3D3622
APPLICATION NOTES
GENERAL INFORMATION
Figure 1 illustrates the main functional blocks of
the 3D3622. Since the 3D3622 is a CMOS
design, all unused input pins must be returned to
well-defined logic levels, VDD or Ground.
The pulse generator architecture is comprised of
a number of delay cells, which are controlled by
the 6 LSB bits of the address, and an oscillator &
counter, which are controlled by the 16 MSB bits
of the address. Each device is individually
trimmed for maximum accuracy and linearity
throughout the address range. The change in
pulse width from one address setting to the next
is called the
increment,
or LSB. It is nominally
equal to the device dash number. The minimum
pulse width, achieved by setting the address to
zero, is called the
inherent pulse width.
For dash numbers larger than 15, the 6 LSB bits
are invalid, and the address loaded must
therefore be a multiple of 64 (ie, 0, 64, 128, 192,
etc). When used in this manner, the device is
essentially a 16-bit generator, with an effective
increment equal to 64 times the dash number.
For best performance, it is essential that the
power supply pin be adequately bypassed and
filtered. In addition, the power bus should be of
as low an impedance construction as possible.
Power planes are preferred. Also, signal traces
should be kept as short as possible.
inherent width, and t
inc
is the nominal increment.
It is very similar to the INL, but simpler to
calculate. For most dash numbers, the relative
error is less than 1.0 LSB at every address (see
Table 1).
The
absolute error
is defined as follows:
e
abs
= t
PW
– (t
inh
+ addr * t
inc
)
where t
inh
is the nominal inherent delay. The
absolute error is limited to 1.5 LSB or 3.0 ns,
whichever is greater, at every address.
The
inherent pulse width error
is the deviation of
the inherent width from its nominal value. It is
limited to 1.0 LSB or 2.0 ns, whichever is greater.
PULSE WIDTH STABILITY
The characteristics of CMOS integrated circuits
are strongly dependent on power supply and
temperature. The 3D3622 utilizes novel
compensation circuitry to minimize the
performance variations induced by fluctuations in
power supply and/or temperature.
With regard to stability, the output pulse width of
the 3D3622 at a given address, addr, can be split
into two components: the
inherent pulse width
(t
inh
) and the
relative pulse width
(t
PW
– t
inh
).
These components exhibit very different stability
coefficients, both of which must be considered in
very critical applications.
The thermal coefficient of the relative pulse width
is limited to
±250
PPM/C (except for the -0.25),
which is equivalent to a variation, over the -40C
to 85C operating range, of
±1.5%
(±9% for the
dash 0.25) from the room-temperature pulse
width. This holds for all dash numbers. The
thermal coefficient of the inherent pulse width is
nominally +20ps/C for dash numbers less than 5,
and +30ps/C for all other dash numbers.
The power supply sensitivity of the relative pulse
width is
±1.0%
(±3.0% for the dash 0.25) over the
3.0V to 3.6V operating range, with respect to the
pulse width at the nominal 3.3V power supply.
This holds for all dash numbers. The sensitivity of
the inherent pulse width is nominally -5ps/mV for
all dash numbers.
It should also be noted that the DNL is also
adversely affected by thermal and supply
variations, particularly at the MSL/LSB
crossovers (ie, 63 to 64, 127 to 128, etc).
PULSE WIDTH ACCURACY
There are a number of ways of characterizing the
pulse width accuracy of a programmable pulse
generator. The first is the
differential nonlinearity
(DNL), also referred to as the increment error. It
is defined as the deviation of the increment at a
given address from its nominal value. For most
dash numbers, the DNL is within 0.5 LSB at
every address (see Table 1: Pulse Width Step).
The
integrated nonlinearity
(INL) is determined
by first constructing the least-squares best fit
straight line through the pulse-width-versus-
address data. The INL is then the deviation of a
given width from this line. For all dash numbers,
the INL is within 1.0 LSB at every address.
The
relative error
is defined as follows:
e
rel
= (t
PW
– t
inh
) – addr * t
inc
where addr is the address, t
PW
is the measured
width at this address, t
inh
is the measured
Doc #06008
5/8/2006
DATA DELAY DEVICES, INC.
Tel: 973-773-2299
Fax: 973-773-9672
http://www.datadelay.com
2
DATADELAY [ DATA DELAY DEVICES, INC. ]
