The 1-Wire fusion of network connection and power offers great economy, but the question of how to provide energy for the occasional indicator or sensor must be addressed.
Dan Awtrey, Dann EL Associates
The 1-Wire net is an economical bus based on an open-drain master/slave multidrop architecture that uses a resistor pull-up to a nominal 5 V supply at the master. The net contains three main elements: the bus master, wiring and associated connectors, and 1-Wire slaves. The protocol uses conventional CMOS/TTL logic levels with operation specified over a supply voltage range of 2.8–6 V. Master and slaves are configured as transceivers, allowing bit-sequential data to flow in either direction, but restricted to one direction at a time, with data read and written least significant bit first. Data are transferred with respect to time slots; to write a logic one to a slave, for example, the master pulls the bus low for 15 µs. To write a logic zero, the master pulls the bus low for at least 60 µs to provide timing margin for worst-case conditions. A system clock is not required, as each slave is self-clocked by its own internal oscillator synchronized to the falling edge of the master. Power for slave operation is derived from the bus during idle communication periods when the data line is at 5 V. Except for a software-controlled 15 mA strong pull-up used to accelerate the rising edge of a time slot, only 5 mA is normally available on the net. It is this limit that hinders the powering of indicators and sensors. The energy supply problem can be solved in any of four ways, all of which take into account the amount of energy required and for how long, as well as to the distance from its source to the bus master:
Sourcing power whenever the line is above 3.5 V
Sourcing power by transferring charge to a capacitor through a blocking diode
Sourcing power with the strong pull-up during idle communication time
External power source using spare conductors in the cable
Tapping the Power Available Between 3.5 and 5 V
Because 1-Wire slaves are designed to operate off a single lithium cell, the energy available between supply levels of 3.5 and 5 V can be tapped. This is equivalent to operating the load in shunt mode and may be used to operate clamp-type loads such as LEDs. This
Figure 1. Whenever the output of the DS2406 is pulled low, the LED is on and the voltage on the bus is ~3.5 V, for forward voltage of the LED.
requires that the total voltage drop across the LED(s) be at least 3.5 V to allow sufficient noise margin for reliable communication. While it is possible in certain installations to connect the shunt load permanently across the bus, the load should generally be operated under bus master control by connecting it between 1-Wire DATA and the output of an addressable switch (see Figure 1). In this mode, 1-Wire communication takes place below 3.5 V and power delivery occurs above that value. Whenever the output of the DS2406 is pulled low, the LED is on and the voltage on the bus is ~3.5 V, the forward voltage of the LED. When the output is turned off, the LED is off and the bus voltage is at its nominal 5 V value. Operational current is supplied by the bus master, which for the DS2480-based DS9097U com port adapter is limited to ~5 mA during normal communications but can be increased to ~15 mA with a software-controlled strong pull-up during intervals when there is no communication activity. As shown in Figure 1, the bus voltage will be clamped to a level that keeps the active pull-up on and supplying 15 mA. If this is not acceptable, a current-limiting resistor (dotted box) shown connected between the data line and the LEDs will allow the active pull-up to turn off. “The LINK” and other bus master circuits are capable of supplying more power than the DS2480 .
Transferring Charge to a Capacitor
Through a Blocking Diode
Figure 2. A Schottky diode and capacitor across the bus generate a local supply on the net at a point of interest.
For some applications, it may be acceptable to use a series Schottky diode and a capacitor across the bus to generate a local supply on the net at the point of interest (see Figure 2). In its wind speed sensor, the 1-Wire weather station uses this technique with a BAT54S for the Schottky diode and a 0.01 µF ceramic capacitor to supply power for the DS2423 counter . During idle communication periods when the bus is at 5 V, the circuit draws power from the line to power the load and charge the capacitor. This is a discrete implementation of the parasite power technique used internally by 1-Wire slaves to derive their own operating power. The value used for C1 depends on the current consumption of its load and how long the voltage must be held above a design value. While simplicity and low cost make this technique attractive, the circuit introduces leakage and capacitive loading that decrease the net’s range and capacity and place an upper limit to the capacitance that may be placed on the net. Furthermore, in the event the capacitor becomes shorted or held in a discharged state by its load, the net will also be shorted and inoperable. No further communication can take place until the shorted capacitor is replaced or the load removed and the line raised above 3.5 V.
This basic technique can be enhanced by using a special-purpose IC to generate a local high-voltage supply on the net. The Supertex HT0440, intended as a MOSFET gate driver, generates two independent and isolated DC voltages that permit isolated 1-Wire control of AC
Figure 3. A sophisticated application of the HT0440 connects the two outputs in series to develop a local source of ~25 V, useful as the reverse bias voltage for photodiodes.
loads . A more sophisticated application of the HT0440 uses the two outputs connected in series and referenced to the voltage developed from the parasite supply to develop a local source of ~25 V that is particularly useful as the reverse bias voltage for photodiodes. As shown in Figure 3, an HT0440 is connected to the parasite power generated by CR1 and C1 with its outputs connected as just described to supply a high reverse bias voltage for photodiode PD1. If a PIN photodiode such as the OPF470 is used, the circuit can function as a detector sensitive to gamma radiation . Some means must then be provided to shield the sensor from optical and IR radiation, such as by adding a metallized Mylar covering. The output consists of random spikes generated whenever a gamma ray passing through the diode creates a momentary increase in leakage current.
Delivering Energy under
Bus Master Control
Figure 4 illustrates how the half-wave rectifier of Figure 2 can be isolated between two addressable switches controlled by the bus master.
Figure 4. The half-wave rectifier in Figure 2 can be isolated between two addressable switches controlled by the bus master.
When the input switch is closed, the capacitor receives charge over the data line in the manner described above, except that because of the greatly increased capacitance the bus must be totally dedicated to charging the storage element during this interval and no communication is possible. The length of the interval is a functioý of the available supply current and the chosen capacitance value. The significant advantage of this arrangement is that when the switch is opened, the capacitor and its charge are isolated from the net and normal communication resumes without the burden of leakage or capacitance loading by C1. Equally important, isolation of the storage element and control of the energy source cycling by the bus master prevents a capacitor/load failure from bringing down the net. If C1 should happen to be shorted when the CHARGE CAP. switch is closed, the DS2406 switch will revert to its default open state when its parasite power dissipates, automatically removing the fault condition from the main bus. When the stored energy is needed, the output POWER DUMP switch is closed and the capacitor is discharged through the load. Important elements of the concept and architecture are the low-level transfer of energy from the bus master to a storage element, and the later use of that energy in a high-energy burst. The principle is somewhat similar to the way the circuitry used in a flash camera develops the energy needed to fire the flashbulb.
Figure 5. In this practical example of Figure 4, a DS2406 is used as the control element and pFETs for the switches.
A practical example of this concept using a DS2406 as the control element and pFETs for the switches is shown in Figure 5. Notice that the MIC94031 FET isolation switches specified are 4-terminal devices with the substrate terminal brought out. This provides for correct biasing of the terminal under all operating conditions. The gate pull-up resistor is integral to the chip but shown for clarity. To ensure that both switches cannot be turned on at the same time and possibly bring down the net, a lockout circuit is constructed using U2, a 74HC126 tri-state gate. Due to the exclusive or connection of U2’s control pins, only alternate enabling of the pass gates to charge and discharge the energy storage element C1 is allowed. As shown in the truth table in Figure 5, if both outputs of the DS2406 are inadvertently commanded to identical logic states, such as during power up, U2 ensures that both pass gates are turned off.
In operation, C1 is charged by commanding output B (pin 6) of U1 the DS2406 to a logic zero. This turns on Q1, connecting the data line to C1 through diode CR2, which prevents C1 from discharging back through the net. If CR2 were not present when pass gate Q1 was turned on, and a new slave device were to be connected on the main bus, its presence pulse would short the net and discharge the capacitor. In the initial state with no charge on C1, the gate of Q2 (the discharge pass gate) is held at a higher potential than the source terminal by pull-up resistor R4, so Q2 is off. When the bus master turns output B of U1 off, the charge stored on C1 is isolated from both the net and the load, and only leakage paths exist to discharge it. When the bus master commands output A of U1 (pin 3) to a logic zero, pass gate Q2 turns on and C1 discharges through the load.
Figure 6. A barometric pressure sensor uses a DS2450 quad A/D converter for the 1-Wire addressable switches.
In a practical implementation of the concept, a barometric sensor was constructed using U1, a DS2450 Quad ADC as the 1-Wire addressable switch control element (see Figure 6). The DS2450 also reads the charge level of the energy storage capacitor C1 and controls a sample-and-hold (S/H) on the output of the sensor. A major design criterion of the circuit was that the barometer required the energy source (C1) to provide up to 10 mA for 22 ms. In this circuit, two of the DS2450 I/O pins are used as the digital outputs that control the capacitor charge and discharge via analog switch U3 and a specialized switched-capacitor voltage regulator U4 used in place of the output (discharge) analog switch.ýThe charge pump, a MAX684, provides a regulated 5 V ±4% output as the energy capacitor discharges down to 2.7 V . Advantageously, the efficiency increases as the input voltage drops, a very welcome feature when using a discharging capacitor as the energy source. The two remaining I/O pins of U1 are used as analog inputs that read the voltage on storage capacitor C1 and the voltage from the S/H (U7) that stores the output from the barometer representing the current barometric pressure. The circuit performed as expected with values up to 0.22 F for C1, the energy storage capacitor, with higher values maintaining the output voltage constant longer. Of course, higher values require longer charge times.
In operation, pulling U1.7 low closes analog switch U3 and allows C1 to charge through CR1. CR1 prevents C1 from discharging back through the 1-Wire net as described previously. The voltage developed by charging C1 can be read as needed by U1.8 to ensure that there is enough energy to operate the load. When U1.7 is turned off, analog switch U3 also turns off and the charge stored on C1 is completely isolated from the net. At the appropriate time, U1.6 is pulled low; this enables voltage regulator U4 and provides a path for C1 to discharge through barometer U5. The MPXA4115 requires 22 ms maximum to turn on and stabilize, after which the output voltage representing current atmospheric pressure is stored on C3, the sampling capacitor . After sampling, U4 turns off to minimize energy loss from storage capacitor C1. The circuit can be improved by replacing the wide sample interval used in the prototype circuit with a narrow pulse immediately after the output has settled.
Using Additional Wires
Normally Available in the Cable
To maximize performance, Dallas Semiconductor recommends CAT 5 UTP (unshielded twisted pair) for the 1-Wire net. However, since CAT 5 typically comes with more than one pair, there is a natural tendency to use a spare pair for routing power. A look at some cable properties will help in understanding how such an arrangement affects 1-Wire performance. In a CAT 5 cable with multiple twisted pairs, on average any given conductor in a pair is adjacent to another conductor in a second pair for half its length. When grounded, this spare wire adds ~30 pF/m to the 50 pF/m between the two conductors of a single pair. Because this increased loading reduces
Figure 7. Shown here is the loading effect of grounding spare conductors in the cable using data from equal lengths of CAT 5 and 6-conductor flat phone cable.
performance, Dallas recommends that unused wires and shields be left unconnected at both ends of the cable. The loading effect of grounding spare conductors in the cable is detailed in Figure 7, with data taken using equal lengths of CAT 5 and 6-conductor flat phone cable for comparison. Notice that while capacitance between two adjacent conductors in the phone cable is somewhat higher than in a CAT 5 pair, the effect of grounding a spare conductor is higher with CAT 5. Still, given a relatively short 1-Wire net with a modest number of slaves, external power can often be successfully routed along with the data and gnd communication pair. Considera-tion must nevertheless be given to current and voltage variations on the power-carrying pair as they can induce crosstalk on data and gnd that disrupts communication.
Since the bus master sees less capacitive loading over flat cable, where the power-routing conductors are isolated from the 1-Wire bus, flat 6-conductor phone cable (Silver Satin) may be used up to about 60 m. This assumes that DATA is separated from the power-carrying conductors by using the outer two conductors next to the 1-Wire GND to carry power. As shown in Figure 7, the wiring sequence should be: NC (no connection), DATA, 1-Wire GND, and then external power and ground on the two outermost conductors. This configuration helps shield the sensitive DATA lead from the additional capacitive load and crosstalk of the external power leads. Notice again that the top two conductors in Figure 7 prior to the DATA line are to be left unconnected. As previously emphasized, unless they are left floating they will substantially increase the capacitive loading seen by the DATA line. An alternative would be to use 4-conductor Silver Satin and assemble the cable using 6-pin RJ-11s with these two slots in the connector empty. Of course, this will have the effect of shifting the 1-Wire communications conductors from using the usuýl red/ green to the yellow/green ones. Unfortunately, a significant disadvantage of flat cable is that it lacks the noise rejection properties of twisted pair cable, so EMI can be a significant performance problem if the net is routed near sources of electrical noise.
A 1-Wire net consists of a bus master communicating with multiple slaves via a single twisted-pair cable over which they also receive operating power. While the fusion of network connection and power provides for great economy, the question of how to provide energy for the occasional indicator or sensor is often raised. Common examples include operating an LED indicator and providing high-voltage bias for a photodiode or moderate current for a pressure sensor. Unless local power is in place, or running a separate cable is an option, some means of transferring these energy requirements over the same 1-Wire communication bus is required. This article has reviewed several ways to provide this energy over the communications twisted pair as well as recommendations for using spare conductors in the cable.
1-Wire and 1-Wire net are trademarks of Dallas Semiconductor.