EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: Twistx77 on April 04, 2014, 05:23:34 pm
-
Hi,
I'm working in a current sense amplifier and I want to get a good precision out of it. I plan to measure form 1mA to 1A with at most 0.5% of error. To "get rid" of the input offset of the amplifier. I plan to use an off the shelf solution form TI or LT or whatever. I've seen that some of them can go as low as 20uV or 10uV. But even with an input offset so low, when measuring currents under 50mA the error is too high for a low value shunt (around 10mOhms).
One solution that I think that might work for keeping a low dropout in the shunt and keeping the error relatively low would be having two low RDSon MOSFETs controlling to amplifiers and two shunts. Then I would read the amplifier for the wider range (100mA to 1A) if the current was lower than lets say 50mA I would turn off the MOSFET for the (10mOhm) and I would turn on the second MOSFET with a 1 mOhm shunt or so. Since the current will be low enough the dropout won't be that high.
Does anyone else have a better idea or solution for my project?
-
I would turn off the MOSFET for the (10mOhm) and I would turn on the second MOSFET with a 1 mOhm shunt or so. Since the current will be low enough the dropout won't be that high.
What's the mosfet's Rds? What if it changes? ...
A great idea on book. Not so much in practice.
Does anyone else have a better idea or solution for my project?
There are reasons why 0.5% or below accuracy current sense amplifiers are atypical. If you don't find out why, you probably will struggle making a better one.
-
The RDSon of the MOSFET I'm planning to use is around 39mOhms and is pretty constant. For the current measurement shouldn't be a problem since I will be measuring directly on each shunt.
-
I'm very interested in this. I have need of a bank (10) of high precision current sensing amplifiers for a project -- high voltage (~100v), low current (0.1 - 10 mA).
-
if the current was lower than lets say 50mA I would turn off the MOSFET for the (10mOhm) and I would turn on the second MOSFET with a 1 mOhm shunt or so.
I assume you mean 100 mOhm :)
Since the low RDSon FETs are expensive I'd use one FET to bypass the 100 mOhm shunt for high current mode and then two FETs to select which voltage to sent to the amplifier.
-
The best operational amplifiers and integrating analog to digital converters have an offset voltage of 0.5 microvolts so even with them, the needed burden voltage to stay within 0.5% would be 100 microvolts which would require a 10 ohm shunt at 1 milliamp full scale. The offset voltage could be calibrated out for an order of magnitude improvement but thermocouple effects will become a problem and I am being conservative.
Using power MOSFETs to switch in different shunts assuming that the measurement is *only* made across the shunt will work but of course the MOSFETs will add to the burden voltage.
If you want a super low burden voltage, what about placing the shunt on the other side of a power MOSFET virtual ground? Then the burden voltage can be arbitrarily small while using a larger shunt. You would still need to switch shunts but this would allow using a much higher shunt resistance.
-
Does anyone else have a better idea or solution for my project?
What do you plan to do with the output of your amplifier, D/A? Integrating current? Integrating power?
You may want a more integrated approach with ICs like http://www.linear.com/product/LTC2942 (http://www.linear.com/product/LTC2942) and http://www.ti.com/lit/ds/symlink/ina219.pdf (http://www.ti.com/lit/ds/symlink/ina219.pdf)
-
If you want a super low burden voltage, what about placing the shunt on the other side of a power MOSFET virtual ground? Then the burden voltage can be arbitrarily small while using a larger shunt. You would still need to switch shunts but this would allow using a much higher shunt resistance.
AFAICS you would need a negative voltage supply for this ... also if you could put the shunt low side you wouldn't be using a current sense amplifier.
-
If you want a super low burden voltage, what about placing the shunt on the other side of a power MOSFET virtual ground? Then the burden voltage can be arbitrarily small while using a larger shunt. You would still need to switch shunts but this would allow using a much higher shunt resistance.
AFAICS you would need a negative voltage supply for this ... also if you could put the shunt low side you wouldn't be using a current sense amplifier.
Or a floating power supply. The application is not real clear.
A current sense amplifier is needed in this case no matter which rail is used because of the low signal level unless an integrating type of analog to digital converter is used.
-
A current sense amplifier is needed in this case no matter which rail is used because of the low signal level unless an integrating type of analog to digital converter is used.
I assumed he was using a current sense amplifier as in a product marketed as such, which are differential and don't make much sense low side, more than the function.
-
A current sense amplifier is needed in this case no matter which rail is used because of the low signal level unless an integrating type of analog to digital converter is used.
I assumed he was using a current sense amplifier as in a product marketed as such, which are differential and don't make much sense low side, more than the function.
They do not have to be differential in the sense of an instrumentation amplifier. A current mirror can be used to create a ground referenced single ended signal from the current shunt:
http://dangerousprototypes.com/2011/04/17/app-note-parallel-led-drivers-for-maximum-current/ (http://dangerousprototypes.com/2011/04/17/app-note-parallel-led-drivers-for-maximum-current/)
High accuracy current monitors replace one half of the current mirror with a single operational amplifier:
http://www.linear.com/product/LTC6101 (http://www.linear.com/product/LTC6101)
-
Hi,
Thank you to all. I haven't explained my self well and I haven't given my requirements neither.
The dropout can be as high as lets say 100mV but I would prefer it not be over 50mV.
I will be powering everything from 2 AA batteries so the simpler the power scheme, the bettter :)
I think that I might be using this amplifier: http://datasheets.maximintegrated.com/en/ds/MAX9934.pdf (http://datasheets.maximintegrated.com/en/ds/MAX9934.pdf) for my low voltage application and then I will choose another one for a higher voltage application since the higher voltage ones hacer higher input offset voltage.
They do not have to be differential in the sense of an instrumentation amplifier. A current mirror can be used to create a ground referenced single ended signal from the current shunt:
That is a good idea.
I've made a simple schematic of the setup I want to use it.
(http://www.yoimg.com/di-TTKS.png)
-
They do not have to be differential in the sense of an instrumentation amplifier. A current mirror can be used to create a ground referenced single ended signal from the current shunt:
http://dangerousprototypes.com/2011/04/17/app-note-parallel-led-drivers-for-maximum-current/ (http://dangerousprototypes.com/2011/04/17/app-note-parallel-led-drivers-for-maximum-current/)
In this case we have a product marketed as a current mirror, but used in the function of a differential amplifier.
-
I've made a simple schematic of the setup I want to use it.
You may want to think about what it takes to turn on those mosfets fully; and in that case, what kind of common mode voltage will the opamp see and if it is designed to handle that.
Current mirrors can be used to level shift but their ability to hold as a current mirror is usually poor at high current levels.
Again, there are valid reasons that 0.5% or better current sense amplifiers are not common.
-
Again, there are valid reasons that 0.5% or better current sense amplifiers are not common.
Can you point us in the direction of a reference paper / note on this? I'm interested in learning more.
-
I think I will just go for 1% or 5% at the worst. I have calculated the error for the current setup and I get less than 1% from 1A to abut 10mA, below that I the error increases a lot because of the input offset voltage.
-
At high currents the accuracy will depend on the temperature coefficient of resistance of the current shunt. 100 ppm/C is 0.01 %/C so assuming a 20 C temperature range, I would design for a maximum of 0.2% accuracy at the low end. To do better, a much more expensive current shunt is needed (10 times more expensive) in which case extra expense for accuracy in the rest of the circuit becomes less important unless you have to use more than one shunt for range switching.