Neutral to Ground Voltage on ADC

[electronerd213]

Hello, and thank you for taking a look at this. I have not been able to come up with a good solution yet, and I may be overlooking it.

I have a system in which I need to read the voltage difference between Neutral and Earth Ground (Vng) on an analog to digital converter (ADC). The ADC is powered by an isolated 3.3V power supply that is powered by mains. What this means, of course, is that the ADC's common is at a level different than the Earth ground (and that level is unpredictable), so I cannot just use a voltage divider directly from Vng into the ADC. Of course, I could just short the DC- to Earth ground, but I am not sure I want to do that.

I have been looking into differential amplifiers and ADC drivers, but I just am not coming up with a good option. Part of the reason for that is that I don't have a DC supply on the board that has the Earth ground as a reference. It's hard for a device to measure a voltage with reference to earth (Vng) when that device does not reference earth.

What I would love is to find a truly differential chip that does not require a supply on the input side, can read up to a 5-10V difference for Vng regardless of the difference between earth ground and the isolated digital ground on the output side, and can reference my isolated digital ground on the output that goes into the ADC. I am also open to I2C or SPI or a lot of other options if those are more available instead of the ADC.

Anybody have any ideas? Things you have worked with?

Thank you!

 

[Nigel Goodwin]

Of course, I could just short the DC- to Earth ground, but I am not sure I want to do that.

Why on earth (no pun intended) not?.

 

[Misterbenn]

I agree with Nigel. Why not reference your isolated supply to ground?

Points to consider:
- check the isolation is adequate, at least 1000V to allow for inductive spikes / lightning strike on the mains
- check the isolated supply output ripple is ok for analogue measurements
- What range of voltages are you expecting between neutral and ground, In some high voltage lab set-ups I've seen several hundred volts between building ground and electrical ground!

In the past I've used AMC1100DUB, an isolated differential amplifier but it does require a 5V supply referenced to your measurement ground. Its a good solution if the measurement needs level shifting.

 

[MikeMl]

I would use a +-thousand V common-mode isolated differential instrumentation amp and put the ADC on the non-isolated side...

 

[strantor]

What's the max voltage difference you expect to see? I assume worst case is full line voltage. If that's the case then you definitely don't want to introduce any risk of shorting earth to neutral. Have a look at the difference between regular 1x/10x/100x ground-referenced oscilloscope probes versus those used on isolated input oscilloscopes (ex fluke scopemeter, Tek THS, et. al.).

Typical probe:

The 9Mohm resistor forms a voltage divider with the 1Mohm internal resistance of the scope, which is referenced to earth.

On an oscilloscope with isolated inputs, the probes are special; they have a different internal ciruit (sorry i can't find a graphic of the probe internal circuit, or create one on my tablet). The important difference is that the probes have TWO internal high value resistors instead of just one. So in the pic above, copy the yellow box on the left and paste it directly below, onto the free wire. Then in the yellow box on the right, delete the connection to earth. So the voltage divider would consist of 3 resistors and the scope input would be taken from either side of the center resistor and fed to an instrumentation amplifier.

If I were going to do this, I'd copy how the 'scope manufacturers do it, and go with the isolated scope probe concept.

 

[gophert]

Hello, and thank you for taking a look at this. I have not been able to come up with a good solution yet, and I may be overlooking it.

I have a system in which I need to read the voltage difference between Neutral and Earth Ground (Vng) on an analog to digital converter (ADC). The ADC is powered by an isolated 3.3V power supply that is powered by mains. What this means, of course, is that the ADC's common is at a level different than the Earth ground (and that level is unpredictable), so I cannot just use a voltage divider directly from Vng into the ADC. Of course, I could just short the DC- to Earth ground, but I am not sure I want to do that.

I have been looking into differential amplifiers and ADC drivers, but I just am not coming up with a good option. Part of the reason for that is that I don't have a DC supply on the board that has the Earth ground as a reference. It's hard for a device to measure a voltage with reference to earth (Vng) when that device does not reference earth.

What I would love is to find a truly differential chip that does not require a supply on the input side, can read up to a 5-10V difference for Vng regardless of the difference between earth ground and the isolated digital ground on the output side, and can reference my isolated digital ground on the output that goes into the ADC. I am also open to I2C or SPI or a lot of other options if those are more available instead of the ADC.

Anybody have any ideas? Things you have worked with?

Thank you!

It seems like an isolation amplifier should work for you. They can be costly, but I have no idea of your budget to goal.

Also, how about a simple volt meter?

 

[dr pepper]

Your concern coonecting the - to earth/neutral is justified, despite as mentioned spikes and lightning strikes, assuming the supply is a smps the designers obviously implemented a insulation barrier to protect against pri-sec shorts, however the design may not have considered the o/p being connected to mains neutral, applying mains noise and spikes to its o/p might cause the transformer insulation to beak down.

Oh yes welcome to the forum.

 

[electronerd213]

Thank you all for your responses, including both the questions and justifications for shorting DC- to Earth. All posts are very good ideas and very much appreciated.

One question that came up a couple of times is regarding normal levels for Vng. I am told they should be no greater than 5VDC (these are usually coming off of 3 phase generators I believe, and any more than the 5V or so is considered unhealthy and needs to be fixed), so I am designing for 10 or so to give some slack but also keep some resolution on my ADC. Of course I am also going to put a clipper on the circuit to protect ... whichever method I use. I have been told that very large voltages (like the several hundred volts mentioned by Misterbenn) will not occur and that I am not to design for them.

It sounds like a differential instrumentation amp and possibly the scope probe concept may lead to some good options. I will look into these.

Thank you all for kindly helping, and I enjoyed some of your humor as well (e.g. "blue to bits", "why on earth not") haha

 

[Misterbenn]

I am told they should be no greater than 5VDC (these are usually coming off of 3 phase generators I believe, and any more than the 5V or so is considered unhealthy and needs to be fixed),

While it may be 5V under normal operation an unbalanced generator or fault on one phase would give you a large voltage on the neutral. While you may not care if the circuit stops working if it sees voltages above 10V I would strongly advise to ensure your insulation ratings are designed to cope with >1000V to ensure protection of equipment users.

You mention using a clipper on the circuit, be aware that the leakage current through the clipper may effect your measurement depending on the design. Also the voltage on the neutral will be AC.

 

[strantor]

While it may be 5V under normal operation an unbalanced generator or fault on one phase would give you a large voltage on the neutral. While you may not care if the circuit stops working if it sees voltages above 10V I would strongly advise to ensure your insulation ratings are designed to cope with >1000V to ensure protection of equipment users.

Agreed. This is my reasoning for suggesting the [isolated oscilloscope input]-like circuit. My suggestion may be overkill, I don't know. If there is a simpler way to offer 1kv isolation then it's definitely at the top of the list.

5V may be "unhealthy" and 10v may be "even more unhealthy " but I have read nothing here that indicates that 5V or 10V or anything even close to that, is the max one could or should expect to see.

 

[electronerd213]

Points taken and I thank you for them.

As I go further down this path (using the scope probe concept plus a diff amp), I think I see that even using a differential amplifier requires a power supply for that amplifier that is referenced to that earth ground, right? The ones I am seeing have absolute maximum ratings for In+ and In- (the differential inputs) that are close to the output side supply + and -. Misterbenn mentions the AMC1100DUB with the same constraint, but are there any that do not have that constraint or have a very generous swing on that constraint?

So I guess I am still looking for a way to do this without another supply/regulator circuit. Or am I reading those datasheets incorrectly? I hope I have a misunderstanding as that would be much more convenient than if I don't.

Thank you again!

 

[strantor]

You don't have to reference the ins. amp to earth and you don't have to utilize such a wide portion of max input differential.

The 3-resistor voltage divider is connected between earth and neutral, and the ins. amp amplifies the voltage across the center resistor. This voltage is not "referenced" to anything but itself. The supply voltage could be in-circuit with neutral, earth, or neither (isolated). It simply measures the voltage drop across the center resistor, and that voltage drop will be a function of the ratio between the center resistor and the outside two resistors. Let's say you design for a max reading of 1,000V and the max differential voltage for the ins. Amp is 10V; you'd set up your voltage divider resistors so that 1,000V across the divider leaves 10V dropped across the center, for example: 100Mohm, 2.02Mohm, 100Mohm. Total = 202.02Mohm. Ratio = 2.02 / 202.02 = 0.009999. 0.009999 * 1000v = 9.999V.

Designing it using the numbers given above would give you a very very small output from the ins. Amp within the typical reading range of <5V (5V * 0.009999 = 49.995mV), so if you were going to send the signal to a 0-10V analog panel meter for example, you would need to feed the ins. Amp output into an opamp to scale it up so that 50mV becomes 10V, and the cutoff is 10V. So if you did it this way, a potential difference of 0.000V would generate a 0% reading on your meter, a potential difference of 5V would generate a 100% reading on your meter, and a potential difference of 485V would also generate a 100% reading on your meter. It would be designed to DISPLAY a useful reading for potential differences between 0V and 5v, but would be designed to safely ACCEPT potential differences up to 1kV. Nobody wants a device designed to accept 5V max, to blow up in their hands when the perfect storm of circuit faults results in unforeseen high voltages being applied to it.

EDIT:
For clarification, I'm describing voltage divider with a ratio of 100:1. Here is a visual:

 

[electronerd213]

Wow. Great explanation! Thanks to you, that's the track I was attempting to go down (I had a very similar LTSpice going), but I was confused about the reference. Now I am much further, and I can do some more tweaking on my own as I do parts selection. Thank you!

 

[ChrisP58]

Do you need to actually measure the value, or just detect something above a given threshold? If the latter, it may be easier to have an isolated circuit that just reports pass/fail through an optocoupler.

Even if you need a measured value, you might consider an isolated microcontroller that measures and drives an optocoupler with a serial bitstream.

 

[strantor]

Wow. Great explanation! Thanks to you, that's the track I was attempting to go down (I had a very similar LTSpice going), but I was confused about the reference. Now I am much further, and I can do some more tweaking on my own as I do parts selection. Thank you!

If that pic was what clued you in about the reference, then it was a misleading clue.
Actually it was probably misleading all around considering where I placed the voltage source. The pic was means to illustrate only the concept of the 3-resistor divider, and what would happen if 1,000V were applied to it.
Let me start over here...

In the pic I placed the GND on one of the signal wires to the Instrumentation Amp.
I did that because LTSpice forces you to put a GND somewhere in the circuit, and depending where you put it, affects the simulated waveforms.
I was lazy and put the GND on that wire because I wanted to see the waveform from top and bottom of the middle resistor.
I should have put the GND at the bottom where it goes, and set up a 2nd reference point.

Here's what it should look like:

The Instrumentation Amp circuit can be powered by an isolated (NO reference) source like you already have:

it can be referenced to "Earth"

or it can be referenced to "neutral":

it doesn't matter where or if you tie it in, because the ins. amp amplifies only the difference between its two inputs.

Hopefully I clarified my clarification and didn't just make it worse.
(sorry )

 

[electronerd213]

Actually, Strantor, I totally got your first explanation. I was seeing that "ground" more as a common, which really fits well with what I want to do if I consider my DC(-) as common. Not that I don't appreciate your additional information - I greatly do! But I want you to know that you covered your (my) bases and then rounded them again. Thank you again.

Do you need to actually measure the value, or just detect something above a given threshold? If the latter, it may be easier to have an isolated circuit that just reports pass/fail through an optocoupler.

Even if you need a measured value, you might consider an isolated microcontroller that measures and drives an optocoupler with a serial bitstream.

I may be misunderstanding, but my problem as I see it is that I already have an isolated micro. I am just looking for a way to bring a differently referenced analog value into the ADC on that micro. But on the first part, I definitely need a value. This is actually a differential voltage input -> magic -> digital display scenario. Thanks for your idea!

 

[electronerd213]

I swear I have done some homework. I even built up a semi working circuit and tested some gains. But I am still stuck. The datasheet for all of the instrumentation amps I am finding actually says that the reference DOES matter (if I interpret correctly). It looks like I need the Vin pins to be within .5V of the supply pins, which tells me for this case that I would need a power supply whose reference is the same as the Vin(-) pin.

I am posting a pic of the data I am looking at for the chip that I selected and have been playing with. Am I reading this completely wrong or maybe looking at the wrong type of part maybe? I am hoping I am just dead wrong.

This is a TI INA332 precision amp, but the more than several chips I have looked at have either been just like this (similar max ratings) or have had pins for 2 supplies - one for input and one for output, which would have the same criteria for the input supply.

 

[strantor]

Come on now, you can't post the specs without the notes :

So it appears that the IC references itself to the supply voltage (via diodes), whether you want it referenced or not.
The way I interpret it, you can't apply more volts than supply volts. So it your supply is 5V, you can't apply more than 5.5V. I am not 100% confident about that though; hopefully someone more knowledgeable will confirm.

Sorry I don't have more time to research the answer. All I can say for sure is that it is most definitely possible, otherwise your DMM wouldn't work unless open up the case and connect the negative terminal of your battery to the circuit under test.

 

[Nigel Goodwin]

So it appears that the IC references itself to the supply voltage (via diodes), whether you want it referenced or not.
The way I interpret it, you can't apply more volts than supply volts. So it your supply is 5V, you can't apply more than 5.5V. I am not 100% confident about that though; hopefully someone more knowledgeable will confirm.

Not at all, it's just got protection diodes like a PIC - so, presumably, (just like a PIC) you simply feed the pin via a current limiting series resistor (value depending on the maximum expected voltage) - and this could actually be part of the divider that feeds the input.

 

[strantor]

I built up a semi working circuit...
I am still stuck....
The datasheets ...

What exactly are you stuck on?
Theory or real world [NOT WORKING]?