CVSD Vocoder with BER analysis for

WiMAX using GNU Radio and USRP

N210
B. Siva Kumar Reddy, B. Lakshmi and Ravi Kishore Kodali
Research Scholar, Associate Professor and Associate Professor
Department of Electronics and Communication Engineering,
National Institute of Technology, Warangal-506004

Abstract-Mobile WiMAX is a broadband wire- Radio Frequency (RF) and signal conversion functions
less solution that empowers convergence of mobile are executed in hardware. All further functions such as
and fixed broadband network through a regular modulation, filtering, demodulation, channel model, etc;
wide area broadband radio access technology and are software defined, maximizing adaptability and em-
adaptable system architecture. An OFDM based powering the development of wireless testbeds that are
WiMAX physical layer combining with Continu- programmable at the PHY, DLC, and MAC. Such SDR
ously Variable Slope Delta (CVSD) speech coding testbeds additionally support protocol engineers optimiz-
strategy has been implemented utilizing GNU Ra- ing higher layers of the stack by adjusting to PHY and
dio and USRP N210. CVSD vocoder is a nonlinear MAC parameters (cross-layer advancement), since these
sampled data feedback framework which acknowl- parameters are easily accessible or controllable. In this
edges a band restricted simple signal and encodes paper, GNU Radio [4] and USRP N210 [5] are utilized
it into binary form for transmission through a as software and hardware platforms separately. Assum-
digital channel. At the receiver, the binary signal is ing that special radio modulation is required, just the
decoded into a nearby estimate of original analog programming must be changed. In this way, SDRs are
signal. We executed a complete transceiver chain adaptable and considerably more efficient than accepted
which we characterize by implementation complex- wireless communication frameworks. This innovation can
ity. Further, we measure the performance of the ideally be utilized within industry and research for fast
testbed by running this chain. The CVSD coding prototyping, since programming advancement is cheaper
system is implemented in GNU Radio and the test and less time-consuming than hardware development. In
outcomes demonstrated that coding has better Bit this paper, we evaluate a particular SDR execution; the
Error Rate (BER) and Symbol Error Rate (SER) GNU Software Radio (GSR) is suitable for developing
values when integrated with FEC (Forward Error testbeds and how it supports the protocol engineer ac-
Correction). FEC has been carried out by link- cording to the criteria.
ing convolutional coder (CCSDS coder) and Reed In digital data communication or storage systems, data
Muller (RM) coders for better BER execution. is transmitted or recorded as series of binary digits. The
keywords: BER, CVSD vocoder, GNU Radio, USRP recipient or player should just recognize between an one
N210. or zero to precisely recover the original data. Digital
speech coding algorithms are judged by their capability to
I. Introduction quantize (digitize) speech accurately for transmission and
IEEE 802.16 is a solution for Broadband Wireless Ac- afterwards perform the reverse at the decoder. Delta Mod-
cess (BWA) commonly known to be Worldwide Interop- ulation (DM) [6] and Continuously Variable Slope Delta
erability for Microwave Access (WiMAX) [1], is a recent modulation (CVSD) [7] are differential waveform quan-
wireless broadband standard that has guaranteed high tization systems. Both utilize two level quantizers (one
bandwidth over long range transmissions. The standard bit). CVSD is essentially DM with an adaptive quantizer.
indicates the air interface, incorporating the Medium Applying adaptive strategies to a DM quantizer takes into
Access Control (MAC) and Physical (PHY) layers of consideration consistent step value change. By changing
BWA. The Mobile WiMAX air interface adopts Orthog- the quantization step estimate, the coder has the ability to
onal Frequency Division Multiple Access (OFDMA) [2] represent low amplitude signals with more excellent exact-
for enhanced multi-way execution in non-line-of-sight en- ness (where it is required) without sacrificing performance
vironments. Scalable OFDMA (SOFDMA) is presented in on large amplitude signals. CVSD is utilized as a part
the IEEE 802.16e and this amendment supports adaptable of strategic interchanges where communication quality is
channel bandwidths from 1.25 to 20 MHz. needed but the choice for security must be accessible. MIL-
Software Defined Radio (SDR) [3] refers to an archi- STD-188-113 (16kb/s and 32kb/s), and Federal Standard
tecture for wireless communication devices where only 1023 (12kb/s CVSD) are cases of tactical communication
systems utilizing CVSD. Notwithstanding point-to-point encoder can accurately process the slope between adjacent
communication, CVSD is ordinarily employed as a part of audio samples and accurately upgrade its inward reference
computerized voice recording/messaging and audio delay quality. Issues with granularity and slope overload could
lines. be drastically lessened by making dynamic adjustments
Channel coding [8] is the method to make the trans- to the quantizer step estimate δ. Adaptive DM (ADM)
mitted signals less vulnerable to errors. Channel cod- algorithms attempt to do this by making δ small for grad-
ing incorporates scrambling, Forward Error Correction ually changing signals and extensive for quickly evolving
(FEC), and interleaving. The field of channel coding is signals. The most promoted ADM algorithm is regarded
related with transmitting a stream of information at as as CVSD. This CVSD algorithm is also called as ”Digitally
high a rate as could reasonably be expected over a given Controlled DM”.
communication channel. After that decoding the original The maximum signal-to-noise ratio (SNR) for a con-
information dependably at the recipient, employing en- sistently quantized signal could be figured by finding the
coding and decoding algorithms that are executable to ratio of the full scale quantization level to the noise voltage
complete in a given technology. The cause for linking two [7].
coding plans is to attain huge coding gains with afford- 2n
SN R = √ . (1)
able decoding complexity. In coding hypothesis, connected 3
codes structure a class of error-correcting codes [9] that
Where, n= number of bits in quantizer.
are picked up by combining an internal code and an
external code. In this paper concatenated coding organized In dB,
as Convolutional coding [8] as external code and Reed SN RdB = 4.77 + 6.02n, (2)
Muller coding [10] as inner code.
equation (2) is an objective measure of value in frame-
II. CVSD VOCODER works using uniform quantization. It ought to be noted
that the SNR figured in equation (2) is a theoretical max-
This block performs CVSD sound encoding in GNU
imum. In practice different variables (e.g. power supply
Radio Companion. Its plan and execution is modelled after
noise) tend to decrease the last SNR. Additionally, the
the CVSD encoder/decoder determinations characterized
human voice is recognized to have in the vicinity of 40db
in the Bluetooth standard. CVSD is a system for encod-
of dynamic range, notwithstanding, during most conversa-
ing speech that looks to diminish the bandwidth needed
tions it is normally about 20db down from maximum. As
for advanced voice transmission. CVSD exploits strong
it were, in general we do not shout on normal conversation.
correlation between samples, quantizing the difference in
Therefore, the normal signal to noise ratio for consistently
amplitude between two continuous samples. This differ-
quantized speech is about 20db less than that might be
ence requires fewer quantization levels as compared with
computed by using equation (2).
different techniques that quantize the real amplitude level,
An alternate objective measure of value which might
reducing the transmission capacity. CVSD utilizes a two
be determined in a comparative way is dynamic range. It
level quantizer (one bit) and an adaptive algorithm that
relates to the determination of a quantization plot. It is
allows for continuous step size adjustment. The coder can
the degree of full scale amplitude to smallest quantized
represent to low amplitude signals with accuracy without
amplitude change,
sacrificing performance on large amplitude signals. This
is a trade off that happens in some non-adaptive mod- DRdB = 6.02(n − 1). (3)
ulations. The CVSD encoder effectively allows for 8-to-1
compression. All the more particularly, each incoming au- An acknowledged method for measuring quality in sig-
dio sample is compared with an internal reference worth. nal processing frameworks is SNR. In the setting of ad-
In the event that the data is greater or equivalent to the vanced coding this might all the more accurately be called:
reference, the encoder yields a ”1” bit. In the event that Signal to Quantization noise Ratio (SQR). Equation (2)
the data is less than the reference value, the encoder yields was determined to calculate the SNR for uniform quanti-
a ”0” bit. The reference quality is then updated likewise zation expecting maximum input signal. To find SNR for
dependent upon the recurrence of yielded ”1” or ”0” bits. input signal amplitudes less than maximum equation (2),
By grouping 8 yields bits together, the encoder basically it must be rewritten as
process one yield byte for each 8 data audio samples. 
Si

This encoder requires that input audio samples are 2- SN RdB = 4.77 + 6.02n + 20log (4)
Smax
byte short signed integers. The consequence bandwidth
conversion is 16 input bytes of raw audio data to ”1” Where, Si = input signal level; SM AX = maximum
output byte of encoded audio data. The CVSD encoder signal level. The point is, when the quantization charac-
module must be prefixed by an up-converter to over- teristic is not uniform, equation (4) is no longer valid. To
sample the audio information preceding encoding. The clarify the execution of companded quantization, the SNR
Bluetooth standard particularly calls for a 1-to-8 interpo- must be figured for a full range of input signal levels. For
lating up-converter. While this decreases the overall com- µ − law companding, the signal to noise ratio is given by
pression of the codec, this is needed with the goal that the the accompanying comparison,
 In general, one can see every equality mathematical
  statement as being created by generating the message
 3(22n )  bits, x, and a generator polynomial, g. In the first sample
SN Rµ = 10log    . (5)
 above, the generator polynomial coefficients are (1, 1, 1)
 2 1.732 1
[ln(1 + µ)] 1 + + 2 2 and (1, 1, 0), while in the second, they are (1, 1, 1), (1, 1, 0),
µSi µ Si
and (1, 0, 1). We indicate by gi the K-component generator
It should be noted that as µ → 0, equation (5) is polynomial for parity bit Pi . We can then compose Pi as
equivalent to equation (4), thus µ = 0 implies uniform accompanies:
quantization. 
k−1

The maximum SNR for DM speech coders is [7]
X
Pi [n] =  gi [j]x[n − j] mod2 (9)
fS3
 
j=0
SN RdB = 10log − 14.0 (6)
fBW f 2 The type of the above mathematical statement is a convo-
Where, fS =data rate (sample rate); fBW =signal BW lution of g and x-thus the expression ”convolutional code”.
(cutoff frequency of LPF); f =signal frequency. Most lit- The amount of generator polynomials is equivalent to the
erature on DM coding agrees on the terms involving amount of produced equality bits, r, in each one sliding
fS , fBW , and f in equation (1) but most do not agree window.
on -14 dB. This term changes from -17.0dB to -11.7dB IV. REED-MULLER CODING
depending upon a variety of different input signals and
In RM Encoder, only the first bit is utilized as a part of
assumptions about environment. Nevertheless, equation
in and output [10]. m must be littler than 31 and r must
(1) is considered as maximum because it assumes no slope
be more smaller than m. Reed–Muller codes are recorded
overload and least granular noise, i.e. ideal step size. The
as RM (d, r), where d is the order of the code, and r sets
adaptive behaviour of CVSD brings about a SNR versus
the length of code, n = 2r . RM codes are partnered to
data level characteristic similar to non-uniform quanti-
binary functions on field GF (2r ) over the components 0, 1.
zation [7]. This non-linear SNR characteristic is because
RM (1, r) codes are equality check codes of length n = 2r ,
of companding, where the quantization level (incline) is r+1 n
changed in accordance with a larger or smaller value with rate R = and least distance dmin = .
n 2
past pitch changes of the input signal. The amount of past
V. EXPERIMENTAL RESULTS
samples (bits) used to make an expectation is regularly
three or four. Four bit companding has turned out to be The GNU schematic of transmitter and receiver sections
best for information rates greater than 32kb/s, three bit of the WiMAX Physical layer with CVSD coder are
companding appears to work preferable for information indicated in Fig. 2. The employed testbed for the execution
rates less than 32kb/s. is demonstrated in Fig. 3. We have implemented only the
mandatory features of the physical layer of WiMAX air
III. CONVOLUTIONAL CODING interface, while leaving the implementation of optional
The example in Fig. 1 shows a set of parity equations, features for future work. Needed OFDM parameters are
which given the way in which parity bits are developed noted in Table 1. As demonstrated in Fig. 2, a triangle
from the sequence of message bits, X. In this example, signal source with sampling rate 8ks/s is given to trans-
the equations are as follows [8] mitter section. Transmitter segment involves Channel cod-
ing (Scrambler, Forward Error Correction, Interleaving),
P0 [n] = x[n] + x[n − 1] + x[n − 2] OFDM Modulation, Voltage Controlled Oscillator (VCO),
(7)
P1 [n] = x[n] + x[n − 1] CVSD encoder and USRP sink. The transmitted signal
is appropriated by the USRP source and decoded by the
An example of parity equations for a rate 1/3 code is receiver segment. To examine BER and symbol error rate
P0 [n] = x[n] + x[n − 1] + x[n − 2] (SER), transmitted signal and appropriated signals are
given as two inputs to the Error Rate block and output
P1 [n] = x[n] + x[n − 1] (8)
characteristics are observed on Number sink (Shown in
P2 [n] = x[n] + x[n − 2] Figs. 4,5). File source in the Fig. 2 is utilized to apply
all the captured signals. Real time signals for various
frequencies can be captured by utilizing GNU schematic
demonstrated within Fig. 6. User can watch the strength
of captured signals on FFT sink and Scope sinks. Needed
GNU Schematic for CVSD coder without FEC is demon-
strated in Fig 7. FFT sink and Scope sink are utilized
to watch the output waveforms in frequency domain and
time dominion individually. Audio sink is utilized to listen
Fig. 1. Convolutional code with paritybit bits per message bit (r=2) the decoded audio signal. The outcomes for different
and constraint length (k=3). frequencies with and without FEC are classified in Table
2. From the outcomes, it might be inferred that CVSD
coder with FEC gives low BER esteem than without FEC.
Figs. 8, 9 show the watched original spectrum and original
waveform separately. Figs. 10 to 13 show the encoded and
decoded spectrum and waveforms.
TABLE I
Simulation parameters defined.

Parameters Values
FFT size (NFFT) 1024
Occupied Tones 768 Fig. 5. CVSD coded Signal BER value without FEC.
Sampling rate 10.66667 M
Center Frequency 2.48 GHz
Convolutional Code 1/2
Cyclic Prefix length 184
Useful symbol duration 91.43 µs
Carrier spacing (1/Tu ) 10.94 KHz
Guard time (Tg=(1/4)* Tu) 11.43 µs
OFDM symbol duration 102.86 µ s
Digital Modulations BPSK, QPSK,
16QAM, 64QAM, 256QAM

TABLE II
Obtained results for BER analysis.

Frequency BER SER With FEC

50 MHz 1.1089999676 0.8029999733 No
50 MHz 0.1112499982 0.0610000007 Yes
710 KHz 1.194499965 1.0000000000 Yes
Fig. 6. GNU Schematic to capture realistic signals.
93.5 MHz 1.4342000217 1.6734602734 Yes
2.4 GHz 2.0000000000 1.0000000000 Yes
6 GHz 0.8914999962 0.9789999723 Yes

Fig. 8. Obtained original spectrum on FFT sink.

Fig. 3. Software Defined Radio development test bed.

Fig. 4. CVSD coded Signal BER value with FEC.

VI. CONCLUSION
Fig. 9. Obtained original spectrum on Scope sink.
We utilize the GNU Radio/USRP platform for eval-
uating WiMAX transceiver segments. In this paper, we
 Fig. 2. GNU Schematic of CVSD vocoder with Forward Error Correction.

Fig. 7. GNU Schematic of CVSD vocoder without forward Error Correction.

Fig. 10. Obtained encoded spectrum on FFT sink. Fig. 11. Obtained original waveform on Scope sink.

have analyzed the BER and SER execution of WiMAX was evaluated and surveyed at FEC coding with CVSD
Physical layer with the usage of concatenated channel vocoder. CVSD has several attributes that make it well
coding technique under different frequency signals over suited for digital coding of speech. CVSD has enough
realistic channel conditions. The execution of WiMAX adaptability to permit advanced encryption for secure
physical layer dependent upon the IEEE 802.16e standard, requisitions. At last, CVSD can work over an extensive
 toll quality. The greater part of these traces make CVSD
attractive to remote telecommunication frameworks. From
the effects we can presume that CVSD coder with FEC
gives better BER execution. Thus, linked coding is utilized
to minimize the error rate in the digital transceiver areas.
References
[1] Andrews, Jeffrey G and Ghosh, Arunabha and Muhamed, Rias,
”Fundamentals of WiMAX: understanding broadband wireless
networking,” Prentice Hall PTR,2007.
[2] Giuliano, Romeo and Mazzenga, Franco, “Exponential effective
SINR approximations for OFDM/OFDMA-based cellular sys-
tem planning,” Wireless Communications, IEEE Transactions
Fig. 12. Obtained decodedl spectrum on FFT sink.
on vol.8, no. 9, pp. 4434-4439, 2009.
[3] Reis, Garcia and Luiz, Andre, “Introduction to the Software-
defined Radio Approach,” IEEE Latin America Transactions,
vol. 10, no. 1, pp. 1156-1161, 2012.
[4] Radio, GNU, “The gnu software radio,” Available from World
Wide Web: https://gnuradio. org, 2007.
[5] Ettus, Matt, “USRP User’s and Developer’s Guide,” Ettus Re-
search LLC, 2005.
[6] Gómez-Estern, Fabio and Canudas-de-Wit, Carlos and Rubio,
Francisco R ”Adaptive delta modulation in networked controlled
systems with bounded disturbances,” Automatic Control, IEEE
Transactions on,vol 56, no 1, Pages 129–134, 2013.
[7] Khalifa, Othman O and Khan, Sheroz and Islam, MD Rafiqul
and Muktar, “Speech coding for Bluetooth with CVSD al-
gorithm,” RF and Microwave Conference, 2004. RFM 2004.
Proceedings, pp. 227–229, 2004.
[8] Reddy, B Siva Kumar and Lakshmi, B ”Adaptive Modulation
Fig. 13. Obtained decoded waveform on Scope sink. and Coding in COFDM for WiMAX Using LMS Channel
Estimator,” Conference on Advances in Communication and
Control Systems,Pages 23-29, 2013.
variety of information rates. It has been effectively utilized [9] Pless, Vera ”Introduction to the theory of error-correcting
from 9.6 kb/s to 64 kb/s. At 9.6 kb/s audio quality codes,” John Wiley & Sons,vol 48, 2011.
[10] Bhattacharyya, Arnab and Kopparty, Swastik and Schoenebeck,
is not particularly good, however, it is intelligible. At Grant and Sudan ”Optimal testing of Reed-Muller codes,” Foun-
information rates of 24 kb/s to 48 kb/s it is judged as dations of Computer Science (FOCS), 2010 51st Annual IEEE
truly satisfactory. And above 48 kb/s it is comparable to Symposium on,pp 488–497, 2010.