Kelvin sensing for current measurement

[anvoice]

No. A 4-terminal resistor gets the copper of the load path out of the sensing path, so it's temperature coefficient doesn't affect measurements.

My understanding is that since the 4-terminal resistor is a monolithic device, it will still get hot where the sense connection is. Thus introducing some temperature coefficient at least. Perhaps it just doesn't get as hot, which is why its apparent temperature coefficient is lower.

It seems you fail to understand what a 4 terminal shunt is all about.

With the sense points of a 4 terminal shunt being right on the resistive element, they are not sensitive to the temperature dependent voltage drops along the copper based leads carrying the load current. Those leads have just as bad a temperature coefficient as a 2 terminal shunt, but it doesn't affect the sensed voltage drop across the resistive element. That is the reason 4 terminal devices are manufactured.

Yes, perhaps I do. That's the reason I'm asking questions in the newbie section.

Thank you for clarifying.

[Eka]

At 30 Amps, that's a 0.15V drop across the resistor, and 9 Watts dissipated. Not many surface mount shunt resistors go above 5W. I'd go to 0.0025 Ohm or less.

There are high side current shunt amplifiers that have 0.1% accuracy amplification stages that are designed to use the really small shunt resistor values to lower energy loss and still provide enough accuracy. Pay attention to the maximum mV the input stages can accept. The one I plan on using only accepts up to 83mV shunt resistor voltage. Using a 0.001 Ohm shunt resistor with it, I get a usable 30 Amps monitoring capacity. I'm putting an input filter and protection circuit between the shunt and current shunt monitoring chip. I loose some accuracy and range of measurement, but I'm still better than 1% accuracy before calibration. It also means I can accidentally short my monitored rail without worrying about blowing the input stage of the amplifier.

[anvoice]

At 30 Amps, that's a 0.15V drop across the resistor, and 9 Watts dissipated. Not many surface mount shunt resistors go above 5W. I'd go to 0.0025 Ohm or less.

There are high side current shunt amplifiers that have 0.1% accuracy amplification stages that are designed to use the really small shunt resistor values to lower energy loss and still provide enough accuracy. Pay attention to the maximum mV the input stages can accept. The one I plan on using only accepts up to 83mV shunt resistor voltage. Using a 0.001 Ohm shunt resistor with it, I get a usable 30 Amps monitoring capacity. I'm putting an input filter and protection circuit between the shunt and current shunt monitoring chip. I loose some accuracy and range of measurement, but I'm still better than 1% accuracy before calibration. It also means I can accidentally short my monitored rail without worrying about blowing the input stage of the amplifier.

Could you clarify how you got that number? The formula I'm using is just I²R, which gives 4.5W of power dissipated, not 9. I agree that I won't find an SMD device for 9W easily.

I have low-side amplifiers in my setup (DRV8323SRTAR), those have a max 40V/V gain but I'm still figuring out the equations for the resistor value.

One of the issues is that I'm still trying to get the motor supplier to give me the stall current of the motors. If he doesn't, I might have to wait until they arrive to measure it myself, so at the moment the 30A is an approximation. If I size the electronics to just barely accommodate 30A and it turns out to be 35, I'd not be in good shape.

[BravoV]

I've had similar problem, its just not possible for that kind of current with tiny small resistor, think of the heat dissipation generated. One of the reason I did buy few low ohm precision resistor like these below, as it has big chunk of metal body to dissipate the heat away and to maintain good accuracy of the resistance. Bought used as I can't afford new one.

An example below, rated 20 Watt on free air, or 50 Watt with big heatsink  , say at 20 Watt at 0.01 Ohm, that is max at 200 Amp without affecting the rated resistor accuracy significantly with tempco of 50 ppm/K and resistance accuracy at 0.1%.

The smaller rod is the Kelvin sense terminal. Shot with a TO-220 sized chip to give an idea how big it is.

[anvoice]

That's an impressive piece of hardware.

I don't have the luxury of using something that massive for my current sense resistors (space constraints, and I need 3) so I'll probably go with SMD. The 20ppm TCR from Vishay look promising, but I'm limited to 1mOhm resistance if I want to measure 30A or so (they dissipate 1W tops).

Another concern is the maximum conduction of the PCB trace, which I'll be reading about shortly. Even if I minimize all track lengths, that's a lot of current to put through the board.

[Eka]

At 30 Amps, that's a 0.15V drop across the resistor, and 9 Watts dissipated. Not many surface mount shunt resistors go above 5W. I'd go to 0.0025 Ohm or less.

There are high side current shunt amplifiers that have 0.1% accuracy amplification stages that are designed to use the really small shunt resistor values to lower energy loss and still provide enough accuracy. Pay attention to the maximum mV the input stages can accept. The one I plan on using only accepts up to 83mV shunt resistor voltage. Using a 0.001 Ohm shunt resistor with it, I get a usable 30 Amps monitoring capacity. I'm putting an input filter and protection circuit between the shunt and current shunt monitoring chip. I loose some accuracy and range of measurement, but I'm still better than 1% accuracy before calibration. It also means I can accidentally short my monitored rail without worrying about blowing the input stage of the amplifier.

Could you clarify how you got that number? The formula I'm using is just I²R, which gives 4.5W of power dissipated, not 9. I agree that I won't find an SMD device for 9W easily.

I have low-side amplifiers in my setup (DRV8323SRTAR), those have a max 40V/V gain but I'm still figuring out the equations for the resistor value.

One of the issues is that I'm still trying to get the motor supplier to give me the stall current of the motors. If he doesn't, I might have to wait until they arrive to measure it myself, so at the moment the 30A is an approximation. If I size the electronics to just barely accommodate 30A and it turns out to be 35, I'd not be in good shape.

Oops, looks like an error in the spreadsheet I wrote in the middle of the night a few weeks ago. Kinda interesting that I had worst case Watts calculated correctly, and nominal Watts wrong. After fixing the inputs, I also get 4.5 Watts now. Still, there should be more than a half Watt of headroom. That resistor will be heating up a lot, and that will effect accuracy.

Now, for over current for a motor, high accuracy isn't as much of a need. A few % off is OK. Is a 4 terminal resistor even needed? Yeah, if you go with a 2 terminal do the traces for the kelvin sensing like the AMD app note suggests to retain as much accuracy as possible, but do you need sub 1% accuracy, or even sub 2%? I see +/-5% talked about for some control resistors in the spec sheet. Also most motor control circuits are usually cost sensitive, not high accuracy sensitive. If they need accuracy, they use some form of feedback loop.

BTW: A 0.005 Ohm 5W 0.1% 4 terminal surface mount shunt resistor is available at DigiKey. Vishay Foil Resistors part number Y14740R00500B0W. Also looks like you can get a 1% 4 terminal surface mount shunt resistor for a fraction of the cost.

On kelvin sensing for 2 terminal resistors. Not all 2 terminal sense resistors are packaged so the terminals are under the resistor element like the one the AMD app note used. Some have them flanking the resistor to each side, see Stackpole Electronics Inc. part number CSSH2728FT5L00. From reading between the lines, they would be best to have the sense wires to come out the inside sides of the pads. That would allow for no vias in the kelvin sense lines.

[anvoice]

Still, there should be more than a half Watt of headroom. That resistor will be heating up a lot, and that will effect accuracy.

Now, for over current for a motor, high accuracy isn't as much of a need. A few % off is OK. Is a 4 terminal resistor even needed? Yeah, if you go with a 2 terminal do the traces for the kelvin sensing like the AMD app note suggests to retain as much accuracy as possible, but do you need sub 1% accuracy, or even sub 2%? I see +/-5% talked about for some control resistors in the spec sheet. Also most motor control circuits are usually cost sensitive, not high accuracy sensitive. If they need accuracy, they use some form of feedback loop.

BTW: A 0.005 Ohm 5W 0.1% 4 terminal surface mount shunt resistor is available at DigiKey. Vishay Foil Resistors part number Y14740R00500B0W. Also looks like you can get a 1% 4 terminal surface mount shunt resistor for a fraction of the cost.

On kelvin sensing for 2 terminal resistors. Not all 2 terminal sense resistors are packaged so the terminals are under the resistor element like the one the AMD app note used. Some have them flanking the resistor to each side, see Stackpole Electronics Inc. part number CSSH2728FT5L00. From reading between the lines, they would be best to have the sense wires to come out the inside sides of the pads. That would allow for no vias in the kelvin sense lines.

My understanding was that the package dissipation spec does include some headroom. e.g. if you're picking 1mOhm with 30A current, dissipating about .9W, the 1W would be up to specs for that. Don't see the point of saying "1W" if it really can't handle that without accuracy deteriorating significantly.

I'd like to do both reasonably accurate torque control and data logging if possible. That's why I think around 1% is a more reasonable number than 5%. Right now I'm designing around Mouser part 71-WSKW06121L000FEA, which has a remarkably good TCR and is cheap. Assuming my current doesn't actually go above 30A, it seems like it'll do a good job.

The cost on that Digi-Key one is kind of prohibitive for 3 resistors per board... Plus they have 4 total that can ship, while I need at least 15 (again, wouldn't be cheap).

[coppice]

My understanding was that the package dissipation spec does include some headroom. e.g. if you're picking 1mOhm with 30A current, dissipating about .9W, the 1W would be up to specs for that. Don't see the point of saying "1W" if it really can't handle that without accuracy deteriorating significantly.

Resistor ratings are quite complex, because the resistor doesn't exist in isolation. If they rate the resistor as 1W, it usually means that it should be able to dissipate 1W out in the open air. Constrict the air around it, and it will be able to dissipate less. Also, check the temperature that the resistor is expected to reach when dissipating 1W. Some spec have the resistor running really hot at the rated power dissipation.

[anvoice]

Resistor ratings are quite complex, because the resistor doesn't exist in isolation. If they rate the resistor as 1W, it usually means that it should be able to dissipate 1W out in the open air. Constrict the air around it, and it will be able to dissipate less. Also, check the temperature that the resistor is expected to reach when dissipating 1W. Some spec have the resistor running really hot at the rated power dissipation.

Makes sense. I can't model my heat dissipation easily since the resistors will be in an (aluminum) enclosure and near other potentially hot components, such as the motor and MOSFETs, so it may be good to have some headroom there. The only issue here is that other resistors I found for a similar price have a much higher TCR, so even if they're larger and run at +30C (unlikely since other components will heat the air around them), their TCR which is over 3 times higher would still make this a wash with the 1W variant running at +100C, if not worse.