Source for low tempco current sense resistors

[Harvs]

Hi all,

I've been using a OAR5R050FLF as the current sense shunt in an eload design.  These are a 50mOhm, 20ppm 5W current sense resistor.

However, by design, they get very hot.  The result is with 5A through it (2.5W dissipation), I'm getting >0.5% shift in resistance, which is results a drift in the set current.  It makes the use of 16DAC and ADC a waste of time.

I've been looking at everything available through the usual outlets (digikey, element14, mouser,)  and there doesn't seem to be much better.  There's a number of power current sense resisters with much higher power ratings, however, they have equally worse tempco.

If I look at something like the Agilent 6060B, they use the same value of resistance per load cell, and run up to 7.5A  through each cell (x8 cells). There specs are 0.1% absolute with 10mA regulation. So I'm taking that as a sanity check that I'm in the ball park for the 50mOhm resistor value.

So how do you get hold of these much larger low tempco shunts?

Are they a custom made part during the device manufacture?  In which case, could I buy the material somewhere and make my own?

Or am I completely off the mark here, and is there a compensation scheme for dealing with this problem?

Cheers

[ve7xen]

I think you should probably just work to minimize the power dissipation in your shunt (or spread it around). Using an in-amp or dedicated current sense amplifier you should be able to get down to the 1mR range for your sense resistor, which can probably just be a length of copper wire and will experience minimal self-heating. Analog has AN105 that covers lots of current sensing topics, which should give you some ideas. The main problem though is your current dependence on the temperature coefficient of this part, when with different design choices you should be able to get similar performance out of a more pedestrian resistor. You should save a lot vs. a precision low tempco resistor.

If you want to stick with this plan, paralleling higher tempco resistors might get you there. High precision is probably not an option unless you're made of money. It's doable with some of what I can find at DigiKey etc., but for $15 a resistor and you'll need 10 of them.

I'm not sure how accurate they are, but there are also hall-effect based sensors that I think are pretty linear with temperature.

[Harvs]

I think you should probably just work to minimize the power dissipation in your shunt (or spread it around). Using an in-amp or dedicated current sense amplifier you should be able to get down to the 1mR range for your sense resistor, which can probably just be a length of copper wire and will experience minimal self-heating. Analog has AN105 that covers lots of current sensing topics, which should give you some ideas. The main problem though is your current dependence on the temperature coefficient of this part, when with different design choices you should be able to get similar performance out of a more pedestrian resistor. You should save a lot vs. a precision low tempco resistor.

If you want to stick with this plan, paralleling higher tempco resistors might get you there. High precision is probably not an option unless you're made of money. It's doable with some of what I can find at DigiKey etc., but for $15 a resistor and you'll need 10 of them.

I'm not sure how accurate they are, but there are also hall-effect based sensors that I think are pretty linear with temperature.

I could drop the impedance, heck I could parallel 10 of these for <$10 that would get me close to my self heating tempco goals (0.1% t 10A).

But the penalty is the sense resolution.  At the moment the DAC and ADC can happily measure and control down to down to 1mA, which is 50uV.  The ADC has a 7.8uV step size with an internal 8x PGA deferentially measuring the shunt in a 4-wire configuration.  Dropping to 5mOhms will significantly reduce that accuracy, which I don't see myself getting back through extra amplification (down to 5uV per mA).

So in the end have I gained any accuracy by dropping the level of the resistance?  Yes at the top end by sacrificing the low end.

I guess I could string 5 of these in the 10mOhm range together, but then the trace resistance and joints between them would probably cause more errors than it's worth.

I just see the answer whenever I look at pics of the comercial eloads.  A large current sense resistor...  Surely they can't be unobtainium?

[Jay_Diddy_B]

Hi Harvs,

I did a little research for you.



The resistors used in the HP6060B are made from an alloy called Cupron.

The temperature curve for Cupron is shown here:

http://cartech.ides.com/ImageDisplay.aspx?E=10&IMGURL=%2fCarpenterImages%2fG-SpecialPurposeAlloyWire%2f16-FW16-CBXCupron%2f02_FW16_Resistance.gif&IMGTITLE=Resistance+Change+vs.+Temperature


It is very flat in the 20 -100 ^oC range.


This doesn't help directly, but you know to search for Cupron.

Jay_Diddy_B

[Harvs]

How'd you find that out?  You've clearly got better research skills than me!  I got the part no from the service manual and searched around, couldn't find a thing!

[Jay_Diddy_B]

Harvs,

This how I did it. Don't tell anyone. This is just between you, me and the other forum members 

From the manual I got the HP (Agilent) Part Number 06060-80014. The fact that it has the Instrument number in the part number means that is a custom HP part for that instrument, as oppose to a regular part with a manufacturer's part number.

I went to the Agilent site and put the part number in their search.

This takes you to the page for the 6060B loads.

Then at the bottom of the page under 'Technical Support' I clicked 'Parts'

I entered the part number again in the search by parts box.

I then clicked on the part number to bring up the description:

The part was described as:

Part Number:  06060-80014 
Part Description:  Cupron Resistor 
Price:  CA$ 27.62 
Quantity on Hand:  Not in stock, but orderable 
Item Status:  Orderable 
Item Type:  Agilent Qualified Part 
Weight:  0.005 KG 
Where Used:  view all products with this partHelp 


I then Googled 'Cupron' and I found information on the alloy.


I have a HP6060A in my collection. The resistors are buried under the heatsink otherwise I might be able to tell you more about them.


Regards,


Jay_Diddy_B

[ftransform]

Fascinating. I recall that dave said that even expensive multimeters are not capable of prolonged accurate measurement at higher currents.

I wonder if you can attach a RTD or thermistor to the current sense resistor and mathematically eliminate the error caused by self heating. That sounds like a nice project.

[Harvs]

Yep DMMs suffer exactly the same problem.  The difference is bench equipment has the room to have much larger shunts to deal with the heat, or in the case of instruments like the 3456A, they expect you to use an external shunt.

You can buy shunts that have a thermistor bonded into the casing and come with full calibration over temperature.  I'd hate to think what that's worth though.

But that's way beyond what I'm looking for.  All I need is a larger version of the low tempco shunt.

Incidentally the tempco of Cupon wire is the same as the shunt I've already got, so it's possibly the same material.  Just those HP ones look much bigger in the board layout from the service manual.

[poorchava]

I suppose one way to solve that would be to actually plot the R=f(T) of the shunt resistor, then attach some garden variety temperature sensor to the shunt. Then you could process that somehow with MCU or maybe even some analog circuit (um... dunno... like a variable gain amp of some sort?).

You could probably also do opposite thing, so stabilize temperature of the resistor (that would give you something like 'oven controlled shunt resistor').

I don't know by how much could you improve your stability over temperature range, but I'm pretty sure that it would improve.

[Rerouter]

look for invar wire, industry grade stuff has a temp coefficient of 1.5ppm, with scientific grade nearly almost under 120ppb

[Jay_Diddy_B]

Harvs,

These guys might be able to help you:

http://www.cshunt.com/dc_ammeter_shunts.html

My thinking is to keep the 50m Ohm resistors in the source of the MOSFET for current sharing and use a single PCB style shunt in the cool side of the air flow.

It is a little tricky to decide the optimum value for the shunt. If you use to too low a value the thermal emfs and offsets in the amplifiers cause problems at low currents. If you use too high a value self heating is a problem.

The other comment is that some resistance wires cannot be soldered. They require special treatment or welding.


Jay_Diddy_B

[Harvs]

Thanks guys, lots of good tips to investigate.

Another way out of this.  I was just reading a application note on calibrating  shunts.  Apparently they should sit for a considerable time at the cal current for the temperature to stabalise, and then the shunt is calibrated at for that particular current.

From what I've seen using the shunt I've currently got in there, it heats up pretty quickly, and stablises in a few seconds.  So I could just calibrate it at a number of points and get the uC to interpolate.

Another idea anyway...

[M. András]

http://www.digikey.com/product-search/en/industrial-controls-meters/accessories/2949204?k=shunt
+-15ppm manganin shunts

[SeanB]

How about heatsinking it, or placing it in a small tin filled with oil to handle the dissipation. Having a heat transfer fluid to remove the heat will keep it at a lower temperature and this will reduce the error.

[Neilm]

Have you considered having two ranges? One for low currents and one for high currents. Are you trying to control the 16A down to 1mA?

[ftransform]

look for invar wire, industry grade stuff has a temp coefficient of 1.5ppm, with scientific grade nearly almost under 120ppb

oooo where can we get some of this stuff? I like the sound of the scientific grade oh so much.

[robrenz]

Invar is actualy an alloy made for mechanical applications needing low thermal coefficient of expansion. Super Invar is a better grade. If you want the best for electrical low TCR you want Carpenter Technology "Evanohm S" notice the "S" not "R".
Look at the "Resistance vs. Temperature Characteristics" link and the "Typical TCR Values of Resistance Alloy - Table" link toward the bottom of this data sheet

[ftransform]

Invar is actualy an alloy made for mechanical applications needing low thermal coefficient of expansion. Super Invar is a better grade. If you want the best for electrical low TCR you want Carpenter Technology "Evanohm S" notice the "S" not "R".
Look at the "Resistance vs. Temperature Characteristics" link and the "Typical TCR Values of Resistance Alloy - Table" link toward the bottom of this data sheet

I thought the electrical properties were related to the mechanical properties?

[robrenz]

Invar is actualy an alloy made for mechanical applications needing low thermal coefficient of expansion. Super Invar is a better grade. If you want the best for electrical low TCR you want Carpenter Technology "Evanohm S" notice the "S" not "R".
Look at the "Resistance vs. Temperature Characteristics" link and the "Typical TCR Values of Resistance Alloy - Table" link toward the bottom of this data sheet

I thought the electrical properties were related to the mechanical properties?

They are related in the sense that the alloying materials that are optimized for the mechanical will inherently influence the electrical properties. Like most things optimizing one property usualy means some other property getting worse.

[Harvs]

http://www.digikey.com/product-search/en/industrial-controls-meters/accessories/2949204?k=shunt
+-15ppm manganin shunts

These will have exactly the same problem that I'm currently having.  Sure they're 15ppm instead of 20ppm, however they reach their max safe operating temp under 2/3 load.  Hence they'll having something like a 75-100 degrees C shift in temp over ambient, so there'll still be a significant shift in resistance.  Also, it's worth noting (like all the shunt manufacturers) they give a figure of 15ppm/C, but that is only loosely related to shift due to self heating since it's a very non-uniform thing.

Have you considered having two ranges? One for low currents and one for high currents. Are you trying to control the 16A down to 1mA?

Yes, and I am implementing two ranges.  I'm looking at a 500mA range, then have 4 load cells for the higher ranges to get a total of 30A.

I don't expect to get accuracy down to the mA, but at the moment that is pretty much what I'm getting (compared to my bench DMM) for relatively low currents like 1A through a single load cell.  However, power in the shunt is proportional to I^2, so as the current moves up it gets hot very quickly in an exponential fashon.  The reality is this will be the case for any small shunt that doesn't have the ability to dissipate a reasonable amount of power without getting toasty.

The most likely solution I see at the moment is to chop it down into more lower powered load cells.  Like 10 of handling 3A each.  It also means having 10x ADC channels to sample the all.

[Jay_Diddy_B]

Harvs,

I am curious to know why you are looking for such high precision in an electronic load?

One reason I can think of is efficiency measurements.

Jay_Diddy_B

[Harvs]

Pretty much spot on there Jay_Diddy_B, the immediate application is efficiency of dc-dc converters particular with an eye to solar powered systems.

There's actually two sides to this story. Obviously the electronic load, but I'm also making a complementary system of just the voltage and current measurement to measure the input power.  Basically the same system without the load mosfets, just pass through.  All isolated and USB controlled.

I'm actually more interested in automation and accuracy than I am total capacity.

I've attached a couple of pics of the test bed at the moment, and a screen shot of a small piece of software I wrote to test it doing something useful.  In this case discharging a NiMH AA cell.

I did some research into that Cupron material.  Interesting stuff, seems to also be referred to as Cuprothal 294 among other things (every maker has a different name.)  It can be easily soldered apparently.  The problem will be getting hold of it in a decent gauge in low quantities, as it's commonly sold as a very fine wire for low temp heating.  I've contacted a couple of sites on the net asking about low quantities of decent gauge, so I'll see what I get back.

http://www.resistancewire.com/Html/Technical/AlloyDataTables/PDFDocs/ADT2001.12.17.294.ENG.pdf

[Jay_Diddy_B]

Hi Harvs,

I understand what you are doing now. Measuring efficiency on power converters is getting tricky now that efficiency >94% are common.

When I am optimizing power converters, I tend to only to look at input current. I try and minimize the input current for a given output condition.

Consider as well that in a typical converter the losses in the converter are generally related to the resistance of copper with a tempco of 3930 ppm and The RDSon of MOSFETs which doubles with Tj from 25 to 125^oC.


Jay_Diddy_B

[SeanB]

If you can only get the finer wire ( I have a reel of it on the shelf) then take 2 mica sheets and a copper core, and wind the wire around it to give a sandwich that has a heat tab out the end. You place a small shim to insulate the core from the edges and place a thick copper wire on each end, then solder the edges to the wires and use that as a shunt. Large surface area and lots of cooling, and adjustable during design and manufacture to the desired value. If you need extra dissipation add 2 thin mica sheets on the outside and sandwich between 2 aluminium heatsinks to make a block.