Patent Application
20110179193
(1) Field of the Invention
The invention concerns a peripheral interface with a control unit, which receives a data stream from a control computer and puts it out sequentially to a peripheral device, a method for transfer of a data stream via a peripheral interface, wherein a data stream is received from a control computer and preferably put out sequentially to a peripheral device, and a data stream containing control data for a peripheral device.
(2) Description of Related Art Including Information Disclosed Under 37 CFR 1.97 and 1.98
A peripheral interface is used to communicate between a control computer and one or more peripheral devices by electronic data transfer. In the prior art, peripheral devices are connected via standard interfaces, such as “USB” (“Universal Serial Bus”), “Ethernet” (IEEE 802) or “Firewire” (IEEE 1394), which provide a separate bus on the peripheral side, while they typically communicate with the control computer by one of its system buses. If a peripheral device is supposed to be controlled by the peripheral interface, the control computer or, more accurately, its central processing unit (CPA) will generate, for example, a stream of control data, which is deposited in the working memory (“random access memory (RAM)”) of the control computer. The transfer to the peripheral interface is generally controlled by driver software, executed by the central processing unit, which initiates accesses of the interface component to the working memory. Access to the working memory occurs in cycles by means of direct memory access (DMA) via the system bus of the control computer. For this purpose, the peripheral interface has a DMA controller. Typically, one data block is transmitted by the DMA controller in each DMA cycle. The size of a DMA cycle as a rule corresponds to a single memory page of the working memory.
The driver software reports each DMA cycle for transfer in the DMA controller of the peripheral device. The control computer then executes other software, such as an application program with user interaction. It later receives a confirmation of the complete transfer in the form of an interrupt request (IRQ). From the interrupt request onward, the regular program execution is interrupted and in its place the driver software is executed, which must continue the data transfer by triggering the next DMA cycle. In order to continually achieve a high data transfer rate, a very short response time of the driver software to such an interrupt request is required, so that the next DMA cycle can be triggered at the proper time. In particular, in a real-time-critical system, a maximum response time must be obeyed in order to prevent a gap in the data transfer. If the control computer is not using a real-time-capable operating system, however, a deterministic processing of an interrupt request for a DMA cycle is not possible. Here, the response time to an interrupt request can increase almost arbitrarily when the system workload increases or when there are many DMA transfer processes at the same time. The number of simultaneous DMA transfer processes increases with the data transfer rate to be achieved and is furthermore determined by the size of the data amount being transmitted. If the response time of the central processing unit exceeds a critical value, there will be interruptions in the data stream.
Yet many peripheral devices depend on a continual, or gap-free data stream. For example, the scanner of a laser scanning microscope (LSM) must be permanently supplied with control data for the movement of one or more galvanometer mirrors while taking a picture. This is typically done by generating the control data stream block by block (corresponding to time intervals) and sending it from the control computer to the scanner. In this way, for example, it is possible to change the scanning movement in the short term, even while taking a picture. But when the data stream is interrupted, each galvanometer mirror performs a free oscillatory movement from its current movement state. Depending on the momentary deflection, velocity and acceleration of the particular mirror at the time of the data stream interruption, damage can occur to the mirror (e.g., by excessive rotation) or the specimen (e.g., by overexposure).
In the prior art, thus, the disadvantage of a costly real-time operating system must be used on the control computer to prevent interruptions in the data stream.
The problem on which the present invention is based is to improve a peripheral device, a data stream, and a method of the aforementioned kind so that a data stream without gaps can be provided for a peripheral device, especially for a non-real-time-capable operating system of the control computer.
According to the invention, the controller of the peripheral interface obtains an instruction from the received data stream for a modification of the output of the data stream. By identifying an instruction in the received data stream for a modification of the output of the data stream and putting out the data stream modified in accordance with the identified instruction, the controller can adapt the data stream to the particular conditions in accordance with the identified instruction, without a real-time-critical interaction with the control computer. Since the control computer produces the data stream, it can influence the behavior of the peripheral interface by an instruction modifying the output of the data stream. The modification, for example, can be carried out one time from the point in the data stream where the instruction is located, or until otherwise instructed, or permanently. It is also conceivable that the modification is instead carried out one time from a time when a given condition (e.g., an external condition) is fulfilled for the first time, or until otherwise instructed, or permanently. The instruction can also be organized, or interpreted by the controller, such that the output of the data stream is only modified if the given external condition is fulfilled. For this purpose, the controller can repeatedly check whether the condition for the modification as contained in the instruction is fulfilled, and only carry out the modification if and for as long as the condition is fulfilled. If the condition is no longer fulfilled, then the output of the data stream can continue unmodified. The modification may consist, for example, in a reordering of the contents of the data stream or a repetition of certain segments of the data stream. This type of modification may exist beforehand or alternatively it can be encoded in the instruction. Additional parameters may be encoded in the instruction in this case.
Thanks to a modification instruction contained in the data stream according to the invention, preferably a conditional one, the makeup of the data stream reaching the peripheral device can be adapted to the particular situation in a flexible manner by the peripheral interface itself, i.e., without time-critical use of the central processing unit of the control computer. For example, the modification instruction can contain a segment of spare data (“emergency segment”) with a marking already known to the controller. If the controller in this example finds that the data stream provided by the controller is broken off, it will seamlessly continue the output to the peripheral device with the spare data. In particular, it can repeatedly carry out the spare output until it once again has sufficient regular data from the control computer. Thus, a gap-free output of a continuous data stream to the peripheral device can be achieved. A real-time operating system is not required. The peripheral interface can be arranged directly at the peripheral device in this case, so that it puts out analog control signals exclusively, for example. Alternatively, an additional data bus can be arranged between the peripheral interface and the peripheral device.
Preferably, the instruction modifying the output contains a conditional or unconditional jump command to continue the output at a position which is determined or can be determined by the jump command within the data stream. A jump command instructs the controller to modify the output by continuing it at another place in the data stream, to be indicated by the jump command. A conditional jump command instructs the controller to continue the output at the other location only if a logical condition to be indicated with the jump command is fulfilled; otherwise, the controller continues the output sequentially after the jump command.
The jump command itself is in no case put out to the peripheral device. The use of jump commands within the data stream allows a flexible changing of the output of the data stream to the peripheral device, without the peripheral interface having to know the internal structure of the data stream. It must only be able to identify jump commands as such and interpret a possible condition for the execution of the jump command. With conditional and unconditional jump commands, the modified output to the peripheral device can occur in loops with slight expense. For example, this can be used to reduce the unneeded transmission bandwidth. Thus, a repeating segment of control data only needs to be transmitted once to the peripheral interface if a counter in the peripheral interface is initialized by means of an instruction embedded in the data stream and a conditional jump instruction brings about an endless repetition from the end to the start of the particular data segment.
Advantageously, it is also possible for the control computer to generate two or more simultaneous control data streams, all of which are transmitted to the peripheral interface, wherein the controller of the peripheral interface at a given time only puts out control data from precisely one of the data streams to the peripheral device. The output can alternate between the control data streams by modification instructions for the output that are embedded in the data streams, such as appropriately expanded jump commands. This alternation can take place in a short time, for example, within microseconds. This rapid switching can be used, for example, in medical experimentation layouts, in which various instrument functions have to be triggered with brief response time. Such a switching between several data streams can also be combined, in particular, with modification instructions for a spare data output. For example, spare data/emergency segments can be embedded in each of the control data streams being transmitted simultaneously.
If the data stream is made up by segments, wherein each segment has a constant length or a length indication, each segment can contain a content indication, such as one in the form of one or more numerical or alphanumerical symbols. With such a content indication, segments with spare data can be marked as such. Such a marking for a spare data segment can be interpreted as a jump command past the spare data segment, so that the spare data are not put out directly to the peripheral device. The jump destination can be determined by the controller using the stipulated or explicitly indicated segment length. As an alternative to a jump command, a jump width or “offset” can be indicated explicitly. The jump destination for continuing the data output then results from the current position in the data stream plus the offset. Alternatively, a jump destination can be indicated directly with the jump command.
In a more general embodiment of the invention, the controller identifies a break in the data stream being received and then by itself puts out a spare data stream. The spare data stream can either be permanently specified for the controller, for example, in a read-only memory (ROM). For example, it might be spare data that control the galvanometer mirrors such that an uncritical specimen region or a region outside the specimen is illuminated (“parking”). As an alternative to memory storage, the controller can generate the spare data stream during the transmission, “on the fly”, for example, from the last data put out regularly to the peripheral device. This can be done, as a precaution, permanently/simultaneously to the regular output, for example, and in event of a break in the regular data output the output is switched only from the regular data stream to the spare data stream. In this way, the response time for output of the spare data stream is minimal. Alternatively, generation of the data stream can be started only after detection of the break.
In an advantageous, alternative embodiment, the controller can receive an instruction for a composition of the spare data stream from the data stream. In this way, the spare output can be organized in flexible fashion, as the control computer in this way can influence the spare output depending on the situation. Even so, no real-time-critical interaction is required between the control computer and the peripheral interface. The instruction for the composition of the spare data stream can be organized, for example, as an instruction for a modification of the data stream in the above-described sense, or be interpreted in this way by the controller.
Advisedly, the controller ends the output of the spare data stream when the control computer continues the data stream to be received, and puts out the continued data stream. In this way, the data stream can be transmitted seamlessly. If a result data stream from the peripheral device is also transmitted to the control computer across the peripheral interface in addition to the control data stream to the peripheral device, the controller can reject result data coming from the period of the output of the spare control data. In this way, the spare data output can be configured transparently to the control computer. Alternatively, it is possible for the controller of the peripheral interface to communicate the break in the control data stream, for example, with beginning or at the end of the spare data output, on a separate signal line or encoded in the result data stream. The control computer itself can then reject a part of the result data stream, for example, a part that only contains result data from the period of the spare data output.
Preferably those configurations of the peripheral interface are preferred in which the controller receives the data stream via a system bus of the control computer. This also refers to embodiments in which the controller actively reads the data stream by direct memory access from the working memory of the control computer. In particular, the data transfer method described in DE 10 2008 009 634 (US 2009/0228618 A1) can be used, whose disclosure content is fully incorporated here.
The benefits of the invention can be used in particular for the control of a component of a laser scanning microscope as the peripheral device. In this way, damage to a scanner or to a specimen can be avoided.
Advisedly, the peripheral interface is configured as a plug-in card for the control computer or as a circuit for use in the control computer. In this way, its controller has access to the system bus with little expense and especially to the working memory of the control computer. Furthermore, a plug-in card requires only little space.
The invention also comprises a data stream, which contains one segment with spare control data to be put out in event of a data stream break and a spare data marking leading the segment. Advantageously, the spare data marking is a conditional or unconditional jump command past the spare control data. In particular, a conditional or unconditional jump command back into the spare control data segment can be disposed in the spare control data, especially at its end.
It is especially advantageous for the spare control data to be configured, prior to a transition to subsequent regular control data, so that it places the peripheral device in the same state as does the regular control data preceding the spare control data. In this way, the spare data output can be made transparent to the control computer.
The invention will now be explained more closely by means of sample embodiments. In the drawings are shown:
In all drawings, comparable parts have the same reference numbers.
In describing preferred embodiments of the present invention illustrated in the drawings, specific terminology is employed for the sake of clarity. However, the invention is not intended to be limited to the specific terminology so selected, and it is to be understood that each specific element includes all technical equivalents that operate in a similar manner to accomplish a similar purpose.
Each of the buffers 16 and 17 consists, for example, of several distributed blocks of the working memory 3 that are assembled into a virtually coherent FIFO (First In First Out) buffer. The buffers 16 and 17 are reserved as resident buffers by the software 14 and have a size of 32 MBytes each. They can also have other and different sizes in other embodiments (not depicted). It is also possible to reserve one or both buffers as a physically coherent block. The cache (not shown) of the central processing unit 2 is deactivated for the address regions in which the buffers 16 and 17 are situated. After the reserving operation, the software 14 informs the controller 11 as to the starting address and size of the read buffer 16 and the write buffer 17. For this purpose, the controller 11 makes available two hardware registers, which are overlaid in the address region of the central processing unit 2.
The peripheral interface 4, for example, comprises a storage buffer (not shown) with a size of 8 kilobytes for each direction of transfer. The first storage buffer is used to temporarily store data that was read out from the sending buffer 16 of the control computer 1, before the further transmission via the peripheral bus 9 to the peripheral device 10. The second storage buffer is used to temporarily store data that was received from the peripheral device 10 via the peripheral bus 9 before being written into the receiving buffer 17 of the control computer 1. In other embodiments, other storage buffer sizes can also be provided.
The controller 10 receives the data stream, block by block for example, in data blocks of 4 KBytes each. Once the first block is fully contained in the first storage buffer, it puts out sequentially the data located in the first storage buffer, i.e., for example, byte-by-byte or word-by-word. During this process, depending on the size of the storage buffer, one or more additional data blocks can already be received and written into the first storage buffer. The output is continued seamlessly with this data block.
If the central processing unit 2 of the control computer 1 embeds an instruction for modification of the output while creating the data stream for the peripheral device 10, the actual modification does not have to be carried out by the control computer 1. Instead, the modification of the output can be done in the peripheral interface 4 by the controller 11, without the need to have any detailed knowledge as to the content of the data stream. A real-time-critical interaction with the central processing unit 2 is not required. The controller 11 must only be able to identify a modification instruction as such in the data stream. For this purpose, a particular bit or byte combination can be stipulated as a marking for a modification instruction, for example. If the controller 11 of the peripheral interface 4 identifies a modification instruction, different response options can be considered. For example, the modification instruction can consist of a block segment that contains a spare data stream that is not put out directly by the controller 11, but instead copied to a special emergency buffer. If the controller 11 later ascertains that the data stream from the control computer 1 is interrupted, it can put out the spare data stream from the emergency buffer to the peripheral device 10 either one time or repeatedly until regular data from the control computer 1 is again available. Other options will be described hereafter on the example of a laser scanning microscope, but without being limited to this special application.
A direct access of the controller 11 to the working memory 3 is also possible in the configuration depicted in
During image recording, for example, on account of a user operation at the control computer 1, there can be a change in one or more scan parameters, such as the scanning speed, so that a change in the mirror movement becomes necessary. As the control computer 1 generates the control data in blocks, the change can already be taken into account in the next block, i.e., in the control data for the next pixel cell, or the one after that. The confocally illuminated and recorded focus volume of the specimen 22 can be moved across the specimen 22 by means of the scanner 30, for example in order to record an image, in which process the scanner 30 rotates the galvanometer mirrors in a specific manner.
At the start of each spare data segment SOS, an unconditional relative jump command JMP is arranged, which tells the controller 11 to continue the output after the spare data segment SOS, i.e., in a following data block Bn+1 (
When a block Bn' is ready, the central processing unit 2 informs the controller 11 of the peripheral interface 4 via system bus 8 that a complete data block is present in the send buffer 16. This can occur, for example, according to DE 10 2008 009 634 by means of a first pointer after the end of the valid data yet to be transmitted and a second pointer at the end of the already transmitted data and thus at the start of the data yet to be transmitted. The controller 11 thereupon copies the data block into its buffer memory and outputs it seamlessly to the peripheral device 10 right after the preceding data block.
If, for example, in a non-real-time-capable operating system, the generating of a data block Bn+1 were delayed on account of high computing workload without the instructions to modify the output of the data stream according to the invention, so that it is not available to the controller 11 in good time for the continuation of the regular output, the control data of the last still fully transmitted control data block Bn would still be put out to the peripheral device 10. But after this, the galvanometer mirrors would be uncontrolled. They would continue their instantaneous movement according to the laws of inertia and harmonic oscillations. In this way, the galvanometer mirrors or the specimen 22 might become damaged. According to the invention, a break in the data stream put out to the peripheral device 10 is prevented by the embedded jump command JMP and the immediately following spare control data SOS. For example, spare data SOS to be put out one time could steer the galvanometer mirrors such that a noncritical specimen region or a region outside of the specimen is illuminated in stationary manner (“parking”) until regular control data Bn+1 is again available.
In an alternative embodiment (
The individual figures of
If only whole blocks (after transfer via the system bus 8) are declared to be valid, a forward jump will always be performed within a data block (
The jump commands can be nested. For example, between a forward jump command and its jump destination there can be a backward jump command with a jump destination before the forward jump command. If the next data block Bn+1 is not yet fully transmitted (
In the example of
Advisedly, but not necessarily, an emergency segment SOS with corresponding jump commands is provided in each data block Bn in a data stream divided into blocks. The particular emergency segment SOS can be located at the end or at any other place within the respective data block Bn. A second alternative embodiment can be realized, for example, when an unconditional forward jump takes the regular output past the emergency segment when there is a backward jump at the end of the regular control data to the emergency segment SOS, at the start of which there is a forward jump to the next block Bn+1 (executed by the controller 11 only when the jump destination is valid).
In further embodiments, the jump commands can involve conditions other than the validity of jump destinations. For example, in connection with an initialization command for a counter kept in the controller 11, jump commands with increment or decrement and a condition for the counter status can be realized. In this way, output loops with a predetermined number of iterations can be programmed. Such loops can contain regular control data, for example. For example, the control data for a two-dimensional image recording in an LSM could fit in only one data block by programming two nested loops for the X and the Y galvanometer mirrors.
In general, a data stream can be divided into segments of constant size, regardless of whether it is generated and transmitted block-by-block. The offset of a jump command in such cases can be indicated as the number of segments to be jumped over. In this way, the required bit width for the offset indication can be reduced as compared to an indication in bytes, for example. With a segment size of 256 bytes and a bit width of the offset indications of 8 bits, even though jumps are then only possible in steps of 256 bytes, but on the other hand over +/−127 segments instead of only over +/−127 bytes.
Modifications and variations of the above-described embodiments of the present invention are possible, as appreciated by those skilled in the art in light of the above teachings. It is therefore to be understood that, within the scope of the appended claims and their equivalents, the invention may be practiced otherwise than as specifically disclosed.
| Number | Date | Country | Kind |
|---|---|---|---|
| DE 102008049887.4 | Sep 2008 | DE | national |
The present patent application is a continuation application of International Patent Application No. PCT/EP2009/006815, filed Sep. 22, 2009, which is based on, and claims priority to, German Patent Application No. 10 2008 049 887.4, filed Sep. 30, 2008, both of which are incorporated herein in their entireties by reference.
| Number | Date | Country | |
|---|---|---|---|
| Parent | PCT/EP2009/006815 | Sep 2009 | US |
| Child | 13074581 | US |
