Advantages of CSL Hardware

Advantages of CSL, Model 4300B, Hardware

An acoustic analysis system requires a powerful hardware environment with specifications suited to the analysis tasks. CSL, Model 4300B, provides a robust, professional-level, no-compromise hardware platform, integrated with the software, to include reliable signal acquisition, analysis, and playback. It provides 16-bit input, analog-to-digital conversion (A/D), 16-bit digital-to-analog (D/A) output, sophisticated signal conditioning and anti-aliasing filters to take full advantage of the 16-bit signal representation, a wide assortment of frequency ranges, and powerful digital signal processing circuitry for fast signal analysis. Unlike systems built around generic plug-in multimedia sound cards (which were designed primarily for sound output, not input), CSL is well-tailored for professional speech processing requirements with fully integrated hardware and software. CSL offers input signal-to-noise ratios typically 20-30dB superior to generic plug-in cards. Some features of special interest are described below.

External Module

The CSL hardware consists of two major modules: an external module with user controls and an internal plug-in printed circuit board installed in the host computer. These two modules are connected with a cable which transmits the high-speed digital sampled data. The external module provides convenient access to input/output controls and jacks (including XLR microphone input). The external module also performs the initial signal conditioning and analog-to-digital conversion. The external module isolates the low-level (e.g., microphone) analog input signals from common sources of noise in PC-based sampled data systems: the switching noise of digital signals, power grounding, and unshielded power supplies.

Input

The CSL includes two- or four-channel inputs, adjustable attenuation (input sensitivity) over a 70dB range, preamplification and a dual-channel 50 kHz, 16-bit A/D converter. The CSL analysis software can import and analyze signal data with any sampling rate, and the hardware can capture or speak this data over a wide variety of sampling rates (2500 to 51200 Hz, in 25 Hz steps). The input and output digital anti-aliasing and analog anti-aliasing filters (120dB/octave roll-off) are automatically selected and track with the sampling rate selected.

The CSL includes a hardware option for four-channel input. This feature may be particularly useful for researchers who wish to view the analysis of multiple signals simultaneously.

Output

The system includes complete output signal conditioning for CD-quality output, 16-bit resolution, and a wide variety of output sampling rates. Appropriate output filters are selected automatically for all sampling rates.

DAT Interface

The external CSL, Model 4300B, module has the option for a Digital Audio Tape (DAT) interface port. This feature allows "pass-through" of DAT recordings directly to the computer’s digital memory without resampling. A pull-down menu selection is provided to low-pass filter and downsample DAT signals for reduction to frequency ranges typically used for spectrographic analysis.

XLR Input

CSL, Model 4300B, includes a professional-level XLR-type microphone and input connector. XLR is a standard in professional applications. It includes balanced input with extra grounding and defined polarity. It provides extra noise immunity from inductance noise and capacitance discharge.

Digital Signal Processors

Once digitized, the input signals are sent to the CSL internal board in the host PC. This board performs high-speed signal processing using two digital signal processors. These DSP processors speed up processing significantly as compared to systems dependent on the host processor for all analysis.

Real-Time Capable

Biofeedback applications (language training and clinical therapy) require programs and hardware which are capable of true real-time operation. The CSL hardware is designed to accommodate this need. Many CSL programs provide unique clinical tools using the real-time capability of the CSL hardware.

The external module with its high-speed, multi-channel, 16-bit A/D and D/A, plug-in board with two high-speed digital signal processing integrated circuits, high-quality microphone, studio-quality speaker and headphones, and the feature-rich software, all provide an effective professional-level environment for speech acquisition, processing, and playback.

NCVS Standards for Voice Analysis

The National Center for Voice and Speech (NCVS) brought together leading voice scientists at a workshop in February 1994, in Denver, in an effort to standardize and improve the acquisition and analysis of voice signals. The published proceedings: "Workshop on Acoustic Voice Analysis" should be read in their entirety by anyone interested in voice analysis (contact NCVS at 319-335-6600). Presented here is a brief analysis regarding CSL, Model 4300B, compliance with the recommendations set forth by NCVS. (The recommendations column includes the paragraph reference designators used in the report.) CSL is substantially better suited to voice analysis than any other commercial system.

NCVS Recommendation	CSL Compliance
"It is useful to classify acoustic voice signals into three types... A spectrogram ... is useful in making the classification." (A1.)	CSL provides robust spectrographic analysis. The user can adjust numerous spectrogram features including filter bandwidth (nine selections), gain, dynamic range, etc.
"All utterances may be customized to an individual’s Voice Range Profile (VRP). This VRP should be obtained first to establish the bounds for further testing." (Test Utterances)	CSL has an option for plotting Voice Range Profiles. Dynamic range, pitch tracker range, and automatic precision attenuation provide sufficient dynamic range (>100dB) for this task.
"A professional grade condenser microphone (omnidirectional or cardioid) with a minimal sensitivity of -60dB should be used." (D1. a.)	CSL is delivered with a professional-level dynamic microphone. However, all voice packages include a condenser-type microphone which complies with the recommendation (i.e., AKG Acoustics’ Model C-410).
"A miniature head-mounted microphone is recommended." (D1. b.)	Microphone, cited above, is head-mounted and can be positioned as recommended by NCVS.
"A 16-bit A/D converter or DAT recorder is recommended, but this must be accompanied by conditioning electronics (amplifiers, filters) that have signal-to-noise ratios in the 85-95dB range." (D1. e.)	CSL includes S/N ratios in excess of 86dB and includes a direct DAT interface. Real-time overload indications and adjustment allow full use of dynamic range.
"Sampling frequencies of 20-100 kHz should be used, depending on the degree of interpolation between samples that the analysis software provides." (D1. f.)	For voice analysis, the CSL option, MDVP, uses 50 kHz sampling and 20:1 interpolation for complete compliance. A wide range of sampling rates is available.
"Manufacturers of workstations for acoustic analysis should be encouraged to provide DC coupling and low-frequency fidelity. For DC coupling, there should be minimal drift and drift should be reported and calibratable." (D2.)	CSL offers switchable AC or DC coupling for all channels. DC drift is very small. CSL autocalibrates for DC drift when loaded. It can also report and calibrate when desired.
"For all input signals real-time feedback for clipping should be provided to avoid overloading the A/D converter." (D2.)	CSL provides real-time feedback on panel and screen. Overloaded samples are flagged and color-coded in waveform display. Voice analysis routines will not analyze overloaded signals.
"Line-level input (on the order of a few hundred millivolts) should be provided as a direct interface to the output transducers, so that the expense of analog preamplifiers can be bypassed." (D3.)	CSL has high-gain professional-level preamplifiers (1 to 200 millivolt peak-to-peak) which comply with recommendation. Overload indicators are linked directly to preamplifier to detect overloading during preamplification.
"A digital audio tape (DAT) recorder should be used to store signals, unless A/D conversion is directly to the computer." (D4.)	CSL can record directly to computer. CSL can also directly (i.e., digitally) interface to DATs for recording and playback.

Summary of CSL, Model 4300B, Hardware Benefits

"A cheap 16-bit sound card...should sound exactly the same as the DAT and DA-88 recorders used in Hollywood. But in reality, you won't get more than 10-bit performance (about 60 dB dynamic range) from the card. The rest are 'marketing bits': They look good described on the package, and they're really there on the chip, but you can't use them. That's because good analog circuits are too expensive for a cheap card. Besides, even the best analog components are comprised by all the electrical noise on a computer motherboard." (Jay Rose. Digital Video, OIctober 1998)	Feature	Benefit
	Direct and easy access to controls before, and *during*, capture and playback	The external module allows the user to easily adjust levels *during* input and output. This avoids the trial and error approach of sound cards only controlled by software. Overloading is also easily monitored and adjusted during operation. Connectors and controls are easily accessible.
	Integration of software and hardware	There is no need to load a different program to control the hardware because all functions of the hardware are directly accessible within the CSL software applications.
	Superior S/N specifications (>85dB) due to isolation of analog input/output from computer	CSL signal quality is far superior to virtually any other PC-based signal acquisition system. The user is assured of accurate noise-free signal acquisition.
	XLR input for microphone. XLR is a balanced input with specified polarity and extra grounding	XLR also helps protect the microphone signal from noise and capacitance discharge. The advantages of XLR input have made it a standard for music applications.
	Digital signal processors on the CSL plug-in card	Allows real-time signal analysis during signal acquisition. Also significantly speeds up most signal processing.
	Circular buffers (CSL records continually while overwriting the oldest data points)	When recording, you don’t want to run out of memory just as you hear the signal of interest. Circular buffers allow you to continually record and decide *after* you’ve heard the desired signal to stop acquisition.
	Wide selection of sampling rates (2500 to 51200 Hz in 25 Hz steps)	Any audio bandwidth signal can be analyzed with the desired sampling rates. Comparisons to methods in the professional literaturecan be exactly replicated.
	High-order, low-pass, anti-aliasing filters with sufficient roll-over to prevent aliasing	These filters make sure that the acquired signal does not contain alias components. A frequency analysis of a stored signal is, therefore, accurate.
	Automatic selection of input and output anti-aliasing filters	You are assured of accurate signal conditioning with any of the 1925 different sampling rates.
	Autocalibration for DC offset during program-loading	Autocalibration minimizes DC drift, and ensures accurate signal acquisition and full use of dynamic range.
	Real-time capable with signal analysis *during* data capture	Speeds up all operations. Real-time is needed when analysis is needed for biofeedback, therapy, or monitoring.
	Real-time flagging of overloading on both amplifier and preamplifier	As you acquire data, you instantly see overloading (i.e., clipping) so that the level can be adjusted accordingly.
	High-gain preamplifier with overload flags	Eliminates the need for external preamplifiers while minimizing overloading.
	AC or DC coupling on input on all channels	DC coupling is needed for some signals (e.g., EGG, airflow).
	Software or front panel control of input and output levels	In some applications, the software will control and adjust calibrated input attenuation. In other applications, the user can adjust on front panel.
	Calibrated signal input level	Software can query input settings of calibrated inputs to allow absolute measurements of signal level.
	Four-channel input	Many applications require multiple-channel simultaneous acquisition and analysis.
	DAT pass-through	Allows direct exchange of digital signal information without the need for resampling.

Kay also has a less expensive CSL hardware system, Model 4150.