Why are opto-couplers used in power supplies?

[powersupply]

I see in Dave's teardown of the medical plug-pack that it contained two opto-couplers. What is their function?

[Lightages]

They are used in switching power supplies. Switching power supplies use a transformer, or should, to isolate the input from the output. The problem is that you need a feedback signal to control the input side to regulate the output voltage. So instead of a direct wire back from the output to the input, the feedback signal is sent back using an opto-isolator. This way the input and out sections of the power supply are kept galvanically isolated and the chances of the mains power getting to the output directly are reduced to almost zero.

[powersupply]

Ahhh of course, that makes perfect sense, thank you for your quick response.

[Ribster]

Is this mostly a digital signal for zero crossing, or analog feedback of voltage ?

[Psi]

Is this mostly a digital signal for zero crossing, or analog feedback of voltage ?

It's an analog signal.

Normally the opto LED is used in conjunction with a zener diode or shunt regulator so it begins to illuminate when the output voltage gets close to the desired level. 

Basically, since the opto LED wont turn on until its forward voltage is reached, you can offset/raise the LED voltage so that it starts to turn on at a specific voltage.
This has the effect of focusing the LED range (from "off" to "on") right at the point your interested in. 11.5V to 12.5V for example if you were trying to regulate 12V

The transistor side of the opto is usually fed into an error amplifier which generates a analog signal proportional to how different the output voltage is compared to what it should be. The drive electronics then uses this to make corrections to the switching PWM.

[coppice]

Is this mostly a digital signal for zero crossing, or analog feedback of voltage ?

Used properly, an opto-coupler is linear over a very wide dynamic range. In most power supplies they are used for analogue feedback. In some recent digital power supplies they may be carrying digital data across the isolation barrier, but I think that is still rare.

[Seekonk]

Most new wall wart power IC's are a wham bam,  That is they have an on/off pin the opto feeds into.  No PWM at 150khz, voltage is a function of the number of pulses allowed in a second.  You could call it a comparator, but then every digital input could also be called a comparator

[Ribster]

Is this mostly a digital signal for zero crossing, or analog feedback of voltage ?

It's an analog signal.

Normally the opto LED is used in conjunction with a zener diode or shunt regulator so it begins to illuminate when the output voltage gets close to the desired level. 

Basically, since the opto LED wont turn on until its forward voltage is reached, you can offset/raise the LED voltage so that it starts to turn on at a specific voltage.
This has the effect of focusing the LED range (from "off" to "on") right at the point your interested in. 11.5V to 12.5V for example if you were trying to regulate 12V

The transistor side of the opto is usually fed into an error amplifier which generates a analog signal proportional to how different the output voltage is compared to what it should be. The drive electronics then uses this to make corrections to the switching PWM.

Thanks Psi for the nice explanation, the analog opto's are starting to make sense now.
I have digital-ish experience, so the older analog stuff isn't always as apparent.
Makes much sense to use an analog opto. Although it would be a challenge to get a nice linear signal through.

[Rerouter]

The other thing to take note of with most of these implementations is what happens when the opto's current gain drops over time, or completly fails, the general design should fail with output to 0V (low side control), however there are many that will fail to full supply or much higher in order to reduce parts count so it will start reliably,

arguably the second approach can be made safe, but with most things the cheapest solution wins :/

[amyk]

The other thing to take note of with most of these implementations is what happens when the opto's current gain drops over time, or completly fails, the general design should fail with output to 0V (low side control), however there are many that will fail to full supply or much higher in order to reduce parts count so it will start reliably,

arguably the second approach can be made safe, but with most things the cheapest solution wins :/

In practice, in a cheap supply the caps and primary-side silicon are going to fail far earlier than the optocoupler, so open feedback is a relatively rare failure mode...