# Overcurrent Sensing

I am designing an overcurrent protection circuit. I am sensing current with a LEM sensor across a sensing resistor of 16.5 ohm. The voltage drop for maximum current is 2 V. In the next stage I am using INA301A1 (CMVR 0 - 36 V, unipolar power supply 0 V / 3.3 V). This IC has a fixed gain of 20 V/V. Now for full scale drop of 2 V, my output will be saturated (40 V), so I need to I need to step down 40 V To 3.3 V.

What is the best way to do it?

I am thinking about using a potential divider between sense resistor stage and overcurrent IC stage but using a lower (10s of ohm) resistance divider may load my sense resistor and higher (kohms) might load input of IC.

Are these apprehensions plausible and what could be a possible solution?

What is the best way to do it?

The best way is to do what every engineer is doing in circuits like this - using as small shunt resistor as possible. 16 Ohm is not a "sensing resistor", it is unnecessary load. Current shunts typically in range of a few mOhm to 1 Ohm, which results in correct voltage drop in mV range (250 mV for INA301A1).

From your numbers the maximum current seems to be 120 mA, so 2 Ohm shunt will give you correct 250 mV voltage drop.

UPDATE

There was a misunderstanding corrected in the comments below. It turns out the OP does not measure voltage on shunt resistor in the actual high power circuit, but rather voltage drop on burden resistor of current transducer.

Taking this new information into account two solutions were suggested:

• adjust burden resistor to bring voltage swing into full-scale range of ADC input and use any OP Amp in comparator configuration for overcurrent signal generation;

• same as above but drop the overcurrent signal altogether and use MCU comparator input to generate interrupt on threshold. This would eliminate all external components except for one burden resistor.

All of the above assumes uni-directional current, implied by original use of INA301 chip. Bidirectional sensing requires more complex circuit, as discussed in @TonyEErocketscientist answer.

• I am using a LEM sensor with 1:2500 turns ratio which gives me a max sec. current of 120mA. So 16.5ohm limits my losses to 250mW. So I don't think it's an unnecessary load. I do agree that I can further decrease the resistance to reduce input voltage drop as you suggested ( also because the input offset voltage is very low and gives me that liberty while maintaining high measurement accuracy) but the sensor I am using is LA305s (LEM hall effect sensor) and it's datasheet specifies a lower limit of 5 ohm for measurement resistance (due to thermal protection of output power stage of transducer) Sep 17 '18 at 22:56
• OK, now I am really confused. What on earth do you need that INA301 for, if LA305 already provides voltage suitable for ADC? Simply adjust your resistor to get full ADC range and be done with it. Sep 17 '18 at 23:09
• Yes, you got it right. Or, better yet, use comparator input if your MCU has one. It can be programmed for interrupt on threshold. I believe this would be simplest, cheapest and easier to tune-up solution. As for OP Amp selection in case you go with external comparator, I think @TonyEErocketscientist would be able to give much better advice. My guess, rail-to-rail inputs with single supply 18V and up should do the trick. Something like LT6015, OPA191, OPA197 etc. Sep 18 '18 at 6:15
• I'm thinking any 1 of quad Op Amp will do as it is low speed with single supply and Rail to Rail. But I would do things differently with bipolar logic OCP using a window detector to detect both polarities or a full wave precision rectifier with one threshold for logic level out and signal condition the Diff Amp to match the signal spectrum of all events needed to be detected. Sep 18 '18 at 6:21
• That would be definitely necessary if bidirectional sensing required. Shifting the drop to ADC range and implementing OCP in both polarities could be quite complex circuit, comparing to simple LA305 + MCU combination. I guess its a choice between user-friendly and budget devices. Sep 18 '18 at 6:31

Rather than using a fixed gain INA, use a gain and offset DIff Amp OA design to scale your input to the desired output.

if 0 to 2V in and 0 to 3.3V out then non-inverting gain = 1.65

If -2V to +2V in and 0 to 3.3V out then gain = 0.825 and offset = +1.65V (=Vss/2)

I suspect you want to measure high current AC primary so you want to measure both polarities > +/- x.xx V so you compute dual thresholds y using Vdd/2 as a null reference.

But keep in mind radiated CM noise. Having 120 dB CM conducted noise rejection is wasted if the cable is only 50dB CMRR with a 500A impulse going thru a Rogowski coil, so choose your cable wisely. that is already 50dB above your current sensor secondary. Getting 100 dB attenuation of Injected H field current impulse near the cable may depend on doubled shielded twisted pair with a Pi filter CM Choke and orthogonal orientation. This is what I have had to use in the past with monitoring 10kA arc welders.

## opinions and design specs?

For the OCP logic output, I would ask if real-time and polarity is important. Because, knowing the high Remanence Currents from saturation on power up due to the phase difference and resulting magnetization V difference between power input stop and restart, start current can be quite large if starting 180 phase shifted and make some Xfmr's "hum" for a long time until unsaturated or return to zero averarge magnetization current. Also Arcs can be polar so a single-sided OCP logic output does not use all the useful information available on different faults.

• I agree with you. I picked INA because it ensures inbuilt impedance matching for differential amplifier and also gives a digital output for overcurrent which can set with a single resistor Sep 17 '18 at 23:11
• Yes but so does a properly designed diff amp with 3 stages in a quad OA. and is not limited to 20 Sep 17 '18 at 23:12
• Ok, got your point Sep 17 '18 at 23:15
• @TonyEErocketscientist I believe INA301 does not work with AC. Even though pins can tolerate +/- 40V the differential input is specified 0..250mV, so current is one direction only. The overcurrent output is open drain Sep 18 '18 at 5:07
• Yes I see now the dual output, but comparator is only +ve polarity inputs Vin*20> DAC input threshold. This could be done with 0.1% R divide gain = 0.825 /20, but OCP assumes symmetrical currents ( unlike random Remanence effects from Xfmr supply outage ) and glitches missed such as an arc. Sep 18 '18 at 5:36