CN107526016B - A kind of detection method and device for semiconductor devices 1/f noise bound frequency - Google Patents
A kind of detection method and device for semiconductor devices 1/f noise bound frequency Download PDFInfo
- Publication number
- CN107526016B CN107526016B CN201710531992.4A CN201710531992A CN107526016B CN 107526016 B CN107526016 B CN 107526016B CN 201710531992 A CN201710531992 A CN 201710531992A CN 107526016 B CN107526016 B CN 107526016B
- Authority
- CN
- China
- Prior art keywords
- frequency
- noise
- scale
- frequency domain
- wavelet function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 239000004065 semiconductor Substances 0.000 title claims abstract description 30
- 238000001514 detection method Methods 0.000 title claims abstract description 16
- 238000005070 sampling Methods 0.000 claims description 14
- 230000005856 abnormality Effects 0.000 claims description 10
- 230000000694 effects Effects 0.000 claims description 6
- 238000005314 correlation function Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 2
- 239000011159 matrix material Substances 0.000 claims description 2
- 230000007423 decrease Effects 0.000 claims 1
- 238000000034 method Methods 0.000 abstract description 12
- 230000009466 transformation Effects 0.000 abstract description 11
- 230000003595 spectral effect Effects 0.000 abstract description 4
- 238000001228 spectrum Methods 0.000 abstract description 4
- 238000007796 conventional method Methods 0.000 abstract description 3
- 230000002745 absorbent Effects 0.000 description 5
- 239000002250 absorbent Substances 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 238000001914 filtration Methods 0.000 description 5
- 239000000463 material Substances 0.000 description 5
- 238000005259 measurement Methods 0.000 description 5
- 230000003321 amplification Effects 0.000 description 3
- 239000002184 metal Substances 0.000 description 3
- 238000003199 nucleic acid amplification method Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000000274 adsorptive effect Effects 0.000 description 1
- 239000004020 conductor Substances 0.000 description 1
- 230000021615 conjugation Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000005669 field effect Effects 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 229910044991 metal oxide Inorganic materials 0.000 description 1
- 150000004706 metal oxides Chemical class 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- JYQUHIFYBATCCY-UHFFFAOYSA-N quinalphos Chemical compound C1=CC=CC2=NC(OP(=S)(OCC)OCC)=CN=C21 JYQUHIFYBATCCY-UHFFFAOYSA-N 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/26—Testing of individual semiconductor devices
- G01R31/2601—Apparatus or methods therefor
- G01R31/2603—Apparatus or methods therefor for curve tracing of semiconductor characteristics, e.g. on oscilloscope
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/26—Testing of individual semiconductor devices
- G01R31/2646—Testing of individual semiconductor devices for measuring noise
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Testing Of Individual Semiconductor Devices (AREA)
Abstract
The invention discloses a kind of detection method and device for semiconductor devices 1/f noise bound frequency, and device includes: shielded box, for shielding outside noise interference;Biasing circuit, for time domain waveform needed for obtaining experiment;Preamplifier, for amplifying noise signal;Digital oscilloscope, for showing the time domain waveform of input signal;PC machine, for handling inputted signal.The device of the invention has lower background noise, higher shielding outside noise interference performance, and the wirewound potential meter itself in device there's almost no 1/f noise, and the data measured are more accurate.Method of the invention can also use the wavelet transformation of other systems, such as use continuous wavelet transform, can effectively improve the accuracy of bound frequency detecting other than using Dyadic Wavelet Transform;It avoids analyzing obtained spectral characteristic curve, this conventional method of bound frequency is found in fitting using by spectrum analyzer.
Description
Technical field
The invention belongs to noise frequency detection fields, special based on the decorrelation of 1/f noise wavelet transformation more particularly, to one kind
Property in low-frequency noise detect 1/f noise bound frequency method.
Background technique
1/f noise is prevalent in the low frequency electrical noise of all kinds of electronic devices such as diode, MOSFET, semiconductor laser
In, one of the striving direction that 1/f noise is device design and research is reduced as far as possible.However the intensity of 1/f noise, frequency range
And relevant parameter is used to detect the latent defect of device at the same time as a kind of highly effective means, passes through analysis device work
When 1/f noise signal may determine that the source of noise, to know position and the type of latent defect;On the other hand, to 1/f
The level of noise, which carries out detection, can effectively assess the reliability of device, generate biggish loss without to device itself.?
Under such background, from accurately detected in low-frequency noise signal caused by electronic device 1/f noise and determine its above and below
The method of frequency limit rate is essential.Nowadays to the research of semiconductor devices 1/f noise, in noise extraction and noise analysis
Two aspects obtain very big breakthrough, but still remain many shortcomings, on the one hand not for the test macro of different components
Together, while the price of test macro is more expensive, therefore there are certain limitations to device itself;It on the other hand is to make an uproar to 1/f
The specific mechanism of production of sound is still less clear.
Summary of the invention
In view of this, to solve the above problems, the present invention provides one kind for the low frequency electrical noise signals of actual measurement
For the detection device of semiconductor devices 1/f noise bound frequency, including shielded box 101, biasing circuit 103, preposition amplification
Device 104, digital oscilloscope 105 and PC machine 106;Biasing circuit 103 is located inside shielded box 101, and biasing circuit 103 passes through BNC
Data are transmitted with the preamplifier 104 outside shielded box 101, digital oscilloscope 105 is connected with preamplifier 104, PC machine
106 are connected with digital oscilloscope 105;
The biasing circuit 103, including DC power supply, pluggable wire-wound resistor 201, pluggable semiconductor devices 202, line
Around potentiometer 203, external impressed current table 205 and sampling output point 206;Wherein, DC power supply includes DC power supply A and DC power supply
B, DC power supply A and DC power supply B are connected to pluggable 202 source and drain branch of semiconductor devices by double-point double-throw switch
It drains, 201 front end of pluggable wire-wound resistor of 203 branch of wirewound potential meter, pluggable 201 rear end of wire-wound resistor connects three-level line
Around potentiometer 203, the grid of pluggable semiconductor devices 202 connects first order wirewound potential meter 203, pluggable semiconductor devices
202 source electrode connects sampling resistor 204, and sampling output point 206 is drawn at 204 both ends of sampling resistor, and sampling resistor 204 passes through switch
207 ground connection, 207 both ends of switch can connect ammeter 205.
The shielded box 101, outside are shielded using double-level-metal, and inside uses electromagnetic wave absorbent material 102, for shielding
External interference noise;
The DC power supply, using storage battery power supply, for providing stable input power for the biasing circuit;
The pluggable wire-wound resistor 201 plays guarantor for limiting the size of current of biasing circuit using wire-wound resistor
The effect of protection circuit;
The pluggable semiconductor devices 202, can be selected MOSFET, triode or diode, and what is selected in the present apparatus is
MOSFET, for amplifying the electric current in biasing circuit;
The wirewound potential meter 203, for adjusting the size of pluggable semiconductor devices source electrode output electric current;
The external impressed current table 205, for monitoring the size of electric current in source and drain branch, by observing the numerical value of ammeter,
To adjust wirewound potential meter 203, change the size of pluggable semiconductor devices source electrode output electric current, and then changes source and drain branch
The voltage at middle sample resistance both ends;
The sampling output point 206 is output to shielded box for exporting the voltage at sample resistance both ends in source and drain branch,
It is input to preamplifier by BNC, is finally output to digital oscilloscope, submits data to computer disposal.
The present invention additionally provides a kind of for semiconductor devices 1/f noise for the low frequency electrical noise signals of actual measurement
The detection method of lower frequency limit, comprising the following steps:
Step 1: obtaining x (n), n=1,2 after sampling to low-frequency noise signal x with fixed sampling frequency ..., 2M;Wherein, M
For the integer greater than 1;
Step 2: being directed to each coefficient matrix, calculate separately the phase of the wavelet coefficient under its each decomposition scale factor j
Function is closed, the maximum value E, E of correlation function is taken to characterize the wavelet conversion coefficient under the wavelet function effect that vanishing moment is N in scale
Correlation under factor j;
Step 3: using vanishing moment N as abscissa, correlation E makes the E- under each scale factor j as ordinate respectively
N curve;
Step 4: being declined using the numerical value of E with the increase of N as standard curve, by the E-N under each scale obtained by step 4
Curve and " standard curve " compare, and recording curve starts scale numerical value j' and j that " abnormality " occur ";
Step 5: calculating wavelet function actual frequency f1 and f2 corresponding to scale j' and j ", that is, surveyed low-frequency noise letter
The lower frequency limit and upper limiting frequency of 1/f noise in number.
Further, step 5 specifically includes following sub-step:
(1), wavelet function ψ (t) is taken as Daubechies (dbN) small echo, wherein N is 2,3 ..., 10;
(2), when calculating scale j=1, i.e. frequency domain window width before wavelet function ψ stretching;
(3), according to the resulting frequency domain window width of step (2), intensity corresponding to bandpass filter centre frequency is calculated separately
When dropping to a quarter and intensity corresponding to centre frequency drop to 1/10th and it is following when frequency domain window width;
(4), it sets the E-N curve under the conditions of scale j >=j' and abnormality occurs, that is, be unsatisfactory for the pass that E increases with N and decayed
When being or having 10 times and the above volume deviation with normalized curve, frequency domain window of the wavelet function at scale (j'-1) and j' is calculated
Centre frequency, and the intensity according to corresponding to bandpass filter centre frequency obtained by step (3) drops to 1/10th and less
When frequency domain window width calculate wavelet function scale (j'-1) and the frequency domain window of j' edge frequency;
(5), the lower boundary frequency that two windows of gained are calculated in step (4) defines 1/f noise in low-frequency noise
Range where lower frequency limit, lower frequency limit should be between two lower boundary frequencies;
(6), set the E-N curve under the conditions of scale j≤j " and abnormality occur, calculate wavelet function scale j " and (j "+
1) the frequency domain window centre frequency at place, and intensity corresponding to the bandpass filter centre frequency according to obtained by step (3) drops to
Frequency domain window width when a quarter calculates wavelet function in the edge frequency of scale j " and the frequency domain window of (j "+1);
(7), according to the lower boundary frequency of the resulting two frequency domain windows of step (6), it can determine the 1/f in low-frequency noise
The upper limiting frequency of noise is between two lower boundary frequencies.
The method of the present invention is described further below, including wavelet transformation to the decorrelation action principle of 1/f noise and
According to the method for scale determination frequency after being influenced in view of frequency domain window width.
Wavelet transformation acts on the decorrelation of 1/f noise:
For 1/f noise signal x (t), the definition of wavelet transformation is
Wherein ψj,k(t)=2-j/2ψ(2-jT-k), ψ (t) is wavelet function, and j is scale factor, and k is shift factor, xj,kFor
The wavelet coefficient of x (t).The cross-correlation function of wavelet coefficient that wavelet coefficient of the scale equal to j and scale are equal to j' is
The power spectral density of 1/f noise x (t) can be expressed as
Wherein A is a constant, indicates spectrum constant, f is frequency, and γ is frequency index.
It can be released according to pa Savall theorem and wiener-khintchine's theorem
Wherein Ψ (ω) is the Fourier transformation of ψ (t), and * indicates conjugation, and ω is angular frequency.
As j=j',
The Fourier transformation Ψ (ω) of wavelet function ψ (t) with N rank vanishing moment is represented by
Ψ (ω)=ωNΨ0(ω),Ψ0(ω)≠0
Therefore x (t) is in the correlation function of the wavelet coefficient of scale factor j
From the above equation, we can see that when the timing of scale factor one, as long as meeting the condition of 2N+1 > γ, (condition generally can all expire
Foot), the correlation E of wavelet coefficient will be with the increase of the vanishing moment N of wavelet function and reduce, and E-N curve exponentially declines
Trend.Wavelet transformation has significant decorrelation effect to 1/f noise as a result,;It sees in turn, the index decreased of E-N curve
Feature can be used as the standard for judging that noise signal is 1/f noise again.Therefore, once actual measurement noise signal becomes by small echo
The E-N curve obtained by a certain scale does not meet the above standard after changing, and " abnormality " phenomenon occurs, indicates that frequency corresponding to the scale
The noise signal of rate range is not belonging to 1/f noise.
However wavelet function actual frequency corresponding under each scale is in fact the frequency domain window of wavelet function
Center frequency value, not in view of the width bring of wavelet function frequency domain window influences, therefore it is only unusual by above-mentioned appearance
Independent scale corresponding to frequency come the 1/f noise frequency limits that judge in measuring signal be it is inaccurate, this is just needed
Further below to frequency domain window width the considerations of.
2. scale-the frequency interpretation method influenced in view of frequency domain window width
Wavelet functionWherein a indicates that scale, b indicate displacement, its Fourier transformation is
If the frequency domain window centre frequency of Ψ (ω) is ω0, window width is Δ ω, then after stretching and translation
ψa,bThe window center frequency of (ω) isWindow width is
By taking wavelet function db3 as an example, its time domain waveform and frequency wave shape is as shown in fig. 6, its frequency domain window width is equal to
2.In view of the sample frequency F of signalsThe center of wavelet function frequency domain window under each scale can be calculated in=4096HZ
Frequency, referring to table 1.
Table 1: frequency domain window centre frequency corresponding to wavelet function db3 under each scale
The width of wavelet function frequency domain window under each scale is calculated, as shown in table 2.
Table 2: frequency domain window width corresponding to wavelet function db3 under each scale
Compared with prior art, there is the device of the invention lower background noise, higher shielding outside noise to interfere
Ability, the wirewound potential meter itself in device there's almost no 1/f noise, and the data measured are more accurate.Method of the invention
It, can be in addition to that can also use the wavelet transformation of other systems, such as use continuous wavelet transform using other than Dyadic Wavelet Transform
Effectively improve the accuracy of bound frequency detecting;It avoids that it is bent to analyze obtained spectral characteristic using by spectrum analyzer
This conventional method of bound frequency is found in line, fitting.
Detailed description of the invention
Fig. 1 is a kind of structural schematic diagram of device for semiconductor devices 1/f noise bound frequency;
In figure: shielded box 101, electromagnetic wave absorbent material 102, biasing circuit 103, preamplifier 104, digital oscilloscope
105, PC machine 106;
Fig. 2 is a kind of bias circuit construction schematic diagram of device for semiconductor devices 1/f noise bound frequency;
In figure: pluggable wire-wound resistor 201, pluggable semiconductor devices 202, wirewound potential meter 203, sampling resistor 204,
External impressed current table 205, sampling output point 206, switch 207;
Fig. 3 is former 1/f noise time domain plethysmographic signal;
Fig. 4 is the 1/f noise waveform for having clear bound frequency by bandpass filtering;
Fig. 5 is the frequency-domain waveform of original 1/f noise signal and the signal frequency domain waveform after bandpass filtering;
Fig. 6 is E-N curve of the noise under each scale after original 1/f noise and bandpass filtering;
Fig. 7 a is the time domain waveform of wavelet function db3;
Fig. 7 b is the frequency domain window figure of wavelet function db3;
Fig. 8 is the process schematic of 1/f noise lower frequency limit in noise after detection filter;
Fig. 9 is the process schematic of 1/f noise upper limiting frequency in noise after detection filter.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached drawing and specific implementation case
The present invention is described in further detail.
A kind of detection method and device for semiconductor devices 1/f noise bound frequency provided by the invention, it is described
Device has the function of that the 1/f noise for detecting a variety of devices such as diode, MOSFET, triode makes by using storage battery power supply
Obtaining biasing circuit has lower background noise;Wirewound potential meter in device uses wire-wound resistor, itself there's almost no 1/f
Noise;Shielded box internal layer uses electromagnetic wave absorbent material, can effectively adsorptive hindrance noise, outer layer be shielded using double-level-metal,
Shield effectiveness is preferable;By directly digital oscilloscope being used to record data, computer disposal is submitted to, is avoided using low frequency frequency
Spectrum analysis instrument, to keep surveyed data more accurate;It is described herein a kind of for semiconductor devices 1/f noise bound frequency
Detection method, in addition to using the wavelet transformation of other systems other than Dyadic Wavelet Transform, can also be used, such as using continuous small
Wave conversion can effectively improve the accuracy of bound frequency detecting;It avoids analyzing obtained using by spectrum analyzer
This conventional method of bound frequency is found in spectral characteristic curve, fitting.
Referring to Fig. 1, Fig. 1 is a kind of structural schematic diagram of device for semiconductor devices 1/f noise bound frequency, packet
Contain: shielded box 101, electromagnetic wave absorbent material 102, biasing circuit 103, preamplifier 104, digital oscilloscope 105, PC machine
106。
As shown in Figure 1, being shielded outside shielded box 101 using double-level-metal, inside uses electromagnetic wave absorbent material 102, can
To increase substantially the ability of shielding external interference noise, surveyed semiconductor devices is inserted into biasing circuit 103, adjusts biasing
Measured voltage signal is input to preamplifier 104 by BNC by circuit 103, and preamplifier is output data to again
Digital oscilloscope 105, last digital oscilloscope transfer data to PC machine 106, with heretofore described one kind for half
The detection method of conductor device 1/f noise bound frequency is handled.
Referring to fig. 2, Fig. 2 is a kind of bias circuit construction of device for semiconductor devices 1/f noise bound frequency
Schematic diagram includes: pluggable wire-wound resistor 201, pluggable semiconductor devices 202, wirewound potential meter 203, sample resistance 204,
External impressed current table 205, sampling output point 206, switch 207.
Pluggable semiconductor devices selects Metal-Oxide Semiconductor field effect transistor M OSFET.
As shown in Fig. 2, A, B are installation's power source in figure, the DC power supply of free voltage can be, the present apparatus is using 1.5V's
Storage battery power supply.Pluggable wire-wound resistor 201 is a wire-wound resistor, and effect is limitation electric current, protects circuit.It is pluggable partly to lead
Body device 202 can place the devices such as MOSFET, triode, diode, and MOSFET is used in the present invention, and effect is that amplification is inclined
Electric current in circuits can detect different semiconductor devices, suitably modified biasing circuit according to insertion.Wirewound potential meter
203 shared three-level amplifications can change pluggable 202 Current amplifier of semiconductor devices by adjusting the value of wirewound potential meters at different levels
Degree.External impressed current table 205 is adjusted by showing the size of electric current, the i.e. size of source-drain electrode branch current on observation ammeter
Wirewound potential meter 203 is saved, after the completion of adjusting, ammeter is pulled up, connects lower switch 207.Output point 206 is sampled to obtain in source and drain branch
The voltage signal at 204 both ends of sample resistance, is connected to the BNC of shielded box 101.
The present apparatus the method comprises the following steps:
1. being stored after the 1/f noise signal sampling that pair actual measurement obtains, sample frequency Fs=4096HZ.Its time domain waveform
As shown in Figure 3.
2. by above-mentioned 1/f noise by bandpass filter, the parameter of bandpass filter is fL=200HZ, fH=1200HZ,
In the low-frequency noise signal of acquisition the lower frequency limit of 1/f noise ingredient be 200Hz, upper limiting frequency 1200Hz, the noise its
Its frequency range is not 1/f noise, the 1/f noise original signal determined using this signal as bound frequency, time domain waveform such as Fig. 4
It is shown.
3. respectively for former 1/f noise and the related letter under the conditions of noise calculation scale j=1 to j=10 after bandpass filtering
Numerical value, taking correlation function maximum value is E.
4. the E value obtained according to step 3, makes E-N curve, as shown in Figure 6.Blue curve and red curve are distinguished in figure
Indicate the E-N curve of former 1/f noise and filtered noise, therefrom it can be observed that when j=2 and j=3 two curves trend sum number
Magnitude is very close, this shows that the signal frequency range that wavelet frequency domain window is covered when scale j=2 and j=3 is between 1/
Between the bound frequency of f noise signal, i.e., noise signal at this time is 1/f noise.On the one hand as j=4 and j > 4, filter
The E-N curve of noise and original signal curve deviation are larger after wave, do not differ only by larger amt grade, and have " abnormality " situation to go out
It is existing, that is, E does not reduce with the increase of N, this shows that frequency determined by j=4 is 1/f noise in filtered noise signal
Lower frequency limit (scale is bigger, and corresponding frequency is higher).On the other hand, as j=1, the E-N curve of filtered noise is also
There is bigger difference with the degree of closeness of two curves when j=2, this shows that frequency corresponding to j=1 can determine that filtered noise is believed
The upper limiting frequency of 1/f noise in number.
5. since in comparison, the difference of N value smaller part curve is become apparent before and after the filtering of E-N curve, therefore selecting N
Point of penetration of the wavelet function as research when=3.Choose wavelet function db3, time domain waveform and frequency domain window waveform such as Fig. 7
It is shown, it is clear that according to its 0dB window width of frequency domain window waveform to be 2, in the frequency domain window of each scale of the calculation shows that
Frequency of heart is as shown in table 1, and the width of frequency domain window is as shown in table 2 under each scale.
6. judging the lower frequency limit of the 1/f noise of filtered noise signal according to the inflection point of j=4 this scale.In view of j
E-N curve is normal when=3, and curve just starts abnormality occur when j=4, and in view of wavelet function under the two scales
The widths affect of frequency domain window, it can be determined that lower frequency limit should be between 180Hz and 240Hz, and this measurement result and reality
Test the f in settingL=200HZ, i.e., the actual lower limit frequency of 1/f noise is very close in noise.The signal of deterministic process
As shown in figure 8, wherein abscissa indicates frequency, the position of wavelet function frequency domain window when two windows are j=3 and j=4 shows figure
Meaning.The upper limiting frequency of the 1/f noise of filtered noise signal is judged according to the inflection point of j=1 this scale.As j=2, E-N is bent
Line is normal, but when frequency increases to j=1, and curve starts abnormality occur, by wavelet function frequency domain window under the two scales
The width of mouth is taken into account, and is able to detect that upper limiting frequency should be between 615Hz and 1230Hz.In fact the 1/ of setting is tested
F upper noise limit frequency is fH=1200HZ, the result and actual conditions of detection are substantially identical.Schematic diagram as Fig. 9 institute shown in,
Wherein abscissa indicates frequency, the position signal of wavelet function frequency domain window when two windows are j=1 and j=2.
Specific embodiments of the present invention are described above.
In conclusion a kind of detection method for semiconductor devices 1/f noise bound frequency according to the present invention
And device, it can be completed substantially by computer software, hardware spending is small, and it is easy to detect quick, realize advantages of simple, and have
Preferable accuracy and reliability has wide application prospect.
Claims (2)
1. a kind of detection method for semiconductor devices 1/f noise bound frequency, which comprises the following steps:
Step 1: obtaining x (n), n=1,2 after sampling to low-frequency noise signal x with fixed sampling frequency ..., 2M;Wherein, M is big
In 1 integer;
Step 2: being directed to each coefficient matrix, calculate separately the related letter of the wavelet coefficient under its each decomposition scale factor j
Number takes the maximum value E, E of correlation function to characterize the wavelet conversion coefficient under the wavelet function effect that vanishing moment is N in scale factor
Correlation under j;
Step 3: using vanishing moment N as abscissa, for correlation E as ordinate, the E-N made under each scale factor j respectively is bent
Line;
Step 4: being declined using the numerical value of E with the increase of N as standard curve, by the E-N curve under each scale obtained by step 3
It is compared with " standard curve ", recording curve starts scale numerical value j' and j that " abnormality " occur ";
Step 5: calculating wavelet function actual frequency f1 and f2 corresponding to scale j' and j ", that is, in surveyed low-frequency noise signal
1/f noise lower frequency limit and upper limiting frequency.
2. a kind of detection method for semiconductor devices 1/f noise bound frequency as described in claim 1, feature exist
In the step 5 includes following sub-step:
(1), wavelet function ψ (t) is taken as Daubechies (dbN) small echo, wherein N is 2,3 ..., 10;
(2), when calculating scale j=1, i.e. frequency domain window width before wavelet function ψ stretching;
(3), according to the resulting frequency domain window width of step (2), the decline of intensity corresponding to bandpass filter centre frequency is calculated separately
When to a quarter and intensity corresponding to centre frequency drop to 1/10th and it is following when frequency domain window width;
(4), set the E-N curve under the conditions of scale j >=j' and abnormality occur, that is, be unsatisfactory for E with N increase and decay relationship or
When having 10 times and the above volume deviation with normalized curve, frequency domain window center of the wavelet function at scale (j'-1) and j' is calculated
Frequency, and the intensity according to corresponding to bandpass filter centre frequency obtained by step (3) drop to 1/10th and it is following when
Frequency domain window width calculates wavelet function in the edge frequency of scale (j'-1) and the frequency domain window of j';
(5), the lower boundary frequency that two windows of gained are calculated in step (4) defines the lower limit of 1/f noise in low-frequency noise
Range where frequency, lower frequency limit should be between two lower boundary frequencies;
(6), it sets the E-N curve under the conditions of scale j≤j " and abnormality occurs, calculate wavelet function at scale j " and (j "+1)
Frequency domain window centre frequency, and the intensity according to corresponding to bandpass filter centre frequency obtained by step (3) drops to four points
The window width of frequency domain for the moment calculate wavelet function scale j " and the frequency domain window of (j "+1) edge frequency;
(7), according to the lower boundary frequency of the resulting two frequency domain windows of step (6), it can determine the 1/f noise in low-frequency noise
Upper limiting frequency between two lower boundary frequencies.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710531992.4A CN107526016B (en) | 2017-07-03 | 2017-07-03 | A kind of detection method and device for semiconductor devices 1/f noise bound frequency |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710531992.4A CN107526016B (en) | 2017-07-03 | 2017-07-03 | A kind of detection method and device for semiconductor devices 1/f noise bound frequency |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107526016A CN107526016A (en) | 2017-12-29 |
CN107526016B true CN107526016B (en) | 2019-08-27 |
Family
ID=60748743
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710531992.4A Expired - Fee Related CN107526016B (en) | 2017-07-03 | 2017-07-03 | A kind of detection method and device for semiconductor devices 1/f noise bound frequency |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107526016B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108258939B (en) * | 2018-01-29 | 2019-06-11 | 河海大学常州校区 | A kind of decompression potted landscape device based on 1/f noise |
CN109884481A (en) * | 2019-02-18 | 2019-06-14 | 上海金艺检测技术有限公司 | Shelf depreciation transient state voltage-to-ground scene noise evaluation method |
CN109669115A (en) * | 2019-02-22 | 2019-04-23 | 江苏云意电气股份有限公司 | One kind being used for automobile-used heavy-duty diode reliability non-destructive testing device |
CN111880019A (en) * | 2020-08-19 | 2020-11-03 | 江苏云意电气股份有限公司 | High-power diode low-frequency noise test system for multi-way vehicle |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS61220582A (en) * | 1985-03-26 | 1986-09-30 | Sony Corp | Separation circuit for luminance signal/chrominance signal |
US6917717B2 (en) * | 2002-03-26 | 2005-07-12 | Microsoft Corporation | System and method for image compression using wavelet coding of masked images |
CN1829908A (en) * | 2003-07-30 | 2006-09-06 | 皇家飞利浦电子股份有限公司 | Integrated 1/f noise removal method for a magneto-resistive nano-particle sensor |
-
2017
- 2017-07-03 CN CN201710531992.4A patent/CN107526016B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS61220582A (en) * | 1985-03-26 | 1986-09-30 | Sony Corp | Separation circuit for luminance signal/chrominance signal |
US6917717B2 (en) * | 2002-03-26 | 2005-07-12 | Microsoft Corporation | System and method for image compression using wavelet coding of masked images |
CN1829908A (en) * | 2003-07-30 | 2006-09-06 | 皇家飞利浦电子股份有限公司 | Integrated 1/f noise removal method for a magneto-resistive nano-particle sensor |
Also Published As
Publication number | Publication date |
---|---|
CN107526016A (en) | 2017-12-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107526016B (en) | A kind of detection method and device for semiconductor devices 1/f noise bound frequency | |
CN102095917B (en) | Test method for current noise of high-resistance device and medium material | |
CN101871975B (en) | System and method for testing cable transfer impedance time domain | |
CN108680955B (en) | A kind of Lithology Identification Methods and identifying system | |
US20120098549A1 (en) | Electrical tomography apparatus and method and current driver | |
CN110389290B (en) | System, method and device for testing noise and evaluating service life of electronic component | |
CN104020365A (en) | Method for testing low-frequency noise of amorphous silicon membrane | |
CN105891761A (en) | Scattering parameter-based partial discharge ultrahigh frequency detection system field check method | |
CN108345039A (en) | A method of eliminating adjacent frequency harmonic wave interference in ground nuclear magnetic resonance data | |
CN102176667B (en) | Power frequency digital notch device and method based on frequency track | |
CN113721071A (en) | System and method for measuring non-intrusive voltage to ground | |
Li et al. | Random noise suppression and parameter estimation for Magnetic Resonance Sounding signal based on maximum likelihood estimation | |
CN103954851B (en) | Noise coefficient measuring method and noise coefficient standard device | |
Prochazka et al. | Statistical analysis and digital processing of the Mössbauer spectra | |
CN106772193A (en) | A kind of current transformer frequency characteristic measurement devices and methods therefor | |
CN104569894B (en) | A kind of partial-discharge ultrahigh-frequency sensor performance evaluation method based on time domain parameter | |
CN106353589A (en) | Coupling detector | |
CN206270482U (en) | A kind of current transformer frequency characteristic measurement device | |
CN106468789A (en) | A kind of sensor performance Quantitative Evaluation System | |
CN105445549B (en) | A kind of weak electromagnetic signal frequency spectrum measuring method | |
CN113049086B (en) | Dynamic database forming method and device for acoustic diagnosis of transformer | |
CN115435895A (en) | Background noise measuring method based on probability distribution function | |
KR101145267B1 (en) | System and method for measuring impedance and noise characteristic simultaneously, and a medium having computer readable program for executing the method | |
Biroth et al. | A low-noise and fast pre-amplifier and readout system for SiPMs | |
CN102230950B (en) | Method for diagnosing apparatus internal electromagnetic interference failure |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190827 |