Improving Modulation Acceptance

Modulation acceptance is the maximum modulation level an AM receiver's detector can handle without distortion. An ordinary diode detector can accommodate upward modulation well beyond 100%. But capacitive loading can prevent a detector from faithfully reproducing downward modulation. At some point it stops and clips the waveform instead of continuing toward zero.

Fig. 1   Detected sine wave

Fig. 1 shows the output of a diode detector for a 100% amplitude-modulated sine wave. The trace is inverted because the detector circuit is arranged to generate a negative output voltage for AVC. 0 V, corresponding to no carrier, is at the top of the screen. The modulation acceptance of this detector is 88%.

Fig. 2   Detected broadcast signal

Clipping distortion on sine waves is easily audible. In fact, you can hear it before you can see it on a scope. But casual tests suggest that clipping on speech and music isn't objectionable as long as the modulation acceptance is greater than about 80%. Fig. 2 shows an AM broadcast signal demodulated and clipped by the diode detector. While the sine wave distortion was obvious, it's not clear from the trace alone whether the broadcast waveform is distorted. In this case the ear agreed with the eye. I was listening to the receiver when I took the photo and I heard nothing amiss.

Modulation acceptance somewhere in the low-80% range is typical of consumer radios from the 1930s and 1940s. But the communications receivers I've tested have been much worse, typically distorting at 60% downward modulation or less. Distortion at such shallow modulation depths is nearly always audible. The audio tends to lose its crispness and clarity, and in severe cases sounds downright grungy. I always modify any receiver I restore for a modulation acceptance of at least 90%. Most approach 100%.

Measuring Modulation Acceptance

To test your receiver's modulation acceptance, you need a signal generator with variable amplitude modulation. If the modulation percentage isn't calibrated, you can measure it by observing the waveform on a scope. Modulate the signal with a 400 Hz tone and increase the modulation percentage until you can just hear distortion in your receiver (the character of the tone will change). You can also use a scope to look for distortion at the detector load (R2 below). The percentage AM at the onset of distortion is your receiver's modulation acceptance.

Causes of Low Modulation Acceptance

Fig. 3   Typical detector

Fig. 3 shows a typical AM detector circuit in a communications receiver. Diode D1 conducts whenever the voltage across it is positive. When the top terminal of IF transformer T1 goes positive with respect to the lower terminal, D1 clamps the terminal near ground. The lower terminal then charges C1 negatively to the waveform peak. C1 partially discharges while the IF waveform reverses sign. R1 and C2 filter out the IF component, leaving a DC level corresponding to the carrier and audio corresponding to the envelope modulation across detector load R2.

The primary cause of low modulation acceptance is back-biasing of D1 due to capacitive loading. The AVC capacitor, the volume control coupling capacitor, and any ANL capacitor are the culprits. Focus on the AVC capacitor alone to see how the damage is done.

AVC capacitor C3 charges to the level of the carrier, and the voltage on it divides back through R3 to R2. R1 and C2 are absent in some receivers, and it simplifies analysis to assume that R1, typically 47kΩ, is zero. In this case the DC voltage on R2 appears at the lower terminal of T1. Now assume that R3 is 2.2MΩ, R2 is 470kΩ, and the detected carrier level is −6 V. Under these conditions a DC bias of about −1 V appears at the lower terminal of T1 due to C3, since R2 ⁄ (R2 + R3) is about 1 ⁄ 6. When the signal drops to about one-sixth of its unmodulated level at 83% downward modulation, it generates +1 V across T1. Added to the −1 V at the lower terminal, this yields 0 V at D1. D1 stops conducting. It becomes back-biased and remains nonconducting for deeper downward modulation. R2 just sits at the −1 V bias level. This creates a flat spot in the detected waveform. The detector has clipped and distorted the demodulated audio.

If you swap R4 and C4, which leaves their current unchanged, you can see that the volume control and its coupling capacitor develop back-bias in the same way. An ANL circuit typically connects an additional filter, represented here by R5 and C5, to the detector load. This is yet another source of back-bias for D1. In each case the value of the offending capacitor matters little. These RC circuits are designed to pass or filter most of the audio spectrum, so the capacitors develop little AC voltage across them. You can think of them as little batteries. What matters is the ratio of the detector load R2 to the parallel combination of resistors that connect to these capacitors. Voltage division by this resistive divider is what causes the back-bias on D1. You can increase modulation acceptance by reducing the resistor ratio.

This analysis ignores the IF passband shape, IF source impedance, and the rectification efficiency of real diodes. These factors actually may increase modulation acceptance by lowering the effective modulation depth at the detector. However, detector back-bias usually swamps these factors in practical circuits. See the detection section of Frederick Terman's Radio Engineers' Handbook for more.

Improving Modulation Acceptance

The easiest way I've found to improve modulation acceptance is to reduce the detector load resistance R2. Surprisingly, dropping the value by as much as a factor of four causes few problems. It does lower the load impedance presented to the last IF stage, and this may alter its passband shape. But I've never found this to be a problem. Often the audio level will drop by a dB or two, but I find this a small price to pay for cleaner audio. Usually I wind up cutting the detector load roughly in half. (Lower audio may be caused by excessive IF ripple on C1. Check the waveform with a low-capacitance ×10 scope probe. If you increase C1 until the peak-to-peak IF ripple is less than 20%, the audio level should be within 1 dB of the maximum possible. Making C1 too large will reduce the treble response.)

AVC resistor R3 typically is 2.2MΩ. I've increased its value to 4.7MΩ without problem. If you try this, it's a good idea to check for leakage current in the AVC circuit, particularly when several tubes are controlled. Any leakage current develops an offset voltage across R3; doubling R3 doubles the offset. Increasing R3 will lengthen the AVC time constant, but I've found that the AM time constant is way too short in most receivers, so this normally is not a problem. You can decrease C3 to compensate if you want.

Some communications receivers (and most consumer radios) replace R2 with the volume control R4, typically 500kΩ. C4 then connects the wiper to a grid resistor, typically 10MΩ. Because this AC-coupled load has much higher resistance than the volume control, this circuit configuration can greatly improve modulation acceptance. This technique applies DC from the detected carrier to the volume control. If the wiper contact is worn or dirty, rotation can generate interrupted DC, causing scratchy audio noise. Make sure the volume control is in good condition if you try this circuit.

R5 typically is 1MΩ. This resistor, along with another 1MΩ resistor often used in series-ANL circuits and not shown, can be major contributors to poor modulation acceptance. There is a way to eliminate the R5-C5 network and any additional resistor, and use the AVC capacitor for the ANL function. Not only does this modification improve modulation acceptance, it makes ANL so transparent that you can leave it permanently engaged. It also automatically implements fast-attack AVC, which helps if you modify your AVC system to minimize selective-fading distortion.

An alternative way to improve modulation acceptance is to use a separate diode for detecting AVC and ANL, leaving the existing diode for audio detection only. The second diode in a 6SQ7, for example, is often unused (or tied in parallel with the first diode). If you couple the anode of D1 to the anode of a second diode through 100 pF or so, you can connect the AVC and ANL resistors to the new diode, along with a second load resistor. D1 is then DC-isolated from the AVC and ANL capacitors, and only the volume control coupling capacitor can degrade modulation acceptance. Better still is to connect the 100 pF directly to the plate of the last IF amp. This feeds a wider passband to the AVC detector and keeps it from boosting the gain as you tune away from a station, eliminating raucous noises. The coupling capacitor must have low leakage if tied to the IF-amp plate. Because of the high-level IF signal at D1, you must take care any time you modify its wiring. Unwanted coupling can result in IF feedback that may alter the gain and frequency response of the IF strip, or even cause it to oscillate.

• Originally published in Electric Radio, August 2003

Updating the Automatic Noise Limiter

A noise limiter clips an audio signal to reduce the amplitude of impulse noise. It prevents the waveform, which includes the desired signal and noise pulses, from exceeding a certain level. This can be effective when the noise pulses are thin and tall, less so when they are wide and short. An aggressive noise limiter may target the latter by using a clipping level well within the normal amplitude range of the desired signal. This will generate audible distortion by limiting the signal itself.

If the clipping level is adjustable with a front panel control, you can trade noise reduction for audio distortion as conditions warrant. This is a nice arrangement, but it requires additional parts, wiring, front-panel space, and operator attention. Many communications receivers settle for an automatic noise limiter, where circuitry automatically sets the clipping level based on the carrier level.

ANL is a good idea, but the implementations I've encountered distort the audio so badly that I never use them. Apparently the designers were so keen to limit noise amplitude that they were willing to sacrifice considerable audio quality. Perhaps they reasoned that since the signal was already corrupted by noise, additional distortion to quiet things down was justified.

But audio clipping has benefits beyond limiting external noise. For example, many receivers generate impulses of their own when you change bands or invoke other functions. I once made the mistake of wearing headphones while changing bands on a Hallicrafters SX-24. The impulse the bandswitch generated was so intense that my ears rang for several minutes. I was afraid I might have damaged my hearing. Later I looked at the audio waveform on a scope. The normal audio level was about 1 V, but the band-change impulse was 100 V or more. I never imagined the receiver could generate so much short-term energy. Had the receiver's ANL been so transparent that I could have left it always engaged, this impulse never would have reached my ears.

In addition to limiting receiver switching impulses, ANL has another benefit for receivers with good bass response. It can eliminate the loud thumps that may occur when tuning rapidly across a band filled with strong signals.

Conventional ANL circuits generate distortion another way by measuring the carrier level with an RC filter on the detector output. As described in the modulation acceptance article, this network can seriously distort the detected waveform by back-biasing the detector diode. This can generate audio distortion even when ANL is not engaged.

Improved Circuit

Fig. 1   Improved ANL circuit

Fig. 1 show the ANL circuit I prefer to conventional circuits. It eliminates the ANL's RC network and uses the AVC capacitor to measure carrier level. R1 + R2 is the detector load, while R2 ⁄ (R1 + R2) determines when clamp diode D2 conducts. D2 limits detected audio to the voltage on AVC capacitor C3. (C2 removes any residual IF signal and isn't essential to ANL operation.)

The sum of R1 and R2 affects modulation acceptance, while their ratio determines when ANL clips. I pick the sum as described in the modulation acceptance article, and the ratio so that D2 conducts only when modulation exceeds 125%. This is the maximum upward modulation permitted for AM broadcast stations. You may find one or two local signals that exceed this limit some of the time on some program material. These signals are useful for testing. To pick R1 and R2, I tune to a local signal that hits modulation peaks of 150%. I adjust R1 and R2 until D2 begins to conduct on signal peaks. Then I check that D2 never conducts on signals with 125% modulation. With the ANL threshold set this way, I never hear ANL artifacts or distortion, and I can leave it on all the time. I used 60kΩ for R1 and 46kΩ for R2 for the ANL circuit in my National NC-57.

I use my spectrum analyzer in nonscan mode with its linear detector to find a broadcast signal with the modulation peaks I want. If I didn't have a spectrum spectrum analyzer, I would find the value of R1 and R2 where most broadcast signals just begin to clip. Then I would reduce R2 until no sign of clipping remained.

When limiting occurs, D2 injects some charge into C3 so that repetitive noise pulses tend to increase the AVC voltage and reduce the audio level somewhat. The effect isn't great unless D2 conducts for a sustained period. If you encounter a strong signal while tuning, for example, D2 will quickly charge C3. Once C3 charges to about 40% of the carrier level, D2 stops conducting and R3 charges C3 the rest of the way. This provides a very useful fast-attack AVC characteristic for strong signals. It permits you to use a long AVC time constant without suffering sustained audio blasts. The next article will explain how a very long AVC time constant can minimize selective-fading distortion.

If your receiver shorts C3 to disable AVC, D2 will clamp the audio to ground. C3 must remain operative when AVC is disabled unless you disable ANL at the same time. A SPDT contact that switches the AVC line between ground and C3 will solve the problem, but sometimes you'll have to improvise. I've installed a SPDT relay in receivers that had only SPST switch contacts.

Although there is no steady carrier in CW mode, the detected BFO signal will keep C3 charged to a level where CW tones do not clip.

With R1 and R2 selected as described above, noise limiting is not as great as when the clipping threshold is within the normal modulation range. When I pick these component values, I trade some noise reduction for clean audio. I don't use ANL to suppress external noise because I keep my neighborhood free of impulse noise sources. Instead, I use ANL to suppress receiver switching impulses and tuning thumps, and to permit a very long AVC time constant. You can adjust R1 and R2 to put the clipping threshold wherever you want it.

Performance on Impulse Trains

Fig. 2   Noise pulses 17 ms apart, ANL disabled

Fig. 3   Noise pulses 17 ms apart, ANL enabled

To demonstrate the ANL circuit, I generated a 70%-modulated AM signal near 10 MHz with one signal generator and combined it with the output of a second generator sweeping 1 to 20 MHz every 17 ms. As it passed the receiver's frequency, the swept signal demodulated to a pulse 300 µs wide. I varied the relative levels of the two generators to obtain the impulse level I wanted. Fig. 2 shows the voltage across R2 in my NC-57 with the ANL circuit disabled. Fig. 3 shows it with ANL enabled. While ANL has greatly reduced the impulse amplitudes, it also has dropped the signal about 3 dB. Pulses this rapid and strong raise the AVC level somewhat. This prevents the clipping threshold from being very close to the demodulated signal level.

Fig. 4   Noise pulses 100 ms apart, ANL disabled

Fig. 5   Noise pulses 100 ms apart, ANL enabled

Figs. 4 and 5 show ANL action for pulses 100 ms apart. Here I used a 100%-modulated signal. Because the pulse density is lower, AVC is less affected and the pulses are clamped closer to the desired signal.

The circuit of fig. 1 permanently engages ANL. If you want to be able to turn it off, add a switch in series with D2. Since ANL is permanently on in my receivers, I use the ANL panel switch to select AVC time constants. The only time I find ANL undesirable is when sweeping the IF during receiver alignment. An internal switch (or unsoldering a lead) solves the problem.

With ANL always engaged, you begin to trust your receiver not to startle you with an unexpected outburst. I find that having it on all the time makes tuning and listening more pleasant.

• Originally published in Electric Radio, September 2003

Minimizing Selective Fading Distortion

Fig. 1   Shortwave voice signal in the 25-meter band, two AVC settings

Fig. 2   Shortwave voice signal in the 31-meter band, two AVC settings

Fig. 1 shows two audio waveforms, each 11 seconds long. This is a voice signal from an AM broadcast station in the 25-meter shortwave band. One waveform was received using fast AVC and the other with slow; other receiver settings were identical. Fig. 2 shows a similar comparison for a signal received in the 31-meter band; these waveforms are six seconds long. Can you tell which AVC rate applies to which waveform?

Automatic volume control is a receiver feedback system intended to minimize variation in demodulated audio level. AVC controls the gain of the RF and IF stages (and sometimes the converter stage). An RC filter extracts a signal proportional to the AM carrier level from the detector output. AVC applies this signal to the tube control grids. The RC time constant determines how quickly the AVC system responds to changes in signal level. Receivers with a fixed time constant invariably use fast AVC (generally 100 to 200 ms) for AM reception. This lets the receiver track rapid variation in signal strength with little variation in audio level. Knowing this much, it's easy to guess that the top waveform in each figure, which shows less level variation, corresponds to fast AVC.

Selective Fading

Fig. 3   Digital shortwave broadcast signal. 20 kHz span, 10 dB/div.

What I haven't told you is that the shortwave signals exhibited selective fading. Selective fading is nonuniform signal-strength variation across the receive passband. Fig. 3 shows some examples of selective fading. These are spectrum analyzer traces of a digital shortwave broadcast signal on 17.620 MHz in the late afternoon. The signal spectrum is 10 kHz wide, apparently and presumably flat. This montage shows the variety of ways selective fading can alter a spectrum.

Fig. 4   HF radar signal. First column, 0.5 ms/div; others, 0.2 ms/div.

Fig. 4 shows the cause of selective fading: multipath propagation. The signal is a uniform pulse train emitted by an HF radar at 17.450 MHz. The radar tracks meteor ionization trails for scientific study. I recorded these traces from a spectrum analyzer in nonscan mode using 30 kHz bandwidth and linear amplitude. In these images the horizontal axis is time, not frequency. The first column shows the pulse train early in the day when a slightly delayed pulse was just starting to appear. This extra pulse may be due to the extraordinary ray, a second ionospheric path caused by earth's magnetic field. The remaining images are closeups taken later in the day when each pulse had become an ever-changing ensemble of many. Mentally superimpose an AM carrier for each pulse you see, recall that time delay is phase shift, and you'll get some idea of the self-interference multipath propagation can cause.

Selective fading can change an AM signal's tonal character by altering the shape of the sidebands, and it can change the audio level by altering the sideband or carrier amplitudes. When it attenuates the carrier, selective fading can cause distortion by raising the effective modulation level beyond 100%. When it changes the relative carrier and sideband phase, it can cause distortion by altering the signal envelope. Distortion from these two mechanisms can be severe.

Now return to figs. 1 and 2. In fact, the upper waveform in each figure was received using very slow AVC (a time constant of about 5 seconds), while the lower trace used fast AVC (about 300 ms). It's clear from the traces that slow AVC was much more effective at stabilizing audio level. What the traces don't show is that slow AVC greatly reduced audio distortion.

Selective fading fakes out an AVC system by giving a false indication that the audio level has changed. When the carrier level decreases but the sideband level remains the same, AVC applies less negative bias to the control grids to increase receiver gain. This causes the audio level to rise when it doesn't need to. Similarly, the audio level will drop when the carrier level alone rises. The AVC does the opposite of what it's designed to do. The problem occurs because carrier and sideband levels are not well correlated during selective fading.


Unwarranted volume variation during selective fading isn't the worst of it though. A misbehaving AVC increases audio level just when it is most likely to be distorted. When a selective fade greatly reduces the level of the carrier with respect to the sidebands, it's as though the transmitter were severely overmodulated. A diode detector can cope well with upward modulation exceeding 100%, but downward modulation so large causes the demodulated waveform to clip and fold. With little or no carrier, which is common during severe selective fades, the audio sounds raucous, irritating, and sometimes unintelligible. Most of the high-amplitude sections in the lower traces of figs. 1 and 2 sound highly distorted.

What's interesting is that when the carrier fades and the audio distorts, the audio level generally drops a few dB if the receiver gain is held constant. This may happen because the selective fade is not so selective and attenuates nearby sidebands along with the carrier. It may also be due to the carrier phase rotating with respect to the sidebands. When the carrier phase is off 90°, the output of an envelope detector will drop 14 dB (and consist of pure second harmonic, which may account for some of the audible distortion). It's unlikely that the phase of the carrier and all of the sidebands will align so exactly, but smaller phase rotations still reduce the detection level. In general, the audio level drops during a selective fade if the AVC system is prevented from raising it. You can take advantage of this natural distortion-suppression mechanism by not undoing it with fast AVC.

Carrier loss during a selective fade often is quite brief. If the AVC time constant is very long, the AVC voltage can ride through brief periods of carrier loss with little change. Such an AVC system measures the long-term carrier level and ignores short-term variations. The receiver gain that results is much more appropriate.

Fig. 5   Sixty seconds of carrier from two shortwave broadcast stations

Fig. 5 shows sixty seconds of carrier from two shortwave broadcast signals. I recorded the upper trace in the 25-meter band in the daytime, while the lower trace is from the 31-meter band at night. I obtained these plots by tuning my Kenwood TS-930S to the carrier in CW mode with the 500 Hz IF and audio-peaking filters engaged to reject modulation sidebands. AVC was off. You can see that carrier loss usually is sharp and brief, often bounding right back to the prior level. An AVC with a time constant of several seconds would have no trouble equalizing the long-term carrier level of these signals while riding through the short-term variations.

Do Your Own Experment

It's easy to see how badly fast AVC behaves using your own receiver. Find an AM signal with pronounced selective fading. Listen to it for a while with fast AVC enabled to get a feeling for the character of the fades and the audio distortion. Then turn AVC off. Adjust the RF gain so that the audio level is about the same as before. Then listen to the audio. If you're like me, you may be surprised not to hear any distortion the first time you try this. I assumed that propagation conditions had changed and selective fading had gone away. Switching AVC back on made clear that it was still there. With AVC off, listen carefully when the audio level drops a bit. You should hear some distortion, though it won't sound nearly as raucous as with AVC enabled. What's unexpected is that if you listen casually, the residual distortion with AVC off may drop below your awareness threshold.

Using manual gain control is a very good way to reduce selective-fading distortion. If you're willing to track longer-term variations in signal strength by riding the RF-gain control, you'll be rewarded with much cleaner audio. But most of the time you won't want to pay such close attention while receiving. This is when the use of a very long AVC time constant makes sense.

Implementing Very Slow AVC

In my receivers I use a time constant of about 300 ms for fast AVC and about 5 seconds for slow. It's impractical to have only slow AVC available, at least AVC this slow. When you tune away from a strong signal, it may take several seconds for the receiver gain to rise enough to hear weak signals. I use fast AVC for tuning and slow AVC for listening.

Don't make the fast AVC too fast. If your receiver's audio system has good bass response, an AVC time constant of less than 200 ms may cause audible distortion on low-frequency tones. Residual audio on the AVC capacitor can change the receiver gain at an audio rate, distorting the signal envelope.

Even with your receiver parked on one frequency, there will be times when signal strength increases rapidly. This may happen when propagation suddenly changes or when a stronger signal appears. It's helpful if the slow AVC has fast attack. Otherwise you may hear a loud burst of audio as the AVC slowly comes to the right level. The automatic noise limiter described in the previous article provides fast-attack AVC as a bonus. This ANL uses the AVC capacitor to clamp audio impulses. When a strong signal appears, the clamp diode quickly charges the AVC capacitor and reduces receiver gain. The ANL and AVC systems work synergistically to limit audio impulses and bursts, and to properly set receiver gain.

I use a 4.7MΩ resistor feeding a 1 µF low-leakage capacitor for slow AVC. The charging current available from the ANL circuit depends on the IF output impedance, diode conductivity, and series resistance. Using a very large AVC capacitor may cause some audible distortion if the ANL can't charge it quickly when a large signal appears. I found it easy to arrive at suitable component values by experiment. Be sure to evaluate your circuit under a wide variety of signal and propagation conditions.

Some receivers connect the main AVC capacitor to the controlled grid circuits through RC filters for RF decoupling. The capacitance of these filters often makes only a minor contribution to the overall AVC time constant. It's best not to modify the AVC time constant by altering the value of a decoupling capacitor. These capacitors often exhibit a small reactance, and altering a value may degrade RF tracking or the IF response.

If no decoupling networks are used and the AVC capacitor functions as an RF or IF grid-circuit bypass, replace the original capacitor instead of paralleling another. The series-resonant frequencies of capacitors of unequal value differ. A parallel combination will exhibit a high impedance at some intermediate frequency where the capacitive reactance of one capacitor equals the inductive reactance of the other, effectively disabling the bypass. Replace the original capacitor using the same attachment points and lead lengths since RF tracking may depend on stray circuit inductances. It's a good idea to check RF tracking and IF alignment after capacitor replacement.

• Originally published in Electric Radio, October 2003

May 7, 202288–108 MHz