DA2002
Theory of Operation
10
The calibration process takes a long time because the node adjustment is interactive (adjusting a
node causes some miss-adjustment at all the other nodes). A single calibration cycle consists of
reading and adjusting of all the nodes. The processor repeats the calibration cycle numerous
times until all the nodes are set properly. Though nodes interaction exists during the adjustment
process, the overall network is guaranteed to converge on a solution by design. Calibrating a
DA2002 for the first time (at the factory) often exceeds 25 minutes. Once calibrated, the settings
are stored in non volatile ram for future startup point of reference, thus all future adjustments are
initialized to the last settings. Therefore the initial tolerance of the components is pre calibrated
already, and each new calibration needs to deal only with component drift under the same given
temperature conditions (ovenized components). The remaining calibration at each subsequent
power on takes less then 2 minutes typically.
Timing and Deglitcher
The conversion from a digital sample value to an analog voltage consists of translating a digital
code to a corresponding setting of analog switches and multiplexers to tap the appropriate
voltage from the analog nodes. Such switching causes unwanted glitch energy to come into play.
The glitch energy is code and signal dependent and can not be removed by filtering. The purpose
of the deglitcher circuit (see diagram) is to block the signal from feeding to the output for long
enough time after each transition, thus allowing the glitches enough time to disappear, and for
each new analog sample to accurately settle to its final value.
The deglitcher circuit is in off state about half a sample time, and on for the rest of the time. The
on time is the critical time and no digital activity takes place anywhere near the analog circuits.
The settled signals are fed to the output filter with minimum disturbance. The deglitcher off time
settling requires the circuit to block as much of the transitions from feeding forwards to the
output filter.
The blocking requirement is very demanding because a transition of many volts between two
adjacent sample values should feed forward less than a microvolt. A single switch can not yield
such blocking performance. The deglitcher utilizes four switches: the first switch shunt (shorts)
the signal to ground. The remaining signal is connected to the second switch that is in open state.
Whatever comes through is shunted to ground by the third switch. The remaining tiny energy is
further blocked by the fourth opened series switch.
During the deglitcher on state, the shunt switches (switch one and three) are opened and the
series switches (switch two and four) are shorted to allow the signal path to the output. The
deglitched circuit utilizes DMOS technology thus providing extremely low resistance during the
“on” state. The remaining problems due to “on” state resistance variations are neutralized by use
of the strong feedback of the deglitcher amplifier.
The main reason for using DMOS transistors is their sub nanosecond switching capabilities. The
jitter critical timing point is all at the deglitcher circuit. Each sample value must exist over the
same time period thus precise deglitcher turn on and turn off are critical for good results. In fact
switching during deglitcher blocking time can be somewhat sloppy, as long as the signals are
well settled prior to turn on. Fighting the jitter wars means feeding the deglitcher circuit a precise
jitter free on / off drive signal.