From what I can remember reading it's totally possible. I believe it requires adding like one extra line of code, but they opted to not go that route because they were limited on memory and needed every bit. As far as the temp sensors go anyway. There is code in there to run the ds2413 and that device only operates on parasitic power as far as I can tell. Perhaps the connection is more stable with the extra power line. But seeing as how some of their devices have a direct power leg and others like the ds2413 don't, kinda points to the fact that perhaps those devices actually have a legitimate need for the separate power leg.
From the DS18B20 datasheet:
In parasite power mode, the 1-Wire bus and CPP can provide sufficient current to the DS18B20 for most
operations as long as the specified timing and voltage requirements are met (see the DC Electrical
Characteristics and AC Electrical Characteristics). However, when the DS18B20 is performing
temperature conversions or copying data from the scratchpad memory to EEPROM, the operating current
can be as high as 1.5mA. This current can cause an unacceptable voltage drop across the weak 1-Wire
pullup resistor and is more current than can be supplied by CPP. To assure that the DS18B20 has sufficient
supply current, it is necessary to provide a strong pullup on the 1-Wire bus whenever temperature
conversions are taking place or data is being copied from the scratchpad to EEPROM. This can be
accomplished by using a MOSFET to pull the bus directly to the rail as shown in Figure 4. The 1-Wire
bus must be switched to the strong pullup within 10µs (max) after a Convert T [44h] or Copy Scratchpad
[48h] command is issued, and the bus must be held high by the pullup for the duration of the conversion
(tCONV) or data transfer (tWR = 10ms). No other activity can take place on the 1-Wire bus while the pullup
is enabled.
The DS18B20 can also be powered by the conventional method of connecting an external power supply
to the VDD pin, as shown in Figure 5. The advantage of this method is that the MOSFET pullup is not
required, and the 1-Wire bus is free to carry other traffic during the temperature conversion time.
The use of parasite power is not recommended for temperatures above +100°C since the DS18B20 may
not be able to sustain communications due to the higher leakage currents that can exist at these
temperatures. For applications in which such temperatures are likely, it is strongly recommended that the
DS18B20 be powered by an external power supply.
Summary: when doing a temperature conversion, the sensors are more power hungry. If you do not provide a strong pullup (mosfet), they will be unreliable. A 4k7 resistor is not a strong pullup, espescially with multiple sensors on the bus.
The result is that the sensors are found in a OneWire search and show in the device manager, but as soon as you do a temp conversion, they fail.
Also note that around boiling temperatures, the sensors are unreliable at parasite power.
For this reason, we detect when a sensor is on parasite power and decide not to show it in device manager. This is usually caused by a loose 5V line and not on purpose.
The sensors have 3 wires, why not just connect the third one to 5V for increased reliability?
We recently refactored the whole OneWire code to be faster and more reliable. This is running in the latest Spark release, but I still have to finish it for Arduino. A new Arduino release will come. We set up the new repository to compile for both the Spark and the Arduino.
