This presentation assumes that interested parties will be, or will refer to, engineers familiar with commutation theory and practice. Accordingly, we turn immediately to a schematic drawing and description.  We recommend that the reader immediately print it for reference throughout this text (for Microsoft Internet Explorer, print page 2 only, in portrait orientation ).

The drawing shows an even bar configuration of the new approach. Gramme-ring windings are drawn for simplicity of illustration. Lap or wave windings are identical in function and are intended in practice.

3.1 The commutator circumference comprises equal arcs of conductors and insulators.

3.2 The twin brushes are insulated from each other, free to move independently, and their arcs of contact are slightly less than the conductor and insulator arcs. It will be seen that no coil can be shorted by a brush.

3.3 It is also evident that D1 and Q1, D2 and Q2, preclude current circulating via the twin brushes and coil under commutation.

3.4 Two coils commutate simultaneously from t = 0 to T/2. No coils commutate from t = T/2 to T. In the latter period, current conditions are restored for the cycle to repeat. Odd bar configurations work equally well. Coils commutate alternately and there is no restoral period involved. An odd bar configuration drawing is available on request.

3.5 Noting that iL = i1 + i2 at all times, and applying the Principle Of Superposition to the symmetrical parallel paths from B1 to B2 and B3 to B4, it can be shown that current through coils lying between B1 and B4, and B3 and B2 = iL/2, and current through commutating coils = (i1 - i2).

Definition: Let Vr stand for the emf of self induction within the coil under commutation, together with, algebraically, emfs induced in the same coil by any and all fields arising from armature reaction, leakage, brush shift, interpoles.

3.6 Considering Stage 1 in the figure and a DC motor as the example, Lenz's Law establishes that the normal directions of the emfs Vr in the commutating coils are as shown. It follows that B1 and B4 are at Vr higher potential than B3 and B2 respectively, and the emfs across the parallel paths (coil groups) from B1 to B2 and from B3 to B4, are equal and of magnitude (Vs - Vr).

The potential difference Vr between adjacent brushes is possible as a result of the inherent characteristic of transistors. Collector to emitter voltage Vce "floats" for any collector current so that, in output circuits around collector and emitter, Vce = supply voltage - S (emfs outside the transistor). For the circuit in the figure comprising the supply, Q1, and B3 to B4, the emf outside the transistor is (Vs - Vr) as just deduced. Therefore, by subtraction, Vce = Vr for all currents iL and i2. This equality is essential for diodes D1 and D2 to remain forward biased. Omission of Q1 or substitution of another diode for Q1 would turn off D1 whenever Vr was present, leaving all of iL to pass through B3 and B4, which would be impossible.

3.7 During Stage 2 in the figure, the same parallel paths as just discussed for Stage 1 are subject to emf Vs, there being no current reversals through coils.

3.8 Since the emfs driving the parallel paths B1 to B2 and B3 to B4 alternate between Vs and (Vs - Vr) with frequency 1/T, current iL/2 through each path will have a ripple component of the same frequency. It can be shown that (a), the ripple component will approximate iL x Lc/Lm, where Lc is the self inductance of a commutating coil and Lm the self inductance of the armature between brushes. In practice, this ripple can be expected to be 1 to 2% of iL; and (b), iL average will equal ((Vs - Vr/2) - E)/R, where E is the speed emf in the active turns in each path and R is the resistance between brushes. The derivation of these results is available on request.

3.9 The following analysis demonstrates that, in theory, i1 and i2 can be caused to vary linearly during Stage 1, recalling that, since Vce = Vr, diodes D1 and D2 are continuously on.

3.10 In conclusion on the theory, it is important to stress that the foregoing analysis and predictions are based on the following view of the physics involved. First, the Principle of Least Action and absence of circulating current, not changes in base currents, give rise to i1 and i2 varying linearly; secondly, i1 and i2 can be thought of as seeking to divide so as to keep surface current density constant, and at the same time, causing Vce and the ib's to automatically fall in with the internal requirements for this.

5.1 COMMUTATOR & BRUSHES.

Whilst it might at first appear that the commutator will be larger than at present, the reverse is the case. Comparison between the existing and new approaches can be drawn as follows.

It is therefore possible to halve the copper content of the commutator, whilst retaining the same diameter and length as would have applied previously. Since the cost and weight of the insulator replacing the copper should be far less, the saving should be significant and broadly proportional to commutator size. It will also be possible to reduce the number of bars on medium and large units as a result of removing the flash-over limitation. The cost of brushes will remain the same, or be lower, if the design relief mentioned earlier proves fruitful in opening the way for a lower cost brush material.

5.2 ELECTRONICS.

There will be an added cost for transistors, diodes and the control of the transistor base circuits. The offsetting savings will be the reduced commutator cost and the deletion of interpole windings. In combination, these should outweigh the added electronics costs. The deletion of interpoles will also remove the labour costs associated with shimming their air gaps during final testing, a generally time consuming operation.

It is to be noted that the transistors and diodes are subject to the same averaging of current as outlined for the brushes. Thus, to handle a line current of iL, components rated at iL/2 will be used.

5.3 EFFICIENCY.

It follows from Section 3.8 that a power loss of iL.Vr/2 is inherent in this new approach, due to the transistors effectively acting as variable resistances. Offsetting this is the reduction in i²R losses resulting from the removal of interpoles. It is likely in practice that the new il.Vr/2 loss will exceed in small measure, the interpole i²R offset. The importance or otherwise of this will vary by application. Where efficiency is a leading consideration, two approaches are open.

With either approach, Vr can be reduced to Vr', where Vr' = Vr - opposing emf, with the loss factor reducing accordingly. Transistors Q1 and Q2 will accommodate any such variation in Vr, in the manner previously described.

It should also be noted that the resistors Rs in Figures 1 and 2 will be well under 1W . Values on larger units will be of the order of 0.01W . Thus iL².Rs loss will have negligible effect on efficiency.

5.4 SERVICE.

It is reasonable to predict that better commutation will result in less maintenance, longer unit life, and therefore lower long term costs to both manufacturer and user.

At the time of lodging the provisional patent application for this invention, a second lodgment was made to cover a new range of D.C. motors. In these motors, the main field is provided by substantial brush shift and neither field windings nor permanent magnets are required. The armature is conventional, except that the use of the new commutator here described is essential to nullify the emf from the relatively strong fringe field close to the pole tips, this emf adding to the emf of self induction.

The motor will provide a series torque-speed characteristic as an important alternative to the separately excited characteristic of permanent magnet motors. A compound characteristic will also be possible by a combination of permanent magnet salient poles and brush shift.