Author Topic: Highside current sensor  (Read 5363 times)

0 Members and 4 Guests are viewing this topic.

Offline MariuszD

  • Frequent Contributor
  • **
  • Posts: 724
  • Country: pl
Re: Highside current sensor
« Reply #25 on: October 12, 2026, 06:45:52 pm »
My meter is the Mastech ms8040, and the measurements are done in the DC mV setting (0-220mV).
So, if you are only measuring DC, you should cut the bandwidth to single Hz. Such a filter before the amplifier will make the amplifier's SR irrelevant.
Quote
I've got the schematic of a similar meter and I'm certain that it measures DC voltages as averages, which is what I need.
The DC meter should respond only to the average value, but in some meters, the AC component interferes with the measurement
By filtering before the meter, you increase the resistance to interference.

 

Offline sourcechargeTopic starter

  • Frequent Contributor
  • **
  • Posts: 289
  • Country: us
Re: Highside current sensor
« Reply #26 on: October 12, 2026, 07:31:55 pm »
The picture that MariuszD posted doesn't have a second supply that is negative.

The "-A" would be on the input, before the fets, and after the negative supply.

The current on each sensor will have to measure the current under conditions of 80-90% power factor from DC sources.

That means current is going both directions, on both sensors.

I added a pic that should clear everything up for you.
Thanks, that helps.
So you want two independent current measurements, one on each of the two rails.
As long as the total supply voltage is within the common mode range of the INA24x (80V for INA240Ax, 110V for INA241Ax), they should work fine for measuring each rail.
Just reference both chips to the negative rail. Their input common mode range extends negative, so one can measure the negative rail current as a low-side shunt.
If you set the reference pin to a positive bias, then you could even get away with only needing one isolated supply (or battery) for both chips, rather than two like in your schematic.

In my first post, the schematic shows a source with max 15khz.

The bandwidth of a this needs to be, as someone else said, 3-10 times greater than the frequency of operation.

The opa2189 has a bandwidth of 800khz with 10x gain and a 20 V/us slew rate.
INA240x has BW of 400kHz with 20x or higher gain, slew rate is 2V/us.
INA241x has BW of 1.1MHz with 10x or higher gain, slew rate is 8V/us.
Still no idea what your slew rate requirement is. I'm assuming a pre-filter can take care of that.

Quote
If you can beat that, with the CMRR that the opa2189 has, great.
Comparing datasheet CMRR between an opamp and a current sense amp is apples-to-oranges because:
1. In your circuit, overall CMRR will probably be dominated by resistor matching, not the opamp itself.
2. The circuit topology is very different, so even with perfect resistors the CMRR of your circuit wouldn't necessarily be equal to the CMRR of the opamp.
But the INA241x boasts typical CMRR of 166dB (150dB min), roughly the same as the OPA2189.

The INA240 has 20x, 50x, 100x, and 200x gain versions, but no 10x.

With the INA241A1, its a 10x gain.

Two things are bothering me about both of them.

First, the Gain Error is specified with the output voltage between (GND + 50mV) < VOUT < (VS - 200mV).

So what this tells me is that the CSAs have the same 4 resistor problem, and can't measure accurately currents less than 50mA.

As the voltage difference bewteen the inputs decreases, the error increases.

This was the reason I went with the simplistic circuit.

The other thing is that the INA241 and INA240 have an input bia current of 35 uA and 70uA respectively. 

That is alot and would significantly offset the measurement, if I'm measuring <10mA.

It's also a source of input to the circuit of which I want to be isolated from any measurement.

The opa2189 has 140pA, but I wish there was a fet based opamp that had the same specs as the opa2189 because those are in the 1 - 2 pA range.

The resistor networks I got are matched by 0.01%, and because it's only 2 resistors, it doesn't have the same problem as the 4 resistor diff opamps.

I think I made right choice, thanks.

My meter is the Mastech ms8040, and the measurements are done in the DC mV setting (0-220mV).
So, if you are only measuring DC, you should cut the bandwidth to single Hz. Such a filter before the amplifier will make the amplifier's SR irrelevant.
Quote
I've got the schematic of a similar meter and I'm certain that it measures DC voltages as averages, which is what I need.
The DC meter should respond only to the average value, but in some meters, the AC component interferes with the measurement
By filtering before the meter, you increase the resistance to interference.

I'm not sure what you mean by "cut the bandwidth to single Hz".

There really is no way to do that with the load that I'm trying to measure.
« Last Edit: October 12, 2026, 07:34:08 pm by sourcecharge »
 

Offline MariuszD

  • Frequent Contributor
  • **
  • Posts: 724
  • Country: pl
Re: Highside current sensor
« Reply #27 on: October 13, 2026, 11:01:16 am »
Quote
I'm not sure what you mean by "cut the bandwidth to single Hz".
You are interested in the average value, so additional averaging won't hurt. I would simply use the lowest cutoff frequency of the filter as much as possible. Because it provides maximum attenuation of high-frequency signals. But too long a time constant will mean that you will have to wait for the meter readings to stabilize. A cutoff frequency of 1Hz will give a time constant of 160ms, while a 12kHz signal will be attenuated by approximately 12kHz/1Hz=12000.
The exact value of the filter's cutoff frequency is not important, 5Hz or 0.5Hz will be fine too, which is why I wrote it in the range of single Hz. Maybe after the translation, it turned out to be confusing.

Once at university, I did an exercise where we checked how the AC component interferes with the operation of a DCV digital multimeter measurement, so I always remember not to assume blindly that it will correctly show the average value under all conditions. The RC filter will always work.
« Last Edit: October 13, 2026, 11:52:04 am by MariuszD »
 

Offline mtwieg

  • Super Contributor
  • ***
  • Posts: 1865
  • Country: us
Re: Highside current sensor
« Reply #28 on: October 13, 2026, 12:30:38 pm »
Two things are bothering me about both of them.

First, the Gain Error is specified with the output voltage between (GND + 50mV) < VOUT < (VS - 200mV).

So what this tells me is that the CSAs have the same 4 resistor problem, and can't measure accurately currents less than 50mA.
Note that these specs also assume single-supply operation. So the only thing that shows is that, like all opamps, performance degrades as the output gets close to the supply rails.
One can use dual supply for these chips to get around that issue (though the datasheets don't suggest this directly). The more common implementation is to use the REF pins to bias the output at a positive voltage. That way bipolar currents can be measured accurately on a single supply without Vout getting near the supply rails.

Quote
As the voltage difference bewteen the inputs decreases, the error increases.
Not sure what you're referring to here.

Quote
The other thing is that the INA241 and INA240 have an input bia current of 35 uA and 70uA respectively. 

That is alot and would significantly offset the measurement, if I'm measuring <10mA.
The input bias current will cause an input-referred offset that scales with the difference in the source resistances presented to the input pins. In the case where the sense resistor is 0.1ohm, this would translate to an input-referred offset of 3.5/7.0uV (or 35/70uA of measured current). This could easily be compensated by adding 0.1ohm of series resistance with just the positive input.

Quote
It's also a source of input to the circuit of which I want to be isolated from any measurement.
Again, not sure how to interpret this.

Quote
The opa2189 has 140pA, but I wish there was a fet based opamp that had the same specs as the opa2189 because those are in the 1 - 2 pA range.
I'm struggling to see how this would significantly impact the performance of the circuit. Even at 140pA, the effect on input-referred offset should be negligible compared to opamp's own offset.

Quote
The resistor networks I got are matched by 0.01%, and because it's only 2 resistors, it doesn't have the same problem as the 4 resistor diff opamps.

I think I made right choice, thanks.
I don't think anyone ever said you made the wrong choice. We're just saying there are other choices which may also meet your requirements (though it's still impossible to be sure since you won't clearly state them).

The only downside I see to your implementation is that it requires that the opamps' supply voltages extend beyond the supply voltages of the half bridge being measured. In the case where the half bridge supply is low (like +/-10V) then that's not a big deal (though you do have to generate the higher supplies for the opamps from somewhere). As the bridge supply voltages get higher (i.e. +/-24V), the opamps for the upper and lower circuits could no longer share the same supplies, and those supplies would need to be referenced to the bridge supply rails rather than ground, so you would need four isolated supplies (or batteries) total. None of this impacts measurement accuracy. They're just matters of convenience and circuit complexity, compared to the INA24x chips which could operate fine without any isolated supplies/batteries (until the bridge supply exceeds their common mode range, that is).

I'm not sure what you mean by "cut the bandwidth to single Hz".

There really is no way to do that with the load that I'm trying to measure.
The DMM is going to only refresh its display every 0.5s, so it can't provide meaningful information on frequency content beyond 1-2Hz. However, having significant AC content can affect the DC readings, leading to measurement error (how much depends a lot on details of the DMM's operation). Therefore it's typically a good choice to filter out such AC content first. I think this is what MariuszD is getting at.
« Last Edit: October 13, 2026, 12:38:55 pm by mtwieg »
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 19282
  • Country: us
  • DavidH
Re: Highside current sensor
« Reply #29 on: October 13, 2026, 05:16:15 pm »
The biggest source of errors in a differential amplifier is resistance asymmetry, which reduces CMRR. The difficulty lies in the fact that a 1% error in the resistor will result more than 1% error of measured value.
There are solutions that are free of this flaw.

Another option which does not require matched resistors are transconductance based difference amplifiers like the AD830, AD8129, and AD8130, or for slow speed precision, the AD8237.  Shown below is an example.

Low frequency common mode rejection ratio for the fast parts extends to 100 dB at low frequencies, and 70 dB at 10 MHz (!).  Keep in mind however that in most low and high side current shunt applications, the common mode swing is small when sensing at the supply voltage, unlike the example shown below, so common mode rejection is less important.
« Last Edit: October 13, 2026, 05:19:44 pm by David Hess »
 

Offline sourcechargeTopic starter

  • Frequent Contributor
  • **
  • Posts: 289
  • Country: us
Re: Highside current sensor
« Reply #30 on: October 20, 2026, 11:45:21 am »
After making the pcb and soldering all of the parts in, I found that the 0.01% matching of the resistors in the resistor networks from Susumu were not within the datasheet's specifications.

Using two different meters, fluke 89 IV, and MS8040 (all three of them), I have found that the networks were matched by only 0.12%, and 0.07%

I have contacted the manufacturer, and asked if they plan on fixing this problem.

They have auto-responded that they will contact me shortly, and I will post their responce if any.
« Last Edit: October 20, 2026, 11:55:07 am by sourcecharge »
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 19282
  • Country: us
  • DavidH
Re: Highside current sensor
« Reply #31 on: October 20, 2026, 03:36:26 pm »
After making the pcb and soldering all of the parts in, I found that the 0.01% matching of the resistors in the resistor networks from Susumu were not within the datasheet's specifications.

Did soldering force them out of specification?
 

Offline sourcechargeTopic starter

  • Frequent Contributor
  • **
  • Posts: 289
  • Country: us
Re: Highside current sensor
« Reply #32 on: October 20, 2026, 03:56:53 pm »
After making the pcb and soldering all of the parts in, I found that the 0.01% matching of the resistors in the resistor networks from Susumu were not within the datasheet's specifications.

Did soldering force them out of specification?

I actually measured them with only one meter, one time, before I put them in, but thought maybe the meter was not right.

After soldering them in, the gain was off, and I double checked them with all of my meters.

They were 5 bucks a piece and the shipping is 5 bucks, not really worth sending them back anyway.

My only hope is that the manufucturer takes pitty on me and sends me new networks that have been QCed.

Edit:  Also, I'm not sure if digikey would take them back anyway, as their return policy says, the products must be unopened, unused, like new condition.

Meaning, you have to open the packaging to determine if the resistors are any good in the first place.
« Last Edit: October 20, 2026, 04:07:55 pm by sourcecharge »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf

 

- - -
-->