DatasheetQ Logo
Electronic component search and free download site. Transistors,MosFET ,Diode,Integrated circuits

ML65245 Ver la hoja de datos (PDF) - Fairchild Semiconductor

Número de pieza
componentes Descripción
Fabricante
ML65245
Fairchild
Fairchild Semiconductor Fairchild
ML65245 Datasheet PDF : 10 Pages
1 2 3 4 5 6 7 8 9 10
ML65245**/ML65L245*
The basic architecture of the ML65245 is shown in Figure
5. It is implemented on a 1.5µm BiCMOS process.
However, in this particular circuit, all of the active
devices are NPNs — the fastest devices available in the
process.
In this circuit, there are two paths to the output. One path
sources current to the load capacitance when the signal is
asserted, and the other path sinks current from the output
when the signal is negated.
The assertion path is the emitter follower path consisting
of the level shift transistor Q1, the output transistor Q2,
and the bias resistor R8. It sources current to the output
through the 75ý resistor R7 which is bypassed by another
NPN (not shown) during fast input transients. The negation
path is a current differencing op amp connected in a
follower configuration. The active components in this
amplifier are transistors Q3-Q7. R3-R6 are bias resistors,
and R1 and R2 are the feedback resistors. The key to
understanding the operation of the current differencing op
amp is to know that the currents in transistors Q3 and Q5
are the same at all times and that the voltages at the
bases of Q4 and Q6 are roughly the same. If the output is
higher than the input, then an error current will flow
through R2. This error current will flow into the base of Q6
and be multiplied by β squared to the collector of Q7,
closing the loop. The larger the discrepancy between the
output and input, the larger the feedback current, and the
harder Q7 sinks current from the load capacitor.
A number of MOSFETs are not shown in Figure 5. These
MOSFETs are used to three-state dormant buffers. For
instance, the feedback resistors R1 and R2 were
implemented as resistive transmission gates to ensure that
disabled buffers do not load the lines they are connected
to. Similarly, there is a PMOS in series with R8 that is
normally on but shuts off for disable. Other MOSFETs
have been included to ensure that disabled buffers
consume no power.
Termination
R7 in Figure 5 also acts as a termination resistor. This 75ý
resistor is in series with the output and therefore helps
suppress noise caused by transmission line effects such as
reflections from mismatched impedances. System
designers using CMOS transceivers commonly have to
use external resistors in series with each transceiver
output to suppress this noise. Systems using the ML65245
or ML65L245 may not have to use these external resistors.
Applications
There are a wide variety of needs for extremely fast
buffers in high speed processor system designs like
Pentium, PowerPC, Mips, Sparc, Alpha and other RISC
processors. These applications are either in the cache
memory area or the main memory (DRAM) area. In
addition, fast buffers find applications in high speed
graphics and multimedia applications. The high
capacitive loading due to multiplexed address lines on
the system bus demand external buffers to take up the
excess drive current. The needed current to skew the
transitions between rise and fall times must be done
without adding excessive propagation delay. The
ML65245 and ML65L245 are equipped with Schottky
diodes to clean up ringing from overshoot and undershoot
caused by reflections in unterminated board traces.
BUFFERING MAIN MEMORY
An example main memory application for the Intel PCI
chipset with the Pentium processor is shown in Figure 6.
This is only intended as a general reference. For details
please refer to the appropriate Intel documentation. This
system has a 66MHz host processor and a 33MHz main
(DRAM) memory bus. The main memory row and column
addresses (RAS & CAS) and write enable (WE) signals are
provided by the PCMC chip (PCI Cache and Memory
Controller) device. The DRAM SIIMMs put a heavy load
on the PCMC and must be buffered. Three buffered copies
of the address signals and write enable are required to
drive the six row array. The ML65245 provides the
buffered signals and gives extra margin to be able to use
slower memory modules instead of the normally required
50/70ns. The burst read (page-hit) performance is typically
7-4-4-4 at 66MHz for 70ns DRAMs or 6-3-3-3 at 66MHz
for 50ns DRAMs. This usually translates to significantly
higher costs. With the speed improvement offered by the
ML65245, a 6-3-3-3 burst with 60ns DRAMs may be
achievable. The extra margin comes from the 1.5ns
propagation delay of the buffer. External resistor arrays are
not necessary. This becomes even more of an issue in
future PCI systems which may operate at 80MHz and
beyond.
This kind of main memory application for the ML65245
could potentially extend to other kinds of processor
systems which do not require latched buffering. Figure 7
shows a main memory design example with the ML65245
for the Mips R4X00 RISC processor based system without
secondary cache. The faster propagation delay essentially
translates to a faster main memory access.
6
REV. 1.0 10/25/2000

Share Link: 

datasheetq.com  [ Privacy Policy ]Request Datasheet ] [ Contact Us ]