Current Draw at Power Supply vs Current Measured across a series shunt

[steamedhams]

I am measuring power consumption of a pcb. The current draw of the power supply reads 0.5 amps.
When I measure across a series shunt on the board (1 mOhm), the voltage across is 2.7 mV which means 2.7 amps are flowing across that resistor.   

What is the deal here?? There are a couple of boost/bucks converters on the board, regulators, inductors...etc Is that jacking up the current??

Thanks in advance   

[SuzyC]

Either your power supply or  your shunt resistor or your mV measuring device is lying to you. Check them out.

[Gyro]

A couple of random thoughts in order of likelyhood...

1. Make sure you are measuring directly across the resistor terminals leads, ie. not sharing solder joints or pads with the current connections, at the 1mOhm level, this can make a huge difference. You need 4 wire aka Kelvin connection. Make separate connections for current and measurement.

2. Your current draw could be very spiky in nature and your measuring device isn't handling it well.

3. Thermocouple effects - check the reading on the meter with the power off and make sure you're getting a zero reading.

My bets are on 1 though.

[SuzyC]

My bet is on 2.

[Gyro]

Yes, it's one or other. Unless the meter has a serious zero offset, thermocouple effect would probably top out at the 10's of uV level. 

[BeBuLamar]

I would say the voltage measured at the shunt resistor is the wrong one. With such a small resistor adding extra resistance to the setup can easily happens. Plus the fact that measure a couple of mV is difficult to get accuracy. I personally do not have a meter that can measure 2.7mV with the certainly of +/- 1mV.

[tggzzz]

What is the tolerance of the 0.001ohm resistor?
What is the tolerance of it and its connections when soldered into the circuit?

If you try to measure 0.001ohms, you will find that it is not an easy measurement to make with any degree of precision and repeatability.

[Gyro]

1mOhm is certainly at the bleeding edge for that sort of current. I can understand that you want to minimise voltage burden but it's probably a step too far. My rule of thumb is to keep the burden at around 10mV in the current range you're interested in, although I do use a 1uV resolution meter. The burden of a standard 100mOhm DMM shunt is too high for most psu applications. Using a separate shunt, as you're doing, is definitely the preferable method.

I would be looking at 10mOhm for 0.5- 2A sort  of measurement range. These are easier to find in 4 terminal configuration with reasonable (1%) accuracy.


Edit: I should qualify all of the above with 'where the burden voltage is a consideration'. Obviously for something like a 1.8 or 3.3V rail, 100mV drop would be excessive. Much less so with regulators being powered from a bench PSU.

[exe]

I found that at least two of my dmms cannot reliably measure millivolts (an8002 and an8008). Like, reading is changing when I wiggle the dial. The error I had was +0.5-1.5mV to the "actual" voltage. So, I'd check capabilities of your gear .

And, as others mentioned, if it's not a dc voltage, then you can't rely on dmm without some hardware filtering. I found that my dmms have problems with DC offset when measuring noisy signals (can't blame them for that). So, an oscilloscope is your friend, unless the signal is below the noise floor

[coppice]

At 1 milli-ohm you need to take some serious care with layout if you are using a 2 terminal resistors. 4 terminal resistors help enormously, but are expensive. https://www.analog.com/en/analog-dialogue/articles/optimize-high-current-sensing-accuracy.html has some good information about suitable layouts for 2 terminal resistors. Note that you really have to get all the copper out of the measured path, as copper has a temp coeff of 0.4% per degree, so even if you compensate for the resistance of the copper being added to your 1 milli-ohm resistor, you still get significant error as the temperature changes.

[BeBuLamar]

For measuring power consumption the burden voltage isn't a problem. Measure the voltage across the PCB and increase the supply voltage to have to correct voltage at the PCB. In this case even the burden voltage is at high as 1V wouldn't be a problem.

[David Hess]

1. Make sure you are measuring directly across the resistor terminals leads, ie. not sharing solder joints or pads with the current connections, at the 1mOhm level, this can make a huge difference. You need 4 wire aka Kelvin connection. Make separate connections for current and measurement.

That is what I think also.  I consider a 4-wire current shunt resistor to be mandatory in this application.

It might be possible to do without one by calibrating the current sense output with a known current source but the temperature coefficient will be poor.

[coppice]

1. Make sure you are measuring directly across the resistor terminals leads, ie. not sharing solder joints or pads with the current connections, at the 1mOhm level, this can make a huge difference. You need 4 wire aka Kelvin connection. Make separate connections for current and measurement.

That is what I think also.  I consider a 4-wire current shunt resistor to be mandatory in this application.

It might be possible to do without one by calibrating the current sense output with a known current source but the temperature coefficient will be poor.

4 terminal resistors are certainly an easy solution. However, you can get great results with cheap surface mount resistors if you take the information in https://www.analog.com/en/analog-dialogue/articles/optimize-high-current-sensing-accuracy.html seriously. I've worked with 2 terminal SMD resistor shunts down to 500 micro-ohms, and achieved good results using a similar approach to the suggestions in that Analog Devices paper.

[David Hess]

4 terminal resistors are certainly an easy solution. However, you can get great results with cheap surface mount resistors if you take the information in https://www.analog.com/en/analog-dialogue/articles/optimize-high-current-sensing-accuracy.html seriously. I've worked with 2 terminal SMD resistor shunts down to 500 micro-ohms, and achieved good results using a similar approach to the suggestions in that Analog Devices paper.

The problem I was referring to is the +600 PPM/C temperature coefficient of copper which quickly dominates the errors.

I notice that Analog Devices didn't show the method I have used in the past with the sense connections coming of of the inside of the resistor pads, but it is not suitable with smaller parts.

[Tomorokoshi]

What is the voltage of the power supply to the board with the switcher?

[Gyro]

It would be really nice to hear back from the OP about now.

It would be helpful to know...

1. What concerns do you have about burden voltage - why the choice of a 1mOhm resistor?
2. Are you making use of the resistor in circuit, or is it just there for external monitoring purposes?
3. What technology are you using? We don't even whether it is through-hole or smd (although I think even zero-ohm resistors are more than 1mOhm?)
4. where exactly are you placing your probes?
5. What instrument (make/model?) are you measuring with?
6. A Photo would be really helpful!

I think we've reached the stage of stabbing in the dark with the information that we currently have.

[exe]

Oh yeah, photos please!

PS never heard of stabbing in the dark (I usually say shooting in the dark). Is it a British saying?

[Gyro]

PS never heard of stabbing in the dark (I usually say shooting in the dark). Is it a British saying?

I guess it must be - you tend to forget the gory derivation of some of these sayings - it must date back to the Victorian era, probably much earlier!

We have another saying: "I'll take a stab at it" as in I will try (not necessarily with expectations of success). A quick web search suggests that it derives from the late 17th century.

[ledtester]

I wonder what they use in the US

The phrase "shoot(ing) from the hip" carries a similar connotation.

[Gyro]

Oops, sorry. I edited that bit out for fear of stirring up something.

[coppice]

4 terminal resistors are certainly an easy solution. However, you can get great results with cheap surface mount resistors if you take the information in https://www.analog.com/en/analog-dialogue/articles/optimize-high-current-sensing-accuracy.html seriously. I've worked with 2 terminal SMD resistor shunts down to 500 micro-ohms, and achieved good results using a similar approach to the suggestions in that Analog Devices paper.

The problem I was referring to is the +600 PPM/C temperature coefficient of copper which quickly dominates the errors.

I notice that Analog Devices didn't show the method I have used in the past with the sense connections coming of of the inside of the resistor pads, but it is not suitable with smaller parts.

Aren't the C and D layouts they show exactly that? Even small amounts of copper can be a disaster for the temperature coefficent, as I said earlier in this thread, but a sloppy approach to the tap off point can easily make a 1 milli-ohm resistor look more like several milli-ohms, which is an even bigger error. I've seen it happen. Try what Analog suggest with 0805 SMD resistors or larger, and the results can be very good. I'd recommend using a 1206 or larger, though.

We researched pretty much what Analog show in the link, and came to the same conclusions about what is practical. They published just as we were preparing our writeup, so we didn't bother publishing it.