Coded data generation or conversion – Digital code to digital code converters – Data rate conversion
Reissue Patent
2001-06-05
2003-06-17
Williams, Howard L. (Department: 2819)
Coded data generation or conversion
Digital code to digital code converters
Data rate conversion
C708S313000
Reissue Patent
active
RE038144
ABSTRACT:
BACKGROUND OF THE INVENTION
Field of the Invention
The invention relates to a comb filter containing series connected integrators, to which is fed a digital data stream at a high sampling rate in order to yield a digital data stream at a low sampling rate.
So called decimation filters are used to convert a digital data stream having a high sampling rate into a data stream having a low sampling rate. The filter function of such decimation filters usually has low pass filter properties and effects an increase in the word width of the data decimated by the decimation filter. A 1 bit stream at a high sampling rate can thus be decimated into, for example, 16 bit words in the voice band.
One field of application is digital telephony. In this case, at least one such decimation filter is connected downstream of a Sigma Delta analog/digital converter.
The decimation factor of a decimation filter is usually determined by a power of two, that is to say 2, 4, 16, 64, etc. The usual structure for such a decimation filter is afforded by a comb filter, since a filter of the latter type can be realized with a relatively low outlay.
SUMMARY OF THE INVENTION
It is accordingly an object of the invention to provide a comb filter which overcomes the above-mentioned disadvantages of the prior art device of this general type, which is suitable as a decimation filter and is distinguished by a simple structure.
With the foregoing and other objects in view there is provided, in accordance with the invention, a comb filter, including:
at least three series connected integrators, having a first integrator and a last integrator, receiving at a high sampling rate a digital data stream having bits including most significant bits in order to yield the digital data stream at a low sampling rate; and
a counter generating a counter reading connected to the first integrator and to the last integrator, the most significant bits of the digital data stream being reset in the at least three series connected integrators, the counter storing a resetting of the bits in the first integrator, and the counter reading of the counter fed selectively into at least one bit of the bits of the last integrator at a reset time.
In the case of the comb filter of the type mentioned in the introduction, the object is achieved according to the invention by virtue of the fact that the most significant bits of the digital data stream can be reset in the integrators, that it is possible to store the resetting of the bits in the first integrator in a counter, and that the counter reading of the counter can be fed into at least one bit, preferably the most significant bits, of the last integrator at the reset time.
The invention has the advantage that a reset error appears as a multiple of a power of two in subsequent stages, with the result that it can be connected in a particularly simple manner in the subsequent stages.
In this case, the number of integrators is not limited in principle, although three integrators are sufficient in many applications. Correspondingly, reference is made below to exemplary embodiments having three integrators, i.e. third order filters.
In the case of the comb filter according to the invention, unlike in the known comb filter of
FIG. 2
, therefore, the reset signal is not passed to the integrators in their entirety, rather it is applied only to the most significant bits. The result of this is that just a simple counter is required as the arithmetic unit for the correction of the reset errors. In particular, the last integrator, and that is to say the third integrator in the case of a three stage configuration, is preset not to zero but to a value determined by the counter.
The reasons for this shall be explained in more detail below:
A third order comb filter with a decimation factor of 256 (2
8
) has a number range from zero to 256
3
(16,777,216). If the representation of 256
3
is then dispensed with and the value zero is accepted for this, then 24 bits are necessary for a representation. 25 bits would be required in the case of a representation of 256
3
, which results in that the following considerations shift by 1 bit. Computational errors modulo 256
3
accordingly are not noticeable and can be ignored. Thus, a computational accuracy of 24 bits suffices in order to obtain an exact result.
If an error f1 occurs in the first integrator at a particular time, which may happen as a result of resetting, for example, then the error is evaluated after 256 clock cycles at the output of the third integrator, where the error is then f3=f1 *256*257/2=f1*128*(256+1). If the error f1 amounts to a*512, for example, then f3=a*512*128*256+a*512*128 is produced. In other words, a*256
2
when considered modulo 256
3
.
This results in the following:
If the third integrator is set not to the value zero but to the value −a*256*256, then the error f1 is compensated for. The same applies correspondingly not just to 256 clock cycles but also to any desired multiples thereof. The reason is that the third integrator simultaneously operates as a differentiator as well, and that when the accumulating values are differentiated, −a*256*256 is always present as the result.
In this case, the value of a is produced as shown next. In the first integrator, the bit with the significance of 512 is reset at the reset time. If the bit previously had the value 1, then the error f1 specified above is produced exactly. This error is added to the preceding errors that occurred correspondingly. Therefore, the error counter of the first integrator is simply incremented in this case. The error counter loads, in turn, the last or third integrator. The negative sign disappears because a is usually negative, since the error arises when a bit is reset.
In order, furthermore, to prevent the occurrence of any further errors, the first integrator must always have a width of 10 bits. The error counter, which is connected downstream of the first integrator, contains multiples of 256
2
and must accordingly have a width of 8 bits, since an overall computational accuracy of 24 bits is presupposed. That is to say that a “small” counter is sufficient.
Therefore, values up to a maximum of 767 come from the first integrator. In 256 clock cycles, this value can accumulate to at most values of less than/equal to 767*256=196352. The start value of the second integrator is less than 216, with the result that a word width of 18 bits is sufficient to avoid an uncontrolled overflow. At the reset time, all bits which symbolize 216 and multiples thereof can be erased because, after 256 clock cycles, the accumulative error in the third integrator has the value 224 or a multiple thereof and, therefore, need not be taken into account. That is to say that the start value after resetting is in the second integrator, in other words has values which are less than 216, as has already been assumed above.
Overall, the comb filter according to the invention therefore operates in the below list manner.
Synchronous resetting of the most significant bits is performed by 1 bit in the first integrator and by 2 bits in the second integrator. In this case, the resetting of the bits in the first integrator is stored in an error counter that is connected downstream of the first integrator and has a word width of 8 bits. The counter reading thus determined is then fed into the most significant bits of the last or third integrator, which simultaneously acts as a differentiator, at the reset time. The less significant bits of the last or third integrator are reset in a customary manner, further computational steps not being necessary. The invention thus provides a comb filter that is constructed in a simple manner, can be realized with a low outlay and can reliably convert a digital data stream having a high sampling rate into a digital data stream having a lower sampling rate.
Other features which are considered as characteristic for the invention are set forth in the appended claims.
Although the invention is illustrated and described here
LandOfFree
Comb filter does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Comb filter, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Comb filter will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3127725