Method of and Apparatus for Voice Communication Storage and Forwarding With Simultaneous Access to Multiple Users

The present invention relates to voice communication storage and retrieval systems as of the type in which a centralized repository is employed to receive and store voice communication messages from multiple users and to be accessed for the retrieval or forwarding of the messages by telephone; the invention being more particularly concerned with the preferred employment of radio-actuated paging; and with permitting simultaneous access to a number of users placing and retrieving messages without subjecting them to delays.

It has been proposed to employ a plurality of magnetic tape recorders for receiving, serially recording and enabling accessing of messages, with complex and expensive multiplexing for achieving simultaneous communication for the above purposes, as described, for example, in U.S. Pat. No. 4,260,854. Such systems are, however, seriously limiting in application by virtue of the fact that recorded messages on tape cannot be accessed randomly and are really useful only when one associates one tape and one recording machine with one individual or one telephone. Further, the recording medium cannot be accessed simultaneously by more than one user; such simultaneous access being a particularly important attribute for a centralized system serving many users at a busy time of the day, and a feature attained by the present invention.

While other approaches have involved disc-based systems with complex storage techniques such as pairs of basic memory buffers for recording on one side while transferring data from the other, as in U.S. Pat. No. 4,371,752, such are limited either by inability to record conversations of random length (from seconds to hours) and/or by expensive construction--limitations overcome by the present invention through its economy of construction residing in the use of dynamically assigned blocks of memory under novel microprocessor control.

In accordance with the invention, it is now possible to create such a system at reasonable cost readily adaptable so that messages may be placed and retrieved by telephone for applications such as the following:

1. Centralized telephone answering capability for a large number of users, typically employees of an organization. The equipment is readily adaptable to allow messages to be forwarded to other system users than the user for whom the message was originally intended; such message forwarding being passive or forced. Thus messages may be left by anyone for subscribers to the system and may be retrieved by subscribers to the system at any convenient time. Such operation is sometimes described as "voice mail".

2. Telephone answering services can be automated using this approach.

3. Radio-actuated paging system can provide automatic recording of voice messages which can than be accessed by the individual pages (tone only) when such individual can reach a telephone.

4. Some specialized dictation systems such as those used to dictate medical test reports or the like can use the system; the information being accessed by listening to the voice recording as soon as the report is dictated, without waiting for the transcription.

An object of the invention, accordingly, is to provide a new and improved method of and apparatus for voice communication storage and forwarding that is not subject to the prior art limitations, above-mentioned and others, but that enables simultaneous access to multiple users with great flexibility of application and low cost.

A further object is to provide a novel message storing and retrieval technique of more broad utility, as well.

Other and further objects will be explained hereinafter and are more particularly delineated in the appended claims.

The method underlying the invention converts the electrical signal which is an analog of the sound pressure of the voice to a digital signal and stores the digital data on a magnetic disc (a Winchester or hard disc), accessable at random such that any communication can be rapidly located and played back. Further, as explained more fully hereinafter, because of the speed with which the data can be transferred to the disc and retrieved from the disc, a buffering system is provided which allows a number of users (up to 32 in the preferred embodiment) to use the same disc at the same time, each with the impression that the disc is totally at the user's disposal (i.e., simultaneous access).

In summary, from one of its important aspects, the invention embraces a method of message communication digitized storage (in recording mode) and forwarding (in playback mode) along a plurality of input/output channels with simultaneous access to multiple users with the aid of a plurality of line interface systems having series/parallel first-in-first-out memories (FIFO) and intermediate memory RAMS under first microprocessor control and with data transfer connection to main memory RAMS the contents of which are data transferable under second microprocessor control with memory discs, said method comprising, reading the message recording or playback status of each of the plurality of channels in order; if in a recording mode, determining the state of filling of the FIFO memories and, under the control of the first microprocessor, transferring the data in as many filled FIFO memories as exist at such time to the first available empty data block of the intermediate memory RAMS, and when such block is filled, transferring the data thereof to the main memory RAMS; if in a playback mode, determining the state of filling of the successive FIFO memories and transferring channel data blocks to fill empty FIFO memories with message data being played back; examining the recording or playback status of each input/output channel of the plurality of line interface systems under control of the second microprocessor; receiving commands identifying which channels are assigned for recording and which for playback; if a recording mode is assigned, and the said examining of said status has recognized such recording mode, determining whether data blocks from such channel still exist in the main memory and transferring such to the memory discs; if the recording mode is not recognized during such examining, communicating such status to the first microprocessor of the appropriate channel line interface system; if a playback mode is assigned, and the said examining of such status has recognized such playback mode, determining whether the channel is calling for more data, and if so, transferring data blocks of such channel from full main memory RAMS to intermediate memory RAMS and successive FIFO memories of the appropriate channel line interface system for playback; and, if the playback mode is not recognized during such examining, communicating an interrupt to the first microprocessor of the appropriate line interface system. Preferred and best mode embodiments and details of construction and operation are hereinafter presented.

The invention will now be described in connection with the accompanying drawings. FIG. 1 of which is a block diagram of a preferred embodiment of the invention illustrating data flow and control of data transfer for a voice or similar storage and forwarding system with simultaneous access for multiple users, shown operating in the recording mode;

FIG. 2 is a similar diagram of the system in playback operation mode;

FIG. 3 is a more detailed diagram of the line interface input/output circuits employable in the systems of FIGS. 1 and 2;

FIG. 4 is a similar detailed diagram of the processor and storage section of each of the line interface systems or boards;

FIG. 5 illustrates circuit details for transferring of data between the line interface systems or circuit boards and the main processor board of FIGS. 1 and 2;

FIG. 6 is a diagram of the details of such main processor boards with address, control, and some circuitry omitted for clarity and to highlight the novel features of the invention; and

FIGS. 7 and 8 are simplified flow charts illustrating the programming operational steps employed by the microprocessors of each line interface board and main processor board, respectively.

Referring to FIG. 1, an overview of a preferred embodiment is illustrated operating to record speech. Data flow paths are shown with solid arrowed lines and data transfer, with dashed lines. The system is provided with hard discs 11, a main processor board 13, and from one to four line interface boards 12. Each of the line interface boards contains eight independent input (or output) circuits 14 for audio signals, one of which is shown in FIG. 1.

In the recording mode, each input 1 is supplied with a voltage signal which is an analog of the voice sound pressure applied to the microphone of a telephone transmitter, for example. This signal is converted to a digital signal by an analog-to-digital converter (A/D) 2. The preferred A/D conversion scheme is the continously variable slope delta (CVSD) modulation method in which the continuously varying audio signal is converted into a bit stream. At the preferred sampling rate of say 20 kHz, the output of the A/D converter consists of a series of voltage pulses, spaced apart by 50 microseconds. Each pulse has a height of 5 V (a "one") or of 0 V (a "zero"), depending on whether the audio signal slope increased or decreased in amplitude within the previous 50 microseconds. This bit stream can be reconstituted into an analog voltage representation of the sound when applied to the input of an appropriate digital to analog (D/A) converter, later discussed.

The bit stream produced by the continously variable slope data modulation A/D converter is stored by a unique storage process in accordance with the invention. The first step in the storage process is the assembly of the bit stream into a series of, for example, 8-bit words. Each of the A/D converters of the line interface boards 12 is shown connected to a series of four 32.times.8 bit serial/parallel, first-in-first-out (FIFO) memories 3. As the bit stream begins to flow it fills the top register 15 of the first 32.times.8 FIFO memory. When the top register of the first FIFO is filled, the contents are shifted ("rippled") immediately to the bottom register of the fourth FIFO memory 16. The top register of the first FIFO memory is then free to assemble the next 8-bit word from the data stream. When the register is filled again, the resultant word is transferred to the second-from bottom register in the fourth FIFO memory 17. In the same way, the top register of the first FIFO memory keeps assembling the bit stream into 8 bit words which are passed down the FIFO memories to the last available unfilled register. At irregular intervals, under control of the program stored in read-only-memory ROM1 18 and scratchpad SPRAM1 27, microprocessor (.mu.p1) 5 (e.g. a Zilog Z80A) determines the extent to which the FIFO memories 3 are filled. When at least the fourth of these FIFO memories is filled, .mu.p1 causes the contents to be transferred to one of the forty 6k.times.8 (6,144.times.8) random access memories (RAM e.g. RAM TI 4164) 6 (the intermediate memory) using conventional data transfer means 4.

The choice of which of the RAMs 6 of intermediate memory (IM) is to be used is determined by .mu.p1 according to the program hereinafter described. At the time a decision to transfer is made, the contents of all filled FIFO memories will be transferred; that is, the contents of the third (if it is full) and the contents of the second (if it is full), in addition to the contents of the fourth. The number of FIFO memories and the interval at which .mu.p1 makes transfers does not allow the first FIFO memory to ever fill up, since if all the FIFO memories were to fill before a data transfer, data would be lost.

Once data from the filled FIFOs is transferred, the data in the partially filled FIFO remaining above the last FIFO to be emptied is rippled to the bottom of the fourth FIFO, and the process of accumulating data in the FIFO registers continues.

The transfer of data to one of the RAMs 6 is organized so that at the first data transfer, the contents of the bottom register 16 in the fourth FIFO memory is placed in the top register 19 of the RAM; the contents of the second-to-bottom register 17 of the fourth FIFO memory is placed in the second register 20 of the RAM/ and so forth. Thus the oldest data is at the bottom of the series of FIFO memories; and the oldest data is at the top of the RAM.

Once an assignment is made of the RAM 6 to be used for recording data from a particular audio input 1, .mu.p1 keeps adding data from that source until the RAM is filled. Thus the RAM accumulates 6k.times.8 bits of data. Since the bits are produced one each 50 microseconds, each RAM will contain 6144.times.8.times.50.times.10.sup.-6 =2.5 seconds of conversation. Subsequent 2.5 seconds of conversation are stored in other RAMs 6 as assigned by .mu.p1 5.

The main processor board 13 provides an assembly point for data from the (up to) four line interface boards 12. As the intermediate memory RAMs 6 on the line interface boards fill up, the .mu.p1 5 on each board queues the filled RAMs for transfer to the eighty 6k.times.8 RAMs 8 on the main processor board--the main memory M M. At irregular intervals, under control of the program stored in ROM2 21 and scratchpad SPRAM2 28, the further microprocessor .mu.p2 9, shown on the main processor board 13, investigates the queues as reported by the .mu.p1 5 of each of the line interface boards 12, and causes the contents of the RAM 6 at the head of a queue to be transferred to one of the RAMs 8 using conventional data transfer means 7. The choice of which of the RAMs 8 is to be used is determined by .mu.p2 according to the later-described program. Once the contents of a RAM 6 are transferred, the RAM is returned to the pool of unused RAMs and may be reassigned by .mu.p1 5.

As the RAMs 8 in the main processor board 13 fill up with data transferred from the audio interface boards 12, the microprocessor .mu.p2 queues the filled RAMs 8 for data transfer to the discs 11. The contents of the RAM 8 at the head of the queue are transferred at irregular intervals by .mu.p2 using data transfer means 10 to a single 6k.times.8 RAM 22. The data is then transferred using data transfer means 25 from RAM 22 to a disc under the control of .mu.p3 23. This microprocessor receives its instructions from ROM3 24 which contains the disc control program and from scratchpad SPRAM3 29. Once the contents of one of the main memory RAMs 8 are transferred to the disc system, the RAM is returned to the pool of unused RAMs and may be reassigned by .mu.p2. System response is initiated by outside control 26 to .mu.p2, illustrated at the bottom of FIG. 1.

The operation of the system to play back speech is shown in FIG. 2, with components similar to those of FIG. 1 identified by prime notations.

In the playback mode, the system is instructed (by external means 26) to play back a message previously recorded. As indicated above, the message has been stored in 2.5 second segments on the discs 11'. The disc control system, under control of .mu.p3 23' and the program stored in ROM3 24' and SPRAM 3 29', finds the segment of disc containing the first part of the 2.5 seconds of the message to be played back, and transfers the data to a 6k.times.8 RAM 22' using data transfer means 25'. At irregular intervals, .mu.p2 9' under control of ROM2 21' and SPRAM 28' assigns one of the available 6k.times.8 RAMs 8' in the main memory to receive this data. The choice of RAM is made by the program later described. Data is then transferred to the RAM 8' from RAM 22' using the data transfer.

At irregular intervals, .mu.p2 9' recognizes the existence of data in one of the RAMs 8' waiting transfer out. The .mu.p2 9' recognizes which of the (up to) four line interface boards 12' contains the audio channel to which the playback is directed and transfers the data from the RAM 8' to one of the available RAMs 6' on the appropriate line interface board, using data transfer means 7'.

At irregular intervals, .mu.p1 5', under control of the program stored in ROM1 18' and SPRAM1 27', recognizes the presence of the first (and subsequent) 2.5 second segments of speech in one of the RAMs 6' and which of the 8 input/output circuits 14' is to be used. Data has been stored in the RAMs such that the first (oldest) data to be played back is in the top register 37 of the RAM; using data transfer means 4', the 8-bit word in the top register 37 of the RAM is transferred to the top register 15' of the first of the four 8.times.32 FIFO memories 3'. As in the case of the record mode (FIG. 1), this word is automatically transferred to the bottom register 16' of the fourth FIFO memory. The next 8-bit word at register 38 in the RAM 6' is then transierred to the top register of the first FIFO memory; this word is then rippled down through the four FIFO memories. As this process continues, the contents of the bottom register of the fourth FIFO memory is removed bit-by-bit (oldest bit first) at the 20 kHz rate and fed to the input of a D/A converter 2'. As soon as the contents of the bottom register 16'have been transferred, the contents of the next higher register 17' are transferred to the bottom register 16' , allowing the D/A conversion to continue at the same 20 kHz rate used to digitize the data.

System timing is so arranged that the series of four FIFO memories is never completely empty, thus avoiding delays in the played back conversations that are long enough to be perceived by the listener. Further, at least the next segment of 2.5 second conversation has been transferred to one of the RAMs 6' by the time the transfer of the previous segment is complete, ensuring continuity.

The D/A converter 2' reconstitutes the analog audio signal using the 20 kHz bit stream, employing the bits to control the slope of a continuously varying audio output signal; the process being opposite to that used to digitize the signal by the CVSD modulation method described in FIG. 1. The audio output appears on the output line so-labelled to the left in FIG. 2.

As with the operation in the record mode (FIG. 1), once the contents of RAM 6' or 8' have been transferred, the RAM is reassigned to a pool of available memories.

The process of recording described in connection with FIG. 1 and of playback described in connection with FIG. 2 may proceed at the same time with up to, for example, 32 channels operating simultaneously with any number n of channels devoted to recording and the balance of the channels (32-n) devoted to playback. The number of channels devoted to each service and the number idle at any moment is determined by the external calls for service at the moment. Once a channel which has been recording or playing back is released, it enters a pool of idle channels which can be reassigned for either recording or playback at the next call for service.

Channels may, if desired, be alternated between playback and record during the course of a conversation, allowing for the insertion of prerecorded prompts which help the user to operate the system. In a typical application, a 32-channel system permits reasonable service for several thousand subscribers.

It is now in order to examine in more detail the operation and programs used to control the system.

The details of the input/output circuit, eight of which are included on each line interface board, are shown in FIG. 3. Clock, some circuitry, and control signals are not shown for clarity and in order not to detract from the essential features of novelty underlying the invention. The most important control signal (provided by .mu.p1) defines whether the input/output circuit will operate in the record or in the playback mode.

When directed to operate in the record mode (FIG. 1), the audio line 1 is routed through a solid state analog switch S to the appropriate input of an antialiasing filter F. In this mode of operation, the filter serves as a bandpass filter with a pass band of from, say, 100 Hz to 3500 Hz, thereby reducing any extraneous noise components on the input signal. The signal is then passed to the A/D CVSD modulator 2 which provides a bit stream at the 20 kHz clock rate. A bus connector is shown at BC1 conected to the FIFO memories 3, being set to a high impedance state; while a further bus connector BC2, connected to the A/D modulator 2 and to the bus connector BC1, is set to pass the bit stream to the first series/parallel FIFO memory 3. The bit stream is formed into 8-bit words as described above, and the FIFO memories 3 are connected so as to ripple each word as it is formed to the bottom-most available register of the series of FIFO memories 3. When one of the FIFO memories is filled, a flag is set which may be passed to .mu.p1 5 through bus connector BC3, upon command. When .mu.p1 5 decides to transfer data, the lower bus connector BC4 connected to the lowermost or fourth FIFO 3, is enabled, placing the contents of the bottom register of FIFO memory onto the data bus DB. The .mu.p1 5 continues to extract data, which keeps rippling to the bottom register until all the contents of the filled FIFO memories 3 are transferred.

When directed to operate in the playback mode, bus connector BC2 is set to the high impedance state and .mu.p1 5 begins to place a series of 8-bit words on the data bus DB. As each word is placed in the top register of the FIFO memory through bus connector BC1, it ripples to the bottom register of FIFO memory 3 and is extracted, bit by bit, at the 20 kHz clock rate, bus connectors BC3 and BC4 being in the high impedance state. The bit stream is passed to the D/A converter of the modulator 2', the resultant audio signal is smoothed by the filter F, and is passed on to the audio line 1' by the switch S which has now been set to the "out" or playback position.

FIG. 4 illustrates details of the microprocessor .mu.p1 5, its associated read-only memory ROM1 18, a direct memory access controller DMA1, the scratchpad SPRAM1 27, and the intermediate random access memory 6 used for storage of 2.5 second segments of speech by each of the (up to) four line interface boards 12 in each system. Address, control, and some connecting circuitry are again omitted for clarity.

When the .mu.p1 5 decides to transfer data from one of the eight channels on its board which is in operation in the record mode, FIG. 1, it first determines how many of the FIFO memories 3 used by the channel are filled. If only the bottom FIFO is filled (FIFO 4 in FIG. 3), .mu.p1 5 transfers the contents to intermediate memory 6 by making 32 transfers through an internal register. If more than one of the FIFOs is filled, .mu.p1 5 transfers the data using a faster method. It first sets up DMA1 to effect a direct memory access transfer, and then turns over the transfer to DMA1 for 64 or 96 transfers, corresponding to a filled condition of FIFOs numbers 3 and 4 or FIFOs 2, 3, and 4 (see FIG. 3). Since it takes longer to set up DMA1 than to transfer the data directly from one FIFO, this approach results in optimum speed of transfer.

The intermediate memory 6 is shown organized as an array of thirty-two 64k.times.1 RAMs. Each of the four rows of RAMs contains eight columns of RAMs, one for each bit in the eight bit words being transferred from the FIFOs. The memory is organized into 40 segments, each of 6k 8-bit words (2.5 seconds of conversation). The .mu.p1 5 identifies an available segment of the intermediate memory in which to place each 2.5 second segment of conversation from the channel providing the input. When a 2.5 second segment of intermediate memory is filled, a new segment is identified and filled with the next 2.5 seconds of conversation. Bus connector BC5 is enabled to take data from .mu.p1's data bus DB and place them into the appropriate RAMs as directed by the address signals from .mu.p1 5 or from the DMA1 controller, whichever is responsible for the transfer.

When the channel being serviced is designated as a playback channel (FIG. 2), the .mu.p1's task is to keep the FIFOs associated with this channel supplied with data from the message which is to be played back. Subsequent circuitry will keep at least one 6k segment of intermediate memory available for transfer to the FIFOs associated with the channel being serviced. If only one of the FIFOs (FIFO 1 of FIG. 3) is empty, transfer will be made through .mu.p1 5. If more than one FIFO of FIG. 3 is empty, transfer will be made by DMA1, thereby optimizing speed of transfer as in the case of data recording. Bus connector BC5 is enabled to take data from .mu.p1's data bus DB as supplied by the appropriately addressed RAMS and supply it to the FIFOs.

The system is fast enough to allow all eight input/output circuits to be served. That is, even if all channels are busy, and regardless of the combination of input/output channels, the FIFOs of any recording channel never fill up before data transfer occurs to clear them out to intermediate memory IM, and the FIFOs of any playback channel are never totally emptied before data is supplied to replenish them. This condition is fulfilled by the architecture described, together with the use of a microprocessor .mu.p1 capable of rapid operation and programming which emphasizes processing speed (for instance using alternative direct or DMA transfer).

Details of how voice data may be transferred from each of the (up to) four line interface boards 12 to the main processor board 13 are illustrated in FIG. 5. The data transfer requires special means because the line interface boards are operating under control of their microprocesors .mu.p1 5 and the main processor board is operating under control of its completely independent microprocessor .mu.p2 9, FIG. 1. Each of the line interface boards is equipped wth two 6k.times.8 RAMs 8" and 8'". The first of the RAMs 8" is used to transfer data in 6k blocks (2.5 seconds of conversation) from one of the blocks of data in intermediate memory IM (FIG. 4) to the main memory MM on the main processor board (record mode, FIG. 1). The second of the RAMs 8'" is used to transfer data in 6k blocks from one of the blocks in the main memory MM to one of the available blocks in the intermediate memory IM of the apropriate line interface board (playback mode, FIG. 2).

Data transfer out to the main processor board 13 (record mode, FIG. 1) is initiated by .mu.p1 5 which queues 6k blocks of data in IM for transfer. The .mu.p1 5 takes the block at the head of the queue and using controller DMA1, FIG. 4, transfers the data to the RAM by placing it on its data bus DB. Bus connector BC7 is opened and bus connector BC9 is closed by .mu.p1 5 to allow the shared data bus SB' to act as an extension of the data bus DB of .mu.p1 5. At the same time, .mu.p1 5 provides a signal to .mu.p2 9 indicating that a block of data is ready for transfer (service request). When .mu.p2 9 is ready to service the request, it makes shared bus SB' an extension of its data bus DB' by turning bus connector BC9 on, and bus connector BC7 off. The .mu.p2 9 then effects a transfer using its DMA2 controller, now to be discussed in connection with FIG. 6.

Data transfer from the main processor board (playback mode, FIG. 2) is initiated by, .mu.p2 9 on the main processor board. The transfer of 6k blocks to RAM 8' is performed by .mu.p2 9 using its DMA2 controller, FIG. 6, and gaining access to the shared bus SB (FIG. 5) by turning bus connector BC8 on, and bus connector BC6 off. Once the transfer is complete, .mu.up2 9 provides a signal to the appropriate .mu.p1 5 and this .mu.p1 transfers the contents of RAM 8' to an available block of 6k IM in its board, using its DMA1 controller, FIG. 4, and gaining access to the shared bus by turning bus connector BC8 off and bus connector BC6 on.

FIG. 6 provides details of the main processor board 13 (13'). Once more, for clarity, address and control signals are deleted, as is some connecting circuitry. The .mu.p2 9 under control of the program stored in ROM2 21 and SPRAM2 28 is responsible for collecting recorded data in 6k blocks from the line interface boards 12 and supplying playback data in 6k blocks to the line interface boards. The .mu.up2 9 transfers data between the line interface boards and an array of eighty 6k.times.8 RAMs 8 using its DMA2 as shown in FIG. 6. RAM 8, the main memory MM, is organized eractly as the intermediate memory described in FIG. 4, except that it has eight rather than four rows of 64k.times.1 devices. The .mu.p2 9 also has the task of transferring data--as usual in 6k blocks--between the main memory and the discs 11. Despite its size, the main memory can hold only 2.5.times.80=192 seconds or 3.2 minutes of conversation. Thus the task of storage must be passed on to the discs promptly to keep the main memory from filling up.

Using the shared bus approach described in connection with FIG. 5 above, and shown diagramically in FIG. 6 as data transfer means 25, data is either placed into or extracted from one of two 6k.times.8 RAMs 22 through .mu.p2's data bus DB'. This data is removed and supplied to the disc system or is supplied from the disc system by a further microprocessor .mu.p3 operating under control of the program stored in a ROM3 and SPRAM3. The special requirements of transferring data between one of the RAMs 22 and the discs is met by the use of disc control circuit 32 and high speed disc data bus 34 rather than the DMA devices associated with previously described data transfers. Circuit 32 represents a special version of circuits used to effect such transfers, however, in that it handles data in 6k blocks rather than the much shorter 256 byte blocks conventionally employed.

Finally, .mu.p2 9 is concerned with communication for control and other purposes with the outside world, represented as "other interfaces"33. For this purpose, ports are provided allowing access to .mu.p2's data bus DB' using the same method as described above with respect to transfer of data between main memory and the intermediate memories of the line processor boards. In particular, external signals are required to inform the system as to which message is to be retrieved, what channel it is to be delivered through, which channel is calling for service to record a message, and how the message is to be identified for later recall. Further it is desireable to back up the contents of the discs on a cassette or other large scale memory device.

A flow chart, simplified for clarity, of the program used to operate the .mu.p1 on each of the line interface boards is presented in FIG. 7. The .mu.p1 maintains a file of eight records in its scratchpad memory, each of which relates to the status of one of the eight input/output channels and, in the event the channel is not idle, the status of the message which the channel is either recording or playing back.

Initially, the .mu.p1 reads the status of the first channel at A. If the channel is idle, .mu.p1 proceeds, after completing the data transfer task indicated in block B later to examine the second channel. The .mu.p1 continues to examine the status of each channel in order. In the event that all channels are idle, .mu.p1 simply cycles through a status check of the channels over and over. The program returns to look at the first channel following a check of the eighth channel; and the idle cycle is shown by the arrow C.

When .mu.p1 reads the status of a channel which is not idle, it determines from the status data whether the channel is to be used to record or play back. This information is initially provided by outside control through .mu.p2 and remains fixed until the channel is returned to idle status, again by outside control relayed by .mu.p2.

Once the channel is designated either as a record or playback channel D, .mu.p1 sets the switch S (and other circuitry) shown in FIG. 3 for the appropriate service. The switch S is shown set at E upon the first cycle through this portion of the program and remains with this setting until the channel's status is changed. The next time .mu.p1 cycles to the line under consideration, if the line is in the record mode, .mu.p1 will examine the status of the FIFO memories at F, and determine if the last FIFO is filled, G. If it is not filled, .mu.p1 returns to examine the status of the next line; if the last FIFO is filled, .mu.p1 transfers the data in as many filled FIFO memories as exist at the time to the first available empty 6k block of IM. The transfer H is effected using either an internal register of .mu.p1 or direct memory access, whichever is faster, as described above. This process continues in subsequent cycles until .mu.p1 notes that the 6k IM block assigned to the channel is filled (at I). At that time, the filled 6k IM block is scheduled for transfer to the main procesor board and a new empty 6k IM block is assigned to the channel for further data transfers J.

Before examining the next channel for status A, .mu.p1 will always check to see if there are any 6k IM blocks of data waiting for transfer at B. If there are (and if the data link to .mu.p2 is available), the 6k IM block at the head of the queue will be placed into the link to await pickup by .mu.p2.

When .mu.p1 cycles to examine the status of a channel which is designated as a playback channel, .mu.p1 examine the status of the FIFOs at K, once the switch S has been properly set at E. If the first FIFO is not empty, .mu.p1 returns to examine the status to the next line. If the first FIFO, however, is empty, as at L, .mu.p1 transfers data to fill as many empty FIFO memories as exist at that time from the 6k block of IM with the message data being played back (M). The transfer is effected using either an internal register of .mu.p1 or direct memory access, whichever is faster, as before described. The process continues in subsequent cycles until .mu.p1 notes that the 6k IM block assigned to the channel is empty, as at N. At this time, the next 6k IM block is assigned for further data transfer to the channel and a request for transfer of a new data block is put out to .mu.p2. The .mu.p1 keeps at least two blocks of 6k data available in IM for transfer to each channel in playback mode.

Unlike the scheduled transfer of data at B for channels in the record mode, .mu.p1 is interrupted by .mu.p2 whenever a block of playback data is placed in the data transfer link by .mu.p2. The .mu.p1 services the interrupt by transferring the data out of the link into the first available block of IM. Changes in the status records initiated by outside commands (begin recording, begin playback, and so forth) are relayed by .mu.p2 by means of interrupts to .mu.p1.

In FIG. 8, a similar flow chart is presented for the program which controls the main processor .mu.p2 on the main processor boards 13. The .mu.p2 maintains 32 records in the illustrative preferred embodiment illustrated and described herein, one for each input/output channel on the (up to) four line interface boards. The program starts by examining the status of the first channel on the first line interface board at A'. If it is idle (B'), the .mu.p2 goes on to examine the second channel on the first line interface board, and so forth. When the system is configured with less than four line interface boards, the status indication for an unpopulated position informs .mu.p2 of its absence and the program immediately proceeds to examine the status of the next channel. After examining the 32nd channel (the eighth channel of the fourth line interface board), the program returns to examine the status of the first channel of the first line interface board. If all channels are either idle or unpopulated, the program continues to cycle as shown by arrow C'.

Commands generated outside the system identify channels which are assigned to record a message and assign the message an identifying number; other commands identify a channel to be used to play back a message and identify the message to be played back by the number assigned to it when it was recorded. The commands are placed in the data link connecting .mu.p2's data bus to a port on the main processor board. Such commands are accompanied by an interrupt to .mu.p2 which then services the command by reading the contents into the appropriate record of the status table.

When the program first examines the status of a channel which has received a command, it first determines whether a record or playback status exists at D'.

If "record" is indicated, the program then determines whether it has recognized this status before or not, at E'. If not, the program puts the new status out to the appropriate .mu.p1 by placing the data into the data link for the correct line interface board and issuing an interrupt to the appropriate .mu.p1 as at F'. If the data link is not available at this point, the program skips to examine the next channel; and when it returns, it will again attempt to inform the appropriate .mu.p1. If the program recognizes that it has successfully informed .mu.p1 on a previous cycle, it determines whether a 6k block of data from the channel exists yet in the main memory, as at G'. If it does, it is queued for transfer to the discs at H' and a new block is assigned to the channel for subsequent recording. The program then returns to check the status of the next channel.

If "playback" is indicated, on the other hand, the program then determines whether it has recognized this status before or not at I'. If not, the program puts the new status out to the appropriate .mu.p1 by placing the data into the data link for the correct line interface board and issuing an interrupt to the appropriate .mu.p1 at step J'. If the data link is not available at this point, the program skips to examine the next channel; and when it returns, it will again attempt to inform the appropriate .mu.p1. If the program recognizes that it has successfully informed .mu.p1 on a previous cycle, it determines whether the channel is calling for more data. If so, and if a 6k MM block is full (steps K' and L'), the 6k block is sent out to the appropriate line interface board at M'. The program then asks if at least two blocks of 6k MM data remain awaiting transfer at N'. If not, the program requests a 6k block data from the discs at O'. The program then returns to examine the next channel's status.

Further modifications will occur to those skilled in this art, including other types of memory, data transfer and control techniques useable to practice the method of the invention, and also other types of communication serviceable by the same; and such are considered to fall within the spirit and scope of the invention as defined in the appended claims.