Voltage sense as feedback but with galvanic isolation feature

Thread Starter

gdylp2004

Joined Dec 2, 2011
101
Hi all,

Is there one such IC chip or device that could:
  1. Sense a voltage at a terminal with a decent accuracy
  2. Provides galvanic isolation between sensed input and output

I was thinking of an optocoupler but the fact that it's o/p voltage does not varies with I/p renders this unsuitable for my application?
 

crutschow

Joined Mar 14, 2008
34,280
Analog Devices makes isolation amplifiers which may work for you.

Edit: Otherwise, if you want to use use an optocoupler in a feedback loop for a switching power supply, you generally can as long as you bias it in the region where it has a relatively linear input current to output current. Here are some opto couplers with a built-in reference voltage specifically for such applications.
 
Last edited:

Thread Starter

gdylp2004

Joined Dec 2, 2011
101
Analog Devices makes isolation amplifiers which may work for you.

Edit: Otherwise, if you want to use use an optocoupler in a feedback loop for a switching power supply, you generally can as long as you bias it in the region where it has a relatively linear input current to output current. Here are some opto couplers with a built-in reference voltage specifically for such applications.
Thank you for your reply. If I would choose an optocoupler from Avago, could I know what is the correct category for such sensor. I did look at the Voltage/current sensor products range but realised the DIP versions are only good for threshold voltage sensing: http://www.avagotech.com/pages/en/optocouplers_plastic/isolated_voltage_current_detector/
 

crutschow

Joined Mar 14, 2008
34,280
Thank you for your reply. If I would choose an optocoupler from Avago, could I know what is the correct category for such sensor. I did look at the Voltage/current sensor products range but realised the DIP versions are only good for threshold voltage sensing: http://www.avagotech.com/pages/en/optocouplers_plastic/isolated_voltage_current_detector/
It would seem that one in the "General Purpose Phototransistor Optocoupler" might do what you want.

If you explain your requirements in more detail I may be able to give you further information on how it would work in your application.
 

Thread Starter

gdylp2004

Joined Dec 2, 2011
101
Your guess is actually right as I am doing a SMPS and have been posting in my thread over here: http://forum.allaboutcircuits.com/showthread.php?t=62817&page=9.

For convenience, I've attached the schematic of my buck driven by an optocoupler gate driver (HCPL3180) to provide galvanic isolation. Of course, for isolation to be made complete, the F/B must have this property as well, hence the question I've posted earlier in this thread.

My o/p voltage is always (or at least very close) to 28V DC and its max load is about 3.6W, hence max load is about 100W basically. My idea is to sense any deviation from the o/p, give this signal to an F/B optocoupler, then in turn produce a proportional and related o/p voltage signal to my MCT which then controls the duty cycle of the PWM that is being fed into the optocoupler driver. The cycle goes round and round.

I am thinking it might be alright even this optocoupler is just basically a voltage follower, that is, if i/p is 28.1V, o/p is also 28.1V and so on. While the MCT input may only allow a smaller voltage range, says 0-5V, I can simply use a pair or high value resistors to "step-down" the sensed voltage signals accordingly (using KVL). Of course, the resistors I used must have accurate resistance values (meaning very low manufacture tolerance). The high resistance should be selected to minimize power loss by reducing current.

Do you think I made sense?
 

Attachments

crutschow

Joined Mar 14, 2008
34,280
To generate the feedback signal, typically a comparison of the output voltage is made to a reference voltage at the output side of the converter. The result of this comparison is then sent through the optocoupler. That way the linearity of the optocoupler signal transfer is not a significant factor in the circuit's operation.

The Fairchild optocouplers I referenced do this by simply placing an accurate reference voltage in series with the optocoupler input. When the output voltage (through the appropriate voltage reduction divider) reaches the reference voltage, current starts flowing through the optocoupler and generating the feedback signal.

Generally you would need a pot in the output voltage divider to allow for component tolerances and accurately set the output voltage.
 

Thread Starter

gdylp2004

Joined Dec 2, 2011
101
To generate the feedback signal, typically a comparison of the output voltage is made to a reference voltage at the output side of the converter. The result of this comparison is then sent through the optocoupler. That way the linearity of the optocoupler signal transfer is not a significant factor in the circuit's operation.

The Fairchild optocouplers I referenced do this by simply placing an accurate reference voltage in series with the optocoupler input. When the output voltage (through the appropriate voltage reduction divider) reaches the reference voltage, current starts flowing through the optocoupler and generating the feedback signal.

Generally you would need a pot in the output voltage divider to allow for component tolerances and accurately set the output voltage.
Thanks. Unfortunately the Fairchild optocoupler you recommended is no longer manufactured. I am looking one similar but with the Avago brand since they are quite a specialist in this area.

If I implement the typical method you described, would the series reference voltage comes together with the IC or I've to build them using discrete components? And what you've mentioned isn't a voltage comparator?

Take my circuit for example, we can set any voltage as our referenced value by playing with the pot resistance values. Assuming I stepped 28V to 5V as my benchmark, any deviation from the o/p away from 28v would also cause a smaller but proportional error to the input of the comparator which has an o/p connecting to the optocoupler?

Is this how it works?
 

crutschow

Joined Mar 14, 2008
34,280
................
If I implement the typical method you described, would the series reference voltage comes together with the IC or I've to build them using discrete components? And what you've mentioned isn't a voltage comparator?

Take my circuit for example, we can set any voltage as our referenced value by playing with the pot resistance values. Assuming I stepped 28V to 5V as my benchmark, any deviation from the o/p away from 28v would also cause a smaller but proportional error to the input of the comparator which has an o/p connecting to the optocoupler?

Is this how it works?
The Fairchild device had it built-in. For other optocouplers you would have to add it externally.

Yes, you could connect the 5V to an analog comparator circuit with a proportional output. Conceptually that could be built with a 5V reference going to the negative input of an inverting op amp and the error signal going to the op amp plus (+) input. The op amp output would drive the optocoupler input diode.

The op amp output would stay at 0V (for a rail-rail op amp powered by a single plus supply) until the error signal reached 5V. At the point the error difference voltage would be amplified by the gain of the op amp as determined by the input and feedback resistors on the negative input.

What is an o/p?:confused:
 

praondevou

Joined Jul 9, 2011
2,942
I successfully used the TL431 in a feedback loop .

Have a look here.

I also used the HCNR201 in a different design. While a circuit with a TL cannot pass an analogue signal over an isolation barrier with high accuracy, the HCNR can.

The disadvantage is that it needs an opamp on the side that measures the output voltage of the your converter, i.e. it needs an additional power supply.
 

crutschow

Joined Mar 14, 2008
34,280
.................

I also used the HCNR201 in a different design. While a circuit with a TL cannot pass an analogue signal over an isolation barrier with high accuracy, the HCNR can.

The disadvantage is that it needs an opamp on the side that measures the output voltage of the your converter, i.e. it needs an additional power supply.
You don't necessarily need high accuracy or linearity across the isolation barrier if it is in a feedback loop for a power supply. Generally such a loop is fairly tolerant of variations in feedback gain as long as the voltage comparison is done at the output of the converter before the error signal is sent across the barrier.

Also having an op amp on the output does not require an additional supply. You can just use the output voltage of the converter to power a rail-rail op amp. As the converter powers up the op amp will come alive and start generating an error voltage by the time the converter voltage reaches the desired value.
 

Thread Starter

gdylp2004

Joined Dec 2, 2011
101
What is an o/p?:confused:
Thank you guys for your replies. i/p = input while o/p = output

I successfully used the TL431 in a feedback loop .

Have a look here.

I also used the HCNR201 in a different design. While a circuit with a TL cannot pass an analogue signal over an isolation barrier with high accuracy, the HCNR can.

The disadvantage is that it needs an opamp on the side that measures the output voltage of the your converter, i.e. it needs an additional power supply.
I've browsed the Avago optocoupler selection guide and had found that the HCPL-7520 could be the one that suits best for me after reading this application note: http://www.avagotech.com/docs/AV02-1847EN.

This is the schematic I saw and it resembles something which is very close to my idea.


I've selected HCPL-7520 instead of HCPL-7800 is because of its single-ended o/p that could be directly connected to the MCT w/o the need for any amplifier. Hence in this configuration, I've eliminated both the i/p and o/p amplifier.

I've drawn a circuit which shows my intention and would like to hear some comments if it would work or not. Please see attached and all comments or advise are welcomed.

The 10 and 10kΩ resistors are selected such that Vin lingers around in the range of 200mV or less so that the corresponding o/p is guaranteed accurate and linear (as recommended by the vendor). VDD1 & 2 are 5V, which will be powered by a linear reg which I've not drawn into my circuit. Lets put that aside first.

Datasheet for HCPL-7520: http://www.farnell.com/datasheets/1266244.pdf
 

Attachments

jimkeith

Joined Oct 26, 2011
540
I have used two Exar 4151 chips in an optically coupled 100mV shunt isolator circuit. The first is a V-F and the 2nd is an F-V converter. This avoids the non-linearity and stability issues of operating opto-couplers in the linear mode. I came up with a neat circuit that avoided using an op amp and could easily tweak the input offset voltage of the comparator. Had to socket the device as I found that it also tends to double as a strain gauge--twisting the circuit board really messed with the offset voltage--never ran across this one in other chips. The main drawback is the requirement for supplying isolated power to power the VCO.

If doing this again, I would check out the Allegro hall effect current isolator.

For isolating high voltage signals such as armature and bus voltage, I have used high impedance differential amplifier circuits (2M or 4M input resistors) with great success--while not offering true galvanic isolation, the subsequent ground currents (e.g. 100ua) are so low that they were never a problem--well below the 5mA ground leakage current allowed by UL/CSA.

Galvanic isolation is not a complete solution either, because in high voltage PWM circuits, there is lots of AC current flowing through capacitances found in pulse & power transformers, opto-couplers, wiring etc--while the average value of such AC currents may be low, the peak current is considerable--this is often detectable without instrumentation--ever feel the tingle through your finger-tip?
 

Thread Starter

gdylp2004

Joined Dec 2, 2011
101
I have used two Exar 4151 chips in an optically coupled 100mV shunt isolator circuit. The first is a V-F and the 2nd is an F-V converter. This avoids the non-linearity and stability issues of operating opto-couplers in the linear mode. I came up with a neat circuit that avoided using an op amp and could easily tweak the input offset voltage of the comparator. Had to socket the device as I found that it also tends to double as a strain gauge--twisting the circuit board really messed with the offset voltage--never ran across this one in other chips. The main drawback is the requirement for supplying isolated power to power the VCO.

If doing this again, I would check out the Allegro hall effect current isolator.

For isolating high voltage signals such as armature and bus voltage, I have used high impedance differential amplifier circuits (2M or 4M input resistors) with great success--while not offering true galvanic isolation, the subsequent ground currents (e.g. 100ua) are so low that they were never a problem--well below the 5mA ground leakage current allowed by UL/CSA.

Galvanic isolation is not a complete solution either, because in high voltage PWM circuits, there is lots of AC current flowing through capacitances found in pulse & power transformers, opto-couplers, wiring etc--while the average value of such AC currents may be low, the peak current is considerable--this is often detectable without instrumentation--ever feel the tingle through your finger-tip?
Hi Jim,

Thank you for your advice once again.

In the datasheet for XR-4151, it says if we were to configure w/o an external amplifier, the linearity suffers about 1%. To obtain an accurate o/p, it must be connected with an amplifier as shown in Fig 3. That, defeats my purpose as I do not wish to introduce any other components.

And if we consider HCPL-7520 and XR-4151 with both configurations having no amp, the former seems to have a better linearity characterstic (only 0.06% typ).

Datasheet: http://www.farnell.com/datasheets/1266244.pdf

For the hall effect current isolator as you recommended, it has been advised to be replaced with newer tech like the HCPL-7520 by reviewers.

No, I did not touch any of the circuit when I am powering up the 100V, haha, so I did not get any tingle feeling YET! The most "painful" experience with my circuit is when I used my index finger to double up as a biological thermometer to sense the temp. If pain, means very hot, if warm means ok. Haha.

That way of testing ended when I start using proper thermocouple. ;)

SO, what do you think of my circuit as attached? From what I see in Fig 5. of http://www.farnell.com/datasheets/1266244.pdf, a 200mV difference at the isolator i/p would translate an about 1.5V o/p difference. And while assuming my error at my buck o/p is <1V, that would mean my error signal seen by the optocoupler would be 1/28 X 28mV = 1mV. Which means the o/p of my isolator would change 1.5/200 = 7.5mV. And I am not sure if this little difference could be responded by the MCT! ha!

Of course, I may increase the 10Ω resistance value so that the amplification is more significant...
 

jimkeith

Joined Oct 26, 2011
540
SO, what do you think of my circuit as attached? From what I see in Fig 5. of http://www.farnell.com/datasheets/1266244.pdf, a 200mV difference at the isolator i/p would translate an about 1.5V o/p difference. And while assuming my error at my buck o/p is <1V, that would mean my error signal seen by the optocoupler would be 1/28 X 28mV = 1mV. Which means the o/p of my isolator would change 1.5/200 = 7.5mV. And I am not sure if this little difference could be responded by the MCT! ha!
Appears to be a simple and good solution
 

Thread Starter

gdylp2004

Joined Dec 2, 2011
101
Hi guys,

I've just realised a 7.5mV signal could be too small for the MCT to react. This voltage signal would be even smaller when the net difference at the Buck converter's output is much less than 1V, that is, if I would like to have an accuracy of 0.05V, the MCT must be able to sense a difference of as low as 7.5mV/20 = ~0.375mV or 375uV.

Is there any good idea that I could amplifier this error value (seen by the MCT input) without adding or perhaps adding as little passive components as possible?
 
Top