Energy and Power Engineering, 2013, 5, 1226-1229
doi:10.4236/epe.2013.54B232 Published Online July 2013 (http://www.scirp.org/journal/epe)
Amalgamated-signal Cymometer Based on TMS320F28335
Z. Y. Xu, W. T. Lv, Z. Y. Li
College of Electrical Engineering, Zhejiang University, Hangzhou ,China
Received April, 2013
ABSTRACT
In our design, an amalgamated-signal cymometer is developed with a TMS320F28335 DSP chip ---- we use its on-chip
12-bit 16-channel AD to conduct a novel dual-sequencer synchronize sampling and averaged to improve the accuracy.
Considering the algorithm, we use TI's C28x_FPU_Lib to conduct a FFT with sampled results and correct spectrum
with Time-Shift Phase Difference Correcting Spectrum Method based on all-phase spectral analysis. Namely, we design
a sampling-frequency self-adaption algorithm which makes operation keep on without outside-hardware command.
Actually, the test result shows that our design's frequency resolution ratio is up to 0.4%, and the maximum frequency’s
D-value between main and minor signal is up to 29.99 kHz.
Keywords: Digital Frequency Meter; Same Precision Measurement; DSP; All-phase Spectral Analysis;
Sampling-frequency Self-adaption
1. Introduction
Today's world is a digital world, in which digital signal
processor playing an important role, and Ti company as
the world's leading semiconductor company; its digital
signal processor plays a very important role in the field
[7]. Discrete Fourier transform (DFT) technology is the
core of digital signal processing technology. In 1965,
Cooley and Tukey proposed the Fast Fourier Transform
algorithm hereinafter referred to as FFT, and it has wide
application in every field of digital signal processing.
Spectrum analysis is an important application of FFT,
and this design is based on spectrum analysis method for
signal frequency and the amplitude of the primary and
secondary in the composite signal.
TMS320F28335 is TI company's new high perform-
ance 32-bit floating point digital signal processor, using
12-bit 16 channel AD to do two-channel synchronous
sampling and take its average in order to improve the
accuracy, and using floating-point runtime library pro-
vided by TI to do FFT computation, and using an ex-
tremely high precision time shift phase difference meth-
od based on all phase spectrum analysis to correct, get-
ting accurate frequency and the amplitude of the primary
and secondary signals, via a serial port to send to upper
computer, and at the same time using the
TMS320F28335's high-precision EPWM to reconstruct
primary and secondary signal waveform.
2. System Solutions
2.1. The Overall Block Diagram
Figure 2.1. The overall block diagram..
2.2. The time Shift Phase Difference Correction
Method based on all Phase FFT Spectrum
Analysis
All phase FFT has better inhibiting spectrum leakage
performance, which has good inhibition of spectral leak-
age performance of amplitude spectrum, and as well can
be directly generated accurate phase spectrum [1]. Based
on this, do simple arithmetic on the main spectral line
with two sequences which have time shift and we can get

Z. Y. XU ET AL. 1227
higher accuracy of traditional phase estimate, have high-
er application value than phase difference correction
method, and have better use value.
Description of all phase FFT spectrum analysis meth-
od:
After all phase data pretreatment and then do FFT to
the data, we can get a new type of all phase of the FFT
spectrum analysis as is shown in Figure 2.2, and its im-
provement in the method of truncation of data greatly
improves the spectrum performance.
The process the Time-Shift Phase Difference Correct-
ing Spectrum Method corrects Spectrum as is shown in
the Figure 2.3 below [8]:
Its Frequency estimation formula is
(1)
The deviation value of main line is
(2)
Its Amplitude estimation formula is
(3)
In the formula, the denominator Figure 2 (dω) is val-
uable when dω is brought into the window function of
the formula of Fourier transform, and the general win-
dow function (such as the hanning window, hamming
window,
Figure 2.2. The basic block diagram of N-order all phase
FFT spectrum analysis.
Figure 2.3. the process of spectrum correction of Time-Shift
Phase Difference method
triangle window) belong to a cosine window, so their
Fourier transform expression are established.
Besides, Time-Shift Phase Difference Correcting
Spectrum Method has no phase difference correction of
phase estimation.
Theoretically, the frequency correction precision of
Time-Shift Phase Difference Correcting Spectrum
Method reaches the level of degrees, and its
phase correction error reaches 10-3degrees, which are
more precise than traditional methods.
Specific analyses on the theory of all phase spectrum
analysis please refer to the reference [2]，[3].
3. Design of the Hardware in the System [5]
3.1. Follower
This part of circuit mainly enhances signal, avoiding
signal attenuation caused by the input and output imped-
ance mismatch. Primary and secondary signal use the
whole pole input with two followers to connect with the
level after circuit, and in this not do change of amplitude
but just joining a resistance whose resistance R is 1 k in
the input stage.
3.2. The Summing Circuit
The module adopts the three input reverse phase summa-
tion operation circuit. because in phase end is grounding，
so for the reverse side its voltage vn = 0 (according to the
use of manual recommended positive input directly to the
ground), and because of broken, there can be a VO =
-R21*V1/R18 - R21*V2/R19 - R21*V3/R20 ，According to
design, the input of two sine signal is damped to 0.5
times of original signal and then superimposed on a dc
bias voltage, and the resistance for R18 = R20 = 22 k,
R19 = 24 k, R21 = 22 k, and the input signal of R19 is
3.3 V dc voltage. So the superposition signal is negative,
the amplitude is 0.5 times of the primary and joining
U=-3.3*11/24V=-1.5125V dc offset signal.
3.3. The Second Order Butterworth Low-pass
Active Filter
Due to the demand of the algorithm is higher to the sig-
nal, the module will filter the high frequency noise signal
Figure 3.1. Input signal conditioning ci rc uit.

Z. Y. XU ET AL.
1228
in the input to the DSP signal, to reduce the interference
signal on the calculation of FFT, and use the negative
input at the same time, will be through the adder origi-
nally negative signal into a positive.
OPA4227 precision itself introduces less noise, con-
sider attenuation rate of a first order butter worth filter
for each frequency doubling is 6 db, every ten times as
much as 20 db; attenuation rate of second order butter-
worth filter for each frequency doubling is 12 db, and
attenuation rate of the third order butter worth filter for
frequency doubling is each 18 db and so on, so we only
need the second order to meet the demand. In this, R22 =
R23 = 100 k, R24 = 22 k, C3 = 180 pF, and cutoff fre-
quency is 33 KHZ, phase reversal 180 degrees when in
120 KHZ.
3.4. Limiter Circuit
This module uses the diode and a resistor in parallel, R25
= 10 k, diode using 1 n5221, positive pressure limiting at
2.4 V, negative for 0.6 V, and according to the calcula-
tion, after the amplitude limiting, output positive maxi-
mum voltage limit is 0.75 + 1.5125 = 2.2625 V, the limit
of minimum forward voltage is 1.5125-0.75 = 0.7625 V,
which are within the control of clipping. While DSP AD
requests input amplitude is in 0 ~ 3.3 V, the design meets
the requirements.
4. The System Software Design
System software design adopts modular design, and it
combines subroutine which completes the specific func-
tions into a functional module, invoked by the main
monitoring program invocation. Software overall block
diagram is as shown in the figure below.
System software consists of main functional modules:
initialization module, AD sampling module, FFT calcu-
lation module, spectrum correction module, SCI sending
module, ePWM waveform reconstruction module, the
watchdog module and interrupt module.
Specific operation modes of System software design
please refer to the reference [4].
Main computational program flow is as shown in the
figure below [6]:
Figure 4.1.general program of the software.
Figure 4.2. Main computational program flow chart.
5. The Core Code
Important structure definition
Type def struct {
Uint16 Main Index;
Uint16 Minor Index;
} SIGSEARCH_F32_STRUCT;
Type def struct {
Float 32 Freq;
Float 32 Mag;
} APCM_F32_STRUCT;
Type def struct {
APCM_F32_STRUCT Main;
APCM_F32_STRUCT Minor;
Uint16 Flag;
} CORRECTION_F32_STRUCT;
2．Spectrum correction function
//===================================
// Function: APCM_F32_STRUCT RFFT_f32_apcm
(RFFT_F32_STRUCT *fft1, RFFT_F32_STRUCT
*fft2,float32 Fs,Uint16 index)
//===================================
APCM_F32_STRUCT RFFT_f32_apcm
(RFFT_F32_STRUCT *fft1, RFFT_F32_STRUCT *fft2,
float32 Fs,Uint16 index)
{
APCM_F32_STRUCT ans;
float32 phi1, phi2;
float32 deltaphi, delta, deltafreqk;

Z. Y. XU ET AL.
1229
float 32 temp, h;
phi1=*(fft1->PhaseBuf+index);
phi2=*(fft2->PhaseBuf+index);
deltaphi=phi2-phi1;
delta=mod(deltaphi,2*PI);
if (delta<-PI)
delta=delta+2*PI;
else if(delta>PI)
delta=delta-2*PI;
deltafreqk=delta/(2*PI);
deltafreqk=deltafreqk+(deltafreqk==0)*EPS;
ans.Freq=(index+deltafreqk)*Fs/fft1->FFTSize;
temp=sin(PI*deltafreqk)+(sin(PI*deltafreqk)==0)*
EPS;
h=2*PI*deltafreqk*(1-deltafreqk*deltafreqk)/temp;
ans.Mag=h*h*(*(fft1->MagBuf+index))/2;
return ans;
}
//===================================
REFERENCES
[1] K. J. Xu, “Theory and Application of TMS320X281X
DSP,” Beijing University of Aeronautics and Astronaut-
ics Press, 2006.
[2] K. Ding, “Discrete Spectrum Correction Theory and
Technology,” Science Press, 2007.
[3] RIFEDG, VINCENTGA. Use of the discrete fourier
transform in the measurement of levels and tones[ J]. Bell
Syst. Tech. J. , 1970, 49( 2): 197- 228.
[4] Z. H. Wang, “Digital Signal All Phase Spectrum Analysis
and Filtering Technology,” Electronic Industry Press,
2009.
[5] “TMS320F28335 Data Manual,” Texas Instruments.
[6] G. Z. Zhao, “Signal Analysis and Processing,” Mechani-
cal Industry Press, 2005.
[7] Z. Z. Li, “Digital Signal Processing and Application with
MATLAB,” Tsinghua University Press, 2008.
[8] “Assembly Good Paper of C2000 Grand Prix of TI
C2000 Senior Embedded Controlled in 2008,” Electronic
Industry Press, 2008.