Background of the Invention
The present invention relates to a digital filtering arrangement and is particularly suited for use in a high speed pulse density modulation to pulse code modulation translation equipment.
British Pat. No. 1,436,878 discloses a pulse density modulation (PDM) to pulse code modulation (PCM) translation arrangement comprising a digital filter to which a PDM signal is applied and logic means to which the output of the filter is applied, the logic means being arranged to select every mth group of n pulses in the digital filter output. Typically the translation arrangement can be depicted schematically as shown in FIG. 1. A PDM signal, which is a continuous stream of binary pulses or bits is applied to a digital filter 10 which is constructed to suppress, as far as possible, high frequency noise. The filtered signals are then applied to a sampling circuit 11 which effectively selects every mth group of n pulses. For example, consider a system in which the PDM rate is 8.064 Mb/s (megabits per second). After filtering this can be regarded as an arbitrary stream of 14-bit words with a word rate of 8.064 Mw/s (megawords per second). If now every 504th 14-bit word is selected, the output becomes a PCM signal of 16 Kw/s (kilowords per second).
Such an arrangement is suitable for a speech channel of 0-4 KHz (kilohertz) bandwidth. The basic rate clock would not be expected to exceed a nominal 8 MHz (megahertz). However, a channel of 56-112 KHz bandwidth as required for frequency division multiplex (FDM) analog-to-digital (A/D) conversion, or 0-50 KHz for a possible music A/D, would need a basic clock rate of about 56 MHz (i.e. this is the presently used 4 MHz for speech scaled up by a factor of 14) and there is a technology problem when working at this speed. In particular the power consumption of integrated circuits becomes unacceptable.
The basic principles of a digital filter coupled with selective logic can be explained with reference to FIG. 2. An N-bit shift register 20 has N tap outputs, each tap output having an individual weighting function X applied thereto. Thus the output of tap 1 is weighted by the function X.sub.1, that of tap 2 by X.sub.2 and so on. For example, if the weighting functions are generally increasing up to tap N/2 and then decreasing, the filter can have an impulse response as indicated by the curve 21 in FIG. 2. The outputs of all the taps are summed in network 22. Now, if the delay between each tap is T equal to 1/f.sub.c, where f.sub.c is the basic input clock frequency, and the total length of the filter is N taps, then the repetition period of the filter is f.sub.s =NXT. If the output of the filter is sampled at a frequency of f.sub.c /NT, then the filter will receive a new set of data during each period and each bit will be uniquely weighted by only one of the respective tap weights according to its position in the time sequence. However, if the sampling rate is required to be four times f.sub.s each bit will be successively weighted by four separate weights during its progress through the filter. This weighting would normally occur at the prevailing basic clock frequency.
Summary of the Invention
An object of the present invention is to provide an improved digital filter arrangement for a serial pulse stream.
A feature of the present invention is the provision of a digital filter arrangement for a serial pulse stream comprising a plurality of shift registers operating in parallel; each of the plurality of shift registers having pulses read out in parallel from selected positions thereof, predetermined combinations of the pulses read out of the plurality of shift registers providing a plurality of address words; first means coupled to the plurality of shift registers to distribute pulses of the pulse stream into the plurality of shift registers sequentially and cyclically; addressable memory means coupled to the plurality of shift registers, the memory means having a plurality of storage locations each having stored therein a digital word, the memory means providing in a non-volatile manner the digital word of each of the plurality of storage locations in response to associated ones of the plurality of address words; and logic means coupled to the memory means for providing a serial output of the digital words read out of the plurality of storage locations.
In a preferred embodiment of the invention the memory means comprises a plurality of separate memories, one for each of the plurality of shift registers, each separate memory having a number of storage locations addressable by the address words from the associated shift register only, and the logic means comprises means for adding the digital words read out of the separate memories once for each input cycle of the serial output.
Brief Description of the Drawing
Above-mentioned and other features and objects of this invention will become more apparent by reference to the following description taken in conjunction with the accompanying drawing, in which:
FIG. 1 is a block diagram of a prior art PDM to PCM translation arrangement described hereinabove under the heading "Background of the Invention";
FIG. 2 is a block diagram and impulse response of a basic digital filter described hereinabove under the heading "Background of the Invention"; and
FIG. 3 is a block diagram of a digital filter arrangement according to the principles of the present invention.
Description of the Preferred Embodiment
In the block diagram of the digital filter arrangement shown in FIG. 3 an input PDM stream is entered serially into an 8-bit shift register 30. After each consecutive non-overlapping group of 8 bits has been entered, the contents of shift register 30 are transferred in parallel to eight further shift registers 31.sub.A -31.sub.H, one bit to each of the further shift registers. The eight further shift registers 31.sub.A -31.sub.H are each operated at a clock rate which is 1/8th of the PDM input clock rate. Selected stages of each of the further shift registers are read out in parallel to form address words for a set of read-only memories (ROM) 32.sub.A -32.sub.H. In the example shown, four equally spaced bits A.sub.1 -A.sub.4, B.sub.1 -B.sub.4, etc. Each ROM carries pre-stored weighted outputs which are read-out in accordance with the address words received from the shift registers 31.sub.A -31.sub.H. Thus, each input bit of the PDM stream will pass through the filter in the period equal to p times 8/f.sub.c, where p is the number of stages in the shift registers 31.sub.A -31.sub.H, and during this period it will be sampled and weighted four times. This agrees generally with the previously described requirements. Each ROM is also addressed by two bits from a common 2-bit binary counter 33. Hence, each ROM contains sixty-four digital words which represent all possible combinations from each of the respective shift registers 31.sub.A -31.sub.H.
The outputs A', B' . . . H' of the separate ROM's are added in adders 34-37 and then in adders 38-39 and lastly in adder 40. The output of adder 40 is fed into a latch 42 via adder 41. Adder 41 and latch 42 together make an accumulator from which a parallel transfer is made into a buffer 43. The contents of buffer 43 are read-out serially at four times the repetition period f.sub.s of the filter to provide the PCM serial bit stream.
It is to be noted that the detail arrangement of FIG. 3 is by way of example only. Thus, the method of distribution of the incoming PDM signals need not necessarily be by a shift register. The eight shift registers 31.sub.A -31.sub.H can be scanned sequentially and the incoming bits being read, one at a time, into the relevant shift register. Also, the particular arrangement of memories and the method of addressing them can be different. Having one memory per shift register as shown is convenient but other arrangements are possible. For example, at one extreme a single large memory could be envisaged, with all the selected bits in the shift registers being combined to form a single address word. This arrangement would have the advantage that no adders would be required but it would have disadvantages in the size of and operation of the ROM. Other various combinations of shift registers, tap and ROM connections can be used. Another example is where the ROM's may be used to store only the weights representing the impulse response of the filter, then it would then be necessary to use multipliers in conjunction with each shift register tap and add the results separately. This would need more adders besides introducing multipliers, but the ROM's could be reduced in size. Yet again, each address word could be formed from taps from more than one shift register. Thus, the address words in FIG. 3 could be for example, A.sub.1 C.sub.2 E.sub.3 G.sub.4, B.sub.1 D.sub.2 F.sub.3 H.sub.4, A.sub.2 C.sub.3 E.sub.4 G.sub.1, and so on, or other combinations. Finally, although the memories in FIG. 3 have been referred to as read-only-memories (ROM) they could be, for example, random-access-memories (RAM) the contents of which can be readily altered when required, e.g. on a day-to-day basis.
While we have described above the principles of our invention in connection with specific apparatus, it is to be clearly understood that this description is made only by way of example and not as a limitation to the scope of our invention as set forth in the objects thereof and in the accompanying claims.