UET considerations
 
                                                             Ravi Kochhar
                                                             Aug.  2,1996
 
        The following is a list of features of  the  present  Unit  Event
   Timer  (UET) and some others that may be desirable in the new version.
   Most of these ideas are not mine, but suggested by Bill.
 
        The present UET (built by Dick and Dan) has the following notable
   features:
 
         (1) Upto 16 input channels (three of which are reserved
             for special functions such as start/stop)
 
         (2) Timing of all channels simultaneously to 1 microsec
             accuracy (not sure if this is strictly true).
 
         (3) UET clock is synchronized with DSS clock to avoid drift
             in cycle histogram computation.
 
         (4) 32k word fifo to offload CPU.
 
         (5) Software or hardware start/stop. The latter is selectable under
             computer control from between two or more inputs.
 
         (6) 24-bit event time accuracy.
 
         (7) A user settable time base (1,2,5 etc. microsecs).
 
        For the design of  the  new  (PC-32  based)  UET,  the  following
   additional features should be considered:
 
         (1) A 32-bit event timer. This would allow longer spike trains
             to be recorded with full 1 usec accuracy. Present 24-bits
             limits us to 8 secs. max. at 1 usec.
 
         (2) An "intelligent" UET. This can take many forms (listed in
             increasing order of complexity):
 
            (a) A continuous monitoring of the "signal/noise ratio" in
                addition to recording of (already discriminated) event times.
 
            (b) A variable and perhaps automatically adjustable threshold
                level.
 
            (c) Automatic spike discrimination from the analog signal.
                A desirable extension is to automatically adjust for
                waxing and waning of spike amplitude, which is often the
                case in real experiments.
 
            (d) Automatic spike discrmination and separation where multiple
                units are picked up by the same electrode.
 
            (e) Automatic spike discrimination and where multiple units are
                recorded by multiple electrodes.
 
        Most or all of this intelligence in the  UET  would  have  to  be
   implemented  via  the  on-board  A/D converter on the PC-32 board.  It
   remains to be answered which of the above tasks can be accomplished by
   the PC-32.
 
        One important restriction is that the A/D may only  sample  at  a
   max  rate  of  100  KHz.   If intelligent spike disrimination is to be
   implemented, this may limit the accuracy of the time base, as well  as
   the number of channels sampled simultaneously.
 
        It is also not clear whether the PC-32 has enough  compute  power
   to  do spike discrimination in (near) real time, a non-trivial problem
   (can we make use of the work of John Oghalai and others ?)