Measuring 36V DC voltage while maintaining isolation from 5V.

[fsky]

Explanation of attached schematic:

I have a battery supply at 36V, which goes trough my power supply board to engines in my application.
36V also powers the switching mode converters, there are two of them, each at 5V. One of them is used to power up board with digital circuits, where also an ADC an uC are located. SMPS's are isolated.

My question is:

What is the best way to measure high voltage, high current noisy power line voltage, without coupling the noise to 5V logic power bus?
Do I necessary need to reference the logic 5V ground to main battery ground?
I have found an isolating amplifier (avago hcpl-7800), however it also needs 5V@30mA power on both isolated sides.
I have thought about serial linear regulators and powering it up directly from 36V, however it doesn't look as a wise solution.

I guess the problem is something about common mode voltage and correct referencing... however, I'm a newbie.

I'll be gratefull for help on this particular problem, also for suggestions for literature
which could help me understand my problem better

Also, hello to everyone on the forum! Till now I was just a reader, however time has come to ask a first question.
So please forgive me if I made any mistake in forum etiquette, I'm open to any suggestions in improving my topic

[PSR B1257]

Do I necessary need to reference the logic 5V ground to main battery ground?

Not if you want to maintain the insulation between them 

I have found an isolating amplifier (avago hcpl-7800), however it also needs 5V@30mA power on both isolated sides.

HCLP7800 is a good idea. Why are you concerned about couple mA supply current?

I have thought about serial linear regulators and powering it up directly from 36V, however it doesn't look as a wise solution.

Well, it does. A 7805 will not be suited in this case. But a LM317 will.

I guess the problem is something about common mode voltage and correct referencing...

Not really. You have to reverence the input of the HCLP7800 to the 36V supply GND (right now you are taping of the sense voltage from the high side of the voltage divider) and the output to the 5V supply.

[max_torque]

A few points of interest:


1) 36Vdc is not "high" in the conventional sense.  yes it's higher than 5, but generally if you use the word high, you mean hundreds of volts, or if you're in the power transmission business, 10 thousand at least!!

2) With just 36v on the power lines, this could power the isolated side of the system.  If you "digitise" the voltage, for example with a V to F convertor, you can then use cheap optocouplers etc to get across the isolation.

3) If the systemm is just to prevent the spread of noise, rather than for any safety reasons, you could also use a suitable filtering architecture to effectively just pass the DC component through to the measurement system.  Again, it depends on the end goal, and how fast and how precisely you want to measure the "high" voltage etc

[fsky]

HCLP7800 is a good idea. Why are you concerned about couple mA supply current?

Actually I was concerned mostly on sole need to supply that power to it, not about the amount, I added amount of current to be precise. Also, I've seen solutions for even higher voltages, micropower converters, but those are able to deliver about 1mA

You have to reverence the input of the HCLP7800 to the 36V supply GND (right now you are taping of the sense voltage from the high side of the voltage divider) and the output to the 5V supply.

Thank you! I didn't notice that I'm probing high side of the divider.

36Vdc is not "high"

I agree, I've exagerrated that a little bit, point taken

With just 36v on the power lines, this could power the isolated side of the system.  If you "digitise" the voltage, for example with a V to F convertor, you can then use cheap optocouplers etc to get across the isolation.

Well, that's a second vote for a regulator, and seems that the only problem left is to design overvoltage protection for LM317, because I expect transients over 60V on those power lines. Maybe I'll just use a Zener diode to pull down LM317 input couple of volts down. But that should not be a major problem. Thank you PSR B1257 and max_torque for your input.

[poorchava]

You could power both sides of the isolation amplifier from your isolated 5v.

Use a tiny transformer (think E8.8 core or something) with turns ratio about 1.2:1 and drive it from some 7400 series logic gate through a capacitor. Then add an LDO on the other side. It will most likely come out cheaper than integrated isolated dc/DC.

Wys?ane z mojego GT-I9100 przy u?yciu Tapatalka

[f5r5e5d]

how much accuracy is always a question - using the cb diode junction photodiode even standard transistor optocouplers are very linear, pretty stable but with a very low CTR few 100s micro which you have to calibrate/trim for - and the LED will age over the very long term

the common mode voltage range is another variable - if it is "low" only few 100s of V and you can tolerate uA "leakage" current then you can make a diff amp on the isolated side with big input R in the dividers http://cds.linear.com/docs/en/design-note/dn025f.pdf
or just buy a monolithic one

[Jeroen3]

Is it a requirement to have electrical isolation?
If yes, multiple options.
- convert to frequency of pwm (or as in sigma delta, but that is difficult af the receiver side)
- put an adc on the measurement side. Isolate spi or i2c (that is not expensive if you use Silabs parts)
- use hcpl or amc1100 analog isolaters, which basically are a sigma delta modulator/demodulator with a isolation barrier in between. Take the extra errors into consideration.

For option 2 there are special IC's, they measure current and voltage of a system with digital output. Iirc, TI's INA family has a few. Do not expect high sample rate. These are often used on sm-bus of a pc.
You'll need less than 100 mA to run an add, easily made with a small smd transformer , sn6501 and mic5205 or such (also good as reference).

Without electrical isolation, if you already share ground (as in a vehicle) you can use a high impance resistor divider. Yes noise will enter the system, but you can filter most of it. Even with isolation, noise is still able to radiate or capacative-couple into the system.

Do you need isolation? What is the nominal range to measure? Are there transients? Common mode transients?

^{typing on an iPad is dissappointing}

[Richard Crowley]

Yes, I agree with @Jeroen3. It is not at all clear from the details revealed so far that you even need full "isolation".  I would question whether separate ground reference is even necessary.  There are other (probably better) ways of dealing with noise.  A simple resistive voltage divider would allow you to monitor the 36V bus with the analog input of any common microcontroller. Add a capacitor to smooth out noise if necessary.  Note that if you use a sufficiently large series resistor (to limit fault current), you can simply use the built-in diode protection on most microcontrollers for over-voltage protection.  Without sufficient justification, I question the whole beginning premise.

[Jeroen3]

I've learned to not rely on the internal diodes, if you need to pass certification tests they'll put a few kV on the the pin as surge. You really want to catch those directly at the connector.

But for starters, it'll work.