Classifications

• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
  • H03H17/00—Networks using digital techniques
  • H03H17/02—Frequency selective networks
  • H03H17/0248—Filters characterised by a particular frequency response or filtering method
  • H03H17/0251—Comb filters
• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03H—IMPEDANCE NETWORKS, e.g. RESONANT CIRCUITS; RESONATORS
  • H03H17/00—Networks using digital techniques
  • H03H17/02—Frequency selective networks
  • H03H17/06—Non-recursive filters
  • H03H17/0621—Non-recursive filters with input-sampling frequency and output-delivery frequency which differ, e.g. extrapolation; Anti-aliasing
  • H03H17/0635—Non-recursive filters with input-sampling frequency and output-delivery frequency which differ, e.g. extrapolation; Anti-aliasing characterized by the ratio between the input-sampling and output-delivery frequencies
  • H03H17/065—Non-recursive filters with input-sampling frequency and output-delivery frequency which differ, e.g. extrapolation; Anti-aliasing characterized by the ratio between the input-sampling and output-delivery frequencies the ratio being integer
  • H03H17/0657—Non-recursive filters with input-sampling frequency and output-delivery frequency which differ, e.g. extrapolation; Anti-aliasing characterized by the ratio between the input-sampling and output-delivery frequencies the ratio being integer where the output-delivery frequency is higher than the input sampling frequency, i.e. interpolation

Definitions

• a digital input signal is converted into an analog signal using a digital / analog converter
• the quality of the signal obtained depends on the noise level of the digital input signal.
• the sampling rate and thus the cutoff frequency are increased by inserting zeros between the data values and by digitally filtering the data sequence.
• the signal rate should be increased by a factor of 20-400.
• the signal rate is first increased by a factor of 4-16 by inserting zeros alternately. The so obtained
• a digital comb filter to increase the signal rate by powers of 2 is through the transfer function
• Figure 1 shows an implementation of such a comb filter according to the prior art.
• a digital input signal 1 with the signal rate f s is first through a first filter stage 2 with the filter characteristic
• the result values of the repeater 3 are a further digital filter stage 4 with the transfer function
• each filter stage has a repeater in the form of a holding element, which outputs each input data value twice and thus doubles the signal rate.
• This repeater is followed by a filter structure, for whose partial transfer function H (z) applies
• k denotes the order of the filter device and z "1 denotes the z-transform of a delay by one sampling pulse.
• the implementation according to the invention has the advantage that it does not use any recursive or feed-backward structures. Neither the repeater nor the filter structure provided on each filter stage provides a feedback of the data values available at the output to the input values. This can be seen from the fact that the partial transfer function
• H (z) has no poles and is therefore a pure feed
• the transfer function of a comb filter can only be achieved with the filter device according to the invention by using feed-forward structures. Recursive structures can be completely avoided, and any bit errors that occur quickly disappear because they are not fed back to the input values.
• the comb filter implementation according to the invention has no error memory, and therefore no circuit for error correction is required. For this reason, the circuitry outlay and thus also the need for silicon area in the construction of the comb filter according to the invention is less than in the known solutions of the prior art.
• each filter structure comprises an adder whose value is formed at the output by adding the data value currently applied to the filter structure and the data value associated with the previous sampling pulse.
• the transfer function H (z) of the digital filter device applies
• k denotes the order of the filter device and 2 n denotes the factor of the sampling rate increase, and where z _1 denotes the z-transform of a delay by one sampling pulse.
• the transfer function H (z) is normalized by a suitable number of right shifts in the bit pattern of the data values.
• Right shifts are easy to implement in digital technology. Any right shift
• the data bus between the holding element and the filter structure of the jth filter stage has a width of at least
• the advantage of this solution is that the data buses only have the minimum bus width required, which is why the circuitry can be kept low.
• the data bus at the input of the first filter stage has a width of WL bits.
• the holding element merely doubles the sampling rate, but the data values themselves remain unchanged. For this reason, the data bus between the holding element and the filter structure of the first filter stage has a width of WL data lines.
• the data is then processed through the filter structure of the first filter stage. This filter structure is due to the transfer function
• the data bus at the output of the first filter stage must therefore have a width of at least
• the method according to the invention for increasing the sampling rate and signal reconstruction of discrete-time input data is characterized by the following two steps, which are repeated n times: First, the sampling rate is doubled by outputting each input data value twice. The data values are then digitally filtered using a filter structure, for whose partial transfer function H (z) applies
• the filter structure required for the digital filtering step is a pure feed-forward structure and requires little circuitry. The invention is described below with reference to an embodiment shown in the drawing. Show it:
• FIG. 1 shows an implementation of a digital comb filter corresponding to the prior art, in which recursive filter structures are used;
• Fig. 1 shows the circuit of a digital comb filter with the transfer function
• the transfer function of the comb filter can therefore also be represented in this factorized notation.
• the signal rate is increased by inserting a corresponding number of zeros between the individual data values.
• the signal rate f s In order to increase the signal rate f s by the factor m to (m • f s ), zeros must be inserted between two data values (m-1).
• the following notation is intended
• the signal rate is to be increased by a factor of 2 n .
• the transformation of the input data caused by the comb filter can thus be represented as follows:
• z "q is used to access the data position behind the q positions.
• the data value lying behind the q positions is used before the insertion of (m-1) zeros, or if, on the other hand, after the insertion of (m-1) zeros on the (m • q) positions previous data value is used, because in both cases the same data value is accessed.
• a frequency doubling ie when a zero is inserted between the data values, arises as a special case
• the implementation according to the invention of this reshaped comb filter characteristic is shown in FIG. 2.
• the comb filter consists of n stages, with a frequency doubling at the input of each stage by inserting each a zero is made between the data values.
• the frequency-doubled signal thus obtained is then fed to a filter unit with the filter characteristic (l + z -1 J) and filtered.
• the input signal 8 with the frequency f s is present at the input of the first filter stage 11.
• this signal first passes through a stage of zero insertion 9, in which the frequency is doubled to (2 • f s ).
• the subsequent filter unit 10 filters the signal, but leaves the signal rate unchanged.
• the signal present at the output of the first filter stage 11 is fed to the second filter stage 12, which in turn causes a frequency doubling.
• the signal 14 After passing through the nth filter stage 13, the signal 14 is obtained which already has the required signal rate (2 n • f s ). Through the attenuator 15, the signal 14 around
• the desired comb-filtered end signal 16 is then present, the power of which corresponds to the power of the input signal 8.
• the weakening is brought about by subjecting the bit patterns of the individual data values (n • k - n) to right shifts, because each right shift causes a shift
• Feed-forward structures are used at each stage of the filter device according to the invention. Therefore, bit errors that occur do not affect the subsequent results, and errors quickly subside.
• a complex error correction circuit as is required in the prior art, can be completely avoided with this implementation of a comb filter.
• a filter arrangement is shown on the left in FIG. 3, which consists of a stage of zero insertion 17 and a downstream filter unit 18.
• the stage of zero insertion 17 inserts a zero between two data values, so that the data sequence ..., 0, c, 0, b, 0, a, 0 appears at the output of stage 17.
• This data sequence is present at the input of the digital filter unit 18, the transfer function of which is given by (1 + z _1 J.
• the filter unit • 18 adds the previous data value to the data value currently present at its input. Because the data sequence at the input of the filter unit 18. .., 0, c, 0, b, 0, a, 0 is present, the data sequence appears at the output of filter stage 18 ..., c, c, b, b, a, a.
• Each of the n filter stages in FIG. 2 contains a filter unit 10, which is characterized by the transfer function fl + z -1 J.
• Each of these filter units can be broken down into two filter units arranged one behind the other; in a first filter unit with a transfer function (l + z _1 and in a second filter unit with the transfer function
• the entire filter device comprises n filter stages, with each filter stage having a repetition stage arranged at the beginning and a downstream filter structure.
• the filter structure then serves as the input of the next filter stage.
• a digital, WL bit-wide signal of the signal rate f s is present at the input 20 of the first filter stage 25.
• This signal is fed to the holding element 21 belonging to the first filter stage 25.
• This hold- Link 21 is sampled at twice the signal rate (2 • f s ).
• Each data value present at input 20 therefore appears twice on data bus 22, which connects holding element 21 to filter structure 23. Since the data values are present at the beginning, a width of WL bits is sufficient for the data bus.
• the filter device according to the invention has the same comb filter characteristic as the solution shown in FIG. 1 according to the prior art.
• Data lines must have that the data bus between the holding element and the filter structure of the jth filter stage has a width of at least
• This filter structure is implemented by an adder 30, to which the current data value 31 and the previous data value 32 are supplied. The result 33 the addition becomes the repeater 34 of the next filter stage
• Each of the n filter stages of the filter device contains one
• an adder 36 is provided, to which the current data 37, the data 39 multiplied by 2 and belonging to the previous sampling pulse, and the data 40 belonging to the previous sampling pulse are supplied.
• the result 41 of the adder can then be fed to the next filter stage.
• the multiplication by 2 is usually brought about by a left shift 38 of the bit pattern of the data value, so that no complex multiplication circuit is required for this.

Landscapes

• Physics & Mathematics (AREA)
• Engineering & Computer Science (AREA)
• Computer Hardware Design (AREA)
• Mathematical Physics (AREA)
• Dc Digital Transmission (AREA)
• Compression, Expansion, Code Conversion, And Decoders (AREA)

Applications Claiming Priority (3)

Publications (1)

Family

ID=7645780

Family Applications (1)

Country Status (5)

Families Citing this family (7)

Family Cites Families (6)

• 2000
  • 2000-06-15 DE DE10029424A patent/DE10029424C2/de not_active Expired - Fee Related
• 2001
  • 2001-05-21 WO PCT/DE2001/001957 patent/WO2001097376A1/de not_active Application Discontinuation
  • 2001-05-21 EP EP01947161A patent/EP1290791A1/de not_active Withdrawn
  • 2001-05-21 JP JP2002511467A patent/JP2004503976A/ja active Pending
• 2002
  • 2002-12-16 US US10/320,126 patent/US7076512B2/en not_active Expired - Lifetime

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events