[SOLVED] flyback - alternative for optocoupler and aux-winding isolated feedback

[avihaio]

Hello,
I'm looking for additional solutions for isolated feedback.
I know the optocoupler and aux-winding ways, but I'm looking for other isolated creative ways to get an accurate output voltage.

Thanks,
Avihai

 

[KlausST]

Hi,

Any informations why you look for alternatives?
What don´t you like with the current solution?
What improvements are you looking for?

Klaus

 

[avihaio]

Hi, thanks for the reply.
Optocouplers is too much affected by environment which reduces the output accuracy.
Aux-winding is a possible solution but requires custom-made transformer which I prefer not to use.
I'm looking for additional isolated feedback solutions to have better output accuracy.

Thanks!

 

[KlausST]

Hi,

You are concerned about accuracy.

What accuracy are you looking for...at which output voltage?

* there are optocouplers with tighter tolerated specifications
* there are special analog signal optocouplers
* there are optocouplers with monitoring diode
* you may use a PI (regulator) circuit

***
There are analog isolators. Better performance but much more expensive.
I assume AD/DA solutions are not fast enough.
Maybe you can use delta sigma modulators.
Maybe you can use PWM or analog modulators.

Klaus

 

[asdf44]

Yeah what accuracy are we talking?

I’m confused by the idea opto’s reduce output accuracy since the opto is within the loop and the integrator properties of the typical loop will cancel gain drift in the opto effectively. On the other hand the variable gain may force compromises in loop design but I’d characterize this as a bandwidth or performance problems, not “accuracy”.

Klaus pretty much hit all the options excep also there are primary side controllers which sense the reflected secondary volatage at the proper time in the cycle. Though this would not be particularly accurate.

 

[avihaio]

Hi,
The accuracy that I'm looking for is smaller than 5%.
I'll check all your suggestions.

As you are so helpful, I'd like to ask another question that maybe you can answer:
I'm looking also for a controller which does not have an internal oscillator in it. The motivation is that I want to provide my own oscillator signal and don't want to be limited by the SYNC pin restrictions compared to the internal oscillator frequency, so I can be flexible with the frequency that I choose (and maybe change in the future).

Do you have an idea for such a component? I prefer of course not to design the whole feedback loop by myself...

Thanks again!

 

[KlausST]

Hi,

What accuracy are you looking for...at which output voltage?

Giving only "5%" is useless, because it makes a big difference if we talk about a 1.8V supply or a 150V supply.

As you are so helpful, I'd like to ask another question that maybe you can answer:
I'm looking also for a controller which does not have an internal oscillator in it. The motivation is that I want to provide my own oscillator signal and don't want to be limited by the SYNC pin restrictions compared to the internal oscillator frequency, so I can be flexible with the frequency that I choose (and maybe change in the future).

I recommend to focus on one problem only. If this one is solved go to the next problem.

Klaus

 

[avihaio]

Hi Klauss,
The output voltage should be 500V. I think that now that I have some ideas I can start to check what are the pros and cons for each and look for relevant application notes (I'll be happy to hear about,if you know some recommended).

I think that we can move to my second problem.

Thank you!

 

[treez]

well, isolated fededback without an opto.....you can use a high frequency transformer and modulate it with the feedback signal...the demodulate it off on the primary side...there are chips that do this.

- - - Updated - - -

you can also do your regulation from the secondary side, without an opto...and then drive the primary fets through an pulse isolation transformer...but note that you need some little extra circuit to push some power to the secondary in the first place to get it all started off.

- - - Updated - - -

...or of course, instead of pulse transformer, you can drive the primary fets from the secondary side via a digital isolator.

 

[asdf44]

More good ideas for eliminating the opto and I’ll add another: Linears timerblox voltage-to-pwm or voltage-to-frequency etc converts are accurate ways to push signals through isolation.

However I’ll note again I don’t see an opto inside the loop as an issue for hitting 5%. Almost all the alternatives are more expensive and/or lower bandwidth.

I’d start by taking a look at a few manufacturer’s parametric tables for controllers. Maybe start with TI amd their UCC line. Also use Digi-Key by finding the right category and using their parametric search. That is a good way to find all the relevant manufacturers.

 

[avihaio]

Thank you all for the isolation suggestions, I'm checking all the options.
I'll post a new thread for the external oscillator issue.

 

[Alan8947]

I did not read the whole thread, 5% is a piece of cake. Like KlausST said, there are more accurate opto couplers than that. I don't have any specific name, we did a whole lot better than that.

For really accurate output, you can use ADC to read the output voltage and send digital data to the primary side as feedback. Done by MPU or whatever you call now a days. In my days, we use Analog Devices little processor with ADC and DAC and Flash and RAM all-in-one. Tons of them on the market. I am sure what I know is all out dated.

 

[Alan8947]

Here is just an idea I came up, just an idea, not a design:



I put the opto link inside the closed loop feedback. You set up a Vref on the secondary side, I do the summing at the output side by U1. The error signal drive through the fiber optic (F/O) transmitter to the F/O receiver on the primary side. This will be the signal to modulate the pulse width ( of driving the control IC) of the SMPS.


Make sure the SMPS do NOT have it's own closed loop feedback. It is literally just the output of the F/O receiver controlling the pulse width. Simple circuit. The closed loop feedback is done on the secondary side by U1 ONLY. With this, you eliminate the accuracy requirement of the F/O link and whatever on the primary side. The ONLY critical part is the gain resistors around U1. You can easily do better than 0.5% accuracy if you are willing to pay for expensive precision resistors and a precision opamp for U1.

Compensation is done by C1 and R2 to form a lag-lead network. R2 is important to form a lead when the pole from the output filter cap kicks in. Don't just use a dominant pole C1 only like most of the SMPS, that' when you get into conditionally stable and noisy SMPS when you have a dominant pole and then a secondary pole formed by the output capacitance with the output impedance of the transformer. You have to either use Bode Plot or something to match up the poles and zeros.


Just a thought.

 

[KlausST]

Hi,

this is the PI regulator I recommedned in post#4.

The ONLY critical part is the gain resistors around U1.

No they are not critical. You just need to set them that the gain is in a stable region. (Can be easily evaluated from the datasheet examples)
But accuracy is not determined by the resistors. (because the integrating part of the regulator)
Accuracy is determined by VRef accuracy and Oamp offset.

Klaus

 

[Alan8947]

Hi,

this is the PI regulator I recommedned in post#4.


No they are not critical. You just need to set them that the gain is in a stable region. (Can be easily evaluated from the datasheet examples)
But accuracy is not determined by the resistors. (because the integrating part of the regulator)
Accuracy is determined by VRef accuracy and Oamp offset.

Klaus

Ha ha, great minds think alike!!!!:-D

I assume that when he ask for accuracy, it's accuracy of setting, not maintaining the voltage. Just simply maintaining the voltage after setting is too too easy.

There are application that require repeatable precision setting with no calibration. In our instruments, we cannot do calibration, when you put in a number, it has to hit the voltage every time, the same setting has to be the same in different instruments made at different time. Like if I enter a number to get 2.925KV, we expect we get 2.925KV within spec every time when we enter the same number even on different units. That's when you need precision resistors.

 

[KlausST]

Hi,

just to confirm there is no misunderstanding.

In the schematic of post#13..
* when you change the vale of R1 (lets say by 10%) --> this won´t change the regulated output voltage
* when you change the vale of R2 (lets say by 10%) --> this won´t change the regulated output voltage

Klaus

 

[Alan8947]

Hi,

just to confirm there is no misunderstanding.

In the schematic of post#13..
* when you change the vale of R1 (lets say by 10%) --> this won´t change the regulated output voltage
* when you change the vale of R2 (lets say by 10%) --> this won´t change the regulated output voltage

Klaus

Those are for compensation only, R1 and C1 set the dominant pole, the R2 is to provide a zero to compensate for the pole formed by the output capacitor and the output impedance. None of these components are critical, you can use 5% resistor and it'll be fine. All those are just compensations.


I am just pulling out of my behind, don't take it too serious, it's just a place to start and address the potential stability problem. This is a suggestion what to look out, not the end design.

I did not even draw any resistor that is critical. The critical things are the Voltage Reference that provide the Vref. The opamp U1 has to be precision. Then the output should be as stable as the Vref.

But if you think OP just mean 5% stability AFTER adjustment, then none are critical, that would be very easy to design.

 

[asimov_18]

There can be few novel ways to make things work but not sure if they would be particularly useful.

I remember making one circuit two digital opto-isolators where the two optocoupler LEDs were in series and driven by one variable analog current source.
Of the the two photo detectors on was used with an Op-Amp to make servo loop eliminating the non-linear response of LED, and the second
photo-detector used in transmitting analog signal which is more or less linear.

The second solution is to use quantum hall effect sensors like IL710, though strictly speaking IL710 is a digital coupler it could be used in varying PWM signal or something similar
to make it work in analog fashion.

Asimov

 

[d123]

Hi,

What about some kind of (presumably shielded) RFID, like for cat tags, type of signal? Or some form of powerline communications/galvanic isolation using capacitors?

 

[treez]

the attached gives you the cascaded opto feedback

- - - Updated - - -

there is also one from linear.com on common collector vs common emitter connection of the opto in feedback loop...there was an edaboard post on it..but i cant find it now.