Cheap current sense

Rotten. The copper thickness isn't going to be all that well controlled, and you'd need a relatively wide, short trace to get any sort of decent precision on the width.

You can buy 0.01R surface mount resistors, from 11 cents (Australian) upwards.

That part is 1%, 100ppm/K.

There were lots around 37 cents

which buys you a 70ppm/K temperature coefficient.

For a dollar you could get a four terminal part with 50ppm/K

There was a $38 part at the op end of the list ...

--
Bill Sloman, Sydney

remember copper is ~0.4% per C

how much is a 10mOhm resistor 5-10 cent?

-Lasse

to be

of

g

%? 10?

The R change due to self heating can be ignored in as far as it only create s some nonlinearity, as long as you dont mind the settling time. So you onl y need consider ambient temp. And that's just a matter of measuring another copper trace. In fact the latter might solve the possible varying thicknes s issue too. A resistor may be cheaper though :) .

NT

Besides all the other stuff, temp coef, etc. How thick is the "1 oz". if it's an outer layer... I'd guess +0% to +40%. Inner layers may be better, (thickness wise) What kind of thickness spec do US pcb places offer?

Manginin, or some other alloy might be a better choice.

George H.

Of course it depends on the process (often boards start with 1/2oz and plate up from there to whatever but I'd expect it to be all over the place. I've had boards come back with traces 1000% of nominal resistance - so bad the boards failed ICT (3ohm traces).

Rohm is an even better choice.

You want to make a resistor from a copper trace? It won't be very accurate unless you calibrate around it, and TC will be bad, about

+0.4% per degree C. Low-ohm surfmount resistors are pretty cheap.

I often include resistance/impedance test traces on boards. A "1 oz" trace can be +10, -25% of calculated resistance, trending towards the thin side of that spread, from a good US pcb house.

They tend to start with thinner copper and plate up, and sometimes skimp on that. It also could be that plated copper has higher resistivity than original foil.

Sometimes I say in the fab notes START WITH 1 OZ COPPER to make sure I have enough copper thickness.

Years ago, I did this with a small piece of constantan wire bent into a U and soldered into the PCB. Now it's far cheaper to just buy a 10m ohm resistor. They're used in switching power supplies for current sense resistors, so pretty cheap and tight tolerance, not to mention avoiding the +0.385%/K tempco.

Eg. CRF0805-FX-R010ELF from Bourns- < 10 cents in 100's, 1% tolerance and better than +/-100ppm tempco.

--sp

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward" 
speff@interlog.com             Info for manufacturers: http://www.trexon.com 
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

Wow! I barely get +/5% accuracy WITH adjusting manufacturing processes to get controlled transmission line impedances.

Outside trace gets extra 'stuff' so that will vary more than an internal trace. 0.01 ?! what exactly is 1 oz. copper's ohms per square? with 0.7 mil to 1.4 mil yields 1 milliohm per square to 0.5 milliohm per sq? Hmmm

*if* that is correct, you need 10 squares, or 20 sq's, minimum width of 5 mils, means 50 mil length, but then you have current carrying capacity, with that small a resistance sounds like might be amperage, so you need 100 mil wide trace for decent ampacity. That leaves you with 1 inch trace to get ten squares. yeah seems reasonable, as long as you can stand the tempco, and the initial tolerances. really, really cheap you don't have any uC or uP on that board to measure somewhere else with some left over intelligence. If you did, you could look at what IC manufacturers do faced with +/- 30%, awful tempco resistors.

Check with Rogers on tolerances, but from memory that varies more than

+/10% which seems tighter than reality. you said cheap? cheaper the PCB base the worse the tolerance too. Do a list yourself: ohms per square +/- 20 manufacturing dimen -5 +10 plate, or no plate -30 +0 that kind of thing, though first thing I would do is call my PCB house and ask THEM. Last time I did that, I was unpleasantly surprised.

I would not expect the accuracy to be very good, But does your board invol ve a microprocessor? Could you measure the resistance and then enter a val ue in the software? Or could you make another pcb board trace vary the gai n so that a high shunt resistance is compensated by lowering the gain of th e circuit amplifying the current signal?

If you are only going to make a few boards, it would be cheaper to just use a resistor and save on the engineering time.

Dan

you'd need a relatively wide, short trace to get any sort of decent precision on the width.

To get .01 ohm resistance you will need a trace that will have the same proportions. If you make it wider , you will have to make it longer to get the same .01 ohm. See ohms per square.

Dan

Right. Thanks. I wasn't thinking hard enough. The point I had in mind was that you wouldn't want to use a narrow trace because the tolerance on the trace width is pretty much independent of the actual width of the trace.

--
Bill Sloman, Sydney

And you have the nerve to criticize others with their use of English.

Jamie

I use a thick track and take 'nicks' out of it to calibrate. If you have the space you can lay down multi tracks in parallel and cut through one or two to get the accuracy.The strips will current share fine batch to batch.

?

unless you put zero value on your time calibration is going to much more ex pensive than a 10cent resistor

-Lasse

Maybe you could program your processor to calibrate it. To reduce R, apply enough i squared to to produce carbon. To increase R, apply enough i squared t to vapourise a little copper :)

NT

Or save a cal factor in the processor!

Calibration (especially if done digitally with an EEPROM or whatever) can be folded into testing, but it's an added complexity.

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward" 
speff@interlog.com             Info for manufacturers: http://www.trexon.com 
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

Den mandag den 7. juli 2014 00.24.59 UTC+2 skrev Spehro Pefhany:

be

ng

10?

ve

e

expensive than a 10cent resistor

yes if done digitally as part of testing it might be almost free, but not if done manually by cutting tracks

-Lasse

Jamie does have delusions about his competence. Happily for the tattered remnants of his reputation, he has refrained from telling us what he thought that he could object to.

--
Bill Sloman, Sydney