LMP2014MT
SNOSAK6B –DECEMBER 2004–REVISED MARCH 2013
www.ti.com
APPLICATION INFORMATION
THE BENEFITS OF LMP2014 NO 1/f NOISE
Using patented methods, the LMP2014 eliminates the 1/f noise present in other amplifiers. That noise, which
increases as frequency decreases, is a major source of measurement error in all DC-coupled measurements.
Low-frequency noise appears as a constantly-changing signal in series with any measurement being made. As a
result, even when the measurement is made rapidly, this constantly-changing noise signal will corrupt the result.
The value of this noise signal can be surprisingly large. For example: If a conventional amplifier has a flat-band
noise level of 10nV/√Hz and a noise corner of 10 Hz, the RMS noise at 0.001 Hz is 1µV/√Hz. This is equivalent
to a 0.50 µV peak-to-peak error, in the frequency range 0.001 Hz to 1.0 Hz. In a circuit with a gain of 1000, this
produces a 0.50 mV peak-to-peak output error. This number of 0.001 Hz might appear unreasonably low, but
when a data acquisition system is operating for 17 minutes, it has been on long enough to include this error. In
this same time, the LMP2014 will only have a 0.21 mV output error. This is smaller by 2.4 x. Keep in mind that
this 1/f error gets even larger at lower frequencies. At the extreme, many people try to reduce this error by
integrating or taking several samples of the same signal. This is also doomed to failure because the 1/f nature of
this noise means that taking longer samples just moves the measurement into lower frequencies where the noise
level is even higher.
The LMP2014 eliminates this source of error. The noise level is constant with frequency so that reducing the
bandwidth reduces the errors caused by noise.
Another source of error that is rarely mentioned is the error voltage caused by the inadvertent thermocouples
created when the common "Kovar type" IC package lead materials are soldered to a copper printed circuit board.
These steel-based leadframe materials can produce over 35 μV/°C when soldered onto a copper trace. This can
result in thermocouple noise that is equal to the LMP2014 noise when there is a temperature difference of only
0.0014°C between the lead and the board!
For this reason, the lead-frame of the LMP2014 is made of copper. This results in equal and opposite junctions
which cancel this effect.
OVERLOAD RECOVERY
The LMP2014 recovers from input overload much faster than most chopper-stabilized op amps. Recovery from
driving the amplifier to 2X the full scale output, only requires about 40 ms. Many chopper-stabilized amplifiers will
take from 250 ms to several seconds to recover from this same overload. This is because large capacitors are
used to store the unadjusted offset voltage.
Figure 29.
The wide bandwidth of the LMP2014 enhances performance when it is used as an amplifier to drive loads that
inject transients back into the output. ADCs (Analog-to-Digital Converters) and multiplexers are examples of this
type of load. To simulate this type of load, a pulse generator producing a 1V peak square wave was connected
to the output through a 10 pF capacitor. (Figure 29) The typical time for the output to recover to 1% of the
applied pulse is 80 ns. To recover to 0.1% requires 860ns. This rapid recovery is due to the wide bandwidth of
the output stage and large total GBW.
NO EXTERNAL CAPACITORS REQUIRED
The LMP2014 does not need external capacitors. This eliminates the problems caused by capacitor leakage and
dielectric absorption, which can cause delays of several seconds from turn-on until the amplifier's error has
settled.
10 Submit Documentation Feedback Copyright © 2004–2013, Texas Instruments Incorporated
Product Folder Links: LMP2014MT