Advice on building a Coaxial Current Shunt

[hazzer]

Morning - I'm looking for advice regarding my build of a Coaxial Current Shunt, it will have 10 ribs each with a 1 ohm foil resistor,
my question regards how the sense lines are to be combined,

ie simply symmetrical circular pattern onto a common copper plane,

or should each sense line have an additional series resistor (10 to 100 times the individual shunt rib resistor value)?

I can try upload a hand sketch but hopefully the description will elicit some advice.

Thanks

Harry

[MK]

How good do you need it to be?
there is a user with neutron in his name on diyaudio who works at one of the US collider faciities, made his own co-axial shunt with pics, ~10-20 pico henries used it on a 30,000Amp pulses.

[hazzer]

Thanks for that - no my build wont be as exotic as that,
so the published work Ive seen so far on these shunts hasn't mentioned the inclusion of the additional series resistors, and its quite awkward to include them, so my thinking is there is little to be gained, but I'm sure there is a more knowledgeable person than me on this subject....

[Doctorandus_P]

I was not familiar with these things, but a simple picture search clarifies things up, and can give you some ideas for DIY. I guess you already did something similar, but different search engines sometimes show different sets of pictures.

https://duckduckgo.com/?q=coaxial+Current+Shunt&iar=images

Several of these (TNP, fluke, Transmille, Tunkia) make them out of a lot of PCB's in a circle, and (both front and back? have multiple solder connections. This will partly be for mechanical stability and low parasitic resistance, but it also makes it easy to add a few extra resistors for averaging the measured voltage drop. Their designs look very similar to each other.

Do these things really have a big advantage compared with simpler low inductance resistors?
Those fluke things only being usable "up to" 100kHz does not seem very impressive on first sight, but my knowledge is lacking here, maybe I'm just babbling.

[Micke]

Hi!

I also had the same thoughts as you what to do with sense when I did my try for a 100mOhm "kind of" coaxial current shunt 
https://www.eevblog.com/forum/metrology/diy-20a-0-1r-current-shunt-metrology-grade()/

[hazzer]

Thanks yes some interesting additions I haven't seen before, yes I'm thinking of a PCB build,
I believe the coaxial configuration is to contain the magnetic field and deliver better AC performance,
interestingly none show evidence of additional sense resistors which is in itself interesting.

Thanks again for the post.

H

[hazzer]

Nice to chat, yes I read your earlier post, a nice build, so I'm thinking of PCB construction with coaxial connectors either end, so quite traditional for this sort of thing, its an easy construction, and I may end up trying with and without the sense resistors, my instinct is to have them on one path only and not symmetrical as in your option 3

I've time yet to decide in case someone says a definite yes or no.

Thanks again for the post

[hazzer]

just for discussion.... my reading of coaxial shunts suggests the current on either side of the PCB struts is key to its design and the struts should be arranged radial face to face.
....  I also suspect a single "sence" resistor per strut would remove the chance of another path for the current if a strut had a different impedance to its neighbour

H

[TizianoHV]

I think that the balancing resistor are needed for "high precision" applications (where thermal EMF matters). For pulse application the voltage will be high and "easy" to measure and you don't need high accuracy (0.01%....).
Once you have assembled the shunt you measure the resulting resistance by passing some DC current. Here a nice video:



I made a pulse shunt following the same design, I used it at >30kA and it works well:

[hazzer]

excellent - thanks for posting - yes that goes some way to explain the coaxial design

H

[laichh]

There are 2 documents that you must read before heading to built one:

1st is the Fluke A40B Series Precision DC and AC Current Shunts's Instruction Manual.

https://media.fluke.com/6884e345-6542-4f7a-a3b1-b10800bfb213_original%20file.pdf



Page 20 - Theory of Operation
Page 31 - Construction of the shunt explained in detail


2nd is the Guildline's patent US8111060B2.

https://patentimages.storage.googleapis.com/9e/5d/dd/7787ef75b3eb4a/US8111060.pdf

[hazzer]

Thank you - that's great - all that fluke doc needs is some Gerbers 


H

[laichh]

Guildline design is easier to construct, only 2 PCBs are needed.

[Doctorandus_P]

I do not understand...

What is the advantage of the long strips of PCB, and then putting the resistors all the way at the back?

The front is already a dual sided PCB with the "in" current on one side and the "out" current on the other side of the PCB, so why not just put all the resistors in a circle on the backside of that PCB?

That would result in something more alike the youtube video from vidduley.

But then maybe turn it around, so the output is taken directly from the front and back of the PCB.
... But that would increase capacitive coupling. So spacing the "output" PCB's at the back makes sense to me.

[laichh]

Well, you simply can't solder two coaxial connectors back-to-back on the opposite side of the PCB, unless both of them are SMT mounting.

The current sourcing side of the shunt is designed to conduct heavy current, and both Fluke & Guildline opt for an LC coaxial connector. For less heavy current, Guildline opts for a UHF connector while Fluke opt for an N-Type connector. While both LC or UHF coaxial connectors are no longer mainstream, the obvious choice would be an N-Type connector. As far as I know, I've never seen an SMT mounting for an N-Type connector.

For the voltage sensing end, not much of an issue for SMT mounting. There are plenty of SMA/SMB coaxial connector to choose from.

[viewtek]

I have spent a great deal of time and dedication on this project. I have built versions ranging from very small currents (1 mA) up to large currents (100 A), using both expensive Vishay resistors as well as low-cost SMD resistors.

The purpose was purely for personal interest, as I do not have access to a high-frequency source or a TVC (Thermal Voltage Converter) to verify the measurement integrity of this product.

[tszaboo]

Ah great. 10 minutes ago I didn't even know this construction exists, now I want to build one. Thanks a lot.

[viewtek]

Additionally, I have an interesting observation to share: the 30 A shunt option in the Fluke 8588A, often regarded as “the best multimeter in the world,” is also built using a coaxial design, with multiple SMD shunt resistors connected in parallel.

[hazzer]

Thanks yes that's interesting, pretty sure I can see "sense resistors" combining the voltage feeds from the 30A shunt elements, and a big thanks to all who posted photos, lots to consider.

H

[hazzer]

I do not understand...

What is the advantage of the long strips of PCB, and then putting the resistors all the way at the back?

The front is already a dual sided PCB with the "in" current on one side and the "out" current on the other side of the PCB, so why not just put all the resistors in a circle on the backside of that PCB?

That would result in something more alike the youtube video from vidduley.

But then maybe turn it around, so the output is taken directly from the front and back of the PCB.
... But that would increase capacitive coupling. So spacing the "output" PCB's at the back makes sense to me.

another way to frame this question might be to ask what defines the dimensions of the Fluke shunts? looking at a range of commercial units it seems the current defines the length the ribs while the radius is defined only by the need to accommodate the number of shunt resistors, so the radius is about heat and physical space, then the length of the ribs is perhaps to reduce the AC coupling (mostly magnetic) between the Current and Voltage connectors, if correct it raises the question "how does the Guildline's design control this source of error"?

H

[tszaboo]

I guess the N connector is because the Fluke calibrators are coming with those connectors for their AC part. Is there any practical reason to use them instead of Banana connector? I don't think that at 30KHz it's going to make a difference, is it?

[Doctorandus_P]

I don't think that at 30KHz it's going to make a difference, is it?

At first I did not find the A40B specification very impressive, with it's stated 100kHz bandwidth. However, it has a linearity of around 25ppm (depending on version, the 100A is 174ppm ) in that band.

[tszaboo]

OK, so I went down the rabbit hole of this, and I'm starting to understand the issue.
Let's take a 10mOhm shunt, in a 2512 package. The WSLT2512 from Vishay has 5nH declared in the datasheet. If I plug this 5nH 10mHohm into a calculator I get 306KHz as the -3dB point.

But this is assuming that the connections going to the shunt have no inductance.

If we have a N connector that has additional ~3nH inductance. This also has capacitance, which for a 50 ohm system compensates for this, but this is not a 50Ohm system, the output is practically 0 ohm. The PCB will probably double this inductance.

10nH 10mOhm has the -3dB point at 150KHz.

To make things worse, they are not going to use a 2512 package resistor, because it's not made from the right metals, so the temperature coeffs are wrong, which makes DC errors. They use Vishay bulk metal foils in a through hole package. They looked like the S series, I don't think the exact package makes a huge difference.

High Frequency Operation
Rise time 1.0 ns at 1 kΩ
Inductance (L)(2) 0.1 μH maximum; 0.08 μH typical
Capacitance (C) 1.0 pF maximum; 0.5 pF typical

Suddenly the inductance is 80nH or let's say 100nH for a single resistor with PCB and connector. The -3dB point is at 15KHz now. But you don't operate at the -3dB point, at that point your AC measurement has 50% error, and your usable bandwidth is likely 1/10th, or even less, because the allowed errors are PPMs

So placing a lot of them in parallel is to reduce the inductance, each parallel path will reduce it.

There is a second component to the construction, about the routing of the current. This is related to the magnetic field of the current.

From what I see, the construction tries to create an area that's free from magnetic fields where the sense line is connected. To understand, first  I opted to understand the separation of these lines. Made a very weak microstrip coupler with 100mm length and 10-20mm separation. Run it for 10-100KHz. I know maybe it's not the right way to analyze this, but this is what I can enter into a simulator. So I got anywhere around -80dB-120dB isolation for this coupler. Which doesn't sound like much, -120dB is 1ppm. But here is what I think makes the issue:
This is I think acts as a current transformer. So let's say you have 1A on the primary, and this makes 1uA on the secondary, your output. You connect your output 1uA to a 10Mohm input of a multimeter, this is suddenly 10V. Not exactly, because they DMM input isn't going to be 10MOhm at let's say 10KHz, and I think the coupling may be weaker, but this demonstrates the issue. When PPMs are measured, this matters.
So the magnetic field for that 4th wire connection must be as small as possible. This is why they don't connect it just "next to it" because this way they can make it symmetric and compensate it on the lengths of the wires.

Right now I don't really understand why they make the shunts so "long". I don't think it's for dissipation. There is this research paper where they made them shorter.
https://www.imeko.org/publications/tc4-2014/IMEKO-TC4-2014-309.pdf
I also don't get the resistor placement/orientation. I think orienting along the current path would make the current path shorter, so why rotate it?