Improved digital input stage for
the Behringer DCX2496
The digital input stage of
the Behringer DCX2496 has three disadvantages:
The maximum sample rate
of the input signal is 96kHz.
With the CS8420 a first
generation SRC was used. It has a minor jitter rejection leading to
jitter components in its digital output.
The PCB-layout around
the CS8420 is not compatible with the recommendations of Cirrus Logic
leading to a high jitter at sample rates of 96kHz.
There is no correction
of emphasis leading to playback errors with emphassis CDs.
This inspired me to
design a new digital input stage (incl. PCB) with the following features:
Low noise power supply
with LM2936 and Black-Gate capacitors.
Usage of the CS8416
AES-S/PDIF receiver to allow input sample rates up to 192kHz.
Optimized PCB layout for
minimal jitter during clock regeneration for the
Usage of a second
generation SRC (AD1896) with delay line and digital PLL. This leads to
a jitter rejection down to less than 5pS!
High SNR (AD18961-
Simple installation in
The PCB was designed on a strict separation of digital and analog
power supplies. Both the analog und digital supply voltages (5V each)
were taken from the CS8420 via ribbon cable and are reduced to 3.3V with
low noise power regulators (LM2936) in combination with Black-Gate capacitors. The AD1896
is configured the way that its output signals reaching the DCX2496, deliver
standard TTL-level (5V). This is necessary to be compatible with the
signals of the
The input to the
CS8416 is fed via two 100nF capacitors. Leaving the original CS8420 in its
place both units operate in sync and stay both full functional. Therefore
a simple A/B-comparison is possible. The layout of the PCB around the CS8416
exactly suits the specification in the datasheet.
The outputs of the CS8416
(26,27,28) are connected with the input of the SRC AD1896. The SRC
operates completely asynchronous. A XTAL-oscillator (>= 28MHz) supplies the
clock for the DSP. The frequency must not be below 28MHz. A lower
frequency causes output errors at an input sample rate of 192kHz. The
output of the AD1896 is synchronized via pin 24 and 25 by the DCX2496.
Die circuit is built-on a double layer PCB with large ground planes. All
signals were supplied via ribbon cable coming from the CS8420 (see
The CS8420 could stay in its place. Only pin 18 must be disconnected from
the PCB of the DCX2496. With a spiky soldering iron and a needle this is
quite easy. Attaching the needle below the pin just heat the pin and bend
the pin up. The now free pad below pin 18 has to be connected to the SDOUT
signal delivered by the ribbon cable. This disables the CS8420 and
delivers the signals from the AD1896 to the DCX2496.
Caused by the better FIR-Filter of the AD1896 the DCX2496 sounds more
fresh and open after the modification. Best you can verify this with an A/B-test.
Therefore insert a switch between the output of the CS8420 pin 18 and the
pad below it. The other input of the switch should be connected to SDOUT
delivered by the ribbon cable. Because both SRCs are running in sync the switch can be
operated at any time.
Really extreme was the difference between both units feeding the DCX2496
with a signal delivered via a telephone line from my working room (see
ways to transport SPDIF Audio signals"). Caused by the sub optimal wire
the jitter was clearly visible in the eye-diagram. In case of the original
CS8420 this minor quality signal was clearly audible. In case of the new
input stage there was no audible difference between the telephone line and
a short 75Ohm line.
With the new input stage also differences between different wires (low
cost - high end) were no more audible. In my opinion the AD1896 with its
delay line and digital PLL delivers the best jitter rejection (<5pS) I
have ever heard.
This mod I can really recommend for those who like to get most out of