figures are approximate). Assuming a random occurrence of events on all channels, the time to perform all operations is 8*1.8+9.35 ~ 24 microseconds. This provides an aggregate counting rate on all channels of 42 KHz. This is not a strict bound; if the events do not occur randomly, the polling loop may execute consistently above or below the average figure with different performance characteristics. However, polling is clearly the limiting time-consuming process and any general performance increase will be gained only if the polling time is decreased.
ADDITIONAL PROBLEMS
1. Design a solution that has a polling process consisting of a branch tree using combinational circuitry at the branch points. That is, a D network would carry out the computations at each of the branch points.
a. What effect does this have on the system performance?
b. Design the polling network so that it polls in a more equitable manner.
c. Write the initialization subroutine to set the event counts and time base.
2. Write a software, n-channel EPUT meter for your favorite computer. What are the performance characteristics of the program?
3. Design an output process for the n-channel EPUT meter; consider using BCD for the output data. How does this process affect the design, its accuracy, maximum frequencies, etc.?
4. Modify the EPUT meter of solution 1 to have independent time base generators for each channel. Assume a single K(clock) and that the values of the time bases are stored in a memory array. Give the performance characteristics of the system.
5. How would multiple DMgpa's affect any of the designs above?
6. Design a polling scheme which will find the selected bit in a constant time by using the binary searching technique. With this technique, the control first examines whether a flag is on in the first 1/2 of the word or not. It then proceeds to subdivide the search space in two each time. Such an approach can be done using either a combinational or register transfer approach.
KEYWORDS: Display, event, histogram, initialization, synchronization, record, waveform analysis, arbiter.
A histogram recorder is a device, principally a memory, which records an occurrence count for each member of a set of discrete events. For the purposes of this problem, an event of this form at a particular time, ti, has a 10-bit value, and is denoted x(ti)<9:0>. A record is defined as the occurrence count, taken over a finite time period, for an event.
An example of a histogram is given in Figure Hist-1. The 11 events represent all of the possible outcomes from a single throw of a pair of dice. The values stored in the histogram are the records for each possible outcome over some time period, e.g., 1000 throws. Histograms are usually presented by such a visual display, hence the recorders described below will have display capabilities.
If the set of events contains M elements, the histogram recorder can be implemented as a memory of M words. The records for events E[0:M-1) can be. uniquely assigned as the value of memory words MW[0:M-1].
209