home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.unix.internals:2057 comp.unix.questions:14708 comp.unix.wizards:5239
- Newsgroups: comp.unix.internals,comp.unix.questions,comp.unix.wizards
- Path: sparky!uunet!netnews!humesdg1
- From: humesdg1@netnews.jhuapl.edu (Dave Humes)
- Subject: Realtime data acquisition under UNIX?
- Message-ID: <BzBGBG.Axw@netnews.jhuapl.edu>
- Summary: To what extent is UNIX capable of realtime processing?
- Keywords: realtime UNIX
- Organization: JHU/Applied Physics Laboratory
- Date: Tue, 15 Dec 1992 19:47:40 GMT
- Lines: 39
-
-
- We have a requirement to process analog data on an existing UNIX system in our
- laboratory. Continuous, gap-free acquisition is required, thus we have a
- realtime processing requirement. The A/D board has dual 64K X 16 bit buffers.
- One fills while the other is available to the host. An interrupt signals buffer
- full and data starts filling in the other buffer. During the interval between
- interrupts, the data must be moved from the A/D buffer to a disk file.
-
- The computer we are using is a Solbourne 5/602 (2 cpu Sun/SPARC) clone. It is
- based on a propietary bus (Kbus), but includes a 6U VME subchassis where we
- intend to install the A/D board. The VME bus is accessible to the host through
- the system I/O board which acts as a bus translator. The operating system is
- Solbourne's OS/MP 4.1.A3, a multiprocessing variant of the Sun operating system.
-
- The worst case occurs while processing one channel of analog data at a sample
- rate of 100KHz. At this rate and using an input buffer of 64K X 16 bits, an
- interrupt will occur every 0.66 seconds. So, every 0.66 seconds we must get
- 128 Kbytes of data out of the A/D, probably by doing a DMA from the input
- buffer to memory on the Kbus, and from there it must be appended to a file.
- The input buffer is mapped to VMEbus addresses.
-
- I realize that it is almost impossible to answer this question without actually
- trying it, but since that requires buying hardware, we'd rather not take that
- approach. I also realize that a number of realtime executives exist to solve
- this problem, but we'd rather not take that approach either if it is not
- strictly necessary. I would greatly appreciate hearing from anyone who has
- actually attempted realtime data acquisition under UNIX. Thanks.
-
- --
-
- -------------------------------------------------------------------------------
- David Humes | Johns Hopkins University Applied Physics Laboratory
- (410) 792-6651 | humesdg1@aplcomm.jhuapl.edu
- -------------------------------------------------------------------------------
- --
-
- -------------------------------------------------------------------------------
- David Humes | Johns Hopkins University Applied Physics Laboratory
- (410) 792-6651 | humesdg1@aplcomm.jhuapl.edu
-