Digital PDFs

XX-A5814-48

2000

4 pages

Original

0.5MB

Document:	SKED systems: Paper tape, Tennecomp, and 0S/8
Order Number:	XX-A5814-48
Revision:
Pages:	4
Original Filename:	10.3758%2FBF03201748.pdf

OCR Text

Behavior Research Methods & Instrumentation
1976. Vol. 8 (2), 76-79

SKED systems: Paper tape, Tennecomp, and OS/8
KEN STEPHENS and AMY BARMEIER
Western Michigan University, Kalamazoo, Michigan 49008

The SKED system has been developed at different levels of complexity and power for differing
hardware configurations. The simplest of these, and the least expensive, is the paper-tape SKED
system. Some of the drawbacks associated with the use of paper tape are absent in the Tennecomp
magnetic tape cartridge system. Users with mass storage capabilities can make use of the OS/8
operating system (copyright DEC) to speed up all aspects of the SKED system (especially editing,
compilation, and loading of programs), and take advantage of a file-oriented system structure for data
storage and analysis.
SKED is a process control language for precise control of psychological experimentation. It is specifically
written for any model of computer from the PDP-8
family, 1 and exists in several forms for differing configurations of hardware options and software support systems. Hardware differences can be categorized according
to processors, peripheral devices, and Interfaces.
Versions of SKED have been written for the PDP-8 and
the PDP·12 to accommodate as little as 4K of core
memory or as much as 32K.
Peripheral options include input/output (I/O) devices
and mass storage devices. The minimum configuration
must have some method of entering programs to core,
such as low-speed Teletype paper-tape reader, some
method of interacting with the programs, such as the
Teletype, and some method of producing output, such
as low-speed tape punch. However, other peripheral
devices can greatly facilitate each of these three I/O
functions. For example, a high-speed paper-tape reader
and high-speed tape punch greatly speed up the input
and output, and other terminals such as the DECwriter
or CRTs improve the interactive I/O functions. The use
of a Tennecomp cartridge system? as a peripheral is one
step ahove a paper-tape configuration in terms of speed
of interactions. The Tennecomp may be used as an input/output device in a manner directly analogous to
paper-tape peripherals. Input programs may be read
from the Tennecomp, and output (data, state tables)
may be written directly onto the cartridges. A variety of
mass storage devices have been adapted to the PDP-8
family, and all of these may be used with SKED to improve the program storage and loading processes. These
storage devices include disk systems (RK8 and RK8E
from DEC and the Diablo disk have all been used),
floppy disks (from DEC, Sykes, or Xebec), and magnetic
tape (the DEC TD8E and TC08 DECtape systems).
A critical feature of any system is the interface connecting the processor with various possible experimental
This work was supported in part by Research Scientist
Development Award K2-MH-70483 from the National Institute
of Mental Health.

stations. State Systems, Inc." markets a low-price interface specifically designed for use in conjunction with
SKED. The DEC UDC and DK8E interfaces are also
good general purpose interfaces which can be adapted
for use with a SKED system. It is also possible to
construct an interface out of standard digital logic
equipment.
SKED software varies accordingly to accommodate
each of the possible hardware configurations, but always
consists of a SKED compiler and a SKED run-time
system (RTS). Other support software, such as an editor
program, is necessary for all systems. However, for
sophisticated hardware configurations such as those
which include a mass storage device, operating systems
and time-sharing systems (e.g., OS/8 and RTS-8 1 ) add
power and convenience for the SKED user.
A BEHAVIORAL ANALYSIS OF PAPER-TAPE,
TENNECOMP, AND OS/8 SYSTEMS
Dingler, Kadden, and Snapper (1975) briefly described the steps necessary for operation of the SKED
system, which are: state diagramming the experimental
procedure, translating the state diagram into a state table,
creating an ASCII version of the state table using an
editor program, compiling the ASCII state table and production of a binary tape, debugging the state table, and
loading the binary into, and operation of, the run-time
system.
Although these six steps are necessary for all hardware configurations, the operations are different in
terms of response effort for the user and immediacy of
reinforcement, or time taken by the computer system to
complete each operation. In this paper we shall compare
paper-tape, Tennecomp, and OS/8 systems in terms of
these critical user variables.
Generation of State Graph
The first step toward ultimately running on the
SKED systems is to translate a verbal description of an
experimental procedure into a state graph. The nota-

SKED SYSTEMS
tional system used for this is identical for paper-tape,
Tennecomp, and OS/8 systems (Lee, Stephens, & Duncan,
1974). In all cases, the final product is an unambiguous
diagram descriptive of the procedure to be implemented.
Translation of State Graph into
Linear Computer Format
Once the experimental procedure has been correctly
diagrammed, the state graph must be translated into a
linear format understandable by the computer. This is
accomplished by using an editor program, and the resulting product is referred to as a state table. The
behavior of typing in the state table will be equivalent
for all three systems: however, they differ in the speed
required to input the editor program into core and output the ASCII product of this typing. The paper-tape
system requires that a paper tape of the SYMBOLIC
EDITOR be read in, either with the low-speed Teletype
reader or a high-speed reader, and then started at 0200,
which requires some interaction with the switch register
of the PDP-8. The output must then be punched, either
on the low-speed Teletype punch or a faster tape punch.
With Tennecomp, the SYMBOLIC EDITOR is read in
from a Tennecomp tape cartridge, which requires about
the same amount of time as the high-speed tape reader.
Then the output is recorded onto one track of an empty
tape cartridge. Both of these operations with Tennecomp require less than a minute. With the OS/8 system,
the user has a choice of editor programs: a file-oriented
version of SYMBOLIC EDITOR, called EDIT, and the
programmable editor language TECO, which offers
greater power and speed to the accomplished programmer. The editor of choice is simply called into core
from the mass storage peripheral device by a command
to the OS/8 keyboard monitor (e.g., R EDIT), and a
command to the OS/8 command decoder then specifies
the name of the input me (if there is one) and the name
of the output me (i.e., *TABLEl.PA < TABLE.PA).
With the concise command language (CCL) option, both
of the preceding commands can be given at once (e.g.,
EDIT TABLEI.PA < TABLE.PA). Depending on which
peripheral devices are used, this step can take from 2 to
20 sec.

machine language object tape (binary) is punched. This
entire process can take up to 10 min, depending on the
length of the state table.
The compiler may also be stored directly on a Tennecomp cartridge and read into core from the cartridge.
Once the compiler has been read in, the cartridge containing the state table is read into the computer. The
user must then enter the number of counters he requires
and depress the "CONTINUE" switch to initiate the
second reading of the state table. The output from the
compiler, if one is using a Tennecomp system, is a binary
paper tape which will later be loaded into the RTS and
the two saved together on a tape cartridge. Thus, the
major advantage of this system over a paper-tape system
concerns the time saved when reading the compiler and
state table into core. With the OS/8 system, the user
calls in the compiler (e .g., R OSCOMP) and specifies the
input file (the ASCII state table) or the device on which
input is to be found, and the output file (the binary object file) or the device to which the binary will be output
(i.e., FILE.BN < FILE.PA, FILE.BN < PTR:, or PTP:
< FILE.PA). The number of counters does not have to
be specified with OSCOMP if the correct number follows
the first "=" in the source file. Compilation with OS/8
SKED, then, requires a minimum of user interaction and
takes only a few seconds.

Debugging
From experience, we have determined that it is the
exception, rather than the rule, for an ASCII state table
to be error free. Each of the SKED compilers, upon encountering an error in syntax or format, informs one of
this error by typing on the Teletype both the line of the
ASCII state table in which the error was found and a
four -digit error code. There are error-code diagnostic
lists for determining the exact nature of the unacceptable ASCII input. This process is identical for the three
SKED systems, but the procedures involved in correcting
the error(s) are markedly different.
The differences between a high- or low-speed papertape reader becomes critical at this stage if one is using a
paper-tape system, since debugging the state table is
largely a matter of input/output time. If an error was
found by the compiler, one must read in the editor proCompiling
gram, read in the ASCII state table, make the necessary
Once an ASCII version of the state table has been corrections on the state table, punch a new source tape,
produced, this state table must be compiled into a read in the compiler, run the two passes, and punch a
binary program. Although there is a version of the SKED new binary object tape. This whole process can be excompiler program available for each of the three tremely time consuming because of the time necessary
systems, the function of these compilers is comparable to read in each of the programs and can take up to 30
and, in all cases, a binary program is the desired output. min.
With a paper-tape system, a tape of the compiler must
The sequence of behaviors required to correct state
first be read into core using a paper-tape reader. A high- table errors are identical for a Tennecomp system. Howspeed reader will greatly speed up this process. The ever, with this system, the input/output time for the
compilation consists of a two-stage procedure, thus the editor and compiler programs is considerably reduced
ASCII state table from the editor program must then be because of the increased speed of Tennecomp cartridges
read into core twice. After the first pass, the user enters over paper-tape readers, thus somewhat reducing the rethe number of counters required for his experiment. sponse cost involved.
After the ASCII tape is read in for the second pass, a
When the OS/8 SKED compiler finds an error, it

STEPHENS AND BARMEIER

prints the same type of error diagnostic as the other systems, and then exits to control by the keyboard monitor
for a new EDIT or TECO command, with specification
of input and output meso However, if either editor program was used immediately before compilation, the me
specification step will automatically be set up, as CCL
will recall the latest used me arguments, just by typing
"EDIT" or "TE." Then the correction can be made
quickly and control returned to keyboard monitor for
the R OSCOMP step. In most cases, this will occur in less
than a minute after the first error was found.
Loading and Other Functions of the Run-Time System
Once an error-free binary version of the desired state
table has been produced, one must first load the appropriate SKED run-time system into core and then load
the binary program for the desired experimental
chambers. This is an extremely tedious process on a
paper-tape system, since the run-time system may take
30 min to read in at low speed. There always exists the
possibility of a checksum error and, if this occurs, the
entire program must then be read in again. The process is
somewhat facilitated if the input device is a Tennecomp
cartridge, due to the increased speed of this device over a
low-speed reader. With OS/8, the run-time system and
state table binaries exist as files on the systems device
and can be quickly and easily called in. In addition, for
the user who regularly uses one set of state tables, these
can be loaded together with the RTS and saved, and
later called into use as one program.
DATA OPERATION
Data CoUection
When the state tables are compiled, memory locations
are set aside as counters. The strategies for data collection with the SKED system have been described extensively elsewhere (Hamilton, 1975), but these will not
vary between the various hardware configurations. The
method of dumping data from core does differ, depending on the transfer medium available. With the simplest
systems, data must be punched on a low- or high-speed
tape punch, which is time consuming. This dumping can
be done by manually typing P and turning on the punch,
or by leaving the punch on and having DODUM statements in the state table determine when the dump takes
place. With Tennecomp, the process is similar, except
that the statement in the state table which can automatically cause the dump is called TODUM, and it can only
dump four stations at a time. The OS/8 run-time system
dumps data into an output me which is stored on a
peripheral storage device. The dump can be initiated
manually, by a DODUM statement, or (at the user's
option) when the last active station finishes, it may
dump automatically.
Data Storage
Once data have been conveniently collected by the
SKED run-time system, provisions must be made for

their failsafe storage. Paper tape is an extremely poor
method for long-term storage, as it is subject to the
ravages of coffee spills, poor folding, inaccurate labeling,
or disastrous rips. Additionally, if an experiment has
been in process for many months and each day's data
meticulously saved, the experimenter may find his
laboratory piled high with reams of paper tape.
The use of a Tennecomp cartridge for a long-term
data storage also is far from optimum practicality. Although Tennecomp Systems makes cartridges ranging
from 4K to 64K, these cartridges generally contain only
four tracks. Most experimenters find it convenient to
store 1 day's data in each of the four tracks, and thus an
experiment lasting for 1 year would involve the use of
nearly 100 Tennecomps for data storage alone! These
tapes are also vulnerable to the same sorts of unplanned
physical accidents as paper tapes. In the OS/8 system, a
great deal of data can reside on the mass storage
medium. However, even with these capabilities, disks
and DECtapes fill up with ASCII data. A more convenient data format has been developed which takes
only one-fourth as much space as ASCII format. The use
of this stripped-down ASCII, or integer format, requires
special programs for transfer, and translation back to
ASCII for data analysis purposes. In the long run, it is
best to transfer data to a permanent storage facility on a
larger computer system, since all magnetic storage media
are subject to their own flaws, such as loss of the format,
physical damage, etc., which can cause loss of all files on
the structure.
Data Analysis
The differences in ease of data analysis routines for
paper-tape, Tennecomp, and OS/8 SKED systems are
due largely to the discrepancies of input/output times
for the data and the analysis programs associated with
each system. For example, a paper-tape system requires
that the FOCAL (1969) language program be read into
core to begin. The experimenter must then read in paper
tapes of his specific analysis program and data. All results of the analysis are subsequently punched or sent
out via the Teletype. In any event, response cost involved in the entire procedure is high, laborious, and time
consuming. Availability of either a high-speed reader/
punch or a Tennecomp cartridge markedly accelerates
the entire procedure by reducing the I/O times.
Although this excess time is somewhat eliminated, the
experimenter is still forced to engage in the chain of
behaviors. There are several data analysis options open
to the OS/8 user: a special version of FOCAL, BASIC,
FORTRAN II, and even FORTRAN IV. These languages can all be stored on the systems device and called
into core by keyboard commands. Then the specific
analysis programs written in these languages can be
called up and run with all the core-saving options of
internal chaining. Of course, without the proper timesharing software, this analysis must be done while the
computer is not controlling experimentation. Some users
prefer to deposit the data on some portable storage

SKED SYSTEMS
medium, such as DECtape or floppy disk. and then
transfer for analysis on a PDP·I I or larger computer.
However, the new time-sharing SKED system (Snapper,
1975) permits analysis of data via BASIC while
experimentation is ongoing.

at least one peripheral storage device and the support
software from DEC, it is more expensive, but the added
power and flexibility make it a much more attractive
system than paper-tape or Tennecomp systems.

REFERENCES

CONCLUSION
The preceding discussion has presented a behavioral
comparison of paper-tape, Tennecomp, and OS/8 SKED
systems regarding the responses necessary to design an
experimental procedure, implement the procedure on a
SKED system, and collect and analyze the actual data
produced by any experimental subject. It is obvious that
the highest response cost as reflected by actual time
spent is associated with a low-speed paper-tape system.
The availability of a Tennecomp device reduces a great
deal of the input/output times for various operations,
but does not actually alter the number of significant experimenter behaviors required. There is a significant response cost involved in learning the OS/8 system, but
the advantages of having a file-oriented system seem to
outweigh this factor. Because the OS/8 system requires

DINGLER, R., lUDDEN, R. M., & SNAPPER, A. G. An
introduction to state notation and SKED. Behavior Research
Methods & Instrumentation, 1976, 8, 69-72.
LEE, D. M.,
STEPHENS, K. R., & DUNCAN, M. D. The
notational system. Kalamazoo, Mich: SKED Users Group,
1974.
HAMILTON, B. E. Data-collection strategies in SKED. Behavior
Research Methods & Instrumentation, 1975. 7, 243-247.
SNAPPER, A. G. Time-sharing SKED and OS/8: A new version.
Behavior Research Methods & Instrumentation, 1976, 8,
80-82.

NOTES
1. Copyright Digital Equipment Corporation, Maynard,
Massachusetts.
2. Copyright Tcnnecomp Systems, Oak Ridge, Tennessee.
3. State Systems, Inc., P. O. Box 2215, Kalamazoo, Michigan.