BEST PRACTICES FOR GROUNDING AND GAIN
STAGING IN ANALOG DEVICES
by Albert Alexander
April 3, 2009
Executive Summary
Grounding and gain staging are widely misunderstood and cost virtually nothing
to do correctly. A robust grounding scheme is rarely more expensive or difficult to
implement, and it contributes to lowering emissions, reducing susceptibility to
interference, minimizing the negative effects of dirty power, and—most importantly—
maximizing signal integrity. Gain staging is a technique used to control distortion and
efficiency both in designing amplifier circuits and in using any analog signal chain. This
paper explores zero-cost implementation of star grounding convention and proper gain
staging both for single devices and signal chain. Using these conventions will increase
the reliability of sensitive systems and maximize the performance of small-signal
circuitry.
Keywords
Grounding, gain, balanced, unbalanced, analog, small-signal, power
Ground Loops
All grounds are not created equal. A common error for young electrical engineers
is to tie grounds together whenever the opportunity presents itself. If all grounding points
were at exactly the same potential, this wouldn’t be much of a problem; however, there is
resistance between grounding points, whether from poor solder joints or connections,
corrosion, or even simply the resistance of the wire, and this causes different ground
points to be at different voltages. Consider a system in which two signals are generated
with reference to 0V (true ground). Let’s say that their outputs are tied to another ground
point which has some resistance (Rg) between it and true ground. As the signals vary, the
sum of their output currents also varies, changing the voltage across Rg. Since the voltage
across Rg is dependent on both output currents, the two signal outputs become dependent
on each other. The larger Rg is, the worse the coupling gets. This problem most
frequently manifests itself in audio systems, where AC power signal is coupled into the
output signal (or, even worse, the input signal, where it is amplified along with the input).
The noise can be heard as a 60Hz hum in the output.
Having star ground points is only one step towards minimizing coupling. Another
way to prevent coupling is to have separate grounds for different power stages. The more
power an amplifier is designed to produce, the less sensitive its input is, so it is not
uncommon to have several stages of amplification between the input and the final output
signal. Since higher-power amplification stages have more current being sunk to ground,
its ground loops tend to be more severe than a lower-power amplifier. Because of this, it
is good practice to give each stage of amplification its own star ground.
Amplifier Grounding Schemes
A block diagram for a single-ended current amplifier is shown in Figure 1. This
type of amplifier is used to drive unbalanced signals. Note that the small signal elements
are all run separately to the power ground. This star ground is then connected to the
chassis at a single point. The chassis may also be connected to the shield if the input and
output signals are shielded. As a general rule, the more current a ground trace needs to
handle, the further up the ground chain the trace should be: small signal grounds are star
grounded to power ground, which runs to the chassis.
The star ground has a lot of electrical inertia, since the chassis has a lot of free
electrons and the power ground is heavy gauge and immediately proximate. This
provides a steady ground reference for all the small signal grounds, which carry too little
current to affect the potential of the star ground, and it provides a safe path for highercurrent transients to dissipate from the power supply, keeping that current away from
sensitive input stage components.
The current flowing from power ground to chassis is the sum of all ground trace
currents in the device, and therefore should be of appropriate gauge and minimum length.
The power-to-chassis ground connection should be located well away from any sensitive
circuitry since it carries so much noise, which should not be difficult since the power
supply should be as far away from the most sensitive electronics as possible. The most
common power-to-chassis ground points are at the IEC terminal or between the largest
filter capacitors in the power supply.
Many manufacturers tie signal ground to shield on unbalanced inputs or outputs.
This is poor design practice. The best way to handle unbalanced signals is to tie cable
shield to the chassis (the default case for any bulkhead jack) and let the signal ground be
fed into the input stage which will ground it to the star ground. Connecting cable shield to
signal ground creates a ground loop, allowing all sorts of noise into the input of the
amplifier.
Figure 1: Single-Ended Amplifier
Figure 1 shows an unbalanced system, commonly seen with high impedance tipsleeve and with low impedance shielded coaxial cable. A balanced signal system
functions quite differently. In this case, the signal ground does not exist. Rather than
having a signal and a reference ground, a balanced signal system carries a signal over a
differential pair. One pin carries the signal as normal while the other pin carries the same
signal with its phase flipped. The input stage of the next device inverts the phase-flipped
signal and then sums the two, canceling common-mode noise picked up along the way.
Balanced signals are obviously much more noise-immune than unbalanced signals, but
they still suffer from the same grounding problems as unbalanced signals. The balanced
equivalent of the grounding scheme shown in figure 1 is shown below in figure 2.
Figure 2: Balanced Differential Amplifier
As with the single-ended amplifier, small signal grounds should be referenced to a
single star ground at the same point as the power ground.
These configurations are extremely robust and do not compromise the
effectiveness of the shield. Cascaded devices should have metal-to-metal contact between
cases to create the largest ground plane possible, reinforcing immunity to electromagnetic
interference.
DC offsets between power supplies in different devices is common. Additionally,
many systems need to be able to interface balanced with unbalanced signals. The most
robust solution to both of these problems is to use transformers on the input and output of
every device. A center tap on the input and output transformers can be used to create a
totally isolated signal ground for both unbalanced and balanced inputs and outputs.
Transformers are also effective filters against RF noise. Unfortunately, transformers are
prohibitively expensive for many applications and are ineffective for passing highfrequency signals. An alternative method is to use a differential transistor pair to balance
and unbalance the signal. This is much more economical and does not have the frequency
limitations of a transformer-balanced system.
Gain Staging
The precision of the ground reference and the linearity of the gain of any
amplifier is inversely proportional to the power of the amplifier. The more power an
amplifier handles, the more harmonic distortion tends to be present. The concept of gain
staging is to maximize signal quality and dynamic range by distributing the total gain of a
signal across multiple amplifiers.
Small-signal amplifiers are the first stage of amplification for any transducer. In
some cases this amplification may be packaged with the transducer, as is the case with
many electret and condenser microphones and active sensors. In these cases, the signal is
so weak that a long cable run would allow noise from interference to overwhelm the
signal. In other cases, a transducer may have hot enough output to drive hundreds of feet
of cable. Dynamic microphones have this characteristic.
The first amplifier stage is often fixed-gain. This allows the designer to optimize
the amplifier for its particular gain setting, keeping distortion at an absolute minimum.
JFETs and MOSFETs are usually used for this first amplification stage due both to their
extremely linear amplification characteristics at low levels and to their near-infinite input
impedance, which prevents loading down the transducer. The role of this amplifier is to
get the signal up to a usable level, such that signal level is very high compared to the
noise added by interference over cable runs or noise added by signal processing.
The second stage of amplification, often called the preamp, tends to have a very
large dynamic range. Whether the application is audio or RF, these amplifiers must be
compatible with a wide variety of input signal levels and impedances. An RF amplifier
may need to handle both highly directional, high-gain antennas as well as unintentional
radiators, which have very little gain. An audio preamp must be able to handle condenser
microphones, which may require as little as 20dB of amplification, as well as fixedmagnet ribbon microphones, which may require up to 90dB of gain. A preamp may make
use of multiple stages of gain in order to reach across this broad dynamic range. Preamps
often have a fixed-gain input like those seen on active transducers. This establishes the
minimum gain of the amplifier while adding very little distortion, establishing a good
signal-to-noise ratio (SNR) for later, noisier stages. The second gain stage is often some
form of op-amp circuit with variable feedback. These circuits are generally bounded at
less than 200kHz but have a broad range of controllable amplification. An op-amp can
provide 40-60dB with very little distortion. In higher frequency applications, simple
transistor amplifiers are more common. These tend to be less linear at low levels, since
most RF applications are about maximizing signal level at every stage of amplification,
while in audio low-level signals must be preserved. In either case, the preamplifier is
what brings a signal up to line level, an appropriate level for processing by filters or
analog-to-digital conversion. At line level there is high enough voltage to maintain a high
SNR but there is not enough current to strain surface-mount components.
The third stage of amplification is the power amp. This takes a line level signal
and amplifies it to extremely high levels for output by another transducer, such as a
speaker or an output antenna. A power amplifier generally amplifies current rather than
voltage to drive an output transducer. Amplification generally ranges from 20dB-60dB of
power, taking a signal of perhaps 1mW and amplifying it to thousands of watts.
Amplification is usually achieved by the use of parallel power transistors. Even with
careful biasing and precision components, power amplifiers add much more distortion
than the earlier amplifier stages. Power amplifier output is usually too hot for any
additional processing since the amount of current would fry most circuit components.
Given these three stages of amplification, each of which may have both a gain
control and an attenuator or pad, the user is faced with picking levels to obtain maximum
fidelity. One standard practice used in professional audio applications is to set the gain of
each device such that it just clips the input of the next device, then (if necessary)
attenuate the signal until the input of the next device is not overloaded. For continuously
variable gain (rather than stepped-gain), no attenuation is needed since the output can be
trimmed to just below clipping by just controlling the gain. This does as much gain in the
cleaner amplification stages as possible, providing a clean signal that will make full use
of the dynamic range of the next input stage, oftentimes a power amp or an analog-todigital converter.
Passive volume controls are by definition attenuators. Their role is usually to
provide a finer degree of control than variable gain, especially since the distortion
characteristics of an attenuator hardly change at all while different gain settings can alter
frequency response. For systems involving a passive attenuation scheme, such as faders
on a mixer, attenuators should be first be set to 0dB, then the gain should be adjusted so
that the maximum signal just hits 0dB on the output meter, then the attenuator should be
used to bring down the signal to the level desired.
In general, attenuation is a negative, since putting resistance in the output line
increases noise and reduces efficiency. In select cases however, it may be a worthwhile
tradeoff. For example, most amplifiers are designed to have maximum stability and
linearity at a fairly high output level, and all amplifiers have different distortion patterns
at different levels (this is especially evident in tube circuits and soft-clip FET amplifiers).
In these cases it may be worth the added noise to crank up the gain and use a lot of
attenuation. But in most cases, especially with measurement devices, attenuation should
be used to trim the signal, not to chop it in half.
Conclusion
These concepts are relevant to a broad range of practices, from circuit design and
troubleshooting to running a radio station. These issues were selected because they apply
“where the rubber meets the road”—even once a good circuit has been designed, the
physical implementation and interfacing with other devices is a nontrivial problem to
address.
While there are other systems of grounding, each with its own unique benefits,
star grounding was selected due to its simplicity and scope. Other grounding schemes
may be more appropriate for multilayer boards and sophisticated mixed-signal systems,
star grounding is a very effective convention for analog amplifiers and simple sensordriven microprocessor systems.