The invention relates to a data processing circuit with a driver and a receiver that are connected by four signal conductors, over which successive signals are communicated so that the sum of currents through the signal conductors are substantially constant as a function of time.

Conventionally, single signal conductors combined with ground lines are used to communicate digital signals. As an alternative, differential signaling has been used, using a symmetrical pair of signal conductors, driven so that a data value is represented a difference between the currents through the signal conductors. In this case the sum of the currents can be kept constant, which improves noise performance of the circuit. To support communication of multiple bits in parallel, a plurality of such symmetrical pairs of signal conductors can be used.

US patent No. 6,359, 931 describes a more advanced apparatus that uses N>2 signal conductors for communicating symbols and uses current differences between all N (N- 1) /2 possible combinations of two signal conductors to encode signals. In the receiver a signal conductor is not just connected to a single comparator, together with another signal conductor with which it forms a pair, rather, each signal conductor is connected to the input of N-1 comparators, in combination with each of the N-1 remaining signal conductors respectively. Mutually different signal currents are supplied to a respective one of the N conductors. Different symbols are encoded by controlling which of a predetermined set of currents is supplied to which conductor. The comparators in the receiver detect which of the conductors in each of the N (N-1) /2 combinations of two conductors carries the most current and the symbol is decoded from the detection results. With this technique a higher transmission capacity (N faculty) can be realized with a given number of N conductors than when a similar number of N conductors is used for transmitting N/2 differential signals (2 to the power N).

In such a system the symmetry of the arrangement of the conductors assumes special importance for avoiding that signal distortion causes signal errors. When the

exchange of signals between conductors leads to different couplings and signal delays detection is compromised. Better symmetry will lead to lower Electromagnetic Interference to the environment (EMI). In the case in where the conductors are organized as N/2 pairs only the mutual symmetry between the conductors in a pair is relevant in this respect, to the extent that an interchange of the signals applied to the conductors of the pair does not lead to different signal propagation. But when all N (N-1) possible pairs are used for detection, it is desirable that the properties of interaction between the conductors do not change when the signals on any two conductors are interchanged. This requires a much higher amount of symmetry than for N/2 pairs.

US patent No. 6,359, 931 describes various arrangements of conductors to realize optimal symmetry. Arrangements are given for three and four conductors. However, the four conductor arrangements merely have symmetry for the interchange of the signals on certain pairs, not for all pairs.

Among other things, it is an object of the invention to provide a circuit of the type mentioned in the preamble that has a higher symmetry of the signal transmission properties under exchange of signals from the conductors.

The circuit according to the invention is set forth in Claim 1. According to the invention four signal conductors are placed in a spatial arrangement in which two signal conductors are arranged in a stack between two shield planes and two other signal conductors are arranged on mutually opposite sides of the stack, each midway between the shield planes.

The width of each particular signal conductor is set so that the particular conductor has substantially the same transmission impedance in the transmission line formed by that particular conductor and the shield planes. In general, the transmission impedance of a conductor between shield planes depends on the width of the conductor and the distance to the shield planes. Thus, conductors at different distances from the shield planes have the same transmission impedance provided that their width is progressively decreased as they are closer to one of the shield planes. As a result, paradoxically, the invention uses conductors of different widths to create a higher symmetry for interchange of signals on the signal conductors.

Preferably, the signal conductors are substantially rectangular in shape, the width being set in a direction parallel to the shield planes. Arrangements of rectangular signal conductors are easier to manufacture.

In an embodiment a shield conductor is included in the stack midway between the shield planes. Thus, direct mutual field coupling between the signal conductors is reduced, with the result that the mutual couplings between different pairs of signal conductors are more closely together (near zero).

In another embodiment, the distances between the stack and signal conductors on opposite sides of the stack are set so that mutual coupling between the conductors is substantially equal. By moving the signal conductors on opposite sides of the stack closer to the stack, coupling between these signal conductors is increased. Thus the distance to the stack can be used to control symmetry of coupling. Preferably, the mutual distance of the signal conductors in the stack is adapted as well, so that their mutual coupling equals that between the signal conductors on opposite sides of the stack.

These and other objects and advantageous aspects of the circuit according to the invention will be described in more detail using the following Figures.

Fig. 1 shows a circuit with a data communication connection, Fig. 2 shows a cross-section of an arrangement of conductors, Fig. 3 shows a further cross-section of an arrangement of conductors.

Fig. 1 shows a circuit with a data communication connection. The circuit contains a driver 10 and a receiver 16 coupled via four signal conductors 14a-d. Driver 10 is supplied from four current sources 12a-d and has a signal input 11. Receiver 16 contains impedances 162a-d in the signal conductors 14a-d and four sense circuits 160a-d for sensing currents via signal conductors 14a-d, through voltage drops over impedances 162a-d. A decoder 164 is coupled to the outputs of the sense circuits. An output of the decoder is an output of receiver 16. Input 11 of driver 10 and output 166 of receiver 16 may be coupled to functional further circuits (not shown) which produce and use the signals transmitted via signal conductors 14a-d respectively.

In operation driver 10 supplies current from current sources 12a-d to signal conductors 14a-d. Driver 10 basically performs the function of a switch matrix, which determines which combinations of current sources 12a-d are connected to which signal conductors 14a-d, under the control of input signals supplied to input 11. Thus, for example, when the currents supplied by current sources 12a-d are all equal to I, current combinations

like (I, I,-I,-I), (2I, O, I,-I) etc and permutations thereof may be supplied to the signal conductors (the currents between a pair of brackets indicating currents to respective ones of the signal conductors 14a-d in a specific combination).

Receiver 16 senses the currents through the various signal conductors and signals the sensed currents to decoder 164, which reconstructs the input signals and supplies reconstructed signals to output 166.

In the circuit shown the net sum of currents through signal conductors is zero at all times. As a result, no common impedance to ground need be connected to the common point 168 of the signal conductors (a high impedance bleeder resistance may be included to cope with inaccuracies). The important point, however, is that the net sum of the currents is constant. This makes it possible to communicate data symbols over signal conductors 14a-d at a high rate. If the sum is non-zero an impedance may be connected to the common point.

Signal conductors 14a-d can be relatively long. To ensure that no significant signal distortion occurs it is desirable that there are little or no reflections over signal conductors 14a-d and that there is little or no cross-coupling, or at least that the cross- coupling is symmetrical in the sense that interchanging the currents of any two signal conductors does not significantly affect transmission properties.

Fig. 2 shows a cross section of signal conductors 14a-d in a plane perpendicular to the direction of propagation from driver 10 to receiver 16. Signal conductors 14a-d are included between conducting shield planes 20,22 which extend in parallel to each other and to signal conductors 14a-d along most or all of the distance from driver 10 to receiver 16 in the direction perpendicular to the plane of cross-section that is shown in Fig. 2.

Shield planes 20,22 are electrically connected to each other, at least for signals in the frequency range used on signal conductors 14a-d. This may be realized for example by connecting both shield planes 20,22 to ground or to a power supply connection Vdd or Vss.

The region of the space between shield planes 20,22 that contains signal conductors 14a-d contains a non-conducting medium surrounding signal conductors 14a-d. The medium may be a conventional material used for printed circuit boards. The medium is preferably homogeneous between planes 20,22 or stratified in layers parallel to the planes 20,22.

Two of the signal conductors 14b, c are stacked symmetrically in a perpendicular column between the planes 20,22. The two other signal conductors 14a, d are provided midway between the planes, symmetrically on mutually opposite sides of the stack of signal conductors 14b, c. The width W2 of signal conductors 14b, c in the stack is smaller than the width Wl of conductors 14a, d on mutually opposite sides of the stack. These widths

have been chosen so that transmission line impedances of all of the signal conductors 14a-d are all substantially equal. These widths have also been chosen so that transmission line impedances of each combination of pairs of two selected conductors from 14a-d are substantially equal.

Each of the signal conductors 14a-d forms a transmission line in combination with shield planes 20,22. The transmission impedances of the transmission line formed by a signal conductor 14a-d depend on the distance between the signal conductors 14a-d and the shield planes 20,22 and the width W1, W2 of the signal conductor 14a-d. For a given distance, the transmission impedance decreases as the width Wl, W2 increases and for a given width W1, W2 the impedance decreases when the distance to the closest shield plane 20,22 decreases. As a result, the same transmission impedance can be realized with a signal conductor 14b, c close to the shield planes as with a wider signals conductor 14a, d at a greater distance from shield planes 20,22. This is used in the arrangement of Fig. 2 by choosing widths Wl and W2 so that each signal conductor 14a-d has the same transmission line impedance relative to shield planes 20,22.

The required width Wl or W2 can be determined experimentally for example by measuring the impedances individually and increasing or decreasing the widths of one pair of signal conductors 14a-d until matching of the impedances is realized. The impedance of a signal conductor 14a-d is a continuously decreasing function of its width: consequently it is always clear in which direction any width should be changed to reduce the difference in impedance. Alternatively, the impedance may be computed using a known computer program for solving Maxwell's equations. In general, the impedances will also depend on the relative location of the various signal conductors 14a-d, the distance between planes 20,22 and the material used for the medium surrounding the signal conductors. This makes it desirable to determine the required widths for each different design of the arrangement of signal conductors 14a-f separately.

Preferably, the distance between the signal conductors 14a-d is chosen to provide equal coupling strength between each pair of signal conductors 14a-d. In principle, there are three different relevant coupling strengths : a first coupling strength between the signal conductors 14b, c in the stack, a second coupling strength between the signal conductors 14a, d on opposite sides of the stack and a third coupling strength between a signal conductor 14a at the side of the stack and a signal conductor 14b in the stack. There are also three parameters of the structure that can be set: the distance between the signal conductors 14b, c in the stack, the distance between the signal conductors 14a, d on opposite sides of the

stack and the widths Wl, W2 which may be varied as long as the transmission line impedances of all signal conductors 14a-d vary in the same way. As a result, these distances and widths can be set so that all coupling strengths become substantially equal. As a further parameter to adjust the coupling strengths relative to one another, the distance between planes 20,22 may be varied.

Signal conductors 14a-d are not shown to scale. However, as shown, signal conductors 14a-d are preferably rectangular in shape, all having the same height.

The required distances and impedances can be determined experimentally, by measuring impedances and cross coupling strength and adjusting the distances can be found by gradually increasing or decreasing the distances until the cross-coupling strengths and the impedances are found to be equal. In general, each cross-coupling strength increases continuously as the distance between the relevant signal conductors 14a-d is reduced.

Consequently, the direction of changes in distance that are required to reduce the difference can readily be determined. In general, the cross coupling strengths will also depend on the relative location of the various signal conductors 14a-d, the distance between planes 20,22 and the material used for the medium surrounding the signal conductors. This makes it desirable to determine the required distances and widths for each different design of the arrangement of signal conductors 14a-f separately.

Fig. 3 shows a cross section of an arrangement of signal conductors 14a-d. In comparison with Fig. 2 a shield conductor 30 has been added, which is located in the stack of signal conductors 14b, c, midway between planes 20,22. Shield conductor 30 is electrically connected to shield planes, at least for signals in the frequency range of the signals on signal conductors 14a-d. Shield conductor 30 serves to reduce the coupling strength between signal conductors 14a-d. Preferably, shield conductor 30 is wider than the signal conductors 14c, d in the stack. The widths of signal conductors 14a-d are chosen so that substantially equal transmission impedances of the transmission lines formed by, on the one hand, shield planes 20,22 and shield conductor 30 and, on the other hand, respective ones of the signal conductors 14a-d are realized.

In addition to the conductors shown in Figs. 2 and 3 additional conductors may be added, for example shield conductors midway between planes 20,22 on opposite sides of the arrangement of the four signal conductors 14a-d. Additional sets of four signal conductors may be added in parallel for transmitting more signals in parallel.

The signal conductors 14a-d may be applied for example on printed circuit boards, or the substrate of a multi component module (MCM) to couple a driver 10 in one integrated circuit chip to a receiver in another integrated circuit chip, or in the packages of BGA ICs for proper signal transfer from the die to the PCB.