Between those two extremes are processors that provide D+/D- but have extra discrete pins for pull-up control and cable power detect (Power Good). These intermediate implementations have more flexibility, but also demand more of an isolation interface.
We regularly get inquiries on how to handle isolation of the intermediate case using the ADI ADuM4160/ADuM3160 USB digital isolators. The isolator has a control features such as the PIN function that allows a processor with a discrete pull-up control pin to pass that signal to the upstream side of the isolator. Some processors include a power-good input as well. The processor can monitor it for when power is applied to the cable. There is not an explicit way to support that feature with the ADI isolators. There are several ways to create an acceptable power-good signal.
The first and easiest way to provide a power good signal is to simply tie it to the processor supply. That makes the processor think that the USB cable is always attached, even if nothing is plugged in. In most cases the USB hardware handles this without causing any issues.
If the application can get power from the USB cable through an isolated DC/DC converter, such as the optional ADuM5000 in the USB evaluation kit, the power-good can be generated from the output of the converter. This is the closest to the implementation that the processor intended. It indicates that there is a real power source out there. On the other hand, if the peripheral had power already and was able to monitor its power good input, it is likely that the application did not require a DC/DC converter to start with, so this case is somewhat moot in isolated systems.
Lastly, an additional isolated signal channel can be added to detect when power is present on the upstream USB connector. Of course I would recommend a part like the ADuM1100 to transfer a signal to indicate that power is present on the cable connector. But that solution is a little bit overkill using a 50-MHz-capable part for a 0.00001 Hz signal. Sometimes when the application demands no performance and has no significant environmental demands, the optimum solution may not be the most elegant solution. As much as it pains me to say it, an inexpensive optocoupler can implement this function adequately and cheaply.
The cheap opto implementation, as shown in Figure 1, can have low power demand from the bus so that standby current requirements are not violated; it needs no speed to speak of, as long as it eventually transfers the signal to the secondary. There is a type of optocoupler that fits this description almost exactly - the high-gain 4-pin Darlington coupler. These can have CTR of over 2000% so they can operate at low power, though painfully slowly, and they are very cheap way to get a small digital signal across an isolation barrier. So here is a tip of the hat to a function that may keep optocouplers from disappearing completely from the scene, at least for now. For now? Analog's offering