High-side current monitoring

[charliecoutas]

Has anybody used a Linear Technology LT6015/6/7 chip for measuring the current in a shunt which is "above" the load - high side? I have struggled with the circuit they suggest but I get massive overshoot. My application is measuring the current during pulses from 10uS to many mS.

Charlie

[top204]

Make sure you have the shunt resistor correct.

For low currents, a higher value resistor is required, otherwise, the Op-Amp cannot amplify the very low voltage across it. The same applies for higher currents, where a low Ohmage resistor is required, otherwise, the amplifier overloads because too much amplification is performed.

I have used the INA219 device in several circuits and for very low currents (10s of uA), I used a 1 Ohm shunt resistor, and for higher currents, I used a 0.1 Ohm resistor. I found that with a 1 Ohm shunt resistor, higher currents overflowed the amplifier within it

[JonW]

What's the load Charlie?  Is it inductive at all?

I have used INA293x1 and think these are one of the fastest dedicated highside monitors around.  As Les has stated the BW will be related to the Gain, so depending on your current range you will be better keeping a lower gain to achieve the fastest response. If you have a high gain, then step response can easily saturate the opamp causing overshoot and will need require time to recover.  The settling time in the data sheet assumes certain conditions for step responses and that the input is at mid point, thus keeping the opamp in its linear region.

[LeonJ]

Hi Charlie,
Have you looked at Allegro's Current Sensor IC's (ACS71240)? It uses Hall Effect and is completely isolated so the current can be measured high-side, low-side, anywhere.
Leon.

[shantanu@india]

Hi Charlie,
My advise would be not to use an OP-Amp with a shunt unless you absolutely have to....too many uncertainties.
Use a Hall Sensor based Allegro or equivalent device as suggested by Leon...ACS family is extremely good & I have used them in the past.Isolated current measurement is much better...all Variable Frequency Drives use this technology.
Shunt based current measurement is advocated only if the current value is small enough that it does not cause a "bump" in the ground plane which is sure to get reflected in your ADC value.

[charliecoutas]

Thanks guys, what a lot of useful help.

The circuit I am having trouble with is from the Linear Tech datasheet. The circuit on the left suffers massive overshoot, My shunt is 0.05 Ohms, load is 1 Ohm to 35 Ohms, current from 0.5 to 10A. I have just tried the one on the right and it looks pretty good, no overshoot. They show a gain of 100 but I only want a gain of between 4 and 8.

I'll look into your excellent suggestions.

Thanks
Charlie

[JonW]

The issue Charlie will have if he uses the hall devices is the BW, these will likely not work with very narrow pulses (10us) 

You can scale the shunt resistor for the gain or the resistors to meet your gain.  Reduce all the resistors by multiplying them by 0.0(gain).  Ex for a gain of 8, multiply them by 0.08. If you are not that bothered about the max CMMR then use the same 1% or 0.1% resistors rather than a POT.

[david]

The attached gives you a lot of flexibility and is certainly fast enough.  I've used it a lot in the past but if you're using a jelly-bean op-amp like an LM358 you need to watch the common mode range.   Input offset voltage can also be an issue if you try to measure very low currents while using a low resistance sense resistor.
It still doesn't help you capture a 10uS pulse - maybe a sample and hold circuit?

David

[JonW]

The LT6015 is a pretty fast amp.  If it is biased in the LT datasheet with the diff amp config, and the split supply should be fast.  Settling is a few us for a step response.

[midali]

Hi Charlie,

Read this AP note from TI , its very usefull. If you'll use a OP , I suggest you to read the high side current only with differnetial OP mode

[charliecoutas]

Thanks Midali, that adds to the options I have. I have just submitted a PCB design for manufacture. Those TI devices look very good. I looked at INA250, a while ago and rejected them for some reason. But it seems that they could also do the job. Many thanks, I'll try one of these on my next revision of the rig.

Charlie

[Giuseppe MPO]

I used this some time ago, and I must say that for what I wanted to do it worked more than well.

https://www.diodes.com/_files/datasheets/ZXCT1009.pdf

[ken_k]

Lots of good suggestions here. I have used a similar circuit to Davids (post #7) at around 400V.

While only suitable for very specific applications I have derived both voltage and current using only resistors and math.

[ken_k]

The circuit I am having trouble with is from the Linear Tech datasheet. The circuit on the left suffers massive overshoot, My shunt is 0.05 Ohms, load is 1 Ohm to 35 Ohms, current from 0.5 to 10A.

Hi Charlie
A few questions:
Is the load inductive?
What is the rise/fall time?
What is the maximum voltage?
I'm assuming it won't do 10A into 35 ohms (350V).

[charliecoutas]

Guiseppe: I looked at those but the 20V max ruled them out. My "shunt" can be anywhere from 0v to 30v above ground.

Ken: The load is not inductive (but it is at the end of a long-ish cable so there will be a bit of L).
Rise/fall times in the order of 5uS.
Pulse lengths from 10uS to 500mS.
Maximum voltage is 30V.

You are right, 10A in 35 ohms won't happen! Max volts 30v, max current 10A, max power <300W

What did you use to measure voltages to 7 figures?

Charlie

[charliecoutas]

I'm pleased with the results show below. Top trace is volts/10 in a 1 ohm load, bottom trace is volts/10 from an LT6016 measuring the current in a 0.05 ohm shunt and multiplying it by 8. But I'm going to try some of your suggestions as well.

[charliecoutas]

Sorry, the previous scope traces were into a 10 ohm load not 1 ohm.

The traces below here are into a 1 ohm load, 5 volts, 5 amps.

Upper trace: load volts
Lower: Current as a voltage from LT6016, gain of 8
Charlie

[JonW]

I'm glad you got it working, Charlie. There is virtually zero overshoot, and it settles almost perfectly, nice one.  Are you using a FET for the low-side switch? If so, you may be able to speed up the rise/fall time further by using a gate driver or a device with a lower gate charge, that is, if the above needs improvement! It does look like the low-side switch is now limiting the slew.    I don't think I have ever tested an LT part equal to the datasheet; they are extremely conservative and are generally much better than the specification. 

[charliecoutas]

The power drive is done by a high-side MOSFET emmiter-follower, quite a big one. Its gate is driven by a small NPN emitter-follower which is driven by the output of the LT6016. There is a PNP transistor to drag the charge from the big MOSFET gate at turn off. Because the load voltage can be as high as 30V, the op-amp must also have this ability, hence the LT6016. I'm pretty sure it is the slew rate of the LT6016 that is slowing things. I think the 5uS rise/fall is going to be OK with the customer.

I can't use a gate driver because this is a proportional system: the voltage on the load is multiplied by the current in the load (using an AD633) to match a preset power level. As the resistance of the load changes, so must the drive to maintain the same constant power level.

The LT6016 does seem to behave nicely compared to some other devices I have tried.

[JonW]

I get it. I thought you were digitally switching the load's low side for current monitoring purposes into a PIC. It sounds like an exciting project. I've never seen the AD633. It sounds like you are building a fast Electronic load  control loop for constant power.