Implantable medical device communication system

[0001] The present invention relates to an interface for communication between primary and secondary devices along a two-wire communication bus. More particularly, the present invention relates to an interface for bi-directional communication between an implantable medical device and a plurality of slave devices such as sensors along a two-wire communication bus.

[0002] Implantable medical devices take many forms to provide therapy to a patient. For example, implantable medical devices provide pacing therapy to assist in maintaining proper heart rhythm. Pacing therapy may be applied to treat a number of conditions, such as atrial fibrillation, atrial tachyarrhythmia, atrial arrhythmias, ventricular fibrillation, ventricular tachyarrhythmia, and bradyarrhythmias caused by heart block or sinus node dysfunction. Implantable medical devices may also take the form of an implantable cardioverter defibrillator to provide therapy for sudden cardiac arrest, ventricular tachycardia, and ventricular fibrillation.

[0003] In order to provide therapy, an implantable medical device (“IMD”) communicates with a plurality of implantable slave devices, such as sensors, to monitor conditions including heart electrical activity and blood oxygen content, for example. Both the sensors and the implantable medical device are implanted within the body, and in order to reduce interference to the body, the implantable medical device communicates with the sensors over a small bus having a minimum number of electrically conductive wires. For example, communication information, along with power and ground, is provided over two conductive wires. At the same time, the implantable medical device operates by way of an internal power source, usually in the form of a battery, which has a limited amount of available power. Moreover, because replacement of the implantable medical device requires surgery to the patient, conservation of power is an important consideration.

[0004] Implantable medical devices are often required to deliver therapy in the form of electrical stimulation to the patient. Accordingly, reliable communication between the implantable medical device and the implantable sensors external to the IMD is important. Furthermore, since physical characteristics such as reduced size and space of the devices are desired in order to decrease patient stress, simplify input procedures, and reduce surgical complications, such characteristics are therefore important considerations that need to be addressed. Since the sensors are often required to be placed within the heart, the amount of available space for additional circuitry, power and memory is limited.

[0005] The present implantable medical device communication system includes an implantable medical device (“IMD”) running a master clock. Slave devices resynchronize to the IMD clock on the rising edge. The sensor interface improves the data communication rate and permits simultaneous or individual communication with a plurality of external sensors. A 2-wire bus provides power, clock, and data from a single master to up to 16 slave devices. Half-duplex communication and pulse width modulation of the voltage on the bus allow a raw data rate of approximately 100 Kbps. The data is encoded as “1's” and “0's” for data transmission between the master and slave devices. An initialization procedure determines the long address of each slave device and assigns a short address to reduce command length.

[0006] A message initiated by the master triggers slave measurements. The master writes data to the slave and the master reads data from the slave. Addressing modes allow individual devices, groups of devices, or all devices to be addressed. The message format includes start, command word, address, data, CRC, and stop. Messages may be selectively acknowledged or selectively not acknowledged by the recipient device, however the master device initiates data communications. Slave devices do not put data on the bus without the master first requesting the data.

[0007] The start of a data bit in a message begins with the master pulling the data wire of the two-wire bus up to a high voltage. A fraction of a bit time later, the master pulls the data wire low, such that the duration of the high voltage pulse determines the data value. The master is the only device on the two-wire bus with pull-up capability. The slaves may include a number of implantable devices such as sensors, actuators or slave implantable medical devices (“IMDs”).

[0065] With reference now to the drawings, wherein like numbers refer to like elements throughout, embodiments of the invention are now described. FIG. 1 is a schematic diagram of an implantable medical device system 100 according to an embodiment of the invention. As illustrated, implantable medical device (“IMD”) 102 takes the form of an internal cardio defibrillator or pacemaker, for example, with implantable pressure reference, right ventricular (“RV”) pressure, and RV oxygen. System 100 includes IMD 102 that communicates with internal pressure reference (“IPR”) 104 communication bus 106.

[0066] IMD 102 is electrically connected with right ventricular coil (“RV coil”) 108, Vsensor 110, Vring 112, and Vtip 114 by way of right ventricular lead (“RV lead”) 116. The RV lead 116 has a true bipolar lead (Vtip 114, Vring 112) for differential sensing, along with a Vsensor capsule 110, and RV Coil 108 for high voltage defibrillation therapy. The Vsensor capsule 110 contains a pressure sensor, an oxygen sensor, and a temperature sensor. The right atrial lead (“RA Lead”) 120 is connected to Asensor 121, and a true bipolar lead (Atip 122, Aring 124).

[0067] In practice, a DC bias voltage of 3.0 V is maintained across RV Lead 116. Data in the form of 200 mV fluctuations are communicated on top of the 3.0 V bias. The communication across the two-wire bus of system 100 seeks to minimize resistance, minimize capacitance, minimize inductance, maximize shielding and maximize reliability. The shielding is most easily maximized in a coaxial implementation with a grounded outer conductor and data/clock/power on the inner conductor.

[0068] IPR 104 has a separate lead whose communication bus 106 is hooked in parallel with the other communication busses in RV Lead 116 and RA Lead 120. These all share a common pair of feedthrus into IMD 102. Pressure sensors are absolute pressure sensors that require temperature measurement capability. Accordingly, pressure changes in a capsule due to temperature effects are calibrated out. The sensors are optionally electrically connected in parallel with IMD 102 through a number of different ways. In this embodiment, each lead is plugged into a corresponding port in a connector block and electrical connections between the sensors are made inside the connector.

[0069] Programmer module 130 communicates with IMD 102 by way of RF Link 176 from RF Head 132. The programmer module 130 programs parameters, read status, read diagnostic data and stored waveforms, as well as real-time waveform data. IMD 102 collects data from IPR 104, an oxygen sensor, and temperature sensors at regular intervals on the order of every couple of seconds. Data is further collected from Vsensor 110 at rates up to 2048 samples per second. Information received from IMD 102 is displayed on programmer module 130 by way of monitor 134.

[0070] FIG. 2 is a block diagram of an implantable medical device system 140 including IMD 102 connected to sensor capsules 142 and 144 by way of two-wire bus 178. The diagram schematically illustrates parallel electrical connection of sensor capsules 142 and 144 to IMD 102 with each capsule containing more than one sensor.

[0071] IMD 102 includes micro-controller unit 146, which has a microcontroller memory (RAM, ROM, EEPROM), telemetry port, and therapy circuitry (for pacing, sensing, defibrillation, and timing). Therapy and diagnostic unit 148 provides an accelerometer and minute ventilation, diagnostic circuitry (electro-gram, waveform compression, battery monitoring, and lead monitoring), and basic infrastructure circuitry (power supply generation, ADC, and references, etc.). Clock 149 is provided within therapy and diagnostic unit 148 to provide timing to IMD 102 and the sensor capsules. Power supply 147 is also provided within therapy and diagnostic unit 148 to provide power to IMD 102 as well as external sensor capsules 142 and 144.

[0072] Communication unit 150 communicates with remote sensors and sends power as well as a synchronizing signal or clock signal to remote sensors. Communication unit 150 also contains a transceiver to transmit and receive data over a two-wire communication bus. Communication unit 150 has protection networks to protect IMD 102 against transient voltages and currents induced on two-wire bus 178 due to electro-surgery, electrical discharge, defibrillation, electro-static discharge, electro-magnetic interference, etc.

[0073] Each sensor capsule 142 and 144 contains circuitry to measure and convert pressure, temperature, and oxygen signals into a digital signal through analog to digital converters (“ADCs”) 152. Each sensor capsule (142, 144) also contains a small non-volatile memory, in the form of EEPROM 154, to store a unique sensor address, serial number, and calibration coefficients, etc. Recovery units 156 recover power from the fluctuating voltages on two-wire bus 178 to power each respective sensor.

[0074] Clock synchronization allows a relatively inaccurate high speed clock recovery unit 157 to be available on each sensor, which is periodically synchronized with clock 149 on IMD 102. IMD 102 thereby controls the sampling rate and sampling time for each of the individual sensors by way of transmitting commands to the sensor(s), thereby instructing the sensor(s) to sample. IMD 102 also controls when each sensor puts each bit of data on the bus. The need for high accuracy clock generation circuits within each sensor capsule is therefore eliminated. A protection network is also included in each sensor capsule to minimize the effect of transient voltages and currents induced on the two-wire bus 178 due to electro-surgery, implantable cardio defibrillator (“ICD”) discharge, defibrillation, electro-static discharge, and electro-magnetic interference, etc. As particularly illustrated, sensor capsule 142 includes temp sensor 158 and pressure sensor 160. Likewise, sensor capsule 144 includes temp sensor 162, pressure sensor 164, and oxygen sensor 166.

[0075] FIG. 3 is a block diagram of implantable medical device system 170 implementing sensor communication between programmer module 130, IMD 102 and sensor capsule 174. Three communication protocols are defined to move data from sensor capsule 174 to micro-controller 146 within IMD 102. The first protocol defines how sensor data is digitized and moved to a small buffer memory within recovery unit 156 in sensor capsule 174. This protocol also defines how data is read/written from/to an EEPROM within sensor capsule 174 and placed into buffer memory. This protocol is defined by hardware within each sensor capsule.

[0076] The second protocol is a sensor communication protocol that defines how data is moved from memory within a transceiver at one end of two-wire bus 178 to memory within a transceiver at the other end of two-wire bus 178. The sensor communication protocol defines an addressing scheme, data encoding, and voltage/current levels on two-wire bus 178. Finally, a third protocol defines how data is moved from a transceiver within IMD 102 to the appropriate locations. Once data is received by IMD 102, it becomes available in memory for signal processing, waveform compression, telemetry, and episode storage. This protocol is defined by hardware and firmware within IMD 102.

[0077] The sensor communication protocol defines how data is moved between remote sensors and IMD 102 via two-wire bus 178. Network protocols generally have multiple layers. The sensor communication protocol is therefore described using the open system interconnection (“OSI”) model for a network. All layers defined in the OSI model are not necessary for the present invention. The OSI model is broken into 7 layers. The physical layer determines how a bit is moved across the wires (voltage, current, timing). The data link layer determines how frames of data are created (bytes, words). The network layer determines how packets of data are assembled. The transport layer determines how packets are moved. The session layer defines how a session is started and ended. The presentation layer determines how data is compressed or encrypted. The application layer determines how data is used at the highest level. The application layer is defined by the circuits, firmware, RAMware, and software running in IMD 102 and/or programmer module 130.

[0078] According to an embodiment of the present invention, the physical layer determines how “1's” and “0's” are moved between IMD 102 and sensor capsules 174 via two-wire bus 178. The physical layer also determines how remote sensors are powered from IMD 102. The two-wire bus 178 provides synchronization between sensors 174 and IMD 102. This allows IMD 102 to control when a sensor makes a sample and maintains synchronous operation of all sensors. The physical layer also defines how the integrity of the wires can be checked.

[0079] The transmission of power from IMD 102 to remote sensor capsules is necessary because the sensors are small enough to pass through the veins and into the heart. The size requirements for the sensors do not allow for a battery within the sensor capsule itself.

[0080] Synchronization is important for applications requiring sensors at different locations (e.g. EGM, RA and RV pressure). If each sensor sends data at a slightly different rate, it becomes more difficult for IMD 102 to create records of the sensor data because each sample is taken at a different points in time. Processing of the data also becomes difficult because the time base for the samples is slightly different. For example, calculating a simple difference between RA and RV pressure is complicated if the sample rates are not the same. Data would then be collected at a higher rate from one of the sensors and interpolation or other filtering techniques would be required to time align the sample data. The synchronization signal also allows the clock circuitry within each sensor to be less accurate. If each sensor has to provide data at the same rate without a synchronizing signal, crystal controlled oscillators would be required, which would add significantly to the size of the sensor capsule.

[0081] Integrity of two-wire bus 178 is continually verified during operation. Numerous faults could occur on the wires, which could have important implications for therapy supplied by IMD 102. If the outer insulation of the wires fails in a way that allows current to flow through the heart muscle, the heart could be stimulated or fibrillation could be induced. If an insulation fault occurs between the two wires of two-wire bus 178, the battery of IMD 102 could be drained or loaded, thereby causing a loss of longevity or loss of function in IMD 102. If a wire break occurs, a loss in communication with the sensor would likely occur.

[0082] EMI filters and/or high voltage protection networks may be optionally installed on IMD 102 and possibly in sensor capsule 174. These networks add capacitance and series impedance, which tend to increase power supply current when driving two-wire bus 178. Furthermore, the use of these networks slows down transitions on the bus and may limit bandwidth.

[0083] The wires used to connect the sensors to IMD 102 have characteristics that limit the performance of the bus in terms of data rate, current drain, and noise immunity. The wire characteristics also affect how signals on the bus couple to adjacent wires on the lead and interfere with these signals. The physical construction of the wires further defines the types of faults in the wires that are most likely to occur and impacts how fault detecting circuitry should operate.

[0084] FIG. 4 is a cross section of a typical two-wire bus lead in the form of co-axial lead 180. Outer coil 182 is connected to the sensor capsule, and to system ground on IMD 102. Inner coil 184 is also connected between a sensor capsule and IMD 102, and is used for communication. A stylet is insertable within inner coil 184 to aid in positioning the sensor at the time of input.

[0085] Bus insulation layer 186 surrounds and protects inner coil 184 and outer coil 182. The outer coil is covered by outer coil insulation layer 183 while inner coil 184 is covered by inner coil insulation layer 188. A gap 185 is formed between inner coil insulation layer 188 and outer coil insulation layer 183. Gap 185 may be air if lead 180 is dry or may gradually fill with water if lead 180 is wet.

[0086] Bus insulation layer 186 and inner coil insulation layer 188 are preferably made from polyurethane (Pellathane 80A). Layers 186 and 188 may optionally be made from a biocompatible insulating material including polyurethanes, ETFE, silicone, or polyamides. Insulating materials with low dielectric coefficients are preferred to minimize capacitance between conductors. The gaps between the insulation and conductors will be air when the lead is dry or may gradually fill with water when the lead is wet. Outer coil 182 and inner coil 184 can be made from biocompatible conductive materials including MP35N, platinum, or silver cored MP35N. Conductive materials with low resistance are preferred.

[0087] FIG. 5 is a sectional view of multi-lumen lead 190. In particular, multi-lumen lead 190 is a three lumen high voltage lead intended for an implantable cardiodefibrillator (“ICD”) and pressure sensing application. Lead 190 supports integrated bipolar sensing, a high voltage RV coil, and uses a coaxial cable for communication with a pressure sensor. Lead 190 includes a plurality of compression lumens 192 defined by lead body 194. Each lumen has an OD=0.029″ while lead body 194 is preferably made from silicone. Lead 190 uses one cable with a coil 196, one coil 198, and a cable for HV coil 200. The diameter of lead 190 is defined by D=0.105″.

[0088] The coil 198 is connected to the tip electrode and is used to allow the stylet to pass through to the tip. The cable within coil 196 is used for sensor communication and is preferably silver cored MP35N with an ETFE insulating layer. Cable 200 is used for a high voltage (“HV”) coil. The resistance of the conductors is low (<<5 ohms) due to the use of silver core/MP35N wire.

[0089] FIG. 6 is an axial view of two-wire co-radial lead 210 that uses a co-radial, multi-conductor construction. Lead 210 uses four individual wires wound side by side. Two wires are connected to signal “a” and two wires are connected to signal “b.” The inner diameter of lead 210 is about 20 mils, with the outer diameter of wires being about 37 mils. An outer insulation of polyurethane is used to provide further protection. The inner diameter of polyurethane insulation is 45 mils with the outer diameter of polyurethane insulation 216 being 57 mils. Wires 212(a), 213(a), 214(b) and 215(b) are wound adjacent to each other rather than one on the outside and one on the inside. The wires are wound together with two wires used for each signal. Polyurethane tubing 216 covers each of wires 212–215.

[0090] FIG. 7 is a sectional view of the conductors used within the co-radial lead 210, including a 3.5 mil diameter core 218 preferably of silver cored MP35N. Wire core 218 is dipped in a polyamide coating 220 to achieve a thickness of about 0.5 mils. Each wire 212 is further insulated with outer insulation 222 ETFE to a thickness of approximately 2 mils to obtain a total thickness for each insulated wire of about 8.5 mils. Four wires are then wound side by side to achieve a lead with an inner diameter (“ID”) of approximately 20 mils. A 55D polyurethane tubing 216 approximately 6 mils thick is then slipped over the lead conductors to provide a final layer of insulation.

[0091] FIG. 8 is an electrical model 230 illustrating electrical interference of sensor leads. Electrical model 230 includes IMD 102, which includes transceiver 234 and stimulus circuits 236. Inner wire 238 and outer wire 240 connect IMD 102 to sensor capsule 242 and sensor capsule 244. Each sensor capsule has an interior transceiver 246. Electrical interference from stimulus circuits 236 to the body tissue are modeled through ring 248, tip 250 and case 252.

[0092] There are several variables that affect performance of the communication link, such as signals on the wires coupling into the body causing sensing errors, signals in the body coupling into the wires causing data errors, EMI coupling into the wires causing data errors, CVF current, and effects due to sensors turning on/off. These factors can change by several orders of magnitude depending on the lead parameters, methods used for stimulating tissue and sensing, as well as methods used for EMI and high voltage protection in IMD 102.

[0093] In electrical model 230, Cio 260 models the capacitance from inner conductor to outer conductor of the coaxial cable. Cio 260 directly affects current drain of the system because current is proportional to capacitance*voltage*frequency (“CVF”). Cob 262 models the capacitance from outer conductor to the body tissue. Cob 262 directly affects crosstalk from external noise sources into the communication lead and allows communication signals to couple into the body, where they could cause sensing problems in IMD 102. Cib 264 models coupling from the inner conductor of the cable to the body. Normally this will be very low in a coaxial lead configuration because the outer conductor shields the inner conductor from the body. Ri 266 and Ro 268 model the resistance properties of the inner and outer conductors. Ro 268 is especially important for rejecting noise from the body and for preventing crosstalk from the communication protocol to the body. Any voltage change in the body relative to ground results in a current flow in the outer conductor due to Cob 262. This current is converted to a voltage by the resistance of the outer conductor Ro 268.

[0094] The tip 250, ring 248, and case 252 of the stimulus/sensing electrodes are modeled as if coupled to the body through resistors. In reality, these electrodes are complex distributed impedances. IMD 102 is shown electrically connected to the lead through the body tissue. Case 252 of IMD 102 is typically the largest electrode in a pacemaker/ICD and is typically the most representative of the voltage seen in the body. In some IMDs, the case is tied to system ground through a low impedance switch, while in others the case is left floating. In this example, Rtr 270 and Rtc 272 are modeled as 830 ohm resistors, while Rrc 274 is modeled as a 415 ohm resistor.

[0095] During a bipolar pace sequence, case 252 is left floating, ring electrode 248 is tied to system ground, and tip 250 is driven to a large negative voltage (0 to −9 V). The case voltage ends up somewhere in between the tip and ring voltage, but typically stays closer to ring 248 than tip 250 because the ring electrode area is typically larger than the tip, thereby creating a lower impedance connection. For a unipolar pace, case 252 is held to ground while tip 250 is driven negative. In this scenario, case 252 stays close to ground.

[0096] In some pacing systems, the stimulus circuits ground the tip, while pulsing the ring or case positive. In these systems, the case typically moves a greater distance from ground thereby inducing a larger voltage in the body relative to system ground. Depending on the method of implementation for the pacing system, voltage pulses on the case may reach 9 V with respect to system ground. In nerve stimulators, this value can be as high as 15 V. Rise times for the leading edges of pacing pulses are typically on the order of 100 ns–1 us.

[0097] Pacemakers and implantable cardiodefibrillators (“ICDs”) typically address high voltage protection differently. In an ICD, the case moves as much as 800 V with respect to ground during discharge. Rise times on ICD pulses are typically very fast with slew rates on the order of 500 V/us at the leading edge. During trans thoracic defibrillation, the case moves as much as 1600 V with respect to ground. Rise times on trans thoracic defibrillation pulses are typically much slower than for ICD's.

[0098] In a pacemaker, voltage on the leads is clamped with protection networks. This reduces the voltage on the leads as seen by the implantable medical device (“IMD”), but the voltage is still very large in the body. Similar voltages can be seen in the body during electro-surgery. A voltage limiting network may be incorporated within IMD 102 and in the sensors.

[0099] Interference into the communication wires can occur if the capacitance on the outer lead conductor is large and the impedance of the outer lead conductor is large. As an example, a large amplitude unipolar pace in a pacemaker uses a pace case architecture with the sensor lead outer conductor grounded. The voltage induced on the lead outer conductor during the pulse rise time is approximately: $\begin{array}{cc}\mathrm{Vouter}\approx \frac{\mathrm{Cob}·\mathrm{Vpace}\left(\frac{\mathrm{Ro}}{2}+\mathrm{Rswitch}\right)}{\mathrm{tfall}}& \mathrm{EQ}.1\end{array}$

[0100] In EQ. 1, Cob is the capacitance between outer conductor and body, Vpace is the voltage on the body relative to ground, tfall is the fall time on the leading edge of the pace, Ro is the resistance in series with the outer conductor and is assumed to be distributed, and Rswitch is the value of the switch on IMD 102 used to connect the lead outer conductor to ground (about 10 Ohms). Thus, it can be seen that interference is minimized by reducing Cob, Ro, Rswitch and increasing tfall.

[0101] From the above example, the lead conductor capacitance and resistance properties play an important part in rejecting interference from pacing pulses. The high-voltage multi-lumen lead construction has significantly better performance than the co-axial lead or the co-radial designs due to lower resistance and lower capacitance properties. Further, the architecture used for the pacing circuits and the rise time of the stimulus pulses will be important in reducing the amount of crosstalk. Finally, using the co-radial construction lead results in enough coupling from the pacing pulses to the sensor signal to disrupt communications, if the pulses occur when the signal line is not strongly driven.

[0102] Just as voltages on the body can be coupled into the wires used for communications, voltage on wires can be coupled into the body causing interference with sensing operation within IMD 102. Three scenarios are set forth below wherein the coupling to the body could be a factor: from lead inner conductor (through Cib), from the lead outer conductor (through Cob) due to the DC current needed to power the sensor, and from lead outer conductor due to transient current flowing in the leads during voltage transitions from communications.

[0103] First, coupling from the lead conductor directly to the body is addressed. Any capacitance between the clock/power/data line to the body may cause an attenuated version of the voltage (Vsignal) on the communication lines to be coupled into the body as a common mode signal to tip 250, ring 248, and case 252. A capacitive divider effect will be formed by Cib, Cob, and Cbody to ground. $\begin{array}{cc}\mathrm{Vbody}\approx \frac{\mathrm{Cib}·\mathrm{Vsignal}}{\mathrm{Cib}+\mathrm{Cob}+\mathrm{Cbody}}& \mathrm{EQ}.2\end{array}$

[0104] This coupling is minimized by reducing Cib and minimizing the voltage transistors on the lead used for communications.

[0105] Next is considered a situation where a sensor is enabled and the current flowing through the lead outer conductor causes a voltage drop across the conductor, which is then capacitively coupled into the body through the lead outer insulation. The voltage induced on the lead outer conductor is dependent on the sensor current (Isensor) and the impedance in the lead outer conductor (Rswitch+Ro). The voltage is then coupled into the body through Cob.
Vouter≈Isensor·(Ro+Rswitch)  EQ. 3

[0106] This coupling is minimized by reducing Cob, Ro, Rswitch, and Isensor. Alternatively, a resistive impedance could be added to the tip, ring, or case to allow the common mode voltage to quickly bleed away.

[0107] Last considered is a coupling from the lead outer conductor. In this case, tip 250, ring 248, and case 252 of IMD 102 float relative to circuit ground. Very fast rise time 200 mV communication pulses with 100 Ohm source impedance are applied to the lead inner conductor with the lead outer conductor grounded. The resistance of the lead outer conductor plus any switch resistances cause the current through the lead outer conductor to result in a voltage drop. This results in a voltage spike that will quickly bleed away. The peak of the spike is given by: $\begin{array}{cc}\mathrm{Vouter}\approx \frac{\frac{\mathrm{Ro}}{2}+\mathrm{Rswitch}}{\mathrm{Rsource}+\frac{\mathrm{Ri}}{2}+\frac{\mathrm{Ro}}{2}+\mathrm{Rswitch}}·\mathrm{Vsignal}& \mathrm{EQ}.4\end{array}$

[0108] The magnitude of Vouter is reduced by minimizing Ro, Rswitch and Vsignal, and maximizing Rsource.

[0109] The time constant of the decay is given by: $\begin{array}{cc}\mathrm{Tau}\approx \left(\frac{\mathrm{Ro}}{2}+\mathrm{Rswitch}+\mathrm{Rsource}+\mathrm{Rswitch}\right)·\mathrm{Cio}& \mathrm{EQ}.5\end{array}$

[0110] Tau can be minimized by reducing Ro, Rswitch, and Cio.

[0111] This voltage is coupled through the capacitance of the lead outer conductor into the body causing a common mode step in voltage on the tip, ring, and case. These glitches appear as common mode spikes on tip, ring, and case electrodes, which are then be filtered by circuits in the IMD. Typically, sense amplifiers have good common mode rejection and high frequency filtering of the input signals. As an example, if a sense amplifier or EGM amplifier had a first order low pass filter at 150 Hz, this results in an attenuation of the glitch amplitude of approximately 10,000×. For coupling through the lead outer conductor, the multi-lumen approach provides the least coupling and the fastest decay time making any filtering more effective as well as reducing the peak amplitude.

[0112] The lead properties have an effect on cross-talk from communications taking place on the sensor wires to the signals on the lead. Also, the effect of turning high current sensors on and off can induce voltages in the body large enough to disrupt sense amplifiers or impedance measuring circuits. The cross talk performance of the high voltage multi-lumen lead is better than the co-axial lead or the co-radial construction lead. Crosstalk can be reduced by minimizing the amplitude of the voltages used on the bus and by reducing the rise/fall times of the signals. The addition of capacitance or resistance to the tip, ring, or case to ground can also help significantly.

[0113] There are several potential wire faults that can affect safety and reliability of the system. Typical causes of damage to leads include: crushing of the lead, environmental stress cracking, scalpel nicks, and work hardening of the metal or insulation due to many cycles of flexing. These lead failures can cause gaps in the insulation layers resulting in impedance paths between adjacent conductors and/or impedance paths between the conductors and the body. The lead failures can also cause high impedance paths in the wires.

[0114] Safety issues include prevention of DC current flow into the body tissue and prevention of inadvertent stimulation of tissue for a scenario where the insulation fails. Another issue is that a lead fault does not cause a high current condition that could cause the battery within the IMD to discharge prematurely. In the presence of a lead fault, the system should not allow a DC leakage path to the body with an average current greater than 100 nA over 1 second or a pulsed current condition in excess of 100 uA for 0.5 ms.

[0115] Options for implementing a leakage detection circuit include issuing a command to the sensors to completely power down the sensors on the bus to allow a leakage check between the conductors and the body at normal supply voltages. Another option is to drop the voltage on the bus to a low voltage level (350–400 mV), which can result in a “0” current state for the sensors on the bus. A preferred mode of implementation is to issue a broadcast command to all bus sensors to go to a “0” current state for a period of time long enough to check for leakage. Circuitry in the IMD may then check for faults between each of the conductors or between each of the conductors and the body.

[0116] An important possible lead fault concerns failure of the outer insulation. To minimize the impact of this possible fault, the outer conductor should be maintained at the same DC potential as the body. Some IMDs use a positive battery terminal as a body ground while others use a negative terminal as body ground. In this case, the outer conductor of the lead could be driven to either the positive supply voltage or negative supply voltage. However, a rectifier function within each sensor would be required to allow the conductor polarity to be different for different IMDs. If the co-radial construction method is used, and there is a lead outer insulation fault, there is no shield layer to prevent current flow to the body. Thus, the possibility of a lead insulation fault that could cause a DC current flow into the body is greater.

[0117] FIG. 9 is a schematic block diagram of a sensor interface system 280 for an implantable medical device (“IMD”) 102 and a sensor capsule 284 for lead fault detection. IMD 102 is capable of switching the transceiver and sensor power source 286 or a leakage detection circuit 288 onto inner conductor 290 or outer conductor 292. The outer conductor 292 of the lead is normally connected to body ground, but can be isolated by a switch. The sensor capsule 284 includes a rectifier & regulator 294 to accept voltages of arbitrary polarity from IMD 102. The sensor capsule 284 includes sensor 298 and also includes isolation circuits 296 for isolating the rectifier and regulator 294 and transceiver 295 from the lead conductors, during a lead integrity check.

[0118] Currently developed sensors have differing operating voltage requirements. With multiple sensors on a bus, it is desirable to provide as large a DC voltage as is practical for the supply, and allow each sensor to regulate the required voltage to a required level. This also reduces the effects of line voltage drops due to changing load currents. The maximum voltage level conveniently available in IMD 102 is the unregulated battery voltage, which can be as high as 3.3 V at beginning of life, typically 2.65 V at an elective replacement indicator, and as low as 1.8 to 2.2 V at end of battery life.

[0119] Each sensor has a different operating current drain. Most have low operating currents (1–10 uA) and will be turned on for a few milliseconds each second or run continuously. Some sensors, and/or the EEPROMs within the sensors, may require currents in the milliamp range. If the source impedance of the battery, supply circuitry, and wire resistance is high, the milliamp current pulses could disrupt communications. A few methods are available to address this issue, such as formulating a design protocol to operate through the high current pulses or suspending communications during high current measurement. Another method includes storing charge on a capacitor in the high current sensor and to isolate the sensor from the bus. The preferable method is to suspend communications during high current measurement.

[0120] A two-wire bus provides power, clock, and data from a single master to up to 16 slave devices. Half-duplex communications, and pulse width modulation of the voltage on the bus allow a raw data rate of approximately 100 Kbps. The “1's” and “0's” are encoded the same for master and slave devices. An initialization procedure is used to determine the long address of each slave device and to assign a short address to keep the commands shorter. The protocol includes messages to trigger measurements, write, and read data. Addressing modes are provided to allow individual devices, groups of devices, or all devices to be addressed. The message format includes start, command word, address, data, CRC, and stop. Some messages require acknowledge by the recipient while others do not. The master device always initiates communications. It is not allowable for a slave device to put data on the bus without the master first requesting the data.

[0121] The start of a message is indicated by pulling the bus low for 2 bit times nominally. This serves to wake up all slaves such that they are ready to read messages.

[0122] FIG. 10 is graph of nominally one bit time for communication across a two-wire bus wherein the bit time is divided into 6 equal parts. One bit time can allow up to an 8% tolerance on the clock in both the sender and receiver.

[0123] FIG. 11 is a graph illustrating one bit time communication between a fast sender and a slow receiver. In this case, one bit time ends up being 10 us−(0.08×10 us)=9.2 us.

[0124] FIG. 12 is a graph illustrating one bit time communication between a slow sender and a fast receiver. In this case, one bit time ends up being 10 us+(0.08×10 us)=10.8 us.

[0125] FIG. 13 is a graph illustrating voltage swings on a two-wire bus. The voltage swings are 200 mV. Signals are driven on the two-wire bus. One wire is labeled “MW_DATA” and is driven between Bplus and Bplus−200 mV. The other wire is labeled “MW_SHEILD.”

[0126] FIG. 14 is a graph illustrating a start bit definition. The start of a message is nominally a bus low for two bit times, however some variation on clock mismatch is allowed between slave and master so that the slave can wake up and detect the start bit.

[0127] The sending of a message from the master to a slave device is first described. The start of a data bit in the message begins with the master pulling the bus up to a high voltage. A fraction of a bit time later, the master pulls the bus back low. The duration of the high voltage pulse indicates the data value.

[0128] FIG. 15 is a graph illustrating a master driving a data “0”. A data value with a time high of ⅓ bit time corresponds to a logic “0”.

[0129] FIG. 16 is a graph illustrating a master driving a data “1”. A data value with a time high of ⅔ bit time corresponds to a logic “1”. In its simplest form, the slave's receiver samples the data on the bus with a time delay of ½ bit time (t3) after the rising edge of the bus.

[0130] FIG. 17 is a graph illustrating a slave driving a data “0”, and FIG. 18 is a graph illustrating a slave driving a data “1”. The master is the only one on the bus with pull-up capability. At the end of the bit time, the master again pulls the bus high starting a new bit time. When the slave drives data onto the bus, the slave can only pull the bus down. Because the slave cannot pull up, the slave uses timing information to determine whether to send a data “1” or “0”. This is essentially the same as when the master is sending “1's” and “0's” but the slave shares driving of the bus with the master.

[0131] FIG. 19 is a graph illustrating a master stop, a master acknowledge request (“ack req”), and a slave acknowledge (“slave ack”). A stop bit is indicated by the bus being held high continuously for 1 bit time. The master can send out commands to multiple slaves. If multiple slaves are responding, then some slaves may be acknowledging the message and others may not. A multi-slave acknowledge therefore is of limited value because if received, the master only knows that at least one slave received the message. The master ends commands with a sequence that the data link layer refers to as Stop-Ack Req-(N)Ack. If the master requires an ACK to the last command, the master will put a logic “0” on the bus in the bit time immediately following the stop bit. If a request for an ACK is given, the slave device will respond with a logic “1” in the next bit time as an indication that the slave received the message.

[0132] FIG. 20 is a graph illustrating a master stop, a master ack req, and a slave non-acknowledge (“nack”). If the slave does not receive the message at all, or if the message is received with an error, no ACK or NACK will be provided and the command will be ignored. Defining what is an error based on the command structure and error checking is also further defined in the data link layer.

[0133] FIG. 21 is a graph illustrating a slave stop-stop-stop. When the slave drives data onto the bus in response to a command, the slave drives “1's” and “0's” out to the master. Upon the last bit of data being sent, the master is still unsure if any more data is to follow. Accordingly, the master allows three chances for the slave to send more data. This is referred to as a STOP-STOP-STOP. The stop bit from the slave is indicated by the lack of data on the bus for a period of 1 bit time for a consecutive 3 bit times in a row. The stop sequence occurs for 3 bit times in a row to avoid conflict on the bus where the sensor is pulling down and the master is pulling up at the same time.

[0134] FIG. 22 is graph illustrating the sending of a message from the master to one or more slaves. Thus, a message from the master to one or more slaves is summarized as a start sequence, some data, and a stop sequence.

[0135] The slave device uses a delay chain triggered off the rising edge of the bus to determine when to sample the bus voltage. The rising edge of the bus is used to reset the falling edge of the slave's clock low and the ½ bit time delay after the rising edge of the bus is used to set the slave's clock high. The clock is then used to sample the data on the bus and to synchronize the data with the clock.

[0136] FIG. 23 is graph illustrating a response with data from a slave to master. First, the master requests data from the slave by sending a complete message. All commands that require data from the slave will be preceded by the ACK request. The slave will then put an ACK on the bus and follow up with data. The master then pulls the bus high for ⅙ of a bit time and releases the bus. The slave puts data on the bus in the same manner as the master did when the master was sending data by pulling the bus low. The duration of the high voltage pulse indicates the data value with a time high of ⅓ bit time corresponding to a logic “0” and a time of ⅔ bit time corresponding to a logic “1”. Once the slave has pulled the bus low, the slave releases the bus at t=⅚ of a bit time. The master continues to pull the bus high at the start of each bit time until the master receives a stop bit by the slave indicating the data transmission is complete. When the master recognizes the stop bit from the slave and determines that the message was received correctly, the master finishes the stop command by putting two more stop bits on the bus.

[0137] To prevent contention on the bus, the bus master releases the bus before the sensor tries to pull it low. The slave also releases the bus before the master pulls the bus back high. To meet this requirement, the longest value possible for ⅚ of a bit time seen in the sensor must be less than the shortest value possible for 1 bit time as seen by the master. This puts an accuracy tolerance on the time delays of about +/−8% if rise-fall times and propagation delays for detecting clocks and turning on/off pull-ups and pull downs are neglected.

[0138] The task of a data link layer is to take a raw transmission facility and transform it into a line that appears free of transmission errors. The data link layer accomplishes this task by breaking up the input data into data frames, transmitting the frames sequentially, and then processing the acknowledgment frames sent back by the receiver. The application layer determines the set of allowed messages and the action taken upon receipt of each.

[0139] Data transmission is based on a master-slave relationship. The master initiates all communication by sending commands to the slave, which then responds with actions or data. This allows more hardware intelligence to be placed in the master IMD, which has more resources such as a battery, a crystal oscillator controlled time base, and more area for integrated circuits and external components. The slave electronics are provided to be simple and low in power. The slave may, of course, still support intelligent operations. Signal processing in the slave allows the sending of processed data rather than raw data back to the master, thereby resulting in fewer transferred data bits.

[0140] FIG. 24 is a general command format for transferring data across a two-wire bus. Commands are preceded by a Start sequence and all slaves on the bus wake up in response to a Start. The fields are labeled Master1 through Master7, implying that the master is talking. The data is sent out from the most significant bit (“msb”) first proceeding to the least significant bit (“Isb”).

[0141] The field formats are configured as follows. The Start field is a unique 2 bit time long sequence issued by the master. The Slave(s) field is a field to identify one slave (unicast) multiple slaves (multicast) or all slaves (broadcast). The Slave(s) field is explained more fully below with regard to addressing. The Quick Trigger (“QT”) field is a special field identifying a Quick Trigger (“QT bit set”) command. All commands except the Quick Trigger (“QT bit set”) have this bit set to zero. When the QT bit is set high, the command is identified as a Quick Trigger command.

[0142] The Master Command Name field provides instruction to the slave. A 5 bit command in the Master Command Name field identifies up to 32 different commands. If the QT bit is set, then this field is not used. The Data field is 8 bits long for most commands. The use of 8 bits allows a byte of data to be written at a time. The data field varies from command to command and can be broken down to sub-fields. The frame check sequence (“FCS”) field provides error checking. The FCS field is used for a cyclic redundancy check (“CRC”).

[0143] The Stop Sequence field is used to tell the slave when the master is talking and that this is the end of the command. When the slave is talking, the stop sequence is used to tell the master that this is the end of the data. The stop sequences takes two forms, namely Stop-Ack Req-(N)Ack and Stop-Stop-Stop. In the Stop-Ack Req-(N)Ack form, the talker is asking the listener if they received the message and if the message was error free. “Ack Req” means Acknowledge Request, “Ack” means Acknowledge, and “(N)Ack” means Not Acknowledge. The Stop-Stop-Stop form is slave data response for the Read and Read Results commands. A bit time is a unit of time (clock cycle) for a “1” or “0”.

[0144] The command format length is 33 bit times for most commands. Because most commands are the same length and format, a simpler form of digital electronics is allowed in a capsule (hermetic body) on a lead. Of course, this area is extremely constrained.

[0145] The sensor interface allows the slaves to sleep, thereby saving power. The bus can have multiple slaves hanging on it. These slaves may be sensors, actuators or slave IMDs. The command structure is implemented with the Master initiating all communication. When the Master starts talking, the Master issues a Start command. All slaves begin listening on a Start command.

[0146] A slave not being addressed by the master may stop listening and interpreting the command to save power. When a particular slave determines that a command is not directed to that slave, the slave may go to sleep until the next Start sequence. Sleep mode may occur at different points in the command format, depending on whether the sensor is mapped or unmapped. A mapped sensor has an assigned short address and unmapped sensor does not have a short address.

[0147] Addressing is provided in two forms: long address, and short address. The long address is 64 bits while the short address in the command is 6 bits. The long address uniquely identifies the slave telling: 1. manufacturer; 2. protocol version; 3. model identifier; 4. unique serial number; 5. manufacturing facility; and 6. slave number (useful for identifying multiple slaves on a single lead). Error checking of the Long Address is performed internally within the sensor via CRC check or other standard checking procedure.

[0148] FIG. 25 is a table providing a slave long address format. The long address contains many bits and the current drain of the data communication is governed by i=cvf, where c=capacitance of the bus, v=voltage swing of data communicating on the bus, and f=frequency of the data on the bus. If the 64 bit long address is used for all commands, the frequency of data on the bus would be exceedingly large for an implantable system. The goal is to have a command structure that allows a map assignment of long addresses to short addresses such that the majority of communication is provided by short addresses. The use of a short address containing only 6 bits will lower the cvf current of the bus. The master communicates to the slaves three different ways: 1. Unicast to one sensor; 2 Multicast (Trigger) up to four slaves simultaneously; and 3. Broadcast to all slaves. A broadcast is sent to every sensor on the bus, and there can be up to 16 slaves simultaneously on the bus.

[0149] The 6 bit address is provided to trigger, i.e. Multicast, multiple slaves simultaneously. The 6 bit format permits identification of 16 sensors uniquely and permits all short addresses sent to the sensors to be of the same format. The 6 bit format allows some multicast capability and a unique broadcast code.

[0150] FIG. 26 is a table of slave short addresses and multicast examples. The table particularly illustrates examples of how the master specifies the short addresses. According to a preferred embodiment, multicasting of more than four slaves is not permitted.

[0151] Addressing at power up is particularly considered as follows. Slaves are powered up or reset and respond only to their long address. In this state they are unmapped, and only three commands are usable in this state: 1. Unlocks; 2. Search Long Address; and 3. Write Short Address. These commands can all use a reserved short address with a code of 3F₁₆. This reserved code is used as a broadcast command going to all slaves on the bus. When the master uses the broadcast short address of 3F₁₆, then all slaves listen. The master may then uniquely identify a particular slave depending on whether the command calls for a long address.

[0152] After power up, the long addresses of the slaves are searched. The sequence is as follows: 1. Use the Unlocks command and ensure all slaves are unmapped—forcing all the slaves to be unmapped; 2. Use the Unlocks command and check for any slaves on the bus “check for unmapped” (if a slave on the bus then proceed to next step); 3. Use the Search Long Address and search for the long address of the slave(s) (once all slave long addresses proceed to next step); and 4. Assign each slave a unique short address via the command Write Short Address.

[0153] The assignment of long addresses are made during manufacturing. Long addresses may be assigned to various slaves on a pacing lead or to a slave IMD, one slave at a time before each slave is added to the bus. A capsule on a lead is a hermetic body that contains the slave's electronics.

[0154] If a capsule contains multiple slaves, the four bit slave number included in the long address can be uniquely hard coded. At capsule test, the master searches the unique long addresses and assign short addresses. The capsule containing multiple slaves then only requires one pair of wires for all slaves. For increased safety, the Unlocks command has a form that “unlocks” the ability to write long addresses. If a lead contains multiple capsules the capsules are assigned unique addresses before they are built into a lead. Each capsule's long address is therefore searched out using the Search Long Address command.

[0155] Data going to or from the master contains a Frame Check Sequence (“FCS”). The FCS is an 8 bit CRC. The generator polynomial G(x) is given as follows:
G(x)=x⁸+x⁵+x⁴+1  EQ. 6

[0156] Polynomial strings are based upon treating bit strings as representations of polynomials with coefficients of “0 and “1”. Therefore, for example, the generator polynomial is the bit string 10011 0001₂.

[0157] The master controls when slaves sample by issuing triggers. The issued triggers ensure that all samples are evenly spaced for a particular slave and ensure that all slaves are sampling at the same time or multiples of each other, i.e. data alignment slave to slave. Furthermore, the issued triggers allow the master to control the data rates. The slave's ADC will require a certain amount of time to perform a conversion.

[0158] The are two trigger commands: Trigger, and Quick Trigger. The Trigger command allows two additional fields of 4 bits each. One field allows time stamping of data via a Count embedded in the trigger command. This allows the master to tell if a sensor missed a trigger(s) and which trigger(s) were missed. Another field will allow Command Codes to be sent down to the sensor. The sensor may perform signal processing of the data stream between two different Command Codes. Command Codes work like a 4 bit write within the Trigger command.

[0159] FIG. 27 is a graph of a triggering example. As illustrated, the read that follows a trigger does not obtain data from the most recent trigger but rather obtains data from 4 triggers previous.

[0160] FIG. 28 is a graph and table illustrating a relationship between triggering of one sensor and reading corresponding data. The graph particularly illustrates an example of triggering four slaves at a same trigger frequency and the staggering of reading the data back to distribute the bus traffic evenly. The Counts that were sent with the triggers are stored in with the data.

[0161] The Quick Trigger (QT bit set) works just like the Trigger command, but is less flexible and allows power savings. The Quick Trigger command does not have a Master Command Name and does not have a Command Code. The inclusion of the Quick Trigger command adds a 1 bit overhead to all the other commands but saves 9 bits on the most frequent operation of triggering.

[0162] FIG. 29 is a graph illustrating re-triggering of data using a count code. On occasion, re-triggering of a sample may occur. Re-triggering is a Trigger or a Quick Trigger command with a count code of Ehex. Re-triggering may be used if the slope of the data is not important or is not adversely effected by a re-trigger. Re-triggering may also be used if the sample period is long compared to the time between the trigger and the re-trigger of a sample (implying the slope is not adversely effected). Further, a re-trigger may be used if the data is non-periodic or if only a single sensor is being triggered.

[0163] When the trigger is broadcast to multiple sensors, the acknowledge request can be meaningless because more than one sensor is acknowledging proper receipt of the trigger command. When the trigger goes out to a single sensor, the acknowledge has more meaning because the acknowledge is specific to a single sensor. Therefore, if the trigger is NACK'd, the trigger can be determined if re-triggering can be performed. A re-trigger may get ACK'd thereby returning some semblance of periodic triggering.

[0164] Some forms of digital signal processing may be performed in the sensor. The signal processing for RV Pressure set forth below may be equally performed in the sensor or in the IMD. If the sensor is area constrained, processing is preferably accomplished in the IMD.

[0165] FIG. 30 is a graph illustrating a collection of data points for an RV pressure waveform. FIG. 31 is a table illustrating Command Codes and measured parameters. The advantage of processing in the sensor is to pass only processed information back to the IMD, rather than every waveform point. This lowers the (i=cvf) current drain. The protocol provides for passing the whole pressure waveform up to the IMD for research, diagnostics and display purposes. Commands are sent out by the IMD at a certain rate and each have a 4 bit Count Code associated with them. With these counts, both time and signal slopes can be derived and the presence of missing Triggers can be determined.

[0166] The Trigger commands optionally have a 4 bit Command Code associated with them. One use of the Command Codes is to determine when to start the signal processing and when to end the signal processing. The Command Codes are essentially used as a short write command embedded within the Trigger command. As an alternative, the Quick Trigger command in combination with the Write command can be used to start signal processing and end signal processing.

[0167] There exists a certain amount of data overhead when retrieving data from the slave. By minimizing the data overhead, the current drain is reduced by reducing the data frequency (f) in the i=cvf equation.

[0168] Data transfer overhead for a single slave using a Trigger command and a Read Results command are analyzed below. For a one slave application generating 12 bit data and sending 4 bit Count codes back with the data, the following information is obtained.

[0169] Each Trigger is associated with 16 bits of data coming back: $\begin{array}{cc}\frac{33\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}+\frac{12\mathrm{Bits}\mathrm{ADC}\mathrm{Data}}{\mathrm{Trigger}}+\frac{4\mathrm{Bits}\mathrm{Count}\mathrm{Data}}{\mathrm{Trigger}}=\frac{49\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}& \mathrm{EQ}.7\end{array}$

[0170] One Read Request and response costs a certain number of bit times: $\begin{array}{cc}\frac{33\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}+\frac{8\mathrm{Bit}\mathrm{Times}\mathrm{for}\mathrm{CRC}\mathrm{in}\mathrm{Response}}{\mathrm{Read}\mathrm{Request}}+\frac{3\mathrm{Bit}\mathrm{Times}\mathrm{for}\mathrm{Stop}\mathrm{in}\mathrm{Response}}{\mathrm{Read}\mathrm{Request}}=\frac{44\mathrm{Bit}\mathrm{Times}}{\mathrm{Read}\mathrm{Request}}& \mathrm{EQ}.8\end{array}$

[0171] Given n triggers to get n samples per Read Request: $\begin{array}{cc}\frac{n\mathrm{samples}}{\mathrm{Read}\mathrm{Request}}& \mathrm{EQ}.9\end{array}$

[0172] For calculation of the number of bit times each sample associated with a trigger costs for a one slave application. $\begin{array}{cc}\frac{49\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}+\frac{\left(\frac{44\mathrm{Bit}\mathrm{Times}}{\mathrm{Read}\mathrm{Request}}\right)}{\left(\frac{n\mathrm{samples}}{\mathrm{Read}\mathrm{Request}}\right)}=\frac{X\mathrm{Bit}\mathrm{Times}}{\mathrm{Sample}\mathrm{Trigger}}& \mathrm{EQ}.10\end{array}$

[0173] If n=4 then we get: $\begin{array}{cc}\frac{49\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}+\frac{\left(\frac{44\mathrm{Bit}\mathrm{Times}}{\mathrm{Read}\mathrm{Request}}\right)}{\left(\frac{4\mathrm{samples}}{\mathrm{Read}\mathrm{Request}}\right)}=\frac{60\mathrm{Bit}\mathrm{Times}}{\mathrm{Sample}\mathrm{Trigger}}& \mathrm{EQ}.11\end{array}$

[0174] As set forth below, the efficiency for one slave is not nearly as great as for multiple slaves. In particular, use of multiple slaves using a Trigger and Read Results command provides increased efficiency. Generalized for a quantity of Y slave applications, each slave generating 12 bits data and transmitting a 4 bit Count up, savings are achieved.

[0175] Each Trigger command is associated with a certain amount of data (16 bits×the quantity of slaves Y) coming back: $\begin{array}{cc}\frac{33\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}+\left(\frac{12\mathrm{Bits}\mathrm{ADC}\mathrm{Data}}{\mathrm{Trigger}}+\frac{4\mathrm{Bits}\mathrm{Count}\mathrm{Data}}{\mathrm{Trigger}}\right)\frac{Y\mathrm{sensors}}{\mathrm{Trigger}}=\frac{33\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}+\frac{16\mathrm{Bit}\mathrm{Times}\mathrm{of}\mathrm{Data}}{\mathrm{Trigger}}\times \frac{Y\mathrm{sensors}}{\mathrm{Trigger}}& \mathrm{EQ}.12\end{array}$
Each slave has its own read request and response that has a certain cost: $\begin{array}{cc}\frac{33\mathrm{Bit}\mathrm{Times}\mathrm{in}\mathrm{Request}}{\mathrm{Read}\mathrm{Request}}+\frac{8\mathrm{Bit}\mathrm{Times}\mathrm{for}\mathrm{CRC}\mathrm{in}\mathrm{Request}}{\mathrm{Read}\mathrm{Request}}+\frac{3\mathrm{Bit}\mathrm{Times}\mathrm{for}\mathrm{Stop}\mathrm{in}\mathrm{Response}}{\mathrm{Read}\mathrm{Request}}=\frac{44\mathrm{Bit}\mathrm{Times}}{\mathrm{Read}\mathrm{Request}}& \mathrm{EQ}.13\end{array}$
Given triggers to get n samples per Read Request: $\begin{array}{cc}\frac{n\mathrm{samples}}{\mathrm{Read}\mathrm{Request}}& \mathrm{EQ}.14\end{array}$

[0176] The number of bit times each sample costs for a quantity of Y slave application is next calculated: $\begin{array}{cc}\frac{33\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}+\frac{Y\mathrm{sensors}}{\mathrm{Trigger}}\times \left(\frac{\left(\frac{44\mathrm{Bit}\mathrm{Times}}{\mathrm{Read}\mathrm{Request}}\right)}{\left(\frac{n\mathrm{samples}}{\mathrm{Read}\mathrm{Request}}\right)}+\frac{16\mathrm{Bit}\mathrm{Times}\mathrm{of}\mathrm{Data}}{\mathrm{Trigger}}\right)=\frac{X\mathrm{Bit}\mathrm{Times}}{\mathrm{Sample}\mathrm{Trigger}}& \mathrm{EQ}.15\end{array}$

[0177] If n=4 and Y=4 as in FIG. 28, then: $\begin{array}{cc}\frac{33\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}+\frac{4\mathrm{sensors}}{\mathrm{Trigger}}\times \left(\frac{\left(\frac{44\mathrm{Bit}\mathrm{Times}}{\mathrm{Read}\mathrm{Request}}\right)}{\left(\frac{4\mathrm{samples}}{\mathrm{Read}\mathrm{Request}}\right)}+\frac{16\mathrm{Bit}\mathrm{Times}\mathrm{of}\mathrm{Data}}{\mathrm{Trigger}}\right)=\frac{141\mathrm{Bit}\mathrm{Times}}{\mathrm{Sample}\mathrm{Trigger}}& \mathrm{EQ}.16\end{array}$

[0178] For 1 trigger resulting in 1 sample for 4 slaves, an average of 32.5 bit times per slave is achieved. $\begin{array}{cc}\frac{141\mathrm{Bit}\mathrm{Times}}{\mathrm{Sample}\mathrm{Trigger}}\times \frac{1\mathrm{Sample}\mathrm{Trigger}}{4\mathrm{Sensors}}=\frac{35.25\mathrm{Bit}\mathrm{Times}}{\mathrm{Sensor}}& \mathrm{EQ}.17\end{array}$

[0179] Accordingly, 35.25 bit times per slave are produced. Considering the acquisition of 12 bits of ADC data and 4 bits of Count data for time alignment or a total of 16 bits, then a factor of approximately 2× for the protocol is efficient. Moreover, efficiency is further increased by reading larger amounts of data, i.e. by increasing “n” in EQ. 14. The cost of reading more data per read command is a lowered specificity to detect errors in the data.

[0180] Using the Quick Trigger command along with the Read Results command also provides increased efficiency as set forth below. Generalized for a quantity of Y slave applications, with each slave generating 12 bits of data and transmitting a 4 bit Count up, efficiency is improved. Each Quick Trigger command is associated with a certain amount of data (16 bits×the quantity of slaves Y) coming back: $\begin{array}{cc}\frac{24\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}+\left(\frac{12\mathrm{Bits}\mathrm{ADC}\mathrm{Data}}{\mathrm{Trigger}}+\frac{4\mathrm{Bits}\mathrm{Count}\mathrm{Data}}{\mathrm{Trigger}}\right)\frac{Y\mathrm{sensors}}{\mathrm{Trigger}}=\frac{24\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}+\frac{16\mathrm{Bit}\mathrm{Times}\mathrm{of}\mathrm{Data}}{\mathrm{Trigger}}\times \frac{Y\mathrm{sensors}}{\mathrm{Trigger}}& \mathrm{EQ}.18\end{array}$

[0181] Each slave has its own read request and response that has a certain cost: $\begin{array}{cc}\frac{33\mathrm{Bit}\mathrm{Times}}{\mathrm{Read}\mathrm{Request}}+\frac{8\mathrm{Bit}\mathrm{Times}\mathrm{for}\mathrm{CRC}\mathrm{in}\mathrm{Response}}{\mathrm{Read}\mathrm{Request}}+\frac{3\mathrm{Bit}\mathrm{Times}\mathrm{for}\mathrm{Stop}\mathrm{in}\mathrm{Response}}{\mathrm{Read}\mathrm{Request}}=\frac{44\mathrm{Bit}\mathrm{Times}}{\mathrm{Read}\mathrm{Request}}& \mathrm{EQ}.19\end{array}$

[0182] Given Y triggers to get n samples per Read Request: $\begin{array}{cc}\frac{n\mathrm{samples}}{\mathrm{Read}\mathrm{Request}}& \mathrm{EQ}.20\end{array}$

[0183] Therefore, the number of bit times each sample costs for a quantity of Y slave application is: $\begin{array}{cc}\frac{24\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}+\frac{Y\mathrm{sensors}}{\mathrm{Trigger}}\times \left(\frac{\left(\frac{44\mathrm{Bit}\mathrm{Times}}{\mathrm{Read}\mathrm{Request}}\right)}{\left(\frac{n\mathrm{samples}}{\mathrm{Read}\mathrm{Request}}\right)}+\frac{16\mathrm{Bit}\mathrm{Times}\mathrm{of}\mathrm{Data}}{\mathrm{Trigger}}\right)=\frac{X\mathrm{Bit}\mathrm{Times}}{\mathrm{Sample}\mathrm{Trigger}}& \mathrm{EQ}.21\end{array}$

[0184] If n=4 and Y=4 as in FIG. 28, then: $\begin{array}{cc}\frac{22\mathrm{Bit}\mathrm{Times}}{\mathrm{Trigger}}+\frac{4\mathrm{sensors}}{\mathrm{Trigger}}\times \left(\frac{\left(\frac{44\mathrm{Bit}\mathrm{Times}}{\mathrm{Read}\mathrm{Request}}\right)}{\left(\frac{4\mathrm{samples}}{\mathrm{Read}\mathrm{Request}}\right)}+\frac{16\mathrm{Bit}\mathrm{Times}\mathrm{of}\mathrm{Data}}{\mathrm{Trigger}}\right)=\frac{132\mathrm{Bit}\mathrm{Times}}{\mathrm{Sample}\mathrm{Trigger}}& \mathrm{EQ}.22\end{array}$

[0185] For 1 trigger resulting in 1 sample for 4 slaves, then this averages to 33 bit times per slave. $\begin{array}{cc}\frac{132\mathrm{Bit}\mathrm{Times}}{\mathrm{Sample}\mathrm{Trigger}}\times \frac{1\mathrm{Sample}\mathrm{Trigger}}{4\mathrm{Sensors}}=\frac{33\mathrm{Bit}\mathrm{Times}}{\mathrm{Sensor}}& \mathrm{EQ}.23\end{array}$

[0186] Therefore, approximately 33 bit times are required per slave. Considering 12 bits of ADC data and 4 bits of Count data for time alignment or a total of 16 bits, then a factor of approximately 2× for the protocol is efficient. Moreover, efficiency is further increased by reading larger amounts of data, thereby increasing n in the above EQ. 20. The cost of reading more data per read command is a lowered specificity to detect errors in the data.

[0187] For a calculation of EQ. 21 for one sensor read every 4 triggers, a value of 51 bit times per sample trigger is determined. Thus, the Quick Trigger command has an efficiency over the Trigger command, which comes out to 60 bit times per sample trigger in EQ. 11.

[0188] FIG. 32 is a table providing a command overview for the sensor interface for an implantable medical device. In general, a unicast command writes to a single slave, a multi-cast command writes to more than one slave, and a broadcast command writes to all slaves.

[0189] Memory inside the slave is partitioned to hold Calibration Constants, Slave Long Addresses, Lead Long Addresses and other variables to control the operation of the device. The memory is preferably non-volatile for Calibration Constants and Addresses so that data is not lost once the bus is powered down. Memory partitioning is specific to each slave and 16 bit addressing is allowed for both EEPROM and internal RAM/Register memory. Examples of commands using this type of memory are given by: 1. LSB RAM/REG Address; 2. MSG RAM/REG Address; 3. LSB EEPROM Address; 4. MSB EEPROM Address; 5. Copy RAM/REG to EEPROM memory; and 6. Copy EEPROM to RAM/REG memory. Data link layer commands are particularly set forth below for a sensor interface for an implantable medical device.

[0190] FIG. 33 is a table illustrating a format for the Unlocks command. FIG. 34 is a table illustrating Unlock command key options. The Unlocks command is used for multiple operations and globally unmaps all slaves. When a slave is unmapped, the slave only responds to its associated long address. The Unlocks command may be sent multiple times because more than one slave is responding and ACKing. Globally, the Unlocks command checks for any unmapped slaves, and is shown as three steps in FIG. 49 below. The Unlocks command also globally checks for any mapped slave, and checks if the Unmap-all command was successfully completed. The Unlocks command allows for copying from RAM/REG address space to EEPROM, and allows copying from EEPROM address space to RAM/REG space. Further, the Unlocks command allows writing of a slave long address. For a command response, the command is either ACK or NACK. The ACK normally indicates that the FCS was valid but if searching for an Unmapped slave, an ACK will only be provided if the FCS is valid and unmapped.

[0191] FIG. 35 is a table of a Master's Search Long Address command. With this command, a binary search for the long address can occur. This search occurs with progressively-longer Long Address Bit Patterns (field Master4). If one or more slaves exist at the long address, then the command pulls the line low via the ACK command. Only unmapped slaves respond to this command.

[0192] FIG. 36 is table summarizing an example search across a sensor for an implantable medical device. The search is performed by the Master implantable medical device for the binary long address bit pattern of the sensors attached to it. If the long address given below is only 4 bits, two slaves are on the bus at addresses (1010₂ and 1001₂), and the search will progress as illustrated. However, slaves having 4 bit addresses is illustrated for example purposes only.

[0193] FIG. 37 is a table illustrating a Master's Write Short Address command. This command writes the short address of a single slave. The Master's Write Short Address Command is given the long address (indicated by the Master4 field) and then a short address defined by the Master2 field is assigned. If the long address matches the slave's long address, then this command marks the short address as mapped. For a command response, the slave will acknowledge the write as long as the FCS is correct. Action will only be taken if the Slave field (Master 2) is Unicast.

[0194] FIG. 38 is a table illustrating a Trigger command. The slave to be triggered depends on the Slave(s) field Master2. The sampling rate depends on the speed that the Trigger command is sent by the master, the frequency of Master and Sensor time bases, and the length of the trigger command. If the Trigger command is sent at 250 Hz, the Trigger command could trigger an RV pressure slave to sample every 4 ms.

[0195] The Command code for the Trigger command is a 4 bit code that encodes information, i.e. events, such that the slave can associate samples with cardiac or other information. Not all slaves will use the Command code field. Some slaves will use the Command code only and some will use the Command code and the Count and some may use neither the Command code or Count.

[0196] FIG. 39 is a table illustrating a Trigger Command Code for a cardiac IMD. The illustrated list is used to communicate cardiac Command Codes to multiple slaves.

[0197] FIG. 40 is a table illustrating an example of a Trigger Command Code for sonomicrometry transmitters and receivers. The illustrated list is used to determine which slave/actuator is making an acoustic ping and which slave/sensor is listening. The Command Code of F₁₆ (1111₂) is a reserved code for cleared data. The Command Code of E₁₆ (1110₂) is a reserved code for normal (not specific Command Code data). Accordingly, a code other than F₁₆ is allowed to be embedded with most of the samples in case the Command Code is sent back with the Data.

[0198] The Count field in FIG. 38 is a 4 bit code that gives a sequence to the triggers that are coming. The Count rotates (in hexadecimal) 0,1,2,3,4,5,6,7,8,9,A,B,C,D,0,1,2,3 . . . . The Counts of E₁₆ and F₁₆ are reserved. The Count of F₁₆ (1111₂) is a reserved count code for cleared data and the Count of E₁₆ (1110₂) is a reserved count code for retriggering of a sample that is NACK'd

[0199] A Slave can determine if a trigger was missed. If sampling on every Count, then the Count should increment 0₁₆ thru D₁₆, if the triggers are emitted to the slave in that order. Time alignment of data between slaves is accomplished in the master. A slave can be told to down sample on even values or odd values of the Count field via the Short Address Field. The Trigger command's Short Address field: G₁G₀A₃A₂A₁A₀ still tells the slave to sample or not, therefore the Count does not dictate if a sample occurs. The Count embedded for this slave's data when read via the Read Results command tells whether any data is missing. Sampling on even Counts have Counts embedded in the data such as 0, 2, 4, 6, 8, A, C, 0, 2. Sampling on odd Counts would Counts embedded in the data such as 0, 3, 6, 9, C, 1, 4, 7, A, D, 2, 5, 8, B, 0, 3. The pattern for odd Counts takes longer to repeat itself.

[0200] Using Counts for signal processing in the slave are performed and slopes can be calculated, even with missing triggers, because the Count value is given for each trigger. Using Counts and Command Codes for signal processing and time measurements is also utilized. For example, the Pre-ejection interval (“PEI”) may be recorded as the time from R wave detection to peak dP/dt. If one of the samples is not one of the reserved Command Code codes: (1111₂ or 1110₂) there could be a code for R-wave detection (for example 0000₂). Signal processing is performed to pick off the dP/dt via the Count codes. Once the speed of Count increments is determined, the PEI may be recorded as the number of Count increments between when the R wave is signaled and when the pressure slave showed a max+dP/dt.

[0201] Not all slaves will use the Count field. Some will use Count only and some will use the Command Code and the Count. Some may use neither the Command Code or the Count. A slave can embed both a Count and Command Code within a data response. Signal processing is performed on the Master end.

[0202] For the Command response, an Acknowledge from one or more slaves is provided. The AckReq gives an edge that allows accurate timing for sonomicrometry operation.

[0203] FIG. 41 is a table illustrating a Quick Trigger Command. The Quick Trigger works like the Trigger command but uses fewer bits to save power. The Slave(s) field Master2, determines which slaves are triggered. The sampling rate depends on the speed that the Quick Trigger command is sent by the master. For example, if the Quick Trigger command is transmitted at 250 Hz, an RV pressure slave may be triggered to sample every 4 ms. The Quick Trigger command is the only one with the QT bit set.

[0204] The Quick Trigger command does not have a Master Command Name Field, therefore the slave must recognize whether the QT bit is set and know that it is the Quick Trigger command. The count field works the same in the Quick Trigger command as in the Trigger command. The count cycles 0 thru D, 0 thru D etc . . . with count codes E and F being reserved. A command response is an Acknowledge from one or more slaves. The AckReq gives an edge that improves timing for sonomicrometry operation.

[0205] FIG. 42 is a table illustrating a Master's Read command. The Address for the read is set up via the Master's LSB RAM/REG Address Command and the Master's MSB RAM/REG Address command. The Quantity of bytes field allows reading up to 256 values from a page. A value of “0” reads 1 byte of data.

[0206] FIG. 43 is a table illustrating a Slave's Read Response. The Read command will leave the address as set by the Master's LSB RAM/REG Address Command and the Master's MSB RAM/REG Address Command, where it pointed to for the first byte read. For a command response, the slave will acknowledge the read as long as the FCS from the master was correct. A response will be given if the command is Unicast to one slave. The response indicated is a response to reading N−1 bytes.

[0207] FIG. 44 is a table providing rules for a Read Results command. FIG. 45 is a pair of tables illustrating a buffer with 8 bit Data and Count and Command Code embedded. FIG. 46 is a table illustrating a Master's Read Results command. The Master's Read Results command is more useful/specific than the general Read command because it performs more for the master, reduces bus traffic and reduces power. The Read Results command reads result data from RAM/Register Address space, and is used in conjunction with slave results being put into this memory typically by an ADC. Data can be put into memory via any format and can be slave specific. As illustrated in FIG. 44, the data format is shown with the oldest data given first.

[0208] The Read Results command works in close association with the Trigger and the Quick Trigger commands. The Trigger command forwards a Command Code and a Count that are values other than F₁₆. Command Codes and/or Count codes that are embedded in the data read by the Read Results command, and that have values of F₁₆, have not yet had ADC data written to their associated data fields. The Quick Trigger works like the Trigger except that it only sends the Count code.

[0209] The Read Results command follows the rules given in the table of FIG. 44. The read command is setup to read from a specific buffer (either “0” or “1”) and depending on which buffer is being read and which buffer the ADC pointer is pointing to the appropriate set of rules are followed. Two buffers have been implemented allowing a retry of reading the data.

[0210] FIG. 47 is a table illustrating a Slave's Read Results example response. The example response is a response to reading N+1 bytes. The indicated response is slave specific, such that slaves may send various data bit widths back as well which may be including Count and/or Command Codes. The response does not have to send back both the Command Code and Count. A response is given only if the Slave field (Master 2) is Unicast. No data is sent and no operations are performed with the ADC pointer if the Slave field is not Unicast.

[0211] FIG. 48 is a table illustrating a Master's Write command. The Address for placement of the value is set up via the Master's LSB RAM/REG Address command and the Master's MSB RAM/REG Address command. Before a value is written to any memory, the frame check sequence (“FCS”) is checked first. The value field indicates the value that is written and the write command leaves the address off at the address where the write occurred. The slave will acknowledge the write with a command response as long as the FCS was correct.

[0212] FIG. 49 is a table illustrating a Master's LSB RAM/REG Address command. The Master's LSB RAM/REG Address command writes the RAM/Register Address Least Significant Byte Pointer. The Value field for the Master's LSB RAM/REG Address command will contain the RAM/Register LSB Address Pointer and for the command response, the slave will acknowledge the write as long as the FCS is correct.

[0213] FIG. 50 is a table illustrating a Master's MSB RAM/REG Address command. The Value field for the Master's MSB RAM/REG Address command will contain the RAM/Register MSB Address Pointer, and the slave will acknowledge the write as a command response as long as the FCS is correct.

[0214] FIG. 51 is a table illustrating the LSB EEPROM Address command. The LSB EEPROM Address command writes the EEPROM Address Least Significant Byte Pointer. The Value field contains the EEPROM LSB Address Pointer and the slave will acknowledge the write as a command response as long as the FCS is correct. The pointer Address will not change on a write that is NACK'd.

[0215] FIG. 52 is a table illustrating a MSB EEPROM Address command. The MSB EEPROM Address command writes the RAM/Register Address Most Significant Byte Pointer. The Value field contains the EEPROM MSB Address Pointer while the slave will acknowledge the write with a command response long as the FCS is correct. The pointer Address will not change on a write that is NACK'd.

[0216] FIG. 53 is a table illustrating a Master's Copy RAM/REG to EEPROM Memory command. This command can copy up to 256 bytes from RAM/Register Address Space to EEPROM address space. The EEPROM addresses are left off at the starting address as dictated by the commands LSB EEPROM Address and MSB EEPROM Address. RAM/Register addresses will be left off at their starting addresses as dictated by the commands LSB RAM/REG Address and MSB RAM/REG Address. The Unlocks command is sent as a guard against writing into the calibration constant and address areas of EEPROM. The slave will acknowledge the write with a command response as long as the FCS is correct. No action will be taken if the appropriate Unlocks command is not performed.

[0217] FIG. 54 is a table of a Master's Copy of EEPROM to RAM/REG Memory command. The Copying command copies up to 256 bytes from EEPROM Address Space to RAM/Register space. The EEPROM addresses will be left off at the starting address as dictated by the commands LSB EEPROM Address and MSB EEPROM Address. RAM/Register addresses will be left off at their starting addresses as dictated by the commands LSB RAM/REG Address and MSB RAM/REG Address. The Unlocks command must be sent as a guard against writing into sensitive areas of RAM/Register space. The slave will acknowledge the write with a command response as long as the FCS is correct. No action will be taken if the appropriate Unlocks command was not performed.

[0218] FIG. 55 is a table illustrating a Master's Quick Read command. This command reads address pointers that are not readable via a normal read using address pointers and also reads status words that are desirable for quick access.

[0219] FIG. 56 is a table illustrating a Master's Which Pointer to Read command and FIG. 57 is a table illustrating a Slave's Read Address Pointers Response command. The generalized Read command does not use these pointers because they are used for the Read itself. This command is useful for debug by checking if pointer manipulation is working properly.