May 2005 6 M9999-052405
MIC4123/4124/4125 Micrel, Inc.
Application Information
Although the MIC4123/24/25 drivers have been specifically
constructed to operate reliably under any practical circum-
stances, there are nonetheless details of usage which will
provide better operation of the device.
Supply Bypassing
Charging and discharging large capacitive loads quickly
requires large currents. For example, charging 2000pF from
0 to 15 volts in 20ns requires a constant current of 1.5A. In
practice, the charging current is not constant, and will usually
peak at around 3A. In order to charge the capacitor, the driver
must be capable of drawing this much current, this quickly,
from the system power supply. In turn, this means that as far
as the driver is concerned, the system power supply, as seen
by the driver, must have a VERY low impedance.
As a practical matter, this means that the power supply bus
must be capacitively bypassed at the driver with at least
100X the load capacitance in order to achieve optimum
driving speed. It also implies that the bypassing capacitor
must have very low internal inductance and resistance at
all frequencies of interest. Generally, this means using two
capacitors, one a high-performance low ESR film, the other
a low internal resistance ceramic, as together the valleys in
their two impedance curves allow adequate performance over
a broad enough band to get the job done. PLEASE NOTE
that many film capacitors can be sufficiently inductive as to
be useless for this service. Likewise, many multilayer ceramic
capacitors have unacceptably high internal resistance. Use
capacitors intended for high pulse current service (in-house
we use WIMA™ film capacitors and AVX Ramguard™ ceram-
ics; several other manufacturers of equivalent devices also
exist). The high pulse current demands of capacitive drivers
also mean that the bypass capacitors must be mounted
very close to the driver in order to prevent the effects of lead
inductance or PCB land inductance from nullifying what you
are trying to accomplish. For optimum results the sum of the
lengths of the leads and the lands from the capacitor body to
the driver body should total 2.5cm or less.
Bypass capacitance, and its close mounting to the driver serves
two purposes. Not only does it allow optimum performance
from the driver, it minimizes the amount of lead length radiat-
ing at high frequency during switching, (due to the large Δ I)
thus minimizing the amount of EMI later available for system
disruption and subsequent cleanup. It should also be noted
that the actual frequency of the EMI produced by a driver is
not the clock frequency at which it is driven, but is related to
the highest rate of change of current produced during switch-
ing, a frequency generally one or two orders of magnitude
higher, and thus more difficult to filter if you let it permeate your
system. Good bypassing practice is essential to proper
operation of high speed driver ICs.
Grounding
Both proper bypassing and proper grounding are necessary
for optimum driver operation. Bypassing capacitance only
allows a driver to turn the load ON. Eventually (except in rare
circumstances) it is also necessary to turn the load OFF. This
requires attention to the ground path. Two things other than
the driver affect the rate at which it is possible to turn a load
off: The adequacy of the grounding available for the driver,
and the inductance of the leads from the driver to the load.
The latter will be discussed in a separate section.
The E-Pad and MLF packages have an exposed pad under
the package. It's important for good thermal performance that
this pad is connected to a ground plane.
Best practice for a ground path is obviously a well laid out
ground plane. However, this is not always practical, and a
poorly-laid out ground plane can be worse than none. Attention
to the paths taken by return currents even in a ground plane
is essential. In general, the leads from the driver to its load,
the driver to the power supply, and the driver to whatever is
driving it should all be as low in resistance and inductance
as possible. Of the three paths, the ground lead from the
driver to the logic driving it is most sensitive to resistance or
inductance, and ground current from the load are what is most
likely to cause disruption. Thus, these ground paths should
be arranged so that they never share a land, or do so for as
short a distance as is practical.
To illustrate what can happen, consider the following: The
inductance of a 2cm long land, 1.59mm (0.062") wide on a
PCB with no ground plane is approximately 45nH. Assum-
ing a dl/dt of 0.3A/ns (which will allow a current of 3A to flow
after 10ns, and is thus slightly slow for our purposes) a volt-
age of 13.5 Volts will develop along this land in response to
our postulated ∆Ι. For a 1cm land, (approximately 15nH) 4.5
Volts is developed. Either way, anyone using TTL-level input
signals to the driver will find that the response of their driver
has been seriously degraded by a common ground path for
input to and output from the driver of the given dimensions.
Note that this is before accounting for any resistive drops in
the circuit. The resistive drop in a 1.59mm (0.062") land of
2oz. Copper carrying 3A will be about 4mV/cm (10mV/in) at
DC, and the resistance will increase with frequency as skin
effect comes into play.
The problem is most obvious in inverting drivers where the
input and output currents are in phase so that any attempt
to raise the driver’s input voltage (in order to turn the driver’s
load off) is countered by the voltage developed on the com-
mon ground path as the driver attempts to do what it was
supposed to. It takes very little common ground path, under
these circumstances, to alter circuit operation drastically.
Output Lead Inductance
The same descriptions just given for PCB land inductance
apply equally well for the output leads from a driver to its load,
except that commonly the load is located much further away
from the driver than the driver’s ground bus.
Generally, the best way to treat the output lead inductance
problem, when distances greater than 4cm (2") are involved,
requires treating the output leads as a transmission line. Un-
fortunately, as both the output impedance of the driver and the
input impedance of the MOSFET gate are at least an order of
magnitude lower than the impedance of common coax, using
coax is seldom a cost-effective solution. A twisted pair works
about as well, is generally lower in cost, and allows use of a