# In an isolated SMPS feedback design, how could optocoupler latency be higher than the switching period?

I often see this in practical SMPS designs. An optocoupler & voltage reference pair is used for informing the controller section about the output voltage level. However, when I read the datasheet of the optocoupler in the circuit, I see that the given signal characteristics "rise time" and "fall time" are higher than the switching period of the controller circuit.

### Example

In the schematics, it says that the switching frequency of this circuit is 250kHz, which means the switching period is 4us. In the optocoupler datasheet, the rise time is given as 4us (typical) and 18us (maximum). The fall time is 3us (typical) and 18us (maximum). The signal has to rise and then fall, so it takes typically 7us (36us at maximum) for a pulse. An engineer must make his design considering the worst case circumstances, so we should take the maximum pulse time as 36us, which is 9 times the switching period.

Wouldn't it mean that the controller will respond to an output change 9 periods late? Is this much latency acceptable, doesn't it cause output voltage level controlling problems? Or, is there anything I'm missing here? Can you please explain me how does this kind of SMPS designs work with all these slow optocouplers?

-

The opto-coupler is not responding to the switching frequency of the fly back converter so therefore, the rise time and fall time specifications are irrelevant. Although the opto-coupler is generally thought of as a digital device, for this type of circuit it is operating linearly; neither being in saturation nor fully turned off. This means it does operate fluently at good speed.

Admittedly the linear region is small but the action of negative feedback keeps it sat in that linear region.

-
I want to study this kind feedback loop solely for SMPS design. How do search it in Google? Is there an online article? – hkBattousai Mar 9 at 9:55
I would consider using LTSpice - linear tech have a vast number of switchers that are simulateable. I think this is your best route. – Andy aka Mar 9 at 10:08

Note that the opto-coupler is powered by the DC after all the filtering components, so the feedback delay will be much slower. The operation of the supply will have a parameter of "Load Response", "Load Regulation" (or something similar) that lists the typical response to a sudden load change, (time wise this is likely in the range of several mS). The extra (or missing) switching pulses generated during this response time will be well filtered by time they reach the DC output. So the "latency" in uS is not very critical in this case.

-
So the optocoupler is operating in analogue mode, coupling the DC output level back to the control? – pjc50 Mar 9 at 9:34
@pjc50, correct - and the extra circuitry on the LED's cathode side helps to control this too. – Nedd Mar 9 at 9:41