Embodiment
In line communication system, signal transmits through the entity circuit, so passage length immobilizes.In other words, the influence that passage length produces each signal all is the same, and perhaps, each receives the imperfect factor that produces because of passage length in the signal all is fixing.Yet, because it is very many influence the factor of passage, if analyze the influence that passage length is caused to received signal, is time and effort consuming and is difficult for carrying out, and influence the accurately indivedual surely quantifications that also differ of each factor that receives signal.Under this situation, the present invention utilizes the notion of autoregression (Autoregressive), analyze current demand signal and early stage the relation between signal, and then judge the effect that environmental factor produces all signals.
Autoregression is a model commonly used in the statistics, and it utilizes previous (one or more) data in the time series, explains current data.And in line communication system, then may be interpreted as with transmission result in the past and represent present transmission result, by this, can judge auto-correlation function, and then estimating channel length.
At first, summary autoregression model.Be noted that, the statistics skill that autoregression model is common, pertinent literature should be those skilled in the art and can obtain easily, so do not detail.
The linear autoregressive models on one k rank is defined as:
and
In
formula 1,
The PARCOR coefficients (PartialAutocorrelation Parameter) that are called as the k phase poor (k-th lag) of Xt,
Be called as partial autocorrelation function (PartialAutocorrelation Function, PACF), and { ρ
i}
1 ∞The auto-correlation function of model for this reason then, it can reflect the degree of correlation of a sequence data between difference sampling value constantly.With line communication system,, make passage length receive the identical factor of influence of signal meeting " contribution ", and this identical factor of influence can be reflected in the auto-correlation function to each because passage length is fixing.Therefore, can have certain corresponding relation between auto-correlation function and passage length, and this corresponding relation can utilize in advance experiment or analog form to judge.In other words, in line communication system, judge the auto-correlation function of a sequential reception signal when receiving terminal after, can with the comparison of the experimental result that obtains in advance, and then estimate and passage length.Wherein, calculate auto-correlation function { ρ
i}
1 ∞Method, algorithm etc. also extensively be disclosed in the pertinent literature, and non-emphasis of the present invention is so will not give unnecessary details.
Above-mentioned is notion of the present invention, can further reduce passage length estimation flow process 10, as shown in Figure 1.Passage length estimation flow process 10 is used for line communication system estimation one passage length, and it comprises following steps:
Step 100: beginning.
Step 102: receive a plurality of signals that this passage transmits continuously.
Step 104: judge a signal in these a plurality of signals and the relation between at least one signal before it, to produce a judged result.
Step 106:, judge this passage length according to this judged result.
Step 108: finish.
In simple terms, after receiving continuous signal, receiving terminal autoregression model capable of using with previous signal description current demand signal, and then is judged the relation between current demand signal and first front signal.Then, with the auto-correlation function and the comparison of preset experimental result of autoregression model gained, then can estimate passage length.
Realization about passage length estimation flow process 10 please refer to Fig. 2, and Fig. 2 is the sketch map of a passage length estimating apparatus 20.Passage length estimating apparatus 20 is used for estimating the length of the formed channel C H of a cable, and it includes a receiving terminal 200, one first judging unit 202 and one second judging unit 204.Receiving terminal 200 is used for performing step 102, i.e. signal S1~Sn of transmitting continuously of receive path.First judging unit 202 is used for performing step 104, promptly utilizes autoregression model, with previous signal description current demand signal, and then judges the relation between current demand signal and first front signal.At last, second judging unit 204 is used for performing step 106, with the auto-correlation function and the comparison of preset experimental result of autoregression model gained, with the estimation passage length.
Be noted that passage length estimating apparatus 20 is embodiments of the invention, in order to the length of estimation channel C H.In fact, the employed mechanics of communication of different line communication systems, channel parameters etc. all maybe be different, so tool common operator in this area does suitable adjustment, with correct estimation passage length when cooperating different system to need according to above stated specification.
For instance, please refer to Fig. 3, Fig. 3 is the sketch map of the embodiment of the invention one receiving system 30.Receiving system 30 is used for a Gigabit Ethernet, and it includes an automatic gain control amplifier 300, a filtration module 302, an analog-to-digital converter 304, one anti-partial-response filtering device 306 and a passage length estimating module 308.Automatic gain control amplifier 300 can be adjusted gain automatically, amplifies the amplitude of the analog signal x (t) that is received by a channel C Hx with appropriateness.Filtration module 302 includes a low pass filter 310, a high pass filter 312 and a multiplexer 314.Low pass filter 310 can carry out the LPF computing to the signal that automatic gain control amplifier 300 is exported, and high pass filter 312 then carries out the high-pass filtering computing to the low pass output signal of low pass filter 310.In other words, the signal exported of high pass filter 312 is a bandpass signal.The control signal CTRL that multiplexer 314 can be exported according to passage length estimating module 308; The bandpass signal that low-pass signal that low pass filter 310 is exported or high pass filter 312 are exported is sent to analog-to-digital converter 304, to simulate to digital translation.In addition, anti-partial-response filtering device 306 is used for compensating the partial-response filtering device of transmission ends, and the digital signal y of its output [n] can supply passage length estimating module 308 to judge the length of channel C Hx.The implementation of passage length estimating module 308 is similar to the passage length estimating apparatus 20 of Fig. 2; Also be to utilize autoregression model; With previous signal description current demand signal; And then judge the relation between current demand signal and first front signal, again with auto-correlation function and the comparison of preset experimental result, with the estimation passage length.Passage length estimating module 308 is that with passage length estimating apparatus 20 differences passage length estimating module 308 handled signals are digital pattern, and passage length estimating apparatus 20 handled signals then are the simulation pattern.Moreover, should comprise a control signal generator in the passage length estimating module 308, be used for producing control signal CTRL according to passage length estimation result.Preferably, before judging passage length, the LPF result of passage length estimating module 308 control multiplexers 314 output low pass filters 310 is to analog-to-digital converter 304.
In receiving system 30, the element beyond the passage length estimating module 308 is the element that has used in the receiving system of known Gigabit Ethernet, and the non-emphasis of the present invention of detailed operation mode is not so give unnecessary details.
Illustrate further notion of the present invention, please refer to Fig. 4, Fig. 4 is the power spectrum density sketch map of 0 and 150 meter following digital signal y [n] for channel C Hx among Fig. 3.Can know by Fig. 4, when the length of channel C Hx more in short-term, receive the noise effect degree less, so signal energy can be evenly distributed in each frequency range; Otherwise, when the length of channel C Hx is longer, receive the noise effect degree higher, so signal energy is concentrated in low frequency.Such result also meets aforesaid notion, that is, when the length of channel C Hx more in short-term, passage length is lower to the contribution degree of the imperfect factor of signal; And when the length of channel C Hx was longer, except noise, passage length also can be contributed the imperfect factor to a certain degree.Wherein, noise supposes that generally it meets the additive white gaussian model, and promptly auto-correlation function is 0 random signal.In other words, arbitrary sampling value of noise is irrelevant with previous sampling value.Therefore, utilize the auto-correlation function of autoregression model gained can reflect the imperfect factor that passage length is contributed signal, can judge passage length by this.
For example, if passage length estimating module 308 can be known with the length of single order autoregression model estimation channel C Hx:
Y [n]=wy [n-1]+e [n], (formula 2)
Wherein, e [n] representes noise, and under the situation of single order, w is auto-correlation function.According to formula 2, can further derive following value equation:
J=E{y [n]-wy [n-1] |
2, (formula 3)
By this, can get the optimum value w of w
Opt, as follows:
(formula 4)
Utilize the lowest mean square algorithm again, can get following formula, wherein μ representes step size:
w
N+1=w
n+ 2 μ y [n-1] (y [n]-w
nY [n-1])), (formula 5)
Or w
N+1=w
n+ 2 μ sign{y [n-1] } (y [n]-w
nY [n-1])), (formula 6).
At last, through testing the relation that can get w and channel C Hx length, as shown in Figure 5.Therefore, utilize Fig. 5, after passage length estimating module 308 calculates w, can compare with it, and then judge the length of channel C Hx.
Be noted that above-mentioned formula 2 to formula 6 is in order to explaining of the present invention one possible execution mode, and be not limited thereto.In fact, the derivation mode of the exponent number of autoregression model, auto-correlation function, experimental result etc. all can be according to the different system demands, and the appropriateness adjustment, correctly to judge passage length.
In addition, receiving system 30 shown in Figure 3 is not limited thereto in order to explain that of the present invention one possibly use, and those skilled in the art do the appropriateness adjustment when required according to system.For example, Fig. 6 is the sketch map of the embodiment of the invention one receiving system 60.Receiving system 60 also is used for Gigabit Ethernet; The all similar receiving system 30 of its framework and function mode; Be used for utilizing the auto-correlation function of autoregression model gained to judge passage length, and the LPF result is exported in control according to this or the bandpass filtering result gives analog-to-digital converter.Receiving system 60 and receiving system 30 differences are to have increased echo and near-end crosstalk arrester 607 between anti-partial-response filtering device and the passage length estimating module of receiving system 60, more accurately to estimate passage length.
Above-mentioned receiving system 30 or 60 function mode can further reduce one and receive signal processing flow 70, and as shown in Figure 7, it comprises following steps:
Step 700: beginning.
Step 702: launch low pass filter, and inactive high pass filter.
Step 704: utilize autoregression model to judge passage length.
Step 706: judge that whether passage length is greater than a preset value.If carry out step 708; If not, carry out step 710.
Step 708: launch high pass filter, with output bandpass filtering result.
Step 710: finish.
Receiving signal processing flow 70 explanation receiving systems 30 or 60 is that autoregression model is judged passage length, and control is only launched low pass filter or launched low pass filter and high pass filter simultaneously according to this, and detailed description can be with reference to aforementioned.
In sum, the present invention utilizes autoregression model, with previous signal description current demand signal; And then judge the relation between current demand signal and first front signal; With the auto-correlation function and the comparison of preset experimental result of gained,, and judge whether to launch high pass filter according to this again with the estimation passage length.In other words, the present invention need not send pulse signal, also need not wait for the eqalizing cricuit convergence, can estimate passage length.Under this situation, estimate the partial parameters that the passage length that can be used for determining eqalizing cricuit, further prevent the generation of error propagation.Therefore, the present invention can effectively estimate passage length, makes receiver launch high pass filter in good time, with the generation probability of reduction error propagation, and improves receiving efficiency.
The above is merely the preferred embodiments of the present invention, and all equalizations of doing according to claims of the present invention change and modify, and all should belong to covering scope of the present invention.