What is the best way to transfer an analog voltage signal through isolation?

[Jdo300]

Hello,

Lets say that you have an analog control voltage that can vary from 0 to 5V and you want to send it through some sort of isolation barrier to another board. How do you do it?

In my case, I have a voltage sensor that will return this range of values and I want to be able to transfer the signal to an isolated digital board to be measured by a microcontroller with built-in A/D converters. My initial thought was to use an optoisolator but quickly realized this would not work since the light the LED emits doesn't cause the transistor on the other side to conduct linearly with the light intensity. Is there some widget out there made specifically for transferring isolated DC measurements?

Thanks,
Jason O

 

[waht]

One solution would be to design a voltage to frequency converter. It would essentially convert your measured voltage to a signal in KHz range and you could send that signal over a medium of choice such as fiber, coax or a twisted pair. That would be isolated via a coupling transformer if you seed isolation.

Then at your receiving end, you would need to employ a frequency to voltage converter.

VF and FV chips available. But this route is sort of redundant.

Is it possible you could just use a separate ADC instead?

 

[berkeman]

Hello,

Lets say that you have an analog control voltage that can vary from 0 to 5V and you want to send it through some sort of isolation barrier to another board. How do you do it?

In my case, I have a voltage sensor that will return this range of values and I want to be able to transfer the signal to an isolated digital board to be measured by a microcontroller with built-in A/D converters. My initial thought was to use an optoisolator but quickly realized this would not work since the light the LED emits doesn't cause the transistor on the other side to conduct linearly with the light intensity. Is there some widget out there made specifically for transferring isolated DC measurements?

Thanks,
Jason O

waht's ideas are good. In addition, you could consider still using the optocoupler approach, with some offsetting of the input signal (so the opto is always conducting somewhat, even for an original signal of 0V), and making a calibration table for the uC to use when digitizing the optocoupled signal. I've done something similar to this for transferring an analog signal over optical fiber, and it worked pretty well. There are limitations to the accuracy, of course, and you will get variations with opto temperature.

Another approach as waht says is to digitize the signal before sending it through the optocoupler, and transfer the information digitally. That will give you the best overall accuracy.

 

[Jdo300]

Thanks Guys,

I thought about the ADC digitizing method and the Voltage to frequency method. Didn't consider the Opto with calibration table though. That sounds like a really simple way to do it though. I want to avoid using ICs on the isolated side if I can because there is a chance that I might blow stuff up and I'd rather sacrifice a $0.25 opto than a $5 ADC

Would you happen to have any recommendations for an optoisolator that is more on the temperature stable side?

Thanks,
Jason O

 

[berkeman]

Would you happen to have any recommendations for an optoisolator that is more on the temperature stable side?

You should look in the datasheets, for graphs of Current Transfer Ratio versus temperature. Also, if you can find a dual opto (I don't know if they exist), then you could run a reference current through one, and the V-to-I current through the other, and digitize both to help you with tracking variations with temperature, etc.

My guess is that an opto with a PIN photodiode output will be better for your application than one with a phototransistor output (but that's just a guess without looking through the datasheets). Photodiode output parts are faster than phototransistor output parts, and may be more stable for your purposes.

Let us know what you come up with!

 

[berkeman]

BTW, one really cool trick that I've used before (a friend of mine thought it up while we were working on a project way back when), is to use the "reference" signal as the Vref input to the ADC, and thus eliminate the need for digitizing both the signal and the reference signal.

ADCs often have an external Vref that you can drive, and the ADC digitized value is with respect to Vref. Vref is usually either full scale or half of full scale, depending on how the ADC is implemented. By using your reference signal as Vref, you are automatically taking out any variations due to drift of the optos (or whatever analog measurement you are making). If you can make a reference signal that goes through the same signal conditioning as your test signal, it is a great candidate for this Vref input trick at the ADC. You may need to scale Vref to get it to be the correct reference for the ADC, but that's a simple matter of an opamp or two.

Great trick!

 

[dlgoff]

"...send it through some sort of isolation barrier to another board."

Am I missing something here. If the distance isn't too far, couldn't you just use a good cable and feed the 0 to 5 volts there?

 

[berkeman]

"...send it through some sort of isolation barrier to another board."

Am I missing something here. If the distance isn't too far, couldn't you just use a good cable and feed the 0 to 5 volts there?

My guess is that he needs either safety isolation or noise isolation. Certainly direct wiring is the siimplest, but there are sometimes reasons why there can be no galvanic connection between two circuits.

Take a typical switching power supply, for example. The input is AC Mains voltage, and is considered hazardous voltages. No galvanic connection is allowed between the power supply circuitry connected to the AC mains, and the low voltage power supply output rails. So a switching transformer (of appropriate construction) is used to isolate the power path (only flux connects the primary and secondary sides of the transformer -- no galvanic connection). And an opto-isolator is typically used for the feedback path that is used by the primary side switching controller circuit to maintain the output voltage in regulation. The flux in the transformer and the light inside the opto are the only connections between the primary input and the secondary outputs of the switching power supply, and so the outputs are safe to touch, whereas the inputs are not.

 

[zeitghost]

Somewhere, somewhen, I saw an isolator that used two optos, one as the isolation, the other in the feedback to an opamp, thus linearising the response.

 

[f95toli]

We use a lot of fibre to transfer analog signals where I work (high precision measurements, we even sell some electronics to other labs) and as far as I know all circuits made in house by our engineers use the DAC/ADC method simply because optocouplers are not precise enough even with calibration tables. Optocouplers are great in certain applications but I would't recommend them in any form of measurement setup if you need reasonable accuracy (people have tried, but it just doesn't work very well).

One "trick" is to use a digital potentiometer instead of a DAC, the latter can be used in circuits that are usually less noisy than circuits based on DACs (since you do no need to constantly clock them) and controlling them via fibre is quite easy.
However, you might still need a calibration table if you need precise values but in this case that is less of a problem since the steps are discrete (the calibration table is merely compensating for errors in the resistor values).

 

[Jdo300]

Thanks everyone, I'll definitely take all these suggestions into consideration.

 

[dsa]

a little late but you may want to look into isolation amplifiers
google iso124

I don't have enough posts to post a direct link: