Experiment 4: DTMF 
DSP Lab Assignment 
 

 
 

Avinab Saha 

9th April, 2018 

   

 
 

Dual Tone Multiple Frequency - Introduction 

Dual-tone multi-frequency signaling (DTMF) is an in-band telecommunication signaling system 
using the voice-frequency band over telephone lines between telephone equipment and other 
communications devices and switching centers. DTMF was first developed in the Bell System in 
the United States, and became known under the trademark Touch-Tone for use in push-button 
telephones supplied to telephone customers, starting in 1963. DTMF is standardized as ITU-T 
Recommendation Q.23.It is also known in the UK as MF4. 

The Touch-Tone system using a telephone keypad gradually replaced the use of rotary dial and 
has become the industry standard for landline and mobile service.  

The engineers had envisioned telephones being used to access computers and automated 
response systems. They consulted with companies to determine the requirements. This led to the 
addition of the number sign (#, ''pound'' or "diamond" in this context, "hash", "square" or "gate" in 
the UK, and "octothorpe'' by the original engineers) and asterisk or "star" (*) keys as well as a 
group of keys for menu selection: A, B, C and D. In the end, the lettered keys were dropped from 
most phones, and it was many years before the two symbol keys became widely used for vertical 
service codes such as *67 in the United States of America and Canada to suppress caller ID. 

Public payphones that accept credit cards use these additional codes to send the information 
from the magnetic strip.Present-day uses of the A, B, C and D signals on telephone networks are 
few, and are exclusive to network control. For example, the A key is used on some networks to 
cycle through different carriers at will. The A, B, C and D tones are used in radio phone patch and 
repeater operations to allow, among other uses, control of the repeater while connected to an 
active phone line. 

The *, #, A, B, C and D keys are still widely used worldwide by amateur radio operators and 
commercial two-way radio systems for equipment control, repeater control, remote-base 
operations and some telephone communications systems.DTMF signaling tones can also be 
heard at the start or end of some VHS (Video Home System) cassette tapes. Information on the 
master version of the video tape is encoded in the DTMF tone. The encoded tone provides 
information to automatic duplication machines, such as format, duration and volume levels, in 
order to replicate the original video as closely as possible. 

DTMF tones are used in some caller ID systems to transfer the caller ID information, but in the 
United States only Bell 202 modulated FSK signaling is used to transfer the data. 

1 
 

Keypad and Generation 

The DTMF telephone keypad is laid out in a 4×4 matrix of push buttons in which each row 
represents the low frequency component and each column represents the high frequency 
component of the DTMF signal. Pressing a key sends a combination of the row and column 
frequencies. For example, the key 1 produces a superimposition of tones of 697 and 1209 hertz 
(Hz). Initial push button designs employed levers, so that each button activated two contacts. The 
tones are decoded by the switching center to determine the keys pressed by the user. 

The DTMF keypad with the frequencies is shown below. 

A DTMF tone representing a single key press on a telephone device consists of two summer 
frequencies that have been chosen such that no harmonics occur i.e., no frequency is a integral 
multiple of another and the sum or the difference of any two frequencies does not equal any of 
the frequencies. 

Decoding 
DTMF was originally decoded by tuned filter banks. By the end of the 20th century, digital signal 
processing became the predominant technology for decoding. DTMF decoding algorithms 
typically use the Goertzel algorithm. As DTMF signaling is often transmitted in-band with voice or 
other audio signals present simultaneously, the DTMF signal definition includes strict limits for 
timing (minimum duration and interdigit spacing), frequency deviations, harmonics, and amplitude 
relation of the two components with respect to each other. 

Typically, a filter bank is used to detect the frequencies associated with a DTMF signal. The filter 
bank consists of 8 band pass filters, each of which passes only one of the 8 possible DTMF 
frequencies. The input signal for all filters is the same DTMF signal. When the input to the filter 
bank is a DTMF signal, the outputs from two if the band pass filters should be larger than the rest. 

2 
 

If we detect or measure which two outputs are the larger ones, then we know the corresponding 
frequencies, These frequencies are then used as row and column pointers to determine the key 
from the DTMF code. A good measure of the output levels is the peak value at the filter outputs. 
As when the band pass filter is working properly, it should pass only one sinusoidal signal and 
the peak value would be the amplitude of the sinusoidal passed by the filter. 

Filter Bank Consisting of 8 Band Pass Filters. 

Band Pass Filter Design 

The L-Point filter design can be used to design a bandpass filter. The Lipoint average filter is a 
low pass filter that is primarily used to design low pass filters. Its pass bandwidth is inversely 
proportional to L. It is also possible to modify the design to obtain a band pass filter centered 
around some frequency other than zero. One simple way to do this is to define the impulse 
response of the L-point FIR Filter as 

3 
 

where L is the filter length of the FIR filter, ŵ​c ​ is

​ the center frequency that defines the frequency 
location of the pass band and 𝛃 is used to adjust the maximum value of the frequency response 
magnitude to be 1. The bandwidth of the bandpass filter is controlled by L, larger the value of L, 
narrower the bandwidth. In our experiment the L is taken to be 101 or the order is kept 100. 

Frequency Response Of the 8 Band Pass Filters with Sampling frequency 4000 Hz 

The Low Frequency Filters 

   

Band Pass Characteristics for 697 Hz  Band Pass Characteristics for 770 Hz 

   

Band Pass Characteristics for 852 Hz  Band Pass Characteristics for 941 Hz 

4 
 

The High Frequency Filters 

   

Band Pass Characteristics for 1209 Hz  Band Pass Characteristics for 1336 Hz 

   

Band Pass Characteristics for 1477 Hz  Band Pass Characteristics for 1633 Hz 

The bandpass filters were designed to have the first side lobe attenuation greater than 10 dB. To 
achieve the maximum magnitude of the frequency response to be 1( or 0 dB), the value of 𝛃​ ​was 
varied. At 𝛃=0.2 we obtain the maximum magnitude of the frequency range along all frequency 
range to be 1 and the minimum attenuation of the side lobes to be 10 dB. 

 
 

5 
 

Objective 
Generate 128 random symbols consisting only [‘1’, ‘2’, ‘3’, ‘4’, ‘5’, ‘6’, ‘7’, ‘8’, ‘9’, ‘*’, ‘0’, ‘#’, ‘a’, ‘b’, 
‘c’, ‘d’]. Depending on the symbol presses on the keypad, 2 tones are generated and combined 
to form 𝑥(𝑡) for each symbol for a fixed time duration of 250 ms.The 8 Bandpass filters of above 
cutoff frequencies were designed earlier were used. Then some AWGN noise was added to 𝑥(𝑡) 
and then retrieved the symbols by filtering the signal 𝑦(𝑡)=𝑥(𝑡)+𝑛(𝑡) where 𝑛(𝑡) is the noise 
component. This experiment was repeated 100 times for a particular SNR and the average 
number of error symbols for some SNR were observed. 

  

Plot of Transmitted and Received Signals 

   

Transmitted And Received Signal SNR -5 dB  Transmitted And Received Signal SNR +5 dB 

Observation 

The noise increases as we decrease SNR value and thus it will lead to increased misclassification 
of symbols. 

6 
 

Results 

After 100 iterations of each case, 

128 symbols are generated in each iteration. The average symbol error is the number of 
mismatching symbols in 1 iteration. 

SNR (dB)  Average 

Symbol Error 

-20  99.37 

-15  48.23 

-10  16.82 

-7.5  10.23 

-5  3.61 

-2.5  0.24 

-1  0.03 

-0.9  0.02 

-0.8  0.01 

-0.75  0.00 

-0.5  0.00 

0  0.00 

1  0.00 

5  0.00 

15  0.00 

7 
 

Variation of Average Symbol Error with SNR 

Average Symbol Error v/s SNR 

Discussions 
The dual tone multi frequency transmission, increases the robustness of the system.From the 
results we can see, the transmission is error free from SNR = −0.80 dB or more. So minimum SNR 
for this case is found −0.80 dB for error free transmission. The bandpass filter we used is simple 
FIR filter by using windowing, but we can use more complex higher order filters. The frequencies 
should be chosen very carefully, so that the signals can be filtered properly. The symbol time can 
be varied, but we need more sampling frequency for smaller duration, so that we have enough 
data.The error increases as SNR decreases. 

8