FEDEVEL
Platform forum

Mahsheed , 07-03-2026, 08:14 PM

I have to measure voltage levels via ADC of different isolated supplies. For that I am using AMC1311 IC which provides input to ouput isloation. Now the input side needs to be powered up from the same supply (common ground) as of which it is measuring. To make the cost effective PCB, I am planning to use LDO instead of isolated DC DC converter. The LDO I used is LM317 which works on the principle of having contant output voltage between output pin and adjustment pin (it does not takes ground pin either from input side or output side). So in order to take 5V for AMC1311 input side VCC, I used resistor divider so that output voltage between output pin of LM317 and Gnd of measuring supply becomes 5V and that virtually made the measuring ground common with that LM317 output VCC voltage and that is fed to AMC1311.

Now I want to confirm from experts, if that would practically make input side grounds common with respect to my measuring voltage and should that circuit be working as expected? (Schematic attached for input side). Thanks in advance.

, 07-03-2026, 08:23 PM

🤖 In the meantime, here’s a response from the [FEDEVEL AI HW Assistant]().
It helps to find relevant resources for your question, but it may make mistakes as it’s still in early development.

This is the response from the AI Assistant:
Let me analyze your isolation approach using the AMC1311 and LM317 configuration:

1. **Key Insight**: The AMC1311 requires complete galvanic isolation between its input and output sides. Your LM317-based power scheme risks creating an unintended ground connection through the resistor divider network, potentially compromising the isolation.

2. **Critical Issue**: The LM317's adjustment pin configuration creates a direct DC path between the input (measuring) ground and the output side through the voltage divider resistors. This defeats the isolation purpose of the AMC1311, as shown in multiple examples of failed isolation attempts from the videos ([Example 1](https://www.youtube.com/watch?v=2oA4X3OTz0M&t=4091), [Example 2](https://www.youtube.com/watch?v=vALt6Sd9vlY&t=1284)).

3. **Practical Consequences**:
- Common-mode currents will flow through the divider network
- Ground potential differences will directly affect measurements
- EMI susceptibility increases due to created ground loops ([EMI Example](https://www.youtube.com/watch?v=vALt6Sd9vlY&t=3403))

4. **Recommended Solution**:
- Use an isolated DC/DC converter as originally intended
- If cost is critical, consider:
* Charge pump isolators (TI ISOW7741)
* Integrated isolated power+data solutions (ADuM5400)
* Reinforced isolation barrier techniques ([Isolation Techniques](https://www.youtube.com/watch?v=vALt6Sd9vlY&t=1660))

5.

, 07-03-2026, 08:23 PM

**Verification Method**:
- Measure continuity between input/output grounds with Ohm meter
- Perform dielectric withstand test (HIPOT) between isolated sides
- Check common-mode rejection ratio in actual operating conditions

The current configuration likely creates an undesired ground path through the voltage divider network, violating the AMC1311's isolation requirements. While clever in theory, practical implementation would suffer from ground loop issues and compromised measurements.

You can continue chatting with the AI Assistant here:
https://fedevel.com/hw-assistant

QDrives , 07-04-2026, 02:48 PM

You cannot use a linear regulator and keep isolation.

Mahsheed , 07-04-2026, 03:34 PM

I am using linear regulator at the input side. The isolation is provided by the IC AMC1311 whose block diagram is attached here. The isolation is achived by having different VDD1 and VDD2 supplies and its grounds. I am using ADC supply and ground at VDD2 side, and using Linear regulator from the same supply (which needs to be measured) at the VDD1 Side. I just have concerns with if it will work on the input side (my specific circuit) where I made the supply ground and Linear reulator ouput's virtual ground common.

QDrives , 07-04-2026, 03:43 PM

It is not the AMC1311, but the linear regulator that is the problem.

Mahsheed , 07-04-2026, 04:10 PM

how?

Mahsheed , 07-04-2026, 04:10 PM

it is only driving the input side of the AMC1311

Mahsheed , 07-04-2026, 04:10 PM

how can it be the problem?

QDrives , 07-04-2026, 04:11 PM

Then what is powering the output side?

Mahsheed , 07-04-2026, 05:10 PM

output side is being powered up from the same supply which is driving the ADC. So my ADC and output side has one isolated supply and input side is being powered up from an LDO which is using the supply that needs to be measured.

QDrives , 07-04-2026, 05:16 PM

"*To make the cost effective PCB, I am planning to use LDO instead of isolated DC DC converter.*" - can you provide schematic or block diagram that shows what you are doing?

Mahsheed , 07-04-2026, 07:53 PM

that's a portion of schematics where I am passing data of 2 isolated voltages V3 and V4.

QDrives , 07-04-2026, 08:10 PM

I see.
A better 'layout' of your schematic will prevent errors.
And do not use ceramic capacitors on the output of the LM317.

Mahsheed , 07-04-2026, 08:42 PM

My design is fine, right?
Specially where I am virtually making the grounds of VDD1 and VIN common (input side)
also, what is the reason of not using ceramic capacitors on the output of LM317?

QDrives , 07-04-2026, 09:47 PM

Fine is the word.
And you are not making a 'virtual' ground. The LM317 has a reference, like V3- which I assume is the reference of V3+.
The LM317 is not stable with ceramic capacitors.

Mahsheed , 07-05-2026, 09:59 AM

It does mention Ceramic capacitors in its datasheet. And LM317 does not have a reference ground. it operates on the principle of input to output differential and maintains a Vref voltage of 1.25V between Output and Ajd pin. We are making V3- its ground by adding a resistor between Adj and V3-

Mahsheed , 07-05-2026, 10:00 AM

@Robert Feranec can you please comment on my design? Thanks in advance

QDrives , 07-05-2026, 08:34 PM

Yes, "**Ci**" can be ceramic, but not "**Co**".
Voltage is always to a reference. The LM317 'pushes' the output to such reference. There is **no** "*input to output differential*".