sensing different voltages isolated from each other

[MrAl]

Hi there,


Reed switches arent a bad idea really, Pretty simple to set up and use. You'd have to evaluate the life of them to see how long they would last in a constant open/close applicatoin.

Someone else had a project similar to this one in the past and here was my suggestion:

Use one micro controller per channel, each one a 'slave'.
Use one more micro controller as the master.
Each slave uC reads the analog input, converts it to digital, and outputs the digital code through a cheap digital opto isolator to the master.
The master reads each channel in turn.
This works as long as you have time to read all of the channels.

Because you have so many channels, you may have to divide them into groups and sub groups. For example, 10 first level uC's send their data to a different master, and those five masters send their data to a the one super master. The super master can either process it or send it to a host terminal.

It's up to you to come up with the timing protocol for each slave and master and the one super master.

 

[abhichaotic]

hi mr. AL,
i had tried this but there is loss of linearity through optoisolator which is creating the problem.

 

[abhichaotic]

Do you need isolated grounds for your measurements? Are the commons of the various voltages possibly at different voltages?

hi mr. crutschow
yes sir you are right. commons are at different voltages

 

[ericgibbs]

hi mr. crutschow
yes sir you are right. commons are at different voltages

hi,
Would a simple drawing help you understand the solution I am suggesting.??
I am assuming 0V will be common to all the dc/dc converters under test.?, if not it will mean switching the 0v lines, which can be done but its more expensive

hi,
I have asked about this point, above was my answer, it easy to do, but costs more.

 

[ronv]

You can use a switched capacitor system to do what you want. Building on Erics reed switches (or opotos) Switch both the common and the voltage to charge a capacitor. Ground reference this capacitor with a differental amplifier and measure with your micro. Visualize it as charging a cap then moving it over to a voltmeter to measure it.

 

[MrAl]

hi mr. AL,
i had tried this but there is loss of linearity through optoisolator which is creating the problem.

Hi again,


No no, you only loose linearity when you use the 'standard' run of the mill opto as a linear device. The system i proposed uses the opto as a digital element, which does not loose linearity.
The AD converter in the slave uC chip converts the analog signal to be measured into a digital word (or words), and that is what is transmitted through the opto isolator, not the analog signal.
For example, if we have a power supply of 5.12 volts and we measure an analog signal of 10mv we would get a digital word of:
00000010
That is then transmitted through the opto isolator and comes out the other (isolated) side as:
11111101
and we simply invert that and get back the original digital code:
00000010
and that is what ultimately gets transmitted to the host, which probably stores that sample.

Note in the above i had shown 8 bits out of 10, the first two are both zeros.

Here's a rough block diagram:

Signal o--->ADC--->DIGITAL--->TRANSMIT--->OPTO--->MASTER_RECEIVER

Note that in the above the OPTO is in the digital signal path, not the analog path. Doing it this way maintains linearity as well as accuracy. The cost for 10 bit accuracy is about 1 dollar US per channel, plus the cost of the opto's and the 5 masters and 1 super master.


A system which does not work looks like this:

Signal o--->OPTO--->RECEIVER

where the OPTO is a common garden variety, and that is *not* the system i proposed.

 

[abhichaotic]

thanks Mr. Al,
we are working now on your idea and ur valuable suggestions are of great help to me. just guide me if i had any problem in this method. i really see a ray of hope now

 

[MrAl]

Hi again,

Ok no problem. As i was saying, this system has already been implemented in someone else's project that had to measure 12 different isolated inputs and it works pretty good. It's not all that complicated anyway.