🚀 go-pugleaf

In article <1992Dec21.065030.17772@cronkite.ocis.temple.edu> john@sg25.aud.temple.edu (John W. Schwegler) writes:
>
>What sort of throughput are you looking for? I'm currently using an
>IRIS 4D/35G for real-time acquisition and processing (soon to move to
>an Indigo R4000/XS24Z), with rates of about 200 kpoints/sec. I use three
>different processes with non-degrading priorities, and it works very well.
>

We have a 464-element silicon photodiode array connected to a 512-channel
A/D converter system. We use this system to record evoked brain activity
using voltage-sensitive dyes (=potentiometric probes). The system is
working EXTREMELY well :-)

The A/D converter system uses two 16-bit A/D units that operate in tandem.
Basically, we DMA pack a 32-bit word on a Motorola VME147 computer (68030-based)
with two 16-bit data points every 6.5 us. (we can bump it to 5.5 us 
under certain conditions). That means we can scan all 464 elements (plus
8 "electrical channels") about every 1.5 ms with a clock speed of 6.5 us.

The operating system we are using is Motorola's VERSAdos -- a REAL real
time operating system. VERSAdos is the "pits" but fast. Unfortunately
Motorola doesn't support it anymore. If my memory serves me, we can
acquire 472 channels x 576 points/channel in about 0.8 s. If my math is
correct, that about 340 kpoints/sec (16-bit points). That's pretty close
to your 200 kpoints/s (assuming your points are 16-bits ;-) We can also
write at this speed directly to a SCSI disk. 

We have an R4000 upgrade on order. It hasn't arrived yet either. Maybe
we'll look into it as a data aquisition system without a "co-processor"
card. 

Thanks.

On each clock "tic", we load a 

-- 
David M. Senseman, Ph.D.     | A man who has never gone to school may steal 
(senseman@lonestar.utsa.edu) | from a freight car; but if he has a university
Division of Life Sciences    | education, he may steal the whole railroad.
UT San Antonio               |             Theodore Roosevelt (1858-1919)