More temperature stable analog Optocoupler

Go To Last Post
20 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I plan to use an optocoupler for isolated analog signal transmission. I looked at the datasheet of SFH617A2. The problem is the current transfer ratio is too much temperature dependant. When Temp rise from 25Degree to 50degree the CTR changes %35-%40.

Linear optocouplers are alternative but they require an isolated power supply on input (because of opamps circuitary).

Is there a less temperature dependant alternatives to SFH617.

Any ideas?

incal99

________________________________ We dream of a world where current does not need the voltage to flow.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

That is probably one of the "holy grails" of circuit design and it really has not been found, yet. I would convert to some time dependent thing, like frequency, PWM, etc, and use either a transformer or opto-coupler in a mode that does not depend on CTR.

Jim

 

Until Black Lives Matter, we do not have "All Lives Matter"!

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

My initial design is based on a PWM duty cycle modulation and transmission over optocoupler, I also thought abaout Voltage to frequency convertion and transmission over optocoupler.

Unfortunatelly they both require isolated power supply on the input side.

I believe carrying the singnal through a frequency modulation is the least CTR dependant one. PWM with duty cycle variation can also be effected by CTR.

Anyhow, it is now more clear that I should choose the longer but more accurate way.

incal99

________________________________ We dream of a world where current does not need the voltage to flow.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Can you explain the power you have available on the input side?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The input side (the side to be measured) is the output of an isolated resonance power converter. (It is an LLC converter) .Output Power (Max) is about 5000Watt. The problem is that the output voltage is variable and can be between 3V and 12V. The converter has 2 modes of operation Voltage mode and current mode, especially in current mode the voltage available can be very small or alternating. (maybe not so fast alternating but surely not stable according to the variable load)

So IMHO, If I want to implement a v/f converter I should also realise an extra isolated power supply to power the V/F converter independant of the voltage to be measured.

If I can find a relative expensive optocoupler that I can *passively* measure and transfer the voltage information it would be the best.

I thought abaout something like IL300 but again, I would need an Opamp to compensate the photodiode dependacy (IL300 has a servo coupling to subtract the temperature drift).

Additionally a good v/f converter is also not cheap (10-12$)

Is there any cheap and available magnetically coupled Ic to transfer analog Dc voltage, Maybe a resisitive divider and then a Hall effect sensor to mirror the voltage data?

Any ideas?

incal99

________________________________ We dream of a world where current does not need the voltage to flow.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Use two (a dual?) opto-couplers with the second in the feedback path to compensate
for non-linearity and temperature changes as detailed in this post.

Stan

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

This is an at least 50 year old principle to measure DC current (from Ta to Tb), or, if Rs removed, DC voltage between Ta and Tb, in an isolated way.

Q1, R1 form a variable load on the transformer's secondary site. The current needed to drive the load is sensed on the primary side over Rt. The oscillator U2 needs to have a low-impedance output to drive the transformer.

Get satisfying linearity and temperature stability requires some effort.

Stealing Proteus doesn't make you an engineer.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Could you use a 1.8V mcu with an adc on the input side? Just stick it behind a simple regulator and have it send packets to the other side using an opto?

oddbudman

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Thank all of you for the original ideas, I should think abaout the Idea of ARnoldB. The simplest and straightforward seems the idea of sten. The opto will be more immune against temperature change. There exists only one problem. I heard that the optos get worse when they alter. If both of them alter equally they will be able to compensate the temperature drift but they will not be able to transmitt the right feedback value. Ex: 2 years later it may be so that their feedback will decrease. When both decrease because of altering , compensation will not help .
Is my logic true?. Or do I miss something?

incal99.

________________________________ We dream of a world where current does not need the voltage to flow.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

oddbudman: Your method is also effective, But I should use PFM (Pulse frequency modulation) to get rid of the sideeffects of CTR completely. Can AVRs have built in PFM? I dont want to use PWM or SPI because I dont trust optocoupler. So the most robust way is the change of frequency (400HZ..100Khz, this should result in 8 bit resolution , it is enough for me)and at the front end measurement of the actual frequency through a counter.

incal99

________________________________ We dream of a world where current does not need the voltage to flow.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Yes, encoding with simple frequency will side-step CTR issues (as long as you design to accommodate both extremes of the CTR) as well as differences in rise time vs fall time.

It is easy to vary frequency of a PWM output. Just choose a mode with TOP controlled by one of the compare registers. Set the other compare register (that sets the pulse width) to be about half the TOP value. You will have a nice square wave with very little change in duty cycle. In fact, you will have more variation in duty cycle due to the opto-coupler than from the frequency generator.

Jim

 

Until Black Lives Matter, we do not have "All Lives Matter"!

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Well really, resonant converters are my own subject and what I write avr programs for.

I use ferrite ring cores to make current transformers.
This is placed directly after the output of the switching devices. Small diodes then rectify this to provide feedback both magnitude and phase.

John

If all else fails, read the instructions.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

incal99,

What i was suggesting is simply sending serial data from the input side via the opto to the secondary side. No modulation, just simple 8N1, manchester, whatever style serial just to get the ADC result to the other side.

No idea of the sampling speed or frequency of what you are looking to measure so i'm interested to hear these parts of your spec.

oddbudman

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

When we talk abaout serial communication or SPI , I feel a little bit unsafe, it exists always the risk of communication failure. It can be because of OPTO or something can go wrong. I would need further CRC or similar, also Serial comm would need maybe at least 5-6V to operate. Winchester coding could be fine, but I am a little bit sceptical because of the fluctuations of the voltage at the output.

The Output voltage is in a range of 2V-12V carrying a load of 5000W-6000W. The load is constant or changes very slowly. So a sampling rate of 100Hz and a resolution of 8bit should be ok.

JohnWalton: For current measurement , I am measuring the Voltage on Cr (resonance capacitor) and then though a capacitive divider (1/10), I get a alternating voltage, after rectification, it gives me a Dc voltage to evaluate the input current. ferrite core or a Hall effect sensor is also a choice.

incal99

________________________________ We dream of a world where current does not need the voltage to flow.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

If you don't have any feed back you have to look at change of the LED (in the opto) after something like 5 years on it's half the ligth.
I would use a 555 (or like it) to make the freq.
offset and gain etc. will allways be a pain in a analog OPTO design :(

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
For current measurement , I am measuring the Voltage on Cr (resonance capacitor)

Then the feedback signal is dependent on frequency and voltage drop in the diodes and assuming series resonant is 90 degrees out of phase with the current.
Using a current transformer avoids this.

I of course dont know your application. In some applications using the capacitor voltage as a feedback signal would be an advantage.

John

If all else fails, read the instructions.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I have also an idea to get rid of the annoying optocoupler. The Hall effect current sensors are evolved, They are cheap and fully isolated and have integrated rectification for alternating currents.

So the measurement of current will not be a problem.I think to switch to hall effect technology.

And for the measurement of Output Voltage, I have the following suggestion:

What if I try to build a transistor based voltage controlled current source. Ex: Vout =2V the current source delivers 1A and Vout=10V current source delivers 5A.

I can measure the resulting current through an isolated hall effect sensor. (It is a simple integrated IC).

I realized it through a resistor. Just serial connected a 2 Ohm resistor with the 1mOhm Hall effect sensor IC. The problem is the power dissipation. At 10V it consumes 50W.

Relavite to the total output power of 5000W it is neglecible but not elegant. So an voltage controlled current source (by using transistors) would be the best alternative).
Using opamps can be a problem, because They need stable supply voltage. But if any opamp can swing its supply voltage between 2V and 12V and work stable it can also be used?

Any ideas of a transistor based voltage controlled current source?

(I had a drawing but I could not upload the picture, where can I upload jpg files)

________________________________ We dream of a world where current does not need the voltage to flow.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

you could use a pulse transformer. One way could be to actually make the tone on the micro side, and just have the sense side load the transformer more or less.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Many commercial inverter welders using a current transformer in the primary measure the current passing through mosfets.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The reason why they choose current transformer can be so that they want to increase the resonance inductance to lower the value of resonance capacitor. A robust high current high frequency capacitor is always more expensive. It is not a bad idea, but I feel myself better with a Hall effect current sensor.

________________________________ We dream of a world where current does not need the voltage to flow.