rick@soma.neusc.bcm.tmc.edu (Rick Gray) (06/14/90)
I am part of a lab group that uses computers for stimulus generation and sampling of neurophysiological signals. We currently use either Masscomps (Concurrent) or DEC LSI-11/23s for this. We will soon have two more setups that need to be computerized and, to keep our options open, are searching for alternative computers for this purpose. Is there hardware for Suns or other UNIX boxes that will do this? We are also hoping that something will become available for the NeXT (other than just stuff for audio). Our requirements are: a) 2 channels of d/a (3 or 4 would be best); b) 4 channels of a/d (we routinely use 5 now, and 8 would be best, but we could get by with fewer if the price was right). c) We need to be able to stimulate through the d/a (i.e., output arbitrary waveforms) and sample through the a/d (1 or more channels) simultaneously (at least within 1 clock tick). We routinely need to sample at rates between ~1 and 20 kHz, but an aggregate sampling rate of at least 100 kHz over 8 channels would be optimal. d) 12-bit a/d's and d/a's with input/output ranges of +/-10V would be best for us. e) We would prefer that the hardware came with a library of routines callable from C so we could integrate it with our current software, instead of having to re-write everything from scratch. f) We would prefer a UNIX-based computer for this. All of our current computers are networked (TCP/IP) and we are used to NFS-mounted directories, rlogins, and X-terminals. g) Most of our data is made up of 1-2K-point transients, repeated once per second (or slower). A typical trace would consist of a 1024- point stimulus output through a d/a channel and simultaneous sampling of 1 or 2 1024-point traces through the a/d, all at 10 kHz. We occasionally have need to sample up to 64K points with no gaps. Hence, our real-time constraints are somewhat soft; sampling rate must be very accurate during a trace, but we can accept a little slop in the inter-trace interval. h) Cost per setup of around $12-15K More on our current systems: Masscomp 5400s (68020, Real Time UNIX) with their EF12M module which has 8 channels of differential a/d (16 single-ended), 2 channels of D/A, 4 programable clocks and a 16-bit parallel I/O channel. It will stimulate/sample at up to 333 kHz. There is a very nice library of routines to talk to the hardware, callable from both f77 and C. They're nice machines, but a little expensive. Each Masscomp has 2 EF12Ms and can run 2 setups simultaneously. Graphics display is either with Masscomp's graphics or NCD X-terminals. The data/acq specs are similar on the 11/23s (running RT-11, with all its limitations). If we could get something similar running on a different computer that is less expensive we would be very happy. We know there are single-user Mac and PC-based systems that will do this, but we prefer to stay in the multi-user, multi-tasking UNIX domain. thanks, Rick Gray uucp: uunet!soma!rick Division of Neuroscience Internet: rick@soma.bcm.tmc.edu Baylor College of Medicine Houston, Tx 77030 Articles to: concurrent@soma.bcm.tmc.edu or uunet!soma.bcm.tmc.edu!concurrent Administrative stuff: concurrent-request@soma.bcm.tmc.edu Stan Barber, Moderator