Fast Audio Peak Limiter

Main Index Projects Index

There have been many attempts to create a Voltage Controlled Amplifier / Attenuator (VCA) that is both fast and linear, and many fine examples exist.  Unfortunately, many of these are relatively expensive or are difficult to get (or both), and the cheaper ones often just don't seem to make the grade for one reason or another.  For those interested in the various gain control topologies that have been used for compressors and limiters, see VCA Techniques.

The majority of simple VCA circuits have a limited input voltage range, with some exhibiting excessive distortion if the input voltage exceeds as little as 10mV.  At such a low signal voltage, noise then becomes a major problem, as well as control voltage 'feed-through'.  This latter effect shows up as very low frequencies being generated by the circuit, and this can easily overload the power amplifier under some circumstances.  It is almost a given that these very circumstances will be present when you least expect or need your subwoofer to 'bottom out'.

A Light Dependent Resistor (LDR) is an excellent (and very low distortion) variable resistance, but most are fairly slow, and allow a maximum attack time of about 15ms.  For many applications, the LDR / LED combination will be quite acceptable (for example with electric guitar or bass or 'full range' music), but for stopping an analogue to digital converter from clipping, you really need something faster.

The circuit devised by Phil Allison still has some input voltage limitations, since it is based on a FET.  Junction FET VCAs create considerable distortion, with the worst of it appearing when the signal is attenuated by 6dB.  The common way to fix this problem is to apply 1/2 of the drain voltage to the gate, along with the control voltage.  Figure 1 shows the conventional way this is done.  The predominantly second harmonic distortion is cancelled by this technique, leaving a much smaller amount of third harmonic distortion.  This is further reduced by ensuring that the signal voltage between drain and source is less than 100mV, but the exact voltage is dependent on the FET used.  In general, try to keep the peak AC voltage below 56mV, or around 40mV RMS.

Figure 1 - 'Conventional' VCA Using a JFET

The arrangement shown looks perfectly reasonable when you first see it, but closer examination reveals that the two 1M resistors form a voltage divider for the control voltage (CV), and this exists until the 100nF cap is charged.  The maximum attack time is limited, with a time constant of 100ms as shown.  Figure 1 is adapted (for comparative purposes only) from a published circuit by a well respected designer. 

The 25k pot is used to adjust the limiting threshold, which is essential because JFETs vary widely in their parameters.

The circuit is only useful if large peaks in signal level are tolerable.  If they are, a peak limiter is probably not needed anyway. 

However, it actually is possible to make this circuit behave itself.  If C1 is made much larger than normal (typically 1-2µF or so, while retaining the 1MΩ resistors R3, R4), the time constant is now so great that the control signal is always attenuated to half its normal value.  This approach has been used in a number of commercial FET based limiters, and while it does couple some low frequency control signal into the audio path, it generally works fairly well.

This audio peak limiter employs a FET as a variable resistance to attenuate the input signal according to a control voltage (CV).  It offers unusually good performance with low cost and component count.  A TL072 dual opamp (U1) provides the circuit gain and full wave peak detection.  The input attenuator (R1, R2) is essential to keep the voltage across the JFET to the minimum possible, otherwise distortion increases dramatically.

The 4.7k resistor and 1µF capacitor (R13 and C5) determine the attack time, which is about 5ms as shown.  R11 and C5 determine the release or recovery time, and as shown this is approximately 1 second.  The attack time can be reduced, but that often results in very audible (and unpleasant) 'click' sounds when the signal changes amplitude rapidly.

R10, C3, C4 and R12 form the distortion cancelling circuit, and as can be seen, the control voltage impedance is very low compared to the distortion cancellation impedance, so the circuit's attack time is not compromised.  The values of resistance and capacitance have been optimised for the least distortion across the audio band, at 0.3% THD typical for frequencies above around 500 Hz, at 1.65V RMS output level.  Below 500 Hz, the distortion rises gently with decreasing frequency, but also falls with lower voltages.  Distortion is negligible at any voltage level below the limiting threshold.

Figure 2 - Fast Audio Peak Limiter

As described above, the attack time with the values shown is 5ms, with a release time of around 1 second.  This is a good compromise for most audio material, but is readily changed by altering the values of R13 (attack) and R11 (release/ decay).  Be careful of values for R13 of less than 1k, as the opamp may be unable to supply the current needed to charge C5.  Ideally, C5 needs to be a low leakage capacitor - either a low leakage electrolytic or a tantalum if you must (although I never recommend tantalum for anything).  A standard electro is probably inappropriate for this circuit, especially if longer release times are desired.  Having said that, most are better than you may have been led to expect.

In addition, keep R11 a minimum of 10 times R13 ... for example, if R13 were to be 1k, the minimum value for R11 will be 10k.  This would be a very fast limiter indeed!  Faster decay is possible, but it doesn't sound nice.  The circuit has been changed so that R11 (decay control) is connected to the diode side of R13, so if a fast decay is needed the control voltage is not attenuated.

Note:   The 2N5459 specified is obsolete, as are most of the JFETs we used to know and use regularly.  While you may be able to buy FETs with that number printed on them, don't expect them to be genuine! The range of FETs has shrunk alarmingly in the last few years.  The 2N5459 might be available in an SMD package (e.g. MMBF5459) but don't count on it.  Many other JFETs can be used in this circuit, but you will need to adjust VR1 to set the threshold, and you may need to try a few different types to get good results.

If you select an SMD part for the JFET, it will be very small (but you knew that already), and hard to mount.  I must leave it to the reader to work out how best to adapt an SMD part if you choose this approach.  No two JFETs (even of the same type) will ever be the same (known as 'parameter spread'), so it may be necessary to evaluate a number of devices to get one that functions properly.  However, the 'adjust threshold' trimpot can provide sufficient range for most JFETs.

Figure 3 shows an optional Schmitt trigger indicator circuit.  This will indicate the limiting is taking place, with the LED illuminating at approximately 1 dB attenuation, which occurs with a control voltage signal of 1.6V.  Make sure that the LED current can't flow in the audio path's earth (ground) circuit, and ideally the indicator's supply will be isolated from that used for the audio.  Fast switching can easily introduce noise into the audio signal.

Figure 3 - Optional Schmitt Trigger Indicator

If desired, a LED VU (or analogue moving-coil) meter may be used here instead, and with proper calibration will give a good indication of the peak attenuation at any time.  This option will require some experimentation from the constructor, and further details are up to the individual to work out.  With most JFETs, the variation of gate voltage may be quite large, but with a significant DC offset.  This makes metering fairly difficult to achieve, and the variation is not linear with gain reduction.

This circuit is a vast improvement on the conventional approach as shown in Figure 1.  With that circuit, any attempt to make the attack time shorter than about 20ms may create nasty clicks in the signal, as the FET only gets half the initial control voltage during the time it takes to charge the distortion cancellation capacitor.  As a result, the attenuation is greatly reduced during this critical period, and the transient is allowed through almost unaffected by the FET.

The resulting 'fight' between the FET and control voltage amplifier circuit can also cause the signal level to be reduced too far initially (after the transient), after which it must then recover.  The overall effect is not at all pleasant, and is best avoided.  (Note the use of gross understatement!)  It is precisely this sort of problem that has given some limiters a bad name over the years.

The descriptive text is a mixture of Phil's original description and some additional information provided by ESP.

Main Index Projects Index