| Homepage |
|
Linear RMS-Sensing Voltmeter and AM Detector
By David Knight.
|
Abstract: The final-stage collector current of a push-pull complementary symmetry (totem-pole) amplifier is a half-wave rectified version of the signal being amplified. By biasing the amplifier in class C, and using a large amount of negative feedback, the rectification process can be made extremely linear. This provides the basis for a high-quality AM detector, or a high-frequency linear RMS voltmeter. Both functions can also be realised simultaneously because there are two collectors from which signals can be extracted. |
|
Introduction: The circuit described below was originally developed for use in an ultrasonic receiver: a device used for monitoring bearing wear in electric motors, insulator breakdown in electricity substations, and other applications in which subtle fizzing and scraping sounds have to be picked out of the general industrial background noise. Such receivers require a linear voltmeter, sometimes used to trigger a level detector, for engineer callout; and an AM detector, so that the engineer can use headphones and a directional transducer to find the source. A different design project carried out shortly beforehand involved push-pull audio amplifiers, and so the relationship between collector current and output current was fresh in mind. Hence, through a random association; a solution was found which bypassed the limitations of conventional diode linearisation schemes, and gave both the voltmeter and the detector in one. The original ultrasonic receiver was designed to work at 40KHz. A quick frequency-response measurement however showed that the circuit worked perfectly well at 1MHz and beyond. Hence the detector was also suitable for connection to the output of a superhet IF amplifier. The reason for wanting to do that relates to the problem of listening fatigue. The half-wave germanium-diode detector is ubiquitous in broadcast receivers; but the translation from modulation level to output is seriously non-linear (and earlier valve / tube detectors are just as bad). The trick in getting acceptable audio quality is to drive the detector hard, but there is no truly linear region in the diode characteristic, and signals too weak to drive the AGC full-on will be detected close to the diode threshold. This does not bode well for the audio quality; and certainly, no one hearing a voice on AM radio will be fooled into thinking that the person is in the room. Still, everyone knows what AM radio sounds like, and so there is no great pressure to do anything about it. Designing detectors however, involves a lot of thinking about detectors; and so the discovery that the new circuit would work at superhet intermediate frequencies provoked a certain curiosity. Also on hand was a Racal RA17C18 general-coverage receiver, which has a 13KHz maximum IF bandwidth, a linear IF phase response, and a handy IF output socket on the back. Hence, another version of the circuit was built (with different design considerations resulting in different component values) and connected as the interface between the RA17 and a hi-fi amplifier. The obvious preconception with regard to broadcast AM is that the quality can never be any good, because the maximum usable audio frequency component is less than 9KHz in the daytime and 4.5KHz at night. In fact, the extreme treble content is not that important to the pleasure of listening, but the absence of high-order distortion products most certainly is. It transpires that a great many broadcast transmitters are hi-fi within the statutory bandwidth restriction, and a tribute to the art of their designers. The sound of the linear detector on tuning around is also somehow clean, and easy on the ear; and the intelligibility of weak stations is greatly enhanced. It is not much of a conjecture to say that the brain must do a considerable amount of processing to extract the intelligence from distorted signals; and the removal of the audio intermodulation products is like the lifting of a veil. A linear rectifier is, of course, not the only solution to the problem of high-quality AM reception. Another way to do it is to phase-lock an oscillator to the carrier and use synchronous demodulation. Such circuits are complicated however, and jump unpleasantly from carrier to carrier when tuning across a band. The linear detector described here is simple (4 transistors), uses a single power supply, and just behaves like an ideal rectifier. The principle of the detector is explained below, in the dour language of the original technical report. The version first tried on the radio was built into an elaborate RA17 adapter, and is described in another article [RA17 adapter part 2]. Bob Batey has subsequently reworked the circuit, and added a notch filter to get rid of adjacent-channel carrier heterodyne interference, and his version is described in the article [Hi-fi AM detector]. DWK, June 2009. |
|
Linear RF voltmeter and AM detector. Conventional rectifier type voltmeter arrangements require high voltage inputs in order to overcome the effect of diode forward threshold voltage on linearity. A bridge formed with germanium diodes will provide a reasonably linear display with 2V RMS input, but it will barely register at all until the input exceeds about 0.2V P-P. Thus, using a conventional diode bridge to measure an input signal of 1μV requires a voltage gain of 126dB in order to achieve modest linearity. Partial biasing of the diodes may be used to improve linearity, but such arrangements are prone to thermal drift. In this design, the signal is fed to a complementary symmetrical push-pull amplifier, with a feedback factor of 1, which delivers power into a load resistor. See fig. 1 below. |
The amplifier is biased for zero quiescent current (class B/C).
The large feedback ratio forces the system into linearity, and
reduces cross-over distortion to a negligible value for the purposes
of this application. Thus the voltage appearing across the load
resistor is the same as the input voltage, and hence: TR1 charges C1, TR2 discharges C1; hence the average current flowing in TR1 collector is half the RMS load current. The meter reading is therefore:  The meter sensitivity is thus defined by the choice of load resistor, and the scale is perfectly linear. Note that the meter can be placed in the collector circuit of either TR1 or TR2. By placing a resistor in the collector of TR2 and shunting with a capacitor to provide a suitable time-constant, the circuit also behaves as a linear AM detector. The audible noise due to the circuit is negligible, but good power supply smoothing is necessary. A working example of the voltmeter / detector is shown in fig 2. |
|
In this case, the load resistor is 100Ω, and the meter
movement is 100μA FSD, giving a sensitivity of 20mV RMS FSD
at the wiper of the input calibration potentiometer. Thus a 1μV
signal only requires 86dB overall voltage amplification in order
to give a full-scale reading. The circuit also features a ×2
(+6dB) scale expansion switch (giving 10mV RMS FSD), implemented
by reducing the load resistor to 50Ω. The LM386 audio amplifier
provides 26dB (×20) voltage gain. The inductor in the LM386
supply feed prevents RF distortion products from the audio amplification
process from appearing on the supply rail. Note that in this
battery powered circuit, the audio is taken from the top of the
totem pole and the meter from the bottom. This arrangement was
chosen to give the lowest possible level of spurious radiation
from the meter wiring, the detector being part of a high gain
ultrasonic receiver with the meter mounted close to the input
socket. In a mains powered instrument, the audio should be taken
from the bottom of the totem pole, since this is the arrangement
giving best supply (mains hum) rejection in the audio output.
The example in fig. 2 is part of a system operating at 40KHz.
The detector input response however, with the 68pF capacitor
at the input removed (and assuming a sensible layout), was found
to be flat within 0.5dB up to 1.1MHz, and -3dB at 1.6MHz. A further
example of this detector operating at 100KHz is given in the
author's RA17 adapter
article. The meter movement can, if desired, be calibrated in dB. The following table gives dB markings against a 0 - 100 logging scale, assuming +3dB FSD. |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
© D.W. Knight. 1981,1987,2000, 2009.
| Homepage |
|
Version 1.00 (Unicode)