ground isolation

[ranganatennakoon]

I planed constant power load control circuit. All parts are fully operational in computer simulation. But I want to Isolate High voltage (180V DC ) DC ground (GND 2) and low voltage (12V/3.3V DC) DC Ground (Gnd) What can be done for this? How to solve this problem?

 

[FvM]

You need gate driver and current sense referenced to GND2 and a respective control voltage supply. Current setpoint and measurement can be isolated by state-of-the-art analog isolators or digital isolators and GND2 side DAC/ADC.

Isolation working voltage and permitted leakage current should be specified. Under circumstances, level shifters or differential amplifiers can be used instead of true isolation.

Driving a high current FET with weak LM324 won't work well.

 

[KlausST]

Hi,

You say "constant power control".
Power: P = V × I.
I see you regulate the current ... but I see nothing about voltage.
--> So indeed it seems to be a "constant current control". Is this true?
(Otherwise you need to measure the load voltage to be able to regulate the power)

*****
The circuit shows:
* DAC --> current control loop--> load current
And
* load current --> ADC
--> To me this is somehow redundant. If you set the DAC .. you already know the current. So why measure it?
In most cases you will read back the expected value ... with (a lot of) tolerance.
There will be errors:
* from DAC to current
* from current to ADC
Thus the measurement of the current does not give much reliable information.
If there is a big deviation you may just assume the the current control loop is "out of regulation".

But why does this matter for the "isolation problem"?
Because in your application now are two analog paths that need to be isolated.
But isolating analog paths is rather difficult and usually generates rather big. Low quality.
Thus - my personal way is - to place the isolation into digital paths.

Without knowing details .. my approach would be this:
ESP32, SPI --> isolation --> cheap SPI DAC --> current control loop --> load current
Current control output --> comparator --> optocoupler --> ESP32

The idea behind the second path:
As soon as the current control loop is out of regulation the current_control_loop_output immediately saturates close at VCC.
This is a good (digutal) indicator of: regulation is O.K. / not O.K.

Additional hints: I personally
* don't like the potis in the analog paths. Its easier, cheaper and more precise and reliable to do the correction in software.
* don't like the ADCs and DACs of the ESP32. They have poor accuracy, poor resolution, poor linearity and poor precision.
( thus using a cheap external DAC improves overall performance)
* would install low pass filters in the analog paths to reduce noise.

The current measurement shunt has high power dissipation. Consider to replace it with a lower value one.
Also the worst case power dissipation of the Mosfet is about 300W. This is huge! You need a big heatsink with fan ... and still it is likely for the Mosfet to explode.
If you could replace linear regulation with switch mode control (PWM) you avoid a lot of heat problems.

Klaus

 

[ranganatennakoon]

You need gate driver and current sense referenced to GND2 and a respective control voltage supply. Current setpoint and measurement can be isolated by state-of-the-art analog isolators or digital isolators and GND2 side DAC/ADC.

Isolation working voltage and permitted leakage current should be specified. Under circumstances, level shifters or differential amplifiers can be used instead of true isolation.

Driving a high current FET with weak LM324 won't work well.

This circuit works very well when one part is taken.
The system works very well when all ground are connected.

This system is flawless.The current controlloing is very smooth and stable.
he current flowing through the shunt could also be measured very accurately.
So there are no errors in it,
I took the help of EEv bolog, Greatscoot, electronoobs YouTube channels for this. There are no errors

--- Updated Jan 30, 2023 ---

Hi,

You say "constant power control".
Power: P = V × I.
I see you regulate the current ... but I see nothing about voltage.
--> So indeed it seems to be a "constant current control". Is this true?
(Otherwise you need to measure the load voltage to be able to regulate the power)

*****
The circuit shows:
* DAC --> current control loop--> load current
And
* load current --> ADC
--> To me this is somehow redundant. If you set the DAC .. you already know the current. So why measure it?
In most cases you will read back the expected value ... with (a lot of) tolerance.
There will be errors:
* from DAC to current
* from current to ADC
Thus the measurement of the current does not give much reliable information.
If there is a big deviation you may just assume the the current control loop is "out of regulation".

But why does this matter for the "isolation problem"?
Because in your application now are two analog paths that need to be isolated.
But isolating analog paths is rather difficult and usually generates rather big. Low quality.
Thus - my personal way is - to place the isolation into digital paths.

Without knowing details .. my approach would be this:
ESP32, SPI --> isolation --> cheap SPI DAC --> current control loop --> load current
Current control output --> comparator --> optocoupler --> ESP32

The idea behind the second path:
As soon as the current control loop is out of regulation the current_control_loop_output immediately saturates close at VCC.
This is a good (digutal) indicator of: regulation is O.K. / not O.K.

Additional hints: I personally
* don't like the potis in the analog paths. Its easier, cheaper and more precise and reliable to do the correction in software.
* don't like the ADCs and DACs of the ESP32. They have poor accuracy, poor resolution, poor linearity and poor precision.
( thus using a cheap external DAC improves overall performance)
* would install low pass filters in the analog paths to reduce noise.

The current measurement shunt has high power dissipation. Consider to replace it with a lower value one.
Also the worst case power dissipation of the Mosfet is about 300W. This is huge! You need a big heatsink with fan ... and still it is likely for the Mosfet to explode.
If you could replace linear regulation with switch mode control (PWM) you avoid a lot of heat problems.

Klaus

• I tried both method optocoupler and ADC and PWM -->OPTO Isolator --> BJT -> lowpass filter . those method works fine .
• esp 32 has internal 8bit DAC its performance well.
DAC --> current control loop -->measure shunt voltage --> non inverting amplifier to gain 6 --> esp 32 ADC >
• This circuit works very well when one part is taken. The system works very well when all ground are connected. But I want to Isolate it . It possible to use ground isolation like B1212 or any other method for this?

 

[FvM]

You are only answering to the LM324 comment. Is the isolation problem already solved or didn't you understand the suggestions?

I should have said, LM324 doesn't work well if you want more than slow DC regulation. It may be sufficient for your purposes. Did you check phase margin in simulation with different load impedances including inductive?