Offline SMPS with digital isolator feedback is OK?

[treez]

Hello,

Do you see any problems with this (non-optocoupler based) isolated feedback method for offline SMPS’s? Why does nobody do it like this?

The attached schematic shows what surely must be a total revolution in isolated feedback for offline SMPS’s? –It allows the feedback signal to cross the isolation boundary without recourse to an optocoupler. (instead , a digital isolator is used, freeing the engineer from the opto with all its inherent problems such as poor parameter tolerance and ageing problems )

With the error amplifier on the secondary side, the feedback signal takes the form of a variable duty cycle square wave, produced by a simple PWM controller IC. This is passed across the isolation boundary by the digital isolator. On the primary side, the PWM error signal is filtered and then fed to the PWM comparator in the primary side PWM controller IC. –Hence the SMPS output is regulated

There is also the LTspice simulation attached.

What’s wrong with this way of doing it?
Why does nobody do this?

SI8410 Digital isolator datasheet:
http://www.silabs.com/Support Documents/TechnicalDocs/si8410.pdf

LT1243 datasheet:
http://cds.linear.com/docs/en/datasheet/1241fa.pdf

 

[Velkarn]

what is the advantage of this method?
compare cost: (TL431+3-4 components+optocoupler) with (LT1243+great number of components+optoisolator)
in addition, i have a doubt about stability of this system.

 

[dick_freebird]

The opto I think still has a better HV isolation and is
a part you don't have to "sell" the old-timers on using.
Sometimes you just don't care enough to fight, and
use what's always worked.

The new type isolator guys I'm sure would like to
help you give it a try. That's the whole point. But
the opto folks aren't giving up spots without a fight,
and that includes the standards game. Avago
literature on this is sort of creepy.

 

[treez]

Advantage is freedom from opto problems such as horrendous poor tolerance on ctr and base capacitance.

 

[Velkarn]

Advantage is freedom from opto problems such as horrendous poor tolerance on ctr and base capacitance.

there is hundreds of ref designs and billions of working PS, so what the opto problems? i mean commercial electronics.
in other areas, such as hi rel, military, space electronics there is another methods of feedback. for example, in military and space devices using of optocoupler based feedback is not allowed (on account of it`s radiation sensitivity and time degradation), so we using transformer based solutions.
anyway, thinking away from stereotypes is good, so you invented something great in future

 

[treez]

well, you say it is great but it didn't feel great when I saw the undershoot on vout...how do I solve that undershoot? (the ltspice sim is above)

 

[Velkarn]

the undershoot occurs in PS with posted shematic (with LT1243 at secondary side)?

 

[treez]

yes, it can be seen in that simulation

 

[Velkarn]

i am sorry, i have no time to simulate your shematic. i think its normal practice to post maximum information: design specification, circuit, diagrams if you want to get help.
when the undershoot occurs?

 

[treez]

ok thanks,, but LTspice is not like any other simulator, LTspice is extremely quick to use, just download, open, open .ascfile, hit running man icon....job done.
The undershoot occurs following a no load to full load transient.

To be honest, In my humble opinion a simulation is far easier and quicker to look at than a page of text, or even a non-simulation schematic

 

[Velkarn]

there is classic simple feedback circuit: TL431 with compensation network and optocoupler. the error signal (instead of voltage information) transferred to primary side.
and now your shem. C6,8,14,10,11,13,12 and a lot of resistors have a fluence on the frequency characteristic and consequently on transient response. did you analysed amplitude-frequency response? why you do not use classic shem? is any special design requirements?

 

[treez]

it is power supply for class d guitar amplifier.
It must have a good no load to full load transient response...we don't want the undershoot or overshoot.

TL431/opto is the cheap way of doing it if you don't care about transient response.
Also, TL431 needs more effort in doing feedback loop toleranceing, due to optocoupler poor tolerance in ctr and base capacitance.

guitar amplifier cost £500, so spending a few more pennys to make the power supply feedback loop more robust is money well spent....the extra components you point out are cheap...(cheap compared to £500)

 

[Velkarn]

and what undershoot is now? what is "no load"? is any internal load resistor?

- - - Updated - - -

no, i pointed this coponents not because of its cost, but because of fluence to AFR

- - - Updated - - -

and most of that capacitors is integrating, so slowering the response

 

[treez]

the undershoot, if you run the simulation , you can see it goes down a volt lower with the digital isolator solution. Yes the caps are integrating, but it should be possible to get a good gain and phase margin with the digi isol...its just a case of whether its worth it from a undershoot point of view.
When "no load" theres say 10mA in the processor , and the divider resistors etc

by AFR I take it you mean frequency response(?)

 

[Velkarn]

1V undershoot with 80V Vout is good, i think. thats 1.25%! what undershoot you want to arcive?

commonly measuring transient from 0.1 of nominal to 1.0. if you need from no load thats special requirement afaik.
there is the practice to add load inside of PS to get better stability. you spend 2W of power.

 

[treez]

to be honest , if the undershoot can be gotten as low as when a standard tl431/opto configuration is used, then we would be happy. (and I am talking about no-load to full load, (absolutely no load))

 

[Velkarn]

i do not understand, with optoisolator you gets more amplitude of undershoot than with optocoutler+TL431?

 

[treez]

I was just using tl431 as an example, really, any configuration using a secondary side error amplifier, whether that's a tl431 or not I am not concerned.
Basically, I am looking to get the digital isolator solution as a firm competitor to an optoisolator solution.
To be honest, I think the filters after the digital isolator are making the bandwidth too low., even though I made these filters as light as possible.

 

[Velkarn]

so its cooked up revolution, now i understand))