain?

[Conrad Hoffman]

I'm still digesting this, and Thanksgiving's turkey, but it seems the tetrajunction is mathematically elegant, and the choice of terminals doesn't matter, but if all you want is a zero ohm standard for a 4-terminal meter, a cross device would be fine. In fact, it seems like a couple bars, connected by a small wire to the midpoints might even work, IOW, a cross with a thinned section on opposite legs. It's just a matter of making sure the sense lead voltage is identical, not much different than the thought process for a single point ground system in an amplifier.

Also thinking of practical implementation, you probably shouldn't screw terminals to a tetrajunction with ordinary nuts, as the contact points are uncertain. I'd make up some old fashioned machine tool nuts, with a perfectly circular area turned on one end. That way you preserve symmetry of the connection point. Or use some small diameter tight fitting flat washers like split rings without the split. It might be even better to press fit some small copper rods in the holes, and solder to the top of those. It's all about preserving electro-mechanical symmetry.

[3roomlab]

i might convince myself to do a PCB version (triangle version) to replace my shorting plug, no wait ... i DO NOT HAVE A REAL SHORTING PLUG  . so in reality, this could be a zero ohm and a shorting plug as well 

zlymex did it hand soldered, so i figured it shouldnt be worse off with some precise mucking around in eagle? using a "fake" pad as a ruler, adjusting for 1000mil or so, it can be used as "arms" referenced from zero/zero XY to mark the tips of the equilateral triangle, the angle can be set very easy. then fine tuned using 0.2mil grid, actual connection pads can be fitted at the 3 tips and the center. then the entire setup shifted into a PCB "blank frame". the result in mycase is somewhat 999.978mil each way, but i think the accuracy of measurement is subject to +/- 0.2mil, it is the smallest grid i could see.

assuming PCBway, seeed, OSHpark, etc all have 1:1 perfect ratios in XY dimensions, things should turn out quite well i guess? the only variances would be the leads connected onto the pads, should they be all strictly of the same length?

then to take things further, a single unit is not just 1 PCB by itself, but could be a sandwich of 2-10 pieces, @ 0.6mm FR4, 10 pc = total of 20x 1oz copper (6mm++). soldering could be a huge challenge, since there cannot be any thermals to preserve the continuous copper pour.

this is getting interesting, vs an actual expensive professional calibrated tetra-junction, esp for peeps who do not need extensive calibrated zero ohms. short thick lead soldered on with Z-serated 4mm banana = it could plug into DMM for shorting or other banana could plug into it.

is this too good to be true? lol

[SeanB]

Would not solder, you then have another uncertain thermoelectric junction. Rather use a large copper block and machine the junction out of it, and also machine the pins ( with a interference fit in the holes) out of the same block. You turn 90% of the bar into scrap shavings, but it will all be the same composition of Beryllium copper throughout so almost no thermal effects.

Know a toolmaker who started with a 40kg block of brass, and the finished bush out of it was under 1kg. Took him a week to machine that out of it.

[Conrad Hoffman]

How about a round or triangular copper piece with holes drilled "banana sized", then just plug in your low thermal emf plugs? I can't see any advantage between round or triangular.

[3roomlab]

i know machined blocks are great, it be nice to hit those nano numbers, but the interesting bit here after seeing the simulation is that, it may not really need a huge block of copper to get those numbers (maybe more layer? i dont know), and it could be done just by accurate PCB fab. and correction could be done by shifting the leads on the solder pad (if allowance is there)
i think problem of thermal EMF already have many ways to mitigate its effects, mainly by thermal insulation to force parts to reach equilibrium quickly, so i wont worry about that. what is out of reach of many hobbyists is ability to access milling equipment and do precision milling, which in this case is mitigated by just using good PCB layout. it is really killing 2 stones with 1 tetra (a short and a zero ohm). and 1 PCB is about 13 grams, but of course, nobody have anything in tetra PCB format to say it works yet 

update

I can't see any advantage between round or triangular.

i was itching to find out too.
here are 2 similar layout. in both cases, i shifted 1 of the nodes off by 0.02 inch to see how much of the field lines are affected.
im not sure how much of this is real in actual, but it seems by suffering a shift off a main axis, the symmetry will suffer badly. but in my case of the simulation scale (=1uV full scale) badly = 20-30nV or so it seems (99 lines, change of about 2-3 lines). so if we look at zlymex's experiment, and compare with this, with a "fabrication" accuracy better than 0.02 inch, anyone should be able to attain very good accuracy effortlessly in theory by placing the circuit entry points smack on the dot.

so i guess this in theory shows the amount of deviation if the circuit pads are in-accurately placed.
A1-2, vertical shift (WRT +ve)
A3-4, vertical shift(WRT +ve)
A5 and 6 horizontal shift

added a7 = crude representation of normal shorting plug. sense side is out of symmetry by 9 -10 lines (90-100nV per uV)

[Conrad Hoffman]

I suspect the symmetry of the part is all that matters, not the resistance. It could be a block of copper or a thin PCB trace, so long as the connection points are reasonably accurate. Armchair thought (my best kind) just take a piece of copper-clad and use a circle cutter in the drill press to create a circle of copper. Don't even cut all the way through. Now lay out the holes and solder in some wires. The resistance of the solder is higher than the copper so even the joints don't need to be identical. If you can measure it well enough, the copper can be corrected a bit with a Dremel or similar to find those last attoohms!

[Gyro]

Just thinking, do these standards suffer from the Hall effect? I know it's not a very strong phenomenon in copper, but if you're looking down at such low levels then might throw the zero reading off. I don't know if the Earth's magnetic field would be enough though.

[Conrad Hoffman]

Ha! Any kind of DIY Hall Effect demo would be really cool. I tried a quick PCB with the circle cutter on the drill press and learned a couple things. It can work well, but the wire attachment is an issue. The solder joints turn out to be critical, in fact solder can be used to fine tune the thing. That explains the lumps on my quick proto. I got about 5 uV errors in any direction with 5 amps of current. Better, I think, to get everything as accurate as possible and use identical screws and washers. I did learn that the zero of my HP3478A is off by about 0.002 ohms. Fun stuff for a Sunday afternoon.

[3roomlab]

i try to simulate a blob of solder (pink corner). could it be the 5uV is more of PSU noise? than a difference in the tetrajunction?

@ conrad *edit : lets see how accurate is this FEMM. i try to guesstimate how much noise is coming out of your PSU, and estimate = 33mV. is there anyway to see if this value sticks on your power supply setup? the value is obtained from scaling up the difference in symmetry.

[Kleinstein]

Measuring the 4 wire resistance is not really sensitive to the quality of the current source or current source noise. Noise in the current would enter linear to the resistance - so only a problem if you need the resistance to 3 or more significant digits - not if it is to approach zero. However thermal effects at plugs could be a problem. Also induction due to moving cables are a possible problem.

The Hall effect is quite small, I doubt on would notice is. I have once done a test / demo for the hall effect in copper. It was quite hard to measure. I used a 17 µm copper PCB (thus half the normal thickness), about 5-10 A of current and a Nd magnet (though small) directly on top of the PCB. Even with a layout more suitable for hall effect, the voltage was really small, hardly a few µV.

For the zero the PCB is not that well suited anyway, as one usually also wants a low 2 wire resistance and a higher sheet resistance increases the requirements for symmetry and the sensitivity to solder blobs.

Having directly holes in the sheet is also not a good idea, as the contact around may not be uniform. The same is true for threaded connections. A thin layer of solder is hard to beat for a good stable contact. No thermal EMF as long as temperature in uniform. The thick copper helps with this. One needs the thin part (wires) to make sure the current enters well defined, independent for a not so well defined plug.

AFAIK the resistance is symmetric to interchanging current and voltage contacts: thus the same value for current between contacts 1 and 2 and current form 3 to 4. So there are only 3 independent resistances to tune. This may help to find good tuning points and simplifies the measurements.

[Conrad Hoffman]

That all makes sense. I did the PCB version using no precision tools, though I have a full shop, just to see what could be accomplished with simple things like a ruler and scribe. I measured the error using a Kepco supply in constant current mode, and a Fluke 845 null meter. The noise level is low, or at least the meter averages a lot of it out. I briefly applied current and watched meter deflection, then made sure it returned to near zero. Based on the comments, a slice off a copper rod, or heavy sheet, precisely drilled, with press-in studs seems like the most direct way to extreme accuracy, though the PCB version was plenty good enough to check my low precision meter.

[VintageNut]

I tried a cross using ROMEX solid core bare copper wire before I fabricated the solid copper disc version. The cross was no better than mico-ohms. Its similar goodness to a Fluke and Keithley PCB zero plug.

When you measure a good tetrajunction, reverse current and then average the two voltage measurements. This should zero out the micovolt offsets that are stable. Otherwise you are just measuring microvolt circuit junctions that are unavoidable.

I don't think that a PCB version will work if you plan to just plug it into a DMM. No DMM has enough current to drive it...and....a tetrajunction needs to be mounted in a magnetic and electrically shielded box. Normal air circulation in a room will prevent the junction microvolts from becoming stable enough to remove via offset compensated current reversal.

[3roomlab]

comparison of excess in material around the 4 points. more material seem to suggest a spread of the density lines around the sense pin making for more/less allowance in construction error (or maybe im interpreting it in reverse? less interference lines? or more?). so bulk/solid copper = more advantage than thin PCB, but it would also suggest mechanical positioning need to be better than PCB fab accuracy in some way.
i do suppose with solder mask, it is possible to control the uniformity of the solder interference with the copper, but then the thickness of solder = uncontrolled (which could become a form of fine adjustment?)
maybe instead of screw bolts with threads, solid pins put in place can be smooth pins set in by friction fit? like a tapered peg in a fitting hole?
if we are looking at this for what it is, then wouldnt the perfect shape for this "device" be a solid sphere?

[acbern]

Guys, 4 short wires drilled together and soldered/crimped well in a single spot will be superior to any triangle or so. The triangle solves a mechanical problem (e.g. banana plugs required in certain positions such as e.g. with the SR1010). If you just want to have 0 ohms, then there is a easier and better way to do it. But from a ohms-nut perspecitive using all kinds of geometries makes perfect sense of course...

[quarks]

4 short wires drilled together and soldered/crimped well in a single spot will be superior to any triangle...

before I build my "Zero Ohm Standard" I have experimented with a crimped wire solution as described
if I remember correctly the "triangle" showed more than an order of magnitude lower values

[Conrad Hoffman]

Everything depends on what you intend to do with this device. I'm not sure it's been explicitly stated, but a "zero" ohm standard built this way may or may not have low actual resistance. Built with something other than copper, it could be quite high and still work. It just fools a 4-terminal ohm meter into thinking it has zero resistance by putting the sense wires on equi-potential points. It simulates a far lower resistance than you can achieve with real physical materials. If you want actual low resistance yes, plant some stout terminals in a small lump of copper, shape not relevant. The junction is a way to insure that two circuits don't influence each other, but have a common connection, as most meters require. The tetrajunction is elegant in that any terminals can be chosen, but there are various ways to accomplish the same thing. The thicker and lower the resistance of the material used, the less sensitive construction will be. The PCB version is certainly overkill for my 5.5 digit meter that resolves 0.1 milliohm. OTOH, if you've got a more serious low ohms meter, it might not be, nor will it be sufficient for the connections in a low value transfer standard. Building one where the connections lie on perfectly equi-potential points is just a matter of mechanical precision. Press-fitted terminals seem like the best way to go, but other ways will certainly work.

[VintageNut]

In my first post of this thread, the sphere is figure 2(c). I discarded the sphere only because of the practical limitations of my machine shop capabilities. The flat massive disc is easy to acquire and not too difficult to drill. I only broke three drill bits before I resorted to Google to find out how to drill copper. Tiny bits at a time and slow. Nobody sells a massive copper triangle.

comparison of excess in material around the 4 points. more material seem to suggest a spread of the density lines around the sense pin making for more/less allowance in construction error (or maybe im interpreting it in reverse? less interference lines? or more?). so bulk/solid copper = more advantage than thin PCB, but it would also suggest mechanical positioning need to be better than PCB fab accuracy in some way.
i do suppose with solder mask, it is possible to control the uniformity of the solder interference with the copper, but then the thickness of solder = uncontrolled (which could become a form of fine adjustment?)
maybe instead of screw bolts with threads, solid pins put in place can be smooth pins set in by friction fit? like a tapered peg in a fitting hole?
if we are looking at this for what it is, then wouldnt the perfect shape for this "device" be a solid sphere?

[VintageNut]

Hello Conrad

DMMs have a limit for the potential between FORCE HI and SENSE HI. This precludes anything with appreciable resistance difference between the two HI terminals. The same goes for the LO terminals. You cannot have an appreciable resistance between these terminals.

Everything depends on what you intend to do with this device. I'm not sure it's been explicitly stated, but a "zero" ohm standard built this way may or may not have low actual resistance. Built with something other than copper, it could be quite high and still work. It just fools a 4-terminal ohm meter into thinking it has zero resistance by putting the sense wires on equi-potential points. The junction is a way to insure that two circuits don't influence each other, but have a common connection, as most meters require. The tetrajunction is elegant in that any terminals can be chosen, but there are various ways to accomplish the same thing. The thicker and lower the resistance of the material used, the less sensitive construction will be. The PCB version is certainly overkill for my 5.5 digit meter that resolves 0.1 milliohm. OTOH, if you've got a more serious low ohms meter, it might not be, nor will it be sufficient for the connections in a low value transfer standard. Building one where the connections lie on perfectly equi-potential points is just a matter of mechanical precision. Press-fitted terminals seem like the best way to go, but other ways will certainly work.

[VintageNut]

Please post pictures and measurements. Everybody can learn from your efforts. It would be good to have some documentation in this thread for different devices that are low ohms.

I will post pictures and measurements of my crossed wires devices. This week I am traveling for work and it may have to wait for next week.

4 short wires drilled together and soldered/crimped well in a single spot will be superior to any triangle...

before I build my "Zero Ohm Standard" I have experimented with a crimped wire solution as described
if I remember correctly the "triangle" showed more than an order of magnitude lower values

[quarks]

Please post pictures and measurements. Everybody can learn from your efforts. It would be good to have some documentation in this thread for different devices that are low ohms.

it is already there, if you follow the link in my first reply to this topic
https://www.eevblog.com/forum/metrology/zero-ohm-diy-4-wire-standard/msg1064308/#msg1064308

what I did not document was the 4 wires experiment, as it is not well made and I do not like it, but att. is a pic.
I used speaker cable 4 mm² all 4 with equal length and crimped with a 16 mm² ferrule and soldered on ring lugs
then screwed Pomona 3770 bindig Posts

meassured/calculated value of this was around 10µOhm
I compared this with identical setup
to my final "Zero Ohm Standard" it was clearly well below 1µOhm

[Conrad Hoffman]

DMMs have a limit for the potential between FORCE HI and SENSE HI. This precludes anything with appreciable resistance difference between the two HI terminals. The same goes for the LO terminals. You cannot have an appreciable resistance between these terminals.

Certainly true, and the question is how much? This is something I've never seen in a meter data sheet unless I don't know what I'm looking for. My WAG is that it might be quite different for different meters, unless everybody happened to settle on the same basic circuit.

[tszaboo]

DMMs have a limit for the potential between FORCE HI and SENSE HI. This precludes anything with appreciable resistance difference between the two HI terminals. The same goes for the LO terminals. You cannot have an appreciable resistance between these terminals.

Certainly true, and the question is how much? This is something I've never seen in a meter data sheet unless I don't know what I'm looking for. My WAG is that it might be quite different for different meters, unless everybody happened to settle on the same basic circuit.

Example:
http://literature.cdn.keysight.com/litweb/pdf/03458-90014.pdf
Appendix A, page 286.

I spare you the looking around. It is +/- 200V.
Seriously, what is the deal with this "tetrajunction"? Is it really hard to make 0 ohm on a multimeter? You need 2 shorting bars, three if you must be proper, and a Pomona shorting bar is 5 EUR on Farnell. Can someone tell me what is the deal with it? Is this being used in turboencabulators?

[Conrad Hoffman]

As a practical matter, it isn't needed. As proof, look how many are offered on the commercial market and for how long. For setting zero on a meter there are certainly simple methods like shorting bars that work fine. That said, the use in the resistance transfer standards is interesting, as is the entire subject of equi-potential lines on a conductor. I read about it and ran to the basement to make an measure one. Not everybody is fascinated by the same things and I suspect those with an interest in sub-ppm measurements would be the target audience.

[Kleinstein]

Just to have a 4 wire zero ohms reference for a meter, it is not needed to have the contacts interchangeable. So no need for a fully symmetric zero. 

A zero ohms with fixed current / voltage path is much easier (and can be lower value) - here the PCB version (like the ones sold by Fluke) are a good solution:  The current contacts (and voltage contacts) connected by a relatively thick line and a thin line from the center connecting the voltage and current pairs. Even with low precision and without adjustment the 4 wire resistance for the right current path will be very low, like not measurable by normal means (e.g.  < pOhms range).

Only if you have a wrong current path the resistance will be quite high.

The fully symmetric one is needed for something like Hamond dividers or similar circuits. Here requirements are not that bad - so the simple 4 wires crimped together can be good enough.

[VintageNut]

DMMs have a limit for the potential between FORCE HI and SENSE HI. This precludes anything with appreciable resistance difference between the two HI terminals. The same goes for the LO terminals. You cannot have an appreciable resistance between these terminals.

Certainly true, and the question is how much? This is something I've never seen in a meter data sheet unless I don't know what I'm looking for. My WAG is that it might be quite different for different meters, unless everybody happened to settle on the same basic circuit.

Example:
http://literature.cdn.keysight.com/litweb/pdf/03458-90014.pdf
Appendix A, page 286.

I spare you the looking around. It is +/- 200V.
Seriously, what is the deal with this "tetrajunction"? Is it really hard to make 0 ohm on a multimeter? You need 2 shorting bars, three if you must be proper, and a Pomona shorting bar is 5 EUR on Farnell. Can someone tell me what is the deal with it? Is this being used in turboencabulators?

For me, it has nothing to do with making zero on a DMM. I wanted to make zero on a nanovoltmeter. That is the challenge.

The DMM shorting plug is easy to replicate with DIY copper but is useless for finding the floor of a nanovoltmeter.

[Vgkid]

Bonus points if you can find this for me, I can't.
Hess, A. E. , Nano·ohm junctions for accurate ratio instruments,
NBS Report No. 9158 (Jan. 1966).

[Assafl]

DMMs have a limit for the potential between FORCE HI and SENSE HI. This precludes anything with appreciable resistance difference between the two HI terminals. The same goes for the LO terminals. You cannot have an appreciable resistance between these terminals.

Certainly true, and the question is how much? This is something I've never seen in a meter data sheet unless I don't know what I'm looking for. My WAG is that it might be quite different for different meters, unless everybody happened to settle on the same basic circuit.

Example:
http://literature.cdn.keysight.com/litweb/pdf/03458-90014.pdf
Appendix A, page 286.

I spare you the looking around. It is +/- 200V.
Seriously, what is the deal with this "tetrajunction"? Is it really hard to make 0 ohm on a multimeter? You need 2 shorting bars, three if you must be proper, and a Pomona shorting bar is 5 EUR on Farnell. Can someone tell me what is the deal with it? Is this being used in turboencabulators?

For me, it has nothing to do with making zero on a DMM. I wanted to make zero on a nanovoltmeter. That is the challenge.

The DMM shorting plug is easy to replicate with DIY copper but is useless for finding the floor of a nanovoltmeter.

Hmm. Does it really go lower (with the zero ohm) than a thick copper wire shorting the input? If so - I'd really love to figure out why....

I ask because to me the concept of looking at the noise floor a high gain amplifier is about getting rid of picked up noise (lots of shielding, shortest wires, no pickup loops, cleanliness) and eliminating spurious voltage sources (thermal equilibrium by waiting a long time, compatible materials for low thermal EMF, etc.)... In RF that may be finding the right Zo short (or low inductance resistor) but in DC it is about "real 0 ohm".

The tetrajunction requires longer wires, creates loops (even within a disk of copper, albeit more eddy current like...), is very hard to generate thermal equilibrium (I'd consider putting it in a big box)... i.e. everything that isn't good for noise floor measurements... Also, the extra wires (unless you can find a tiny 0 ohm shape that perfectly fits the layout & configuration of the input jacks of the Keithley) means that the tetrajunction 0 Ohms is virtual rather than real (so the 2 wire resistance measurements - done at the wire ends would be somewhat higher than just a short instead of wires).

Lastly, there is something oddly suspicious in pushing 5A across a device that you want to measure 0 ohm. Not something that we haven't done before (think of a Peltier device cooling a input device pair to lower noise - high current that helps drop noise like a brick)... just that the high current/voltage were always separated galvanically from the input circuit....   

[VintageNut]

Shorting the input of an amplifier is a completely different problem than providing a zero for a 4-wire measurement system.

The tetrajunction was developed for connecting 4-wire junctions of transfer standard resistor arrays. The tetrajunction is not just a trick for fooling an instrument. It is a useful device where nearly zero resistance is required for an array of metrology resistors.

Keithley had a common plug for nanovoltmeters and microvoltmeters from the 1960s through the 1990s and there is a  short, the model 1488. It is a simple copper wire in a can. It is used to zero the nanovoltmeter before connecting a measurement cable.

[Assafl]

Shorting the input of an amplifier is a completely different problem than providing a zero for a 4-wire measurement system.

The tetrajunction was developed for connecting 4-wire junctions of transfer standard resistor arrays. The tetrajunction is not just a trick for fooling an instrument. It is a useful device where nearly zero resistance is required for an array of metrology resistors.

Keithley had a common plug for nanovoltmeters and microvoltmeters from the 1960s through the 1990s and there is a  short, the model 1488. It is a simple copper wire in a can. It is used to zero the nanovoltmeter before connecting a measurement cable.

Your original statement: "For me, it has nothing to do with making zero on a DMM. I wanted to make zero on a nanovoltmeter. That is the challenge. The DMM shorting plug is easy to replicate with DIY copper but is useless for finding the floor of a nanovoltmeter."

My assumption was the "finding the floor" meant figuring out the noise or "measurement floor" of the Nanovoltmeter. In reading this again I think you meant to "zero the 4w zero using a nanovoltmeter" (floor referring to "Flooring" the 4w-zero, not the nanovoltmeter).

BTW -  since you are measuring such low signals - why not up the current? Say at 10A or 20A it should be easier to measure the offsets since the offset will be 2-4 times as high. If you do it quickly so as not to heat the copper...

[Conrad Hoffman]

Now I'm a bit baffled, not being familiar with nanovoltmeters. I'm not aware of any potential measuring device that uses a 4-wire connection, or any reason why it would or should. The tetrajunction does not have a lower resistance than any similar lump of copper. It doesn't have any lower noise than any similar lump of copper. A tetrajunction in a box labeled "zero ohm standard" isn't really, in fact one could say it isn't even a resistance standard, as the fundamental source-input hook-up for making a 4-wire resistance measurement isn't being observed. It's just one of several ways of setting zero on that type of meter. (Are we talking a nanoampmeter here, not nanovolt, as there is that other thread on measureing nanoamps?)

[3roomlab]

a strange idea came to me, about extending the "accuracy" of a PCB or flatten tetra thingy. but i am sleepy, so i might be wrong.
see pic
top 2 pin are force.
centre pin = sense +ve or -ve
the bottom thingy, has a wave trace sitting within some zone of the neutral "zero", but we dont know where is this zero
between A-B, is about 9mm. however the height of the trace is 22mm approx (i just randomly made it so), and it is (6x 22mm++) worth of traces connected to get from A to B. (practically expanding 9mm into 130mm+++)

would it be true to say if 1 is to tap the -other sense probe on the exposed wave trace, at some point, you should find a zero? i suppose wire can be soldered there to make this accuracy extension even longer to 1:50? 1:100?
in conrad's case, the neutral zone is somewhere "there", by soldering 2 long wires to 2 spots just outside this zone, then "dowse" out the zero? hopefully the trace doesnt become an antenna ?
i hope i didnt bend the laws of physics too much?

(or after connecting bottom pin, you know you are off zero by X, by connecting a parallel "dowsing" pin into the "wave" trace, to find the counter-potential and offset the error? so instead of shaving material off the copper block, find an additional offset spot?)

[Conrad Hoffman]

Not too much bend. Me thinks you've been looking at the old lead compensation boxes for old bridges. They did similar things long ago, just without the tetra part. This isn't that different from the way I used to balance the grounds of two KVDs I wanted to compare. I connected the grounds with a length of copper wire, then a clip lead to the center point. By moving the clip a bit, I can insure a null at zero, which is harder to do with separate leads. It's been a while, but it was something like that. No reason you can't do it in 2D.

[VintageNut]

Shorting the input of an amplifier is a completely different problem than providing a zero for a 4-wire measurement system.

The tetrajunction was developed for connecting 4-wire junctions of transfer standard resistor arrays. The tetrajunction is not just a trick for fooling an instrument. It is a useful device where nearly zero resistance is required for an array of metrology resistors.

Keithley had a common plug for nanovoltmeters and microvoltmeters from the 1960s through the 1990s and there is a  short, the model 1488. It is a simple copper wire in a can. It is used to zero the nanovoltmeter before connecting a measurement cable.

Your original statement: "For me, it has nothing to do with making zero on a DMM. I wanted to make zero on a nanovoltmeter. That is the challenge. The DMM shorting plug is easy to replicate with DIY copper but is useless for finding the floor of a nanovoltmeter."

My assumption was the "finding the floor" meant figuring out the noise or "measurement floor" of the Nanovoltmeter. In reading this again I think you meant to "zero the 4w zero using a nanovoltmeter" (floor referring to "Flooring" the 4w-zero, not the nanovoltmeter).

BTW -  since you are measuring such low signals - why not up the current? Say at 10A or 20A it should be easier to measure the offsets since the offset will be 2-4 times as high. If you do it quickly so as not to heat the copper...

I was unclear. Not just the noise floor of the nanovoltmeter. The floor of the nanovoltmeter while measuring a device under power.

The net result is that the device really has to be as close to true zero as possible. Not just an apparent zero.

The Ohm Labs device has a specified limit of 5A. Beyond that damage may occur. The manufacturer tests the device at 5A for the calibration report.

[VintageNut]

I do not understand you assertion. What fundamental hookup is not being observed?

A tetrajunction in a box labeled "zero ohm standard" isn't really, in fact one could say it isn't even a resistance standard, as the fundamental source-input hook-up for making a 4-wire resistance measurement isn't being observed. It's just one of several ways of setting zero on that type of meter. (Are we talking a nanoampmeter here, not nanovolt, as there is that other thread on measureing nanoamps?)

[Assafl]

I can't answer for Conrad but for me the fundamental difference is that in a "true" Kelvin Connection the Force and Sense are connected at the DUT. In a tetrajunction, symmetry means that there is no such connection point.

Dealing with nomenclature, one could say that for devices that measure trans-resistance (like 4w DMM), the tetrajunction could, in theory, demonstrate the lowest trans-resistance. But switching to a resistance measuring mode (e.g. 2w DMM) - I can't see a tetrajunction beating a thicker slab of copper.

Given short and thick leads, a real standard should probably measure close in both resistance and trans-resistance measurement modes...

Obviously, a physical device with 0-ohm doesn't exist in a tetrajunction. So, as an example, you can't use it for levitation (instead of superconductors in cryogenic experiments). 

[Assafl]

The net result is that the device really has to be as close to true zero as possible. Not just an apparent zero.

If you measure a real (not apparent - e.g. a superconductor) "zero" in both 2w (resistance) and 4w (trans-resistance) measuring instruments, they should come out as "0" (give and take cabling - so hook them directly to the input jacks).

A tetrajunction will come out with a substantial difference between 2w and 4w measurement as a result of the onus to be symmetrical. It won't "fool" the 2w mode, but the trans-resistance 4w mode will show an apparent reading as close to "0" as construction symmetry allows. 

[Assafl]

Some here ask "what is the use for this" outside of a Hamon ladder. I ask that too.

I guess the question I have is: As a source for an "apparent" zero - what does that mean for the DMM? If one were to calibrate 4w zero using a tetrajunction - would the stored calibration figures be different (ignoring the hookup wires - bend the tetrajunction to fit the front plate - but gently so the symmetry doesn't change) than the Fluke 4w zero?

I tried drawing this. The main difference is an additional "bridge" overlaid between the 4 wires.


So let's assume the HP34401A. My guess is that if there is a difference - it will be due to the 30pA bias current which will unbalance the legs of the "bridge". But 30pA out of 5A??? For the HP34401A the Force current is 100mA. Even then, 30pA is sub-sub-sub 1ppm....

 

[VintageNut]

If you want an accurate resistance measurement, you use 4-wire below something like 20k ohms. Your discussion of being able to measure low ohms the same 2W or 4W is a non-sequitur. AND you have to use offset compensated ohms procedure to remove the always present microvolts that are created by wires+connection neodes required for the circuit to take the measurement.

You cannot measure low ohms with 2W. Lord Kelvin was using 4W before the invention of electronics.

I can't answer for Conrad but for me the fundamental difference is that in a "true" Kelvin Connection the Force and Sense are connected at the DUT. In a tetrajunction, symmetry means that there is no such connection point.

Dealing with nomenclature, one could say that for devices that measure trans-resistance (like 4w DMM), the tetrajunction could, in theory, demonstrate the lowest trans-resistance. But switching to a resistance measuring mode (e.g. 2w DMM) - I can't see a tetrajunction beating a thicker slab of copper.

Given short and thick leads, a real standard should probably measure close in both resistance and trans-resistance measurement modes...

Obviously, a physical device with 0-ohm doesn't exist in a tetrajunction. So, as an example, you can't use it for levitation (instead of superconductors in cryogenic experiments).

[Assafl]

Of course you measure low ohms with 2 wires. Every resistance measurement has an excitation (a Force current or voltage) and a reaction (a reaction voltage or current, respectively) that happen together on the same device (a resistor or other DUT). Always 2 wire! (Edit: this truism is just Ohm's law - in a lumped version or the Maxwell equation version)

Kelvin is about separating the sense lines from the excitation lines as far as possible so that the only thing measured is the DUT. Eventually they MUST connect.

They have a few options to connect: they can connect on the resistor leads. Or they connect at the kelvin probe. Or they connect using stacked bananas. (Edit: they have to connect outside - encompassing - the DUT). If you run the 4w mode without connecting the force and sense - you'll get odd readings.

In the tetrajunction they never connect. Because it is a unique device whose trans-resistance is 0 no matter what its resistance is (within bounds inflicted by the measurement instrument - of course Edit - irrelevant since the transresistance will be 0 even if the DMM can't measure it... ). You can build a tetrajunction using Manganin wire - its resistance would be in many-many Ohms. Its Trans-resistance (if balanced) will be 0.

[VintageNut]

In theory only. In practice 2-wire low ohms would require zero resistance cables between the measurement instrument and the device under test. Practical cables have a total of a very significant fraction of an ohm which is many orders of magnitude larger than what is being measured.

Of course you measure low ohms with 2 wires. Every resistance measurement has an excitation (a Force current or voltage) and a reaction (a reaction voltage or current, respectively) that happen together on the same device (a resistor or other DUT). Always 2 wire! (Edit: this truism is just Ohm's law - in a lumped version or the Maxwell equation version)

Kelvin is about separating the sense lines from the excitation lines as far as possible so that the only thing measured is the DUT. Eventually they MUST connect.

They have a few options to connect: they can connect on the resistor leads. Or they connect at the kelvin probe. Or they connect using stacked bananas. (Edit: they have to connect outside - encompassing - the DUT). If you run the 4w mode without connecting the force and sense - you'll get odd readings.

In the tetrajunction they never connect. Because it is a unique device whose trans-resistance is 0 no matter what its resistance is (within bounds inflicted by the measurement instrument - of course Edit - irrelevant since the transresistance will be 0 even if the DMM can't measure it... ). You can build a tetrajunction using Manganin wire - its resistance would be in many-many Ohms. Its Trans-resistance (if balanced) will be 0.

[Conrad Hoffman]

Actually, force and sense only have to connect well enough to stay within the common mode range of the sense device. When I do a precision resistance measurement, I connect my standard in series with the unknown and measure across each with a floating meter. "Floating" in the sense that it can't be more than 500V or so away from the force circuit. 4-wire DMMs may be more limited! They also generally use a constant current source, though that's not (I think) relevant to the matter.

The nanovoltmeter problem seems to be one of common mode. If you fasten the leads together on a copper wire stub, you should be able to touch the stub anywhere on the high current line with no change in zero. Obviously this is only going to work with all copper connections, ideally crimped and not soldered, and every other effort made to eliminate thermal effects. Tetrajunction shouldn't be necessary to do this. Just curious, what's the input impedance of the nanovoltmeter? Bias currents? The surplus world has not yet blessed me with a nanovoltmeter. 

IMO, though you can look at the tetrajunction modeled as bridges and resistors, the potential lines of the FEA are far more revealing in how it functions.

[Assafl]

IMO, though you can look at the tetrajunction modeled as bridges and resistors, the potential lines of the FEA are far more revealing in how it functions.

I am okay with how it functions and am having a blast with your FEA program. It is a hoot to see voltage (~E - Electric Field) and current lines (J - current flux)... It is looking at Ohm's law (the more generalized Paul Drude version J=?E) for free!

Bridges and resistors - I wanted to change the POV from the tetrajunction to the other side of the binding posts - and to answer a different question:

1. How would it affect the calibration parameters if one were to use a tetrajunction to calibrate their DMM?
2. Would the calibration parameters provide improved accuracy over a regular zero (piece of copper stub, or a Fluke zero)?

I can't make any inroads on that. Feel stupid (a bit) since theoretically speaking it is simply Ohm's law... But then again it isn't...

(This is in theory only - I think thermal EMF will get you long before you realize any upside, cabling to the tetrajunction will also get you, and lastly, even if all the rest was taken care of - my DMM - an HP34401A - zeros all ranges at once - so one can't use the tetrajunction just for the 4-w resistance).

[Assafl]

In theory only. In practice 2-wire low ohms would require zero resistance cables between the measurement instrument and the device under test. Practical cables have a total of a very significant fraction of an ohm which is many orders of magnitude larger than what is being measured.

When you do 4w Kelvin, the force and sense are shorted at both sides of the DUT. At that point, be it 2w or 4w, electrons are forced to flow by creating an electric field in the DUT that results in the current required (by the regulated current source), the potential is then measured by the A/D, and the ratio between the two is calculated and labelled resistance.

The flow of electrons is because of and in the direction (or opposite) of the electric field lines (not the equipotential lines - the flux lines). 

But a tetrajunction measured 4w is not physically a resistor: In no place in the tetrajunction is there an electric field that results in an current flow in the direction of measurement (except maybe that due to Bias currents and imbalance of the tetrajunction).

Ohms law (as J=?E or as I=V/R) is not just not applicable here - it is wrong to use it! You are not measuring the electron flow in the direction of the electric field. You are inducing current that is perpendicular to the potential measurement (so keep magnets away or have to deal with hall effect).

It is a curious device nonetheless...

The fluke zero has no wires. There are internal wires (PCB tracks) from where the current is injected into the 2w terminals (in the HP34401A I think this point is near the big Coto reed relay). The 4w vs 2w compensates for that extra length of PCB.

The fact that the DMM measures trans-resistance is interesting since it would suggest that one could use a DMM to balance a bridge. The two force contacts to the top and bottom, and the two sense to L/R. minimize the Ohm reading and your bridge is balanced....

[VintageNut]

In theory only. In practice 2-wire low ohms would require zero resistance cables between the measurement instrument and the device under test. Practical cables have a total of a very significant fraction of an ohm which is many orders of magnitude larger than what is being measured.

The fact that the DMM measures trans-resistance is interesting since it would suggest that one could use a DMM to balance a bridge. The two force contacts to the top and bottom, and the two sense to L/R. minimize the Ohm reading and your bridge is balanced....

No. For most cases, the requirement for maximum resistance between SENSE HI and HI is violated. The returned measurement is crap.

[Assafl]

I had never gave the max lead resistance any thought. I always kind of thought of it as a "safety margin", for abnormally high resistance leads.

I completely missed the fact that the spec is actually about the ability to resolve the transimpedance of the 4w as "Ohms" (i.e. as if they were a physical property of a conductor - which happens to be true only for Kelvin connections).

Looking at the specs for the HP34401A, the maximum lead resistance is 10\$\Omega\$ for 100\$\Omega\$ range, 100 for the 1k range, and 1k for the rest. So in theory balancing a 10 bridge should be doable (at the highest sensitivity) - or a 1k bridge at the lowest.

[VintageNut]

There are more things to consider. The open circuit voltage is limited on DMMs. You will be limited to bridge arms that develop a voltage less than or equal to the DMM ohms current multiplied by the bridge arm resistance.

Also, you normally want the galvanometer to be able to accurately measure 1uV. If you use your proposed method, you may not be able to see that microvolt with enough accuracy.

Give it a try and post your results.