High dynamic range current measurement with ADC

[Axk]

I'm thinking of doing a power meter for an e-bike.

The whole thing will be power from the battery which is 54V to 39V depending on the charge level so I'll need a voltage regulator for everything but the current sense amp which can work up to 60V.

This is most probably an overkill and not needed but I wonder if I can make it high dynamic range, e.g. from 5 mA up to 20A.
There's this high dynamic range current sense amp LTC6102 that converts a sense voltage to a sense current, it is recommended that the sense current is not greater than 1 mA full scale (which translates to just 60 uA of current for 1 LSB of the ADC).
The current waveform is 100 KHz at most accroding to my measurements.
So I'm thinking of oversampling it and the voltage and integrating (just summing) the product.
I'm looking at LTC1407A for the ADC which is 14 bit 1.5 with 2 channels and simultaneous sampling.
Thought I could buffer the current sense and voltage (from a resistive divider) signals with OPA2365 which has a low offset voltage which would allow to measure the lower end into the milliamps (1LSB would be about 100 uV for the ADC).

There are probably a ton of other thins that would prevent me from measuring the low end and getting this high of a dynamic range but one issue that I'm considering now is providing a negative rail for the opamp for the output to get near ground (1LSB is about 100uV).
The datasheet says that the negative rail need not be lower than -0.3V but still how do I do this with low noise.

I've found that one can do a virtual ground with a resistive divider and an opamp.
In my understanding this virtual ground will have to be shared by all the chips: the current sense amp, the opamp, the ADC and the FPGA so I'm worried that it would spread create noise and I won't get the dynamic range.

On the user hand at the low currents the motor is not working so I think the current will be pretty much DC, but there will be the ADC and the FPGA with their switching still.

Is this dynamic range that I'm aiming at not achievable this easily and I should just give up the idea?

If it is how should I go about the negative -0.3V for the negative rail for the opamp?

[mvs]

Your approach with current sense amp, buffer, fast external ADC and FPGA sounds a bit like over-engineering to me.
I would use low side sensing to eliminate CSA and MCU with fast ADCs (like STM32F302) to eliminate external ADC and FPGA.

In my understanding this virtual ground will have to be shared by all the chips: the current sense amp, the opamp, the ADC and the FPGA so I'm worried that it would spread create noise and I won't get the dynamic range.

Virtual ground is signal reference point, not a normal ground (power return connection) for your ICs.

[Axk]

So if I measure on the low side then I can get away without a current sense amp, but I'll still need an amp not as a pure buffer (because the source impedance of the shunt is low and fine for the ADC) but as an amplifier because of the low sense voltage, right?

So if I want to measure the current near 0 to get a hight dynamic range I still need the opamp to have its negative rail slightly below ground?

So I'll still need to generate a slightly negative vlotage relative to my ground?

[mvs]

So if I measure on the low side then I can get away without a current sense amp, but I'll still need an amp not as a pure buffer (because the source impedance of the shunt is low and fine for the ADC) but as an amplifier because of the low sense voltage, right?

Yes, you will need an amplifier.

So if I want to measure the current near 0 to get a hight dynamic range I still need the opamp to have its negative rail slightly below ground?

You can set signal ground 50-100mV above power ground and compensate offset digitaly, after ADC conversion is done. For many ADCs and R2R opamps it is enough. For such small rise you may not need even a buffer, just 2 resistors for divider (see attachment).
However if you need to measure also negative current (battery charging), then you might go for VCC/2 or use even differential input ADC.

[Axk]

What if I add an additional small 1.5V battery to pull the voltage down?

Will this work?

Any drawbacks with this besides the need for an additional battery and the battery eventually discharging?

[Doctorandus_P]

I think low loss is more important than high dynamic range.
ACS712 might be suitable.

The UNI-T 210E is another option.
With this DC capable clamp meter you do not have to interrupt your existing circuit.
It also has a DC resolution of 1mA (Not Accuracy!).

It can be hacked with an analog output upto about 2kHz.
The chip in mine has real pins. It's not a glued down blob and the chip itself (I believe a DM0660) has serial output capability, so you could for  example add an Bluetooth module.

[mvs]

What if I add an additional small 1.5V battery to pull the voltage down?

Will this work?

For R2R or single supply opamp it will work, but ADC might be also not very linear near 0V input.

Any drawbacks with this besides the need for an additional battery and the battery eventually discharging?

You will need double pole power switch.

[mvs]

I think low loss is more important than high dynamic range.
ACS712 might be suitable.

Sensitec CMS3000 are better on accuracy and bandwidth... unfortunately more expensive and EOL.
https://www.sensitec.com/products-solutions/current-measurement/cms3000-family