TUTORIAL DESIGN AND OPERATION OF AUTOMATIC GAIN CONTROL LOOPS FOR RECEIVERS IN MODERN COMMUNICATIONS SYSTEMS his article is intended to provide insight VGA TYPES into the effective operation of variable There are two major classes of VGA in use Tgain amplifiers (VGA) in automatic gain today. The first is the so-called IVGA (input control (AGC) applications. Figure 1 is a gen- VGA), which can be regarded as a passive eral block diagram for an AGC loop. The in- variable attenuator followed by a fixed-gain put signal passes through the VGA to produce amplifier. The second type is the output VGA the output level to be stabilized. The detec- (OVGA), which is essentially equivalent to a tor’s output is compared against a setpoint fixed-gain amplifier followed by a passive at- voltage to produce an error signal, which is tenuator. then integrated to produce a gain control volt- An IVGA is the preferred choice for a re- Fig. 1 VGA-based AGC age. This is applied to the control input of the ceive AGC system because the available out- loop block diagram. ▼ VGA. The attenuator shown be- tween the VGA and the detec- VGA (LINEAR IN dB) SIGNAL tor is used to align the maxi- This article is intended OUTPUT mum output level of the VGA SIGNAL INPUT with the maximum input level to provide insight into the of the detector. effective operation of variable RF ATTEN DETECTOR In the course of this article several key issues will be ad- gain amplifiers (VGA) dressed, including VGA types, C integ Rstab Rin loop dynamics, detector types, in automatic gain control the operating level of VGA and (AGC) applications. − the operating level of the detec- + tor. Then an example applica- ERROR INTEGRATOR tion revolving around an AD8367 VGA will be presented DANA WHITLOW SETPOINT INPUT RSSI OUTPUT for further discussion of the Analog Devices material. Beaverton, OR TUTORIAL vide considerable reduction of the law influences the dependency of the LOG LINBNV ripple. loop’s equilibrium level on the input’s RMS SQLAW One of the benefits of using an waveform or crest factor. +20 IVGA in an AGC loop is that the Four detector types will be consid- +10 VGA’s gain control voltage bears an ered here: envelope detector; square- 0 accurate logarithmic relationship to law detector; true-RMS detector; and −10 the input signal level when the loop is log detector. (dB) − o 20 in equilibrium. This means that the P −30 gain control voltage may also be used ENVELOPE DETECTOR −40 as an excellent received signal (RECTIFIER) −50 strength indicator (RSSI). The output voltage of the enve- 0 5 10 15 20 25 30 lope detector is proportional to the TIME (ms) LOOP DYNAMICS magnitude of the instantaneous RF ▲ Fig. 2 Simulated response of AGC loop Response time is an important is- input voltage. Assuming that suffi- to large amplitude steps for various detectors. sue when designing any AGC loop. cient low pass filtering is applied at There is usually a compromise be- its output to eliminate RF ripple, this put level at low distortion is relatively tween having the loop respond to un- detector produces a voltage propor- independent of the gain setting. This desired input level fluctuations as tional to the envelope amplitude of is the desired trait for an AGC sys- rapidly as one would like, and having the RF signal. tem, whose very object is to maintain it undesirably modify amplitude mod- Assuming that the loop’s band- a constant output in the face of vary- ulation on the signal. Additionally, width is made sufficiently small as to ing input signal amplitude. large and/or abrupt changes in the in- avoid significant gain pumping, the The OVGA is generally ill suited put level may lead to unacceptable effect of the loop using an envelope to AGC applications because of its re- recovery behavior, necessitating fur- detector is to stabilize the average duced output signal handling capabil- ther adjustments of the response rectified voltage of the signal. The re- ity at low gain settings and therefore time. sulting power is therefore dependent will not be discussed further here. The issue of excessive loop band- on the RF signal’s envelope wave- When a single IVGA is used in a width deserves a bit more explana- form. Such a loop acting on a con- situation in which the VGA sets the tion. If the loop responds too quickly, stant-envelope signal such as GSM system noise floor, the output SNR is it will introduce undesired gain mod- will produce an average output power essentially independent of the input ulation arising from the loop’s efforts which is different than that for a signal; it does not improve as is often to stabilize the output level of a signal heavily-amplitude-modulated signal, preferred. Occasionally, it is desirable containing legitimate amplitude mod- such as CDMA or 64QAM. to cascade two VGAs in order to ame- ulation. This is referred to as “gain The output of the envelope detec- liorate this behavior or simply to ob- pumping.” In the context of digital tor cannot go negative no matter how tain more gain control range. Doing modulation, the presence of apprecia- weak the input signal, but may reach so requires proper coordination of ble gain pumping can result in signifi- extreme positive values in response to the gain control inputs of the two de- cant modulation errors and perhaps very strong signals. Starting with the vices. even noticeable spectral re-growth in AGC loop in equilibrium, a sudden If the gain control of only the sec- extreme cases. A tolerable value of large increase in input amplitude ond stage VGA is manipulated in the gain pumping would generally be only causes a very large initial increase in weak signal regime, the signal level to a fairly small fraction of 1 dB. detector output, which very rapidly the first stage VGA’s amplifier in- drives the loop towards lower gain. creases with increasing input level, so DETECTOR TYPES (DETECTOR On the other hand, an abrupt reduc- the output SNR improves with in- LAW) tion of the input signal level (no mat- creasing input level. It is necessary to One convenient aspect of an AGC ter by how many dB) cannot reduce hand off the gain control from the loop is that the detector need not nec- the detector output below zero, and second stage to the first stage only essarily have a very wide dynamic the loop’s best response is to slew to- when overload of the first stage’s am- range over which it obeys any particu- wards equilibrium at a fairly low rate plifier is imminent. lar law. This is because the detector until the detector output begins to Alternatively, the two gain control operates at a constant average level change by a significant fraction of the inputs may simply be driven in paral- when the AGC loop is in equilibrium; reference voltage, at which point the lel, in which case the output S/N (ex- thus, the detector should only need to recovery trends towards an exponen- pressed in dB) improves at half the cope accurately with the level range tial decay. In the slew rate limited re- rate at which the input level (also ex- associated with a modulated signal. gion, the gain of the signal path is pressed in dB) rises. In cases where However, as mentioned earlier, the varying at a constant number of dB the VGAs used have residual ripple in detector’s response law (that is linear, per second. their gain control functions, an addi- log, square law, etc.) can play a signifi- Figure 2 shows the behavior of tional benefit of this approach can be cant role in determining the loop’s dy- such a loop for a large input level obtained if the two gain control input namic response during large, abrupt step (note that curves for all four de- signals are intentionally offset by half changes in signal level. Perhaps more tector types are superimposed on this the period of the ripple. This can pro- importantly, the detector’s response plot). These results were obtained TUTORIAL from simulations in which the VGA of the signal path that is square law has a shallow slope for high input lev- has representative limits on the gain brings forth the possibility of the els, resulting in a diminished re- range and on the maximum output large-step response being different sponse rate to sudden increases in level. The detectors contrived for from that of the simple envelope de- signal level. At the other extreme, the these simulations have no particular tector, which can indeed be seen in square-law detector’s small slope near limits, on the grounds that in most the figure. Note that the RMS detec- zero input level gives it a very slug- practical situations the designer will tor has a slightly slower recovery from gish response to large decreases in in- scale the circuit so that the detector a large downward amplitude step put amplitude. Conversely, the does not limit appreciably before the than does the standard envelope de- square-law detector exaggerates the VGA does. tector, but a slightly faster recovery response to large signals, giving the (and a bit of overshoot) from a step fastest response to increasing signals. SQUARE-LAW DETECTOR up in input amplitude. In common The envelope and RMS detectors, This type of detector has an in- with the square-law detector, the having intermediate characteristics, stantaneous output which is propor- true-RMS detector will make the give response speeds in between.