DSS-III Specifications and Features
-----------------------------------
(RK 4/15/97)
(rev. 4/22/97)
The following is an attempt to list some features of the new
DSS (DSS-III), along with the new UET and A/D converter. The list is
subject to change, and should be considered more as a starting point
for discussion.
DSS-III is based on two microprocessor boards (the PC44 and PC32)
working in tandem. These boards plug into the ISA bus of a Pentium
based host CPU. Some additional hardware that is external to the above
consists of (a) 25-bit DACs (b) Attenuators (c) Programmable amplifier
and filters (d) A front panel and (e) A remote control panel.
DSS-III is currently being constructed by Mike Rosing under the
direction of Bill Rhode, with the assistance of Ravi Kochhar, and with
helpful advice from the rest of the auditory group (its not too late to
get your .02 in!)
(1) Two channels of sound stimulus (possibly 4 channels (see below)).
(Choice of 25 or 16-bit DAC's)
(2) Upto 25-bit accuracy in DAC output ( >110 dB signal/noise)
(3) Programmable filters at output stage.
(10 KHz, 20 KHz, 60 KHz ?)
(4) 127 dB attenuators, settable in 1 dB steps.
(5) 2 MWord (32-bit) onboard memory for general waveform playback
(this memory is shared with the A/D converters)
(6) Two high-speed (500,000 samples/sec) 16-bit A/D converters.
(The PC32 also has two independent 100,000 samp/sec 16-bit A/D's)
(7) 16 channels of UET input (i.e. upto 16 events can be timed
simultaneously). The timing accuracy is 1 microsec (?), and times
are stored as 32-bit words (this allows timing of upto 2000 sec
long sequences at full resolution).
(8) 16 general-purpose TTL output lines for turning on lights etc.
(9) All timing done at 1 microsec accuracy with flexible 32-bit timers.
(upto 2000 sec per timer with 1 usec accuracy).
(10) Variable length rise/fall window applied to output waveforms.
(11) Programmable rate generators ?
New Features
------------
There are many capabilities of the new system that were not
available before. The following is a partial list:
(1) 25-bit DAC. This allows sound to be presented with unprecedented
signal to noise ratio. In theory 25 bits = 150 dB signal/noise, but
practical considerations may limit this to 110 or 120 dB signal/noise.
This compares very favorably with the 80 dB S/N achieved by the current
DSS (DSS-2).
In 2-tone experiments (both tones to same ear), it should no
longer be necessary to use two phones just to combine (add) the tones.
The higher signal to noise ratio should allow both tones to be combined
computationally and presented via the same phone, while still retaining
a reasonable signal to noise ratio for each tone. For minimum
intermodulation distortion it may still be necessary to use two separate
phones.
In some cases we can correct for the phone calibration
arithmetically. This is very useful in expts. where attenuator switching
transients have been a problem, or where high speed phone correction
is needed (e.g. a constant-SPL FM sweep).
Another application of this capability is to extend the
range of the attenuators beyond 127 dB.
(2) Use of 32-bit timers means that all timing will be done with
1 microsec accuracy. DSS-2 uses 16-bit timers, which allows a max. of
65 millisecs at 1 us (650 msec at 10 us etc.). This is no longer a
limitation.
(3) The timers are much more flexible. In DSS-2, we have three basic
timers, delay, duration, and repetition, plus a repetition count. In
DSS-III, one can also ask for arbitrary timing sequences, e.g. 50 msec ON,
100 msec OFF, 30 msec ON, 20 msec OFF, then repeat the whole thing N times.
(4) The 2 MWord memory for each DSS allows for larger (or more)
general waveforms to be played out. This memory is shared with the A/D
converters, so all of it is not available for waveform playback.
(5) The programmable filter stage at the output should lead to
reduction in high frequency noise.
(6) A more complete front (and remote) panel should allow for
better feedback and control. In particular, there will be knobs and
buttons on the remote panel which could be used in psychoacoustic
experiments for data entry. The design of the front/remote panels is
an area that can benefit greatly from user feedback at this time.
(7) There are 16 TTL output lines accessible from the front panel that
can be used to control lights, select speakers etc. (This is similar to what
is presently done with a separate DRV11 digital I/O board on some systems).
(8) The PC44 has four 16-bit DAC's, which can be used to produce
the stimulus instead of the 25-bit DAC's. This should make possible
certain experiments in which 4 channels of sound are necessary (or
just to save cost by eliminating the 25-bit DAC's).
(9) The new DSS/UET system should be smaller and lighter than the
current system (or at least lighter, we hope ?)
(10) The remote panel will allow manual control of parameters during
a search for units. It can also be taken into the sound proof booth to
search while watching the preparation. In addition, it could perform
as a response box during psychoacoustic experiments with the subject in
the sound-proof booth.
In addition to above, there are other features which are less apparent
but should prove advantageous in the long run:
(1) Many of the core functions (timing, stimulus generation) are
microprocessor based, i.e. they are programmable. This should lead to
easier implementation of new experimental paradigms. As an example, it
should be easy for the new DSS to play out a very long noise sample without
repeating the pattern. Another example: the DSS could monitor events
via either the A/D or UET and modify its output in near real time based
on some preset criteria.
(2) The use of the popular PC platform with an industry standard
bus (ISA) should allow us to upgrade the host hardware while still
retaining the DSS/UET. By comparison, the Q-bus has been phased out, and
it is no longer possible to upgrade the computers. It may have been
somewhat preferable to use DSP boards based on the PCI bus (mainly for
faster data transfer to/from the host computer, but at the time the
decision had to be mage (1995), suitable boards based on the PCI bus
were not commercially available.
Some Questions
--------------
While the new design has many benefits, some features of the old DSS
are either missing or implemented differently. This has an implication
for certain experiments. The following list is in the form of questions
because I am not clear how some things are (will be) implemented.
This is the right time to ask similar questions about the use of
DSS-III in your current or future experiments.
(1) A/D sampling. How are the rate generators implemented ?
For example, if we want to sample at 123400.5 samples/sec, how accurately
can rate be set ?
(2) The old DSS was designed with several discrete boards which
could function independently. In addition, the UET and A/D were
independent subsystems. The new design uses two boards (PC32 and PC44)
to implement all those functions. This has an effect on how many things
can be done in parallel. The following are some sample paradigms that
may be commonly encountered. Are all these possible with DSS-III ?
(a) The DSS produces a sound (e.g. clicks or tones) and the
response is sampled by the A/D converter. How is signal
averaging implemented ?
(b) The DSS produces a sound and the response is sampled via
the A/D converter and also the UET, simultaneously. Is
some feedback possible to the host computer for display of
intermediate results before all the repetitions are over ?
(c) The DSS monitors some expt. variable (for example, eye
position, or whether a bar or button has been pressed), and
responds with sound or lights, and also notifies the host
computer, while A/D and UET sampling is not interrupted.
(3) It should be possible to interrupt (clear) the DSS at any time
from the host computer. In some cases we need to be able to interrupt
and switch to a search stimulus to adjust the oscilloscope or electrode,
then resume from the point of interruption.
(4) How are external start/stop of the DSS/UET/A-D implemented ?
(5) General Waveforms (speech, noise) need to be played back with
a precise interval between samples. How is this done, and with what
accuracy ?
(6) Sometimes general waveforms need to be played out in succession
(e.g. to simulate a moving source). This means in practice that we
play out a different waveform for each new repetition.
(7) What is the timing accuracy (time base) of the UET ? What sort
of things (data transfer to host, new rep etc.) can make it worse ?
(8) What are the links between the PC32 and PC44 ? Are they using the
same clock to eliminate drift in UET timing ?
(9) Can we select between the 16-bit DAC's and the 25-bit DAC's
under software control ?
(10) How would the PC44 16-bit DAC's be connected to the output stage
(attenuators etc.) ? If the four 16-bit DAC's can operate truly
independently (i.e. playing out different waveforms at different rates
for different durations) then we may have to leave the possibility
open for adding extra attenuators/filters/amplifiers at a later stage,
should it become necessary to so a 4-channel experiment. Allow extra
space on the main front panel also for output etc.
(11) It is sometimes necessary to start and stop timers during data
collection. For example: Sound presentation plus UET and A/D are started,
subject presses button at arbitrary time, timer is started to trigger
some other event exactly XX millisecs after button press.
Are there timers on the PC32 or PC44 that are accessible from
the host computer, and can they turned on/off even while those boards
are busy with stimulus generation or data collection ?
(12) Intelligent UET ?
(13) Approx. cost of the system (DSS/UET/A-D + front panel) ?
With and without 25-bit DAC ?
(14) How much power can the 25-bit DAC deliver to a speaker ?
16-bit DAC ?
(15) Should there be a manual control for the filter(s) on the front
panel ?
(16) Could there be some "self test" capability built into the system
so common hardware failures could be spotted quickly ?
(17) How is "interaural delay" implemented ? How accurately can it
be specified ?
Back to Top