AN2384 Freescale Semiconductor / Motorola, AN2384 Datasheet - Page 10

• Manufacturer Part Number: AN2384
• Description: Generic Tone Detection Using Teager-Kaiser Energy Operators on the StarCore SC140 Core
• Manufacturer: Freescale Semiconductor / Motorola

DTMF Detector on StarCore
The implementation of the TK algorithm for DTMF detection (see Figure 5) requires additional processing to pass
Q.24 requirements. The algorithm is based upon the following principles:
4.2 Adaptations to Recommendation Q.24
This section covers issues pertaining to frequency, including power level per frequency, tolerance, power level
differences between frequencies, signal reception timing, and signal and pause duration.
4.2.1 Power Level Per Frequency
The ITU-T Q.24 standard recommends that a key with a power level per frequency between 0 dBm0 and –25dBm0
must be recognized. The power given in dBm0 is obtained in the following equation [6]:
P
range [–1, 1]).
The AGC is calculated at the beginning of sample processing by finding the maximum sample value to be
processed and then selecting the proper value for the gain. The amplitude modulation introduced by this process
between consecutive sample blocks causes a transient period that is negligible when the sample block is large
enough (about 40 samples, which correspond to 5 ms of data at a sampling rate of 8 KHz). Without the AGC, 16-bit
rounded sample values limit the detector to a power level per frequency of about –18 dBm0; with the AGC, the
power level per frequency can be smaller than –25 dBm0 (value recommended in the Q.24 norm).
10
t
is the average power of a linear PCM signal computed in fractional format (that is, with normalized values in the
•
•
•
•
•
Generic Tone Detection Using Teager-Kaiser Energy Operators on the StarCore SC140 Core, Rev. 1
Automatic gain control (AGC). A device added to improve dynamic range. It finds the peak value of
the signal and amplifies the input samples by shifting variables (the gain varies between 1 and 2
before entering the sample processing section.
Sample processing section. Explained in Section 1, Tone Detection Basics, so the corresponding block
is not represented in detail in Figure 5. This figure emphasizes that the calculation is bypassed when
signal power is too weak.
frequencies, respectively, that must be compared with reference values.
Difference between a digit and a key. An intermediate value is found after the energy detector and
before the decision logic. This “digit” results from the analysis of a sample. In this document, “digits”
are the outputs of the sample processing section, and “keys” are the outputs of the digit processing
section. Thus, a digit is produced each time a sample is processed. A key is the result obtained by the
analysis of a set of consecutive and coherent digits with specific conditions of duration and
interruption that make it compliant with Q.24 norm. Every signal not interpreted as a key is a “non-
valid” signal. A non-valid signal may be a high-energy sound with characteristics not Q.24 compliant
(noise, too short signal, etc.). A “pause” is a non-valid signal that lasts at least 40 ms (Q.24 norm).
Signals of very short or negligible duration are “interruptions.”
A-Law encoding: Power(dBm0) = 10log
-Law encoding: Power(dBm0) = 10log
L
and
10
10
H
(P
(P
t
are the frequency detectors of the low and high
t
) + 6.18
) + 6.15
Freescale Semiconductor
8
)