10 ohm 1W 0.1% resistor

At this impedance and accuracy level you should look for a four-terminal shunt. 0.1% will mean 10 milliohms, and that is difficult to handle with two-wire connections.

How about a smaller resistor and a chopamp?

Cheers

Phil Hobbs

Sometimes people have a problem and they need to sense current and they say "Well I'll current-sense on the low side."

Now they have two problems...

yCMF5510R000BEEK

Vishay used to do some really classy - if non-cheap - precision resistors i n high-dissipation packages. I used then at Cambridge Instruments in the sc an amplifiers for the electron-beam microfabricator (which were supposed to be 18-bit accurate over the full field, though you had to wait about 1msec when stepping from one sub-field to the next.

is an example.

covers the range.

At least two problems...

I like to use multiple surface-mount thinfilm resistors, which are cheap and super stable. Like for a 50 ohm terminator, four 50 ohm

Surface mount resistor cooling is dominated by the heat sinking from the pads, and heat transfer into the PCB ground or power planes. An

U could get a sense resistor with a +3300 ppm tempco (or do the math to make a series/parallel arrangement that had approximately that) and then buffer the sense voltage to the base of a BJT fed with a current calculated to give the same temperature rise as in the resistor

Hmm... I didn't think about that. My total error will be bigger than 0.1%. I can make the interconnecting traces wider.

George H.

You know I've never used a chop amp. At the low end I've got 20 mA,

Then Tauno's 'stray' resistance effects are larger.

George h.

Care to share? GH

I guess I'm semi-crippled 'cause I'm only doing two layers. Do you put the large pours on the ends on the top layer, or couple it to one of the inner layers? I've got ~1/2" by 1" of space on the pcb for the sense resistor.

George H.

Just big pads will help a lot. Add copper pours to the top and bottom, and connect with vias to move the heat. Use all the area that you have, 5 or even 10 paralleled cheap 0805's maybe. Of course, my bias is towards using reel-fed surface-mount parts, not hand-soldered leaded things.

Soldered to big copper pours, the central hot-spot temperature of a surface-mount resistor, at a given power dissipation, is almost independent of resistor size. That's because they are usually all 20 mil thick alumina, so the sheet thermal resistance is constant, so theta from center to ends is constant.

You can make 4-wire PCB connections to standard 2-terminal surface-mount resistors, or even to an array of resistors. Just use two fat traces for the current, and pick off close to the bodies with two sense traces.

Can you sketch that circuit?

Depends on the layout. 1-oz copper is about 0.5 milliohms per square, so at that level it's generally fine to take the sense lines off the pads.

Cheers

Phil Hobbs

The two big ones with sensing on the low side are that if the sense resistor is in series with the load, a naive sensing arrangement won't detect a fault condition where the load is shorted to ground from either the top or bottom. "Wow my load seems to have gone really high impedance! Better jack the current up..." And depending on the load/application there may be no good way to distinguish a fault condition from normal operation using low side sensing.

Also the ground loop problem where one must ensure both the ground of the sense amp and the load are always at very close to the same potential. No sense in having a super-accurate resistor if the relative ground between the two is varying all over the place. That usually implies having the sense amp very close to the load which again application-depending isn't always practical...

Appeal to authority: IIRC Bob Pease was very "down" on low-side sensing in general too ;-)

The common-mode voltage will be way lower if low-side sensing is used, and isn't subject to power supply changes. That makes the downstream electronics much easier.

It is easier and does work pretty well sometimes