A Survey on Hamming Codes for Error Detection

Vraj Kishan Dugar, Manan Godha, Vishal Thakur, Harsh Khandelwal

School of Information Technology
VIT University,
Vellore, India
[email protected]

Abstract—Information sent gets damaged many-a-times II. LITERATURE SURVEY

during signal transmission. Therefore, the signals received by the
receiver are erroneous and different to what the sender had 1. Error detection and correction by hamming code, Anil
originally sent. This paper provides a comparison between 3 Kumar Singh [1]
different types of error correcting codes, all based on the basic
Hamming code principle. Each code has its own efficiency in the Error detection:
way it functions, its power/area consumption etc. In this paper we
Suppose the signal received has even parity bits. To decode
focus on the working of these 3 error correcting codes and try and
find out the more effective one.
he codeword, first check the parity bits. Identify the parity
bits as those bits present in the sequence at powers of 2.
Keywords— Hamming Code, error correction, error detection, Find the value of each parity bit in the following way :
bit error, parity check, 3D parity, Diagonal hamming code, error
control, comparison of error correction codes. • To find value of parity bit 1 : Check each digit one
by one. If total no. of ones in these bits is even,
I. INTRODUCTION parity value is 0, else 1.
The main purpose of achieving error detection is to add some • To find value of parity bit 2 :Check 2 digits, skip 2
extra overhead with the signal, which helps the receiver to check digits, check 2 digits and repeat. If total no. of
whether the received data has been damaged during ones in the bits you check bits is even, parity value
transmission or not. There are various error detecting codes is 0, else 1.
which are capable of detecting errors of different sizes and have
their own advantages and disadvantages based on their • To find value of parity bit 4 : Check 4 digits, skip
efficiency, bit overhead and code rate they give. In this paper we 4 digits, check 4 digits and repeat. If total no. of
look at three such codes based on the basic, original Hamming ones in these bits is even, parity value is 1, else 0.
code and compare their working and characteristics.
Match the values of the parity bits you get with the bits
Hamming code is well known for its ability to correct single- received in the signal. If they match, no error exists,
bit errors. Hamming code is derived from the principle of otherwise you have an error.
including ‘r’ redundancy bits into ‘n’ data bits such that 2r n + r
+1.For 5 redundancy bits for a 11-bit data item and 6 redundancy
bits for a 56-bit data stream. These redundancy bits are to be To find which bit has the error, add the location values of
added into the data bits at bit positions 2n (n = 0, 1, 2, 3…) with all those parity bits which did not match with those
the original data bits to provide a redundancy measure. calculated. Suppose if parity bits 2 and 8 were found to be
incorrect, error exists at bit 2 + 8 = 10th bit.
Error correction codes take data bits at the source and create
redundant bits based on the data bits which is then included in Error Correction:
the data that is to be sent. Redundant bits are forwarded to the
receiver with the actual group of data bits at fixed positions and
Now that we now which bit in the incoming message has
with fixed values. The receiver of the data utilizes the redundant
bits to check for errors that may have been generated during an error, we correct it by changing its value.
sending or storing of the data. When the data is going to be
stored, redundant bits are evaluated and stored alongside the If the value is 0, change it to 1. If the value is 1, change it
original bits and later, when the data is read from the memory it to 0.
is checked to check whether any errors in the transmission of the
data have taken place between the timespan of storage and Here, we get the corrected signal. So, we find that if the
retrieval. calculated values of parity bits in the received signal don’t
match with the values in the signal itself, an error exists at
bit position equal to the sum of positions of incorrect
parity bits.

Example :
 Figure 1.0 shows the received 14 bit value.

Figure 2.1 Grouping to obtain Hamming parities

Figure 1.1 shows the Parity bit calculation.

Here Incorrect Parity bits are 2,4 and 8. Therefore error

position = 2 + 4 + 8 = 14th bit.

Figure 1.2 shows the corrected signal. Figure 2.2 Hamming parity locations w.r.t grouped data bits

The number of groups that are so formed according to

2. Horizontal – Vertical Parity and Diagonal Hamming,[2] Figure 2.1, should be the same as the number of rows in
Raha, P., Vinodhini, M. and Murty, N the encoding matrix. As you can see, there is a specific
diagonal pattern in the way data bits are grouped. First
The HVPDH method requires an encoder along with a
group : ( D16, D9, D2, D6, D13, D20, D27, D31). Second
decoder. An encoder to produce the parity bits and encode
group : ( D8, D1, D5, D12, D19, D26, D30, D23). And so
the signal with them. And a decoder to verify or change the
on.
received signal according to the parity bits. Three different
The Hamming bit calculation is as shown in the figure
parity bit sets need to be computed, vertical parity, compare
below:
diagonal parity and horizontal parity bit sets.
Encoding:
Consider a 32-bit dataword. Arrange It into a N x M matrix.
For 32-bits, let N=4 and M=8.

Figure 2.3 : Hamming parity bits calculation

For evaluating 32 bit data, we will get 8 vertical, 16

diagonal parity bits and 4 horizontal Hamming parity bits,
i.e. the three parity it sets have been calculated.
Figure 2.0 : The encoding matrix
Even parity is used. A bitwise XOR applied on each row Decoding (at receiver end):
gives that particular row’s horizontal parity bit, and an
XOR on the column gives that particular column’s parity When the receiver receives the codeword, the decoder
bit. This way, two parity bit sets of the 3 are produced. again calculates the parity bit sets for error detection.
Then the bit sets are merged again with the precious
To find the diagonal hamming code, the encoding/data matrix, to again compute both H and V parity sets.
matrix is grouped in a special way.
Error Detection:
 Observe the recalculated grouped diagonal Hamming diagonal, vertical and horizontal. These parity bits are once
parity bits. These bits help locate errors in the dataword again ordered in vectors and these vectors are encoded in
received. Hamming code by utilizing the HVHC encoder. Hamming
parity bits are then calculated and parity bits are stored along
with data bits as vectors in memory.

Encoding:

3D parity bits are used as error correction codes for data bits
stored in a matrix using the 3D Parity Check encoder. 3D
parity bits are evaluated using even parity technique, by
using bitwise XOR on corresponding bits. Again, if we take
an example of a 32-bit data word, an effective way to order
them as a matrix getting a low number of parity check bits
will be either a 4 x 8 or a 8 x 4 matrix. Going with the 4 x 8
again (Figure 3.0), we will get 4 horizontal parity bits ht[0]
Figure 2.4 : The received, decoded dataword to ht[3] (one for each row), 8 vertical parity bits vt[0] to
vt[7] (one for each column) and 11 diagonal forward parity
If single bit error exists in every diagonal parity bit bits as shown in Figure 3.1.
group, that bit can be easily located using the diagonal
Hamming parity bits recalculated by the decoder. If the
number of error bits in every diagonal parity set is > 1,
the error position then is found out with the help of the
vertical and horizontal parity bit sets, which are also
calculated again by the decoder when the new dataword
is merged into the matrix representation again.

Syndrome bits are calculated in a simple manner: just  Figure 3.0: Input data bits arranged as a matrix
apply bitwise XOR on the vertical and horizontal parity
bit sets received, with the vertical and horizontal parity
bits calculated again. If any syndrome bit is not 0, the
bits of that row or column are considered to proceed
with error correction. Location of the error bits is found
by comparing the above bits with the Hamming parity
results. Correction is performed by flipping the value
of the error bit.

Cycle :

• Correct the error bit detected at the beginning.

• Data bit is again sent to decoder to recalculate
parity bits
Fgure 3.1: Input matrix with parity bits placed
• Repeat process until all parities received are
zero.
To generate the code word systematically, these parity check
bits are appended to the data bit vector at the last index. The
3. 3D Parity Check Code with Hamming Code, Tambatkar, data bits and 3D parity check bits are then sent as input to the
S., Menon, S., Sudarshan, V., Vinodhini, M. and Murty, N HVHC encoder. The matrix is optimized if there are 8 parity
[3] bits in every vector

This method is a unique hybrid of the error – correcting

aptitudes of Hamming & 3D parity check code, in a way so
as to increase/better the reliability by compensating in the
number of parity bits.

The encoder converts data into a binary data stream and

arranges it in a matrix to reduce overhead, the size being m
x n. These bits are encoded in 3D parity check code, Figure 3.2: The systematic code-word
calculating three different sets of parity bits – Forward
So, we get 3 vectors incorporating the 23 parity bits.
Each of these vectors is encoded by the HVHC
encoder. Finally, data bits with Hamming code and
this 3D parity check code are stored in memory.

Figure 3.5: Detection of two-bit error in data bits

Figure 3.3: The encoded vector

Decoding:

Encoded data bits are passed to HVHC decoder and

the 3DPC decoder. HVHC decoder splits out 3DPC
bits and Hamming bits. These are arranged in vectors
and syndrome vectors are calculated for each of these
vectors. These syndrome vectors are used to decode
Hamming code used for the 3DPC bits and likewise Figure 3.6: Detection of three-bit error in data bits
and error bit in each Hamming code row vector can
be rectified. This can be done by finding decimal
equivalent of location of error bit. III. EXPERIMENTAL RESULTS
From the graphs shown next, we can deduce that both the
If the error is a single bit, the bit that is inverted is HVPDH method and the 3DPC Hamming Code method are
shown by the syndrome. Therefore, error detection almost equally efficient. Both have similar overheads too. The
goes up to 6 bits while 3 error bits in the 3DPC bits HVPDH method comes out to be the better one as far as the bit
can be corrected. The HVHC decoder can detect only overhead is concerned and as that is a major factor to look upon,
two error bits in each row. the graphs clearly depict that among these 3 error-correction
codes, the HVPDH method is most efficient, especially when
After detection, data bits and 3DPC bits are directed used in semiconductor memories.
to the 3PC decoder. It splits the data bits from the
3DPC bits that are used as guidelines. For data bits
that are more likely to contain errors, 3DPC bits are
re-evaluated. These reference parity check bits are
matched with these bits to look for any similarities or
dissimilarities either in the row-wise, column-wise or
diagonal parity bits. If a dissimilarity exists, the data
bit referring to that diagonal/row/column is flipped
and hence corrected.
Figure 4.1: Bit overhead and Code Rate for 3DPC Hamming
Code procedure

Figure 3.4: Detection of one-bit error in data bits

Figure 4.2: Bit overhead and Code Rate for HVPDH method
 Hamming codes. The normal Hamming code can only
correct single bit errors.

Compared to the previous method, The HVPDH Method is

successful is detecting errors of up to 8-bits and succeeds in
correction of 1-bit, 2-bit and some 3,4,5 – bit errors.

The Hamming Code with 3D Parity Check Code method is

capable of correcting any error combination up to 3 bits.
Thus, this method is one of the few ones capable of detecting
errors in both parity and data bits. It attains greater
reliability but compensate in power/area consumption.

Figure 4.3: Parity Bit Comparison

REFERENCES

[1] Singh, A. (2016). Error detection and correction by hamming code. 2016
International Conference on Global Trends in Signal Processing,
Information Computing and Communication (ICGTSPICC).
[2] Raha, P., Vinodhini, M. and Murty, N. (2017). Horizontal-vertical parity
and diagonal hamming based soft error detection and correction for
memories. 2017 International Conference on Computer Communication
and Informatics (ICCCI).
[3] Tambatkar, S., Menon, S., Sudarshan, V., Vinodhini, M. and Murty, N.
(2017). Error detection and correction in semiconductor memories using
3D parity check code with hamming code. 2017 International Conference
on Communication and Signal Processing (ICCSP).
[4] M. Kishani, H. R. Zarandi, H. Pedram, A. Tajary, M. Raji, B. Ghavami,
“HVD: horizontal-vertical-diagonal error detecting and correcting code to
protect against with soft errors", Design Automation for Embedded
Figure 4.4: Code Rate % Comparison Systems, Vol. 15, Issue 3-4, pp. 289-310, Dec. 2011
[5] Pranjali Pothare, Prajakta Ambatkar, Payal Patre, Karishma Padole,
Shilpa Lende, Ankita Belekar, “Hamming Code For Single Bit Error
Detection & Error Correction With Even Parity Using Vhdl”,
International Journal of Advanced Research in Computer Engineering &
Technology (IJARCET), Volume 4 Issue 1, January 2015, pp. 262-265,
2278 – 1323
[6] Varinder Singh, Narinder Sharma, "A Review on Various Error Detection
and Correction Methods Used in Communication", American
International Journal of Research in Science Technology Engineering &
Mathematics, vol. 15–200, pp. 252-257, 2015.
[7] Md. Shamimur Rahman, Muhammad Sheikh Sadi, Sakib Ahammed, Jan
Jurjens, "Soft Error Tolerance using Horizontal - Vertical Double - Bit
Diagonal Parity Method", 2nd IEEE International Conference on
Electrical Engineering and Information and Communication Technology
(ICEEICT), 21–23 May 2015.
[8] Comley, R. (1978). Error detection and correction for
Figure 4.5: Bit overhead % Comparison memories. Microprocessors, 2(1), pp.29-33.
[9] Ehrenborg, R. (2006). Decoding the Hamming Code. Math Horizons,
IV. CONCLUSION 13(4), pp.17-18.
The first error correction code which we studied is the [10] Ma, R. and Cheng, S. (2011). The Universality of Generalized Hamming
Code for Multiple Sources. IEEE Transactions on Communications,
original Hamming Code. The next two are more complex 59(10), pp.2641-2647.
ones, but codes which have their roots sill based on