Digital Filter

Oversampling and interpolation is performed here. As an option digital emphasis is available, controlled by the input receiver. The SM5842 digital filter of Nippon Precision Circuits is placed between the input receiver and the DAC chips.

The input of this filter is a 16 bit resolution signal, according to the CD standard.

The outputs of the filter are two data streams which are eight times oversampled, and have 20 bit resolution. Internally 32 bit arithmetic is used to obtain more than 117dB damping in the stop band, and the ripple in the passband is 0.0002dB. The filter has linear phase response.

As a special feature of the SM5842, the output data can be (slightly) asynchronously clocked with respect to the input data. This enables the easy use of an external, low jitter clock, to clock data to the DA converters. A special circuit in-between the filter and the DAC chips is used, to reclock all digital signals entering the DAC chips. This reclocking circuit is described elsewhere (clock regeneration and jitter).

Implementation considerations

Audio input data is in 2-complement, MSB first format, hence the data output of the Crystal input receiver can be connected to the input DI of the digital filter directly (INF1N should be low, 16 bit after packing, alternating left and right channel). The length of the input word is determined by IW1N and IW2N, and are both set to high to set it to 16 bit input word length (CD standard). We haven't made a provision for other input word lengths (for new formats like DCC and DAB), as a lot of extra circuitry is needed to extract the input word length from the SPDIF data (the channel data block must be assembled, and a mapping from the category bits to an input word length code must be established). Audio data output is 20 bits (OW1N low, OW2N high), in order to be compatible with the PCM63. When CKSLN is low, the bit clock rate BCKO equals 256.fs, and the data word lenght equals 24.tsys, according to the input format of the PCM-63K used for DA-conversion (see corresponding chapter for details). Pre-emphasis is a result of bad engineering of Sony. At the time CD was introduced, they were not capable of producing decent DACs, and since then each recordable media must support pre/deemphasis. The current audio standards w.r.t. deemphasis are as follows. DAB is 48kHz with no pre-emphasis. In practice we don't expect recordings with pre-emphasis on 32 and 48kHz. The digital filter needs information about the input rate of the data such that the right deemphasis can be applied. To extract the frequency information from the input receiver requires some extra complex circuitry (because the Crystal input receiver shares the frequency status information at the same pins as other important status information). However, in our main application (CD) deemphasis will only be necessary at 44.1kHz, and hence we have decided not to support deemphasis for other sample frequencies. Therefore, FSEL1 and FSEL2 are both low.

The digital filter has a pin which allows to add dither to the signal. We've tried this feature, but couldn't detect any audible difference.