Select Page

Maximizers in the construction of essb signal

Maximizers in the construction of essb signal
Maximizers in the construction of essb signal
Boosting the efficiency of the Radio Station’s transmitter

MAXIMISER is the final, completed plug-in in our digital station. In this article, I’ll use information from izotope A. With. Lukin with my additions.

Since I created an entire article on this topic, because this VST plug-in in our field of ESSB Radio is in essence the most important way to generate a signal to use the power of your transmitter 100%.

I also want to draw your attention if you use the maximizers, you will have to turn off the compression on your transceire to avoid distortion of the signal. In fact, the maximisors at the level of software but higher quality create the compression of your signal. And do not forget if the transceire is turned off compression your transceiver works much cleaner in the construction of your signal. Any transceiver compressor increases all the noise of your sound channel. The computer maximizer makes compression but does not increase channel noise.

The purpose of the appointment – no matter what voice is broadcasting into the microphone, on the professional station AM or FM – at the entrance to the transmitter is an
ethereal processor.Its main function – as much as possible to reduce the peak factor of the signal, with the aim of higher efficiency of use of the transmitter power!

Maximizer (digital peak limiter) is a dynamic processing device that increases the level of sound signal when mastering or broadcasting. Sometimes the maximizers are other types of devices, such as psychoacoustic processing of the type BBE Sonic Maximizer – we will not consider them here. We will get to know the principles of dynamic processing devices and compare some popular models of software maximisors.

Loudness and levels

The volume of sound depends not only on the level of sound (or sound pressure), but also on its spectral-temporal composition. If the frequency balance of the phonogram is already defined and it is undesirable to change it, then to increase the volume of the phonogram you need to raise the signal level. Why raise the volume? There are two reasons for this. The first is that loud music often seems “beautiful” than quiet, and attracts more attention. Therefore, most producers are trying their best to increase the level of phonogram when mastering: because it may depend on its commercial success. The second reason for the volume increase is the desire to make the most of the dynamic range of the audio media, whether it’s a CD or an analog tape. It is also important to make the most of the dynamic range of the reproducing device so that the recording does not sink in noise.

When recording a sound, the media usually limit the amount of overload to the peak signal level rather than its medium-square power (this is a somewhat simplified but near-reality model for most analog and digital sound carriers). The ratio of peak level of the phonogram to its average level (RMS) is called the peak factor (crest-factor, cross-factor). The rectangular wave (meander) has a single peak factor of 0 dB. The peak factor of sine waves is 3 dB. Phonograms with broad dynamics or sharp peaks have a high peak factor (20 dB or more), and highly compressed phonograms – low peak factor (10…15 dB). It is clear that with limited peak power, a phonogram with a smaller peak factor can reach a higher volume. In order to reduce the phonogram peak factor, it is passed through dynamic processing devices (Figure. 1). Let’s take a look at their how they work.

Maximizers in the construction of essb signal
Fig. 1. Phonogram before dynamic processing and after. Reducing the peak factor. Clippering is not here, and the sound is quite acceptable for radio broadcasting.

Dynamic processing devices

The main devices for working with phonogram levels are dynamic processing devices. The principle of these devices is to analyze the level of the audio signal they enter and change this level by some law. The main parameters of dynamic processing devices are the transmission characteristic and attack/recovery time.

The transmission characteristic (not to be confused with the amplitude-frequency characteristic) is the dependence of the desired output level of sound from the input level. According to the transmission characteristic, the dynamic processing device determines the gain factor that needs to be applied to the input signal at any given time. An example of the transmission characteristic is shown in the pic. 2. This dynamic processing device is called a compressor; it passes without change sounds with amplitude up to -20 dB and reduces the amplitude of all sounds above -20 dB. Thus, the compressor makes loud sounds quieter, narrowing the dynamic range of the phonogram.

Maximizers in the construction of essb signal
Fig. 2. The transmission of the compressor. The threshold is -20 dB, the compression degree is 2:1.

A fracture in the transmission characteristic is called the knee. The entry level corresponding to the knee is called the threshold. The angle of the transmission characteristic above the threshold determines the degree of compression (ratio, degree of compression). The compression rate of 2:1 means that if the input level increases by 2 dB above the threshold, the output level will increase by only 1 dB. If the compression rate is one, the sound level will not change when you pass through the device. If it strives for infinity, the device will limit the amplitude of the output sound to the threshold. These devices are called limiters, they limit the dynamic range. If the compression rate is less than one, such as 1:1.5, it means that when the entry level exceeds the threshold, the device will raise the output level compared to the input level. These devices are called exploiters, they extend the dynamic range. There are other types of dynamic processing devices: gates, daqers, leveers, etc., with their specific transmission characteristics and performance parameters.

Sometimes the transmission characteristic is smoothed out so that there are no sharp angles (Figure. 3). This mode is called soft knee or soft threshold. The compressor with a soft threshold begins to slightly reduce the signal level even before it reaches the threshold.

Maximizers in the construction of essb signal
Fig. 3. Soft threshold.

The work of the dynamic processing device can be described as the following scheme. The device monitors the entrance level and adjusts the output level, i.e. applies to the input signal some amplitude (gain factor) that change over time. Several conditions need to be met for a good sound of the resulting signal. The most important of these is that the amplitude bend should be smooth, without ruptures and, if possible, without kinks. Indeed, if the amplitude envelope has ruptures, the output sound will also have bursts in the form of a wave, audible as clicks and crackling. The kinks in the amplitude enveloping will also lead to output distortions.

There are two parameters for smoothing the amplitude of the dynamic processing devices: attack time, trigger time, and recovery time (release). They determine the speed at which the device reacts to changes in the entry level. The timing of the attack shows how long the device reacts to exceeding the threshold (attack), and the recovery time shows how long the device reacts to returning the entry level back to the threshold.

Maximizers in the construction of essb signal
Fig. 4. Sound before and after compressor treatment.

Let the compressor enter first a weak signal that does not exceed the threshold, and then an attack that exceeds the threshold (Figure). 4). According to the transmission characteristic, the compressor must miss a weak signal without change, and the level of loud signal (attack) – to weaken. The attack time indicates how long the compressor will change its gain factor from single to resultable, prescribed transmission. If, following a loud signal, the entry level falls below the threshold again, the compressor goes into the recovery stage and again increases its gain factor to a single one. The time during which the gain factor will return to a single value, and will be the time of recovery.

The timing of the attack and recovery may vary from manufacturer to manufacturer. In some devices, recovery time is not understood as the full time of the gain factor return, but the time of its return, say, up to half way back. Often the gain factor returns to the original value of the exhibit, in which case only the second definition makes sense. In some devices, the attack time is set at the rate at which the gain factor (dB/s) changes, or, conversely, the time it becomes that the gain factor will change by 6 dB.

Attack time and recovery time are measured in milliseconds and can vary widely for different dynamic processing devices and depending on specific tasks. For example, in compressors the usual time of attack is about 10… 100ms, and the typical recovery time is about 100… 1000ms. It is clear that the longer the attack and recovery time, the slower the amplitude wrap will change over time, the smoother it will be. However, in a high time of attack, the compressor will miss short attacks that exceed the threshold, because won’t have time to react to them. This can be undesirable, for example, for limiters.

Another option found in dynamic processors is the release delay or hold delay. This setting sets the time through which the recovery stage begins after the entry level decline below the threshold. In other words, this option allows you to postpone recovery for a while. This can be useful when exceeding the threshold in the signal go periodically, one after another. In this case, the recovery delay will help to avoid constantly switching the compressor between the attack and recovery modes and reduce the breakdown of the amplitude envelope.

Now about how dynamic processing devices determine the level of the input signal. This is usually done in one of two ways and is similar to the functioning of level indicators: peak and medium-square. The first way is to detect instant peaks in the input signal. The second is the averaging of power in time, i.e. RMS calculation. The peak method is often used in limiteds, where it is necessary to limit the peak values of the signal to some threshold (for example, before issuing a signal to the radio line or recording on a CD). RMS is more commonly used in compressors to align the volume of audio, as it is used. volume is more associated with medium-square than with peak power.

Peak power exceeds RMS, and this should be taken into account when setting up the device. It is also clear that calculating RMS power requires some time interval to integrate power, and therefore the device’s reaction time to input changes may not be much less than this integration time. In other words, the RMS compressor can miss short-term signal peaks, almost before it has time to reduce the transmission rate.

Another common feature of dynamic processing processors is side-chain, an additional control input for the beep. When this function is activated, two signals are received to the device: through the main and control entrances. In this case, the “control” signal is used only to determine the input level and to control the level of the main signal in accordance with the transmission characteristic.

With side-chain, you can achieve some interesting effects. If the side-chain sends the same signal as the main entrance, the device will behave as usual, without side-chain. If another signal is sent to the side-chain, the device will handle the main signal, guided by the amplitude profile of the control signal. For example, if a side-chain sends a signal that is passed through an equalizer with a frequency characteristic, the Fletcher-Manson reverse curve (the curves of equal hearing volume), the amplitude of the control signal will more correctly reflect the real volume of the main signal. And the dynamic processing device will be guided by the real volume of the original signal, rather than its amplitude, when processing the main signal. With this technique, you can more plausibly align the volume instead of the amplitude.

Let’s emphasize that the signal sent to the side-chain does not affect the timbre (frequency balance) of the main processing signal. It only controls the amplitude of the envelope.

When working with stereo records, dynamic processors typically operate in linked channels mode, i.e. apply the same amplitude enveloping to the left and right channels. Otherwise, the stereo panorama is disrupted.

The extender cancels the compressor.

In conclusion, the general part about dynamic processing should be noted that although the best recordings of the world’s sound engineering were created with the help of compressors, careless handling of the compressor can irrevocably spoil a good record. It is a mistake to think that the action of the compressor can be undone by an expander. If the dynamics are lost, there is nothing to expand. In addition, both compressors and expanders have some inertia, which makes it impossible to accurately restore the dynamics.

ak, our task – to raise the level of the finished mix to the maximum possible values, without making significant distortions. The easiest way to achieve this is to normalize the level, when the peak of the maximum level in the phonogram is sought and the entire phonogram is amplified by the magnitude of this peak, so that the peak takes the value of 0 dB. Further increase in the level of the phonogram will lead to clipping (clipping, amplitude restriction) – overload, causing unwanted distortion.

It is obvious that dynamic processing can be used to further improve the level of the phonogram. If you skip the phonogram through the compressor or limiter, the peak values of the phonogram will decrease and you can still raise the overall level without overload.

What to use to increase the volume: compressor or limiter? Renowned mastering engineer Bob Katz recommends using a compressor when a change in the nature of sound is required, a noticeable decrease in its dynamics. Limiters, on the other, use when you don’t want to make any changes to the sound, except for the volume.

The maximizer is a dynamic processing device that consists of a limiter and a subsequent signal amplifier. Often the system of reducing discharge is also embedded in the maximisers, but here we will not consider this part.

The main control elements of the maximizer are the threshold of the trigger (threshold) and the limiter settings (attack, release). In some maximisers there is also a “ceiling” regulator, which allows after limitation to strengthen the signal not to 0 dB, but a little weaker, to leave a little “headroom” in case of a little further processing. For example, if the phonogram is encoded in mp3 after the maximiser, the shape of the wave will change slightly when encoded, and clippings may occur. Even if the signal is not supposed to be further processed or compressed with quality loss, a little free space may be required for the dittering noise added when the discharge is reduced.

The lower the trigger threshold, the stronger the limiter limits the dynamic range and the stronger it will be possible to raise the volume after the limiter. Thus, lower threshold values lead to a louder sound at the output of the device.

If you limit the peaks to more than 3 dB, the sound will spoil.

The amplifier does not cause any questions, so let’s stop at the device limiter. The task of the maximizer is to maximize the level of the signal, but to prevent clipping, i.e. not to allow instant power to go beyond the 0 dB level. It follows that only the peak method is suitable for the maximum supfore as a method for determining the entry level. The maximizer should track the peaks of the signal and build an amplitude envelope so that after its application to the signal peaks are below the threshold level. When the entry level is below the threshold, the maximizer misses the signal unchanged. And when the entry level exceeds the threshold, the limiter should weaken the signal so that there is no exceeding the threshold.

Since we want the amplitude to be smooth, without breaks and kinks, we conclude that the limiter should know what amplitude profile the sound wave will have in the coming moments of time. Indeed, if the limiter did not have such an opportunity, if there was a sharp attack at the entrance, exceeding the threshold, the limiter would have to immediately lower the level of gain to prevent the threshold from exceeding. Instantly lowering the level of gain – this is a gap or kink in the amplitude enveloping, which is desirable to avoid. So, to build a smooth amplitude envelope limiter, you need to know the values of the signal with some advance. Since there is no reliable predictor of the signal on past values, the look-ahead function is implemented with a slight delay of the output signal relative to the input. Thus, when issuing a output signal corresponding to the time t, the limiter actually already has an input for the moments of time up to t’T, where T is the delay time. It’s like a news channel that “relays” another news agency with a delay of 10 minutes. At any given time, the channel’s staff already know what will happen in 10 minutes, and can accordingly modify the news issued to viewers, achieving better quality of the material due to delay.

Keep in mind that the delay made by the limiter may be undesirable in some situations. For example, if you insert a limiter into the console line break, this line will be delayed in time relative to others, which can lead to a distortion of the timbre when mixing lines. Fortunately, the maximizers usually apply to the ready-made mix in the mastering process, in which case the delay does not play a role. If the real-time maximisor does not delay the signal, it means that it either allows for exceeding the threshold, or its amplitude envelope is broken. The third is not given.

You should also keep in mind about the delay when you want to synchronize channels in real-time sound processing programs. If processing is not done in real time, the program that performs the processing (host application) can most often compensate for the delay, i.e. “align” the output signal of the maximizer on time. Usually the delays made by the maximizers are small, up to 10 ms, but there are exceptions.

Guided by future peak levels, the limiter can build a smooth amplitude, starting to weaken the amplification in advance, before the attack in the input signal. In other words, the limiter should build an amplitude enveloping around the peaks in the form of pits, where the depth of the pits will be determined by the amount of exceeding the threshold peak level, and the width – the time of attack and recovery (Figure. 5). It is clear that the wider the pits, the larger areas of the signal will be suppressed, and the less will be the final volume of the phonogram. Thus, the volume of the phonogram depends not only on the set threshold, but also on the time of the attack/recovery, as well as on the shape of the amplitude enveloping during the attack and recovery.

Maximizers in the construction of essb signal
Fig. 5. The original sound (shown by the limiter threshold) and the amplitude envelopes built by various maximizers. From above – rapid recovery, in the middle – slow, from below – the algorithm of auto-control of recovery.

Time management of attack and recovery time

When the signal is multiplied by an amplitude envelope, additional harmonics may appear in the signal spectrum. The less time the attack and recovery of the maximizer, the louder the resulting sound is, but the more broken the amplitude is and the more intermodulation distortions occur.

With little attack time and recovery, intermodulation distortions become especially noticeable when the signal has bass tones of a large amplitude with a period greater or equal to the attack/recovery time. This can be demonstrated on test signals, which are sums of sinusoid with different frequencies (standard test of intermodulation distortions, rice. 6).

Maximizers in the construction of essb signal
Fig. 6. Intermodulation distortions that occur when two harmonics pass through the maximizers. In the top picture, the aggressiveness of the maximizers is higher.

With a high time of attack and recovery, the “pumping” effect begins to show. Volume failures occur around short-term peaks in the signal (Figure. 5). Around each of these peaks, the amplitude envelope has the form of a wide pit, failing the entire signal at volume. By ear it is perceived as a fallout, a tremor of volume.

Thus, the timing of the attack/recovery is a trade-off between intermodulation distortion and the volume failure effect. For further considerations, we will introduce the concept of aggressiveness of the maximizer. Let’s say that one maximiser is more aggressive than the other, if at equal threshold values the first maximizer gives a louder (rmS) sound on the way out. It is clear that aggressiveness depends on the time of the attack/recovery and on the shape of the amplitude enveloping during attacks/recovery.

For most maximisors, the user sets the time to attack and recover manually after setting the threshold. If intermodulation distortions are heard, aggressiveness decreases (attack/recovery time increases). If you are not audible – you can try to increase aggressiveness in the hope of achieving greater volume and reducing the effect of volume failure. Normally, deep limitation requires more attack/recovery time.

There is a way to automatically adaptively control the aggressiveness of the maximizer based on input analysis. Indeed, if there are sharp peaks in the phonogram, it is desirable to establish a higher aggressiveness, so that there is no effect of failure of volume. Intermodulation distortions in this case will not arise, because. if the peaks are isolated, there will be no significant periodics in the amplitude, leading to intermodulation distortions. In addition, our ear has the property of reduced sensitivity to short-term, up to 6 ms, distortions. Thus, the maximizer will “respond quickly” to single peaks, immediately returning to a single gain factor.

If the entrance receives a periodic signal, with constant, periodically following excesses of the threshold, it is desirable to lower the aggressiveness of the maximizer (i.e. increase the attack/recovery time to avoid intermodulation distortions.

If such regulation is carried out adaptively, constantly adjusting to the input signal, it will significantly increase the average aggressiveness of the maximizer (i.e. volume output), without increasing distortions.

One of the first maximisers implementing such a strategy is Waves L2 in ARC (Auto Release Control). It should be noted that the literal understanding of the term “automatic recovery time management” does not accurately describe how Waves L2 works. This maximizer uses a slightly more complex method of constructing an amplitude envelope, based on a combination of two types of amplitude envelopes: aggressive and non-aggressive. At single peaks of the input signal is used aggressive envelope, and at periodic, group excesses of the threshold – a certain combination of two rounding. This achieves a louder and better sound than simple management of recovery time. A similar algorithm is implemented in the maximiser i’otope Ozone 3.

 To check the presence of the function of auto-control aggressiveness, we will conduct a simple test. In the left channel of the test file, we will create the next test signal. In the first second, let there be one short-term pulse peak. In the second second, put there a sine wave with a frequency of 100 Hz and leave at the end of a little silence. In the right channel, put a constant current (DC) throughout the file. The amplitude of signals will be taken so that both the pulse and the sinusoid e
xceed the threshold of the maximizer, and the constant current – did not exceed. applied the same amplitude enveloping to both channels. Then at the exit of the maximizer in the right channel will contain an amplitude envelope, which the maximizer built on the left channel (Figure. 5). Looking at the shape of the amplitude wave can draw many useful conclusions about the functioning of the maximizer. If the recovery time after the peak is significantly less than the recovery time for the sinusoid, the maximizer uses auto-control of recovery time.

Overload between counts

The vast majority of maximizers are digital devices. Indeed, in the analogue it is almost impossible to do “looking forward” and therefore analog peak limiters are doomed either for the instant an instant time of attack (which leads to breaks amplitude envelope), or to pass some overload (clipping). Digital maximizers are able to “look ahead” and respond to an attack in advance, milliseconds before the peak.

  Analog limits are better than digital ones.

The purpose of most digital maximizers is to prevent the digital wave from exceeding the threshold, i.e. to limit the value of all signal counts to the size of the threshold. However, this digital restriction does not ensure that the threshold is not exceeded in the analog wave restored by the digital signal. Indeed, the analog wave, smoothly oscillating between discrete counts, can exceed the value of digital counts of up to 3 dB and above (Figure. 7). How can this affect the sound? First, there may be an overload of DACs. Typically, DAs use recrestyrate – a digital increase in the frequency of signal sampling. In this case, the restored digital wave values between the calculations of the original digital wave can overwhelm the dac network (which is often the case). Thus, the clipping of the “similar” wave led to distortions of sound even before the signal became analog. But even if the CAP correctly restored the wave above the level of 0 dB FS, the rest of the components of the audio circuit (e.g., operating amplifiers) may not be as resistant to overload.

Maximizers in the construction of essb signal
Fig. 7. Analog and digital peak levels may not be the same.

It turns out that it is possible to make a digital limit so that the restored analog wave also does not contain excess thresholds. It is enough to use recresmeration to algorithmically restore the analog wave and to detect peaks not by digital counts, but by analog wave. Further limitation of the digital wave is carried out as usual, but using new, “similar” information about signal peaks.

The traditional tool for dealing with the problem of analog clipping is to understate the ceiling (gain after limitation ratio) by the decibel share. As we can see from our reasoning, such a measure is totally insufficient. For real audio analog excess threshold is often 1…1.5 dB, not a fraction of the decibel.

 Here’s a simple test to determine how the device is healthy to eliminate analog clippings from the maximizer. We generate a sinusoid in a digital file with a frequency of equal to a quarter of the sampling frequency and the initial phase of 45 degrees (figure. 7). For such sinusoids, the analogue wave exceeds the value of digital counts by 3 dB. Let’s skip this sinusoid through the maximizer. Let’s set the threshold as low as possible. If the maximizer does not allow to disperse the level of digital counts any noticeably higher than -3 dB, it correctly determines the peaks of analog wave. If it habitually dispersed digital counts to 0 dB, then the detection of peaks in it is carried out on a digital wave.

Smoothing the amplitude of the envelope

Explosions and kinks of amplitude enveloping – a common occurrence for the vast majority of maximizers. They lead to similar ruptures and kinks in the form of a weekend sound wave. In spectrum terms, this means that the range of emerging intermodulation distortions becomes wide, covering all frequencies. At the same time, the audibility of distortions increases many times. Indeed, with smooth amplitude enveloping intermodulation distortions are usually grouped around peaks in the signal spectrum, where they are likely to be psychoacoustically disguised by these peaks. If there are kinks and ruptures amplitude envelope spectrum of distortion expands and can go beyond the threshold of camouflage. Distortions become audible as crackling. On rice. 6 examples of the work of maximizers with smoothing amplitude and without smoothing.

 To illustrate these distortions, we will conduct the next simple experiment. Let’s take any audio recording and filter out all frequencies above 3 kHz. After that, let’s skip the recording through the maximizer. We set the threshold so that the limiter does not do nothing, and we will listen to the result. If we have chosen a maximizer with ruptures or breaks of amplitude, the recording will be heard a noticeable crackling (figure. 8).
Why did we filter out everything above 3 kHz? To make the crackling more noticeable. If we hadn’t filtered out the RF, the crackling would have been the same, but it would have been somewhat masked by the high frequencies of the original sound.
Maximizers in the construction of essb signal
Fig. 8. Cod spectrogram in a record that appears after processing by the maximizer. The upper passage was processed by the maximizer without smoothing the amplitude. Lower – with smoothing amplitude envelope (and even with more aggressiveness).

Tips for using maximisors

The maximiser should be the last link in the mastering chain. After that, only the audio discharge is reduced (often combined with maximization). All other sound processing and conversions, including sampling frequency conversions, must be performed before the level is maximized, as they can change the peak values of the sound wave and lead to clipping or incomplete use of the dynamic range.

When setting up the maximisor settings, you should start from the volume level of the phonogram you need to get. Set the maximizer threshold to meet the desired volume increase and then move on to setting up the aggressiveness. If intermodulation distortions are noticeable (such as wheezing on bass notes), reduce aggressiveness (for example, by increasing recovery time). If the distortions are not noticeable – try to increase the aggressiveness to reduce the effect of volume failure (pumping).

If the threshold of the maximizer is already very low, and the volume is still not enough – refer to other dynamic processing devices. Treat the sound with a compressor. If even after the compressor and the maximiser “dispersal is not the same” – try a multi-band compressor. If that’s not enough, check to see if the phonogram has already turned into a pink noise.

Limiters and maximizers can easily “kill” the microdynamics of the phonogram. If compression is usually an artistic technique, then limitation is rather a technological solution. And the technology is better left for mastering specialists, who often have better equipment and means of controlling the result.

Comparison of the quality of maximisors

  With equal attack/recovery time, all maximizers sound the same.

And now – an overview of popular maximisors implemented in the form of software modules (plug-ins) for PC. Since there is no single adequate criterion for comparing maximisors, several possible tests are proposed to identify the features and design disadvantages of the maximizers.

The following plug-ins are included in our review:

  • Waves L1+, L2
  • Voxengo Elephant HQ 1.3, Voxengo Elephant 2.0 (EL-2 mode), Voxengo Elephant 3.0 (EL UNI mode)
  • Sonalksis MaxLimit
  • TC Native Limiter
  • Digidesign Maxim
  • Flux Pure Limiter
  • iZotope Ozone 2 Maximizer (Brickwall mode)
  • iZotope Ozone 3 Maximizer (Intelligent mode)
  • Sonic Foundry Wave Hammer
  • DSP-FX Optimizer
  • Steinberg PeakMaster
  • DB Audioware Mastering Limiter
  • Kjaerhus Audio Classic Master Limiter
  • Wave Arts FinalPlug 4.6
  • Anwida Soft L1V 1.5
  • 4Front Mastering Bundle XLimiter
  • built-in limiter from Nuendo 2
  • built-in limiter from Logic 5.5.1
  • built-in VST Dynamics limitedr from Cubase SX 2
  • built-in Hard Limiter from Cool Edit Pro (Audition).

It is immediately necessary to stipulate that not all existing models are considered here. Maximizers are chosen in two ways: sound quality and popularity. The popular Timeworks Mastering Compressor was deliberately excluded from testing because it is not a maximizer with a strict trigger threshold and misses the clipping. Also not included in the Steinberg Loudness Maximizer comparison, as in most modes it intentionally introduces overload distortion into the signal.

Comparison criteria

What is the best way to compare maximisors? Listen to them at work! Let’s try to offer some conditions of experiments for higher objectivity of comparison when listening. But first of all, let’s turn to some of the comparisons recommended in other articles and explain their shortcomings.

In the article “Maximizers” by Alexei Seitsev (“Musical Equipment,” June 2001) the approach is such. A test sound of two sinusoids (intermodulation distortion sinerate test) is passed through the maximizers. After that, the spectrogram analyzes harmonic and intermodulation distortions (harmonic distortions, apparently, the level of harmonics only test tone with a higher frequency). All test ers are given “equal conditions”: recovery times and thresholds. As we already know, equal recovery time at equal thresholds does not ensure the same aggressiveness of maximizers. Therefore, the conclusions about the quality of maximizers based on the levels of KGI and CNI measured in this test are not entirely fair.

Another common type of test is to bring the input and exit waves to the same amplitude and subtract them. Such a test is popular when they want to show how the maximizers spoil the sound. Indeed, most of the received wave is zero, but in those places where the maximiser was triggered, strong clicks, cracking and scraps of the original melody break out. Perhaps this test is only useful to determine the moments when the maximizer is triggered. He does not allow to judge the real sound of the maximizer, just as such a “different test” does not allow to judge, for example, about the noise-cancelling system. The visibility of distortions cannot be estimated in isolation from the main sound.

Another popular test identifies the “best” frequency of sampling, at which the maximizer gives the least distortion. The 1 kHz sine is taken and is passed through the maximizer at various sampling frequencies (44.1 kHz and 96 kHz). On the graphs of distortion spectrums (Figure. 9a) It is concluded that the 96 kHz maximiser sounds much cleaner, without distortion. Such illustrations are, for example, in Bob Katz’s book “Mastering Audio”. In fact, a smaller number of distortions at 96 kHz sampling frequency is a pure coincidence for this particular test signal. The thing is that 96,000 is divided into 1000, and 44,100 is not divided into 1000. Therefore, the amplitude profile of this 1 kHz sinusoid at 44.1 kHz is complex, forcing the maximizer to follow its curves and make additional harmonics. If you take the sinusoid with a different frequency, say – 1260 Hz, then it can get the exact opposite results (Figure. 9b).

Maximizers in the construction of essb signal
Fig. 9 (a, b). The level of distortion of the maximizer at different sampling frequencies is incorrect to determine one test tone.

In fact, the level of distortion of maximizers really depends on the frequency of sampling. However, addiction is not always easy. The fact is that not all devices correctly adapt the time of attack/recovery to different frequencies of sampling, and they are different aggressiveness. But if you throw away the questions of aggressiveness, it often turns out that at higher sampling frequencies the level of distortion of dynamic processing devices turns out to be less (both on test sinusoids and on real signals). For sinusoidal signals, this is because at high sampling frequency the amplitude profile of the digitized sinusoidis is closer to the constant, and the maximizer does not have to follow its bends and make distortions. For real musical signals, this is because the intermodulation distortions that occur during processing are distributed across a wider band of frequencies, and some distortions are not audible. (With lower sampling frequencies, distortions that would be in ultrasound are reflected in the audible part of the spectrum, so-called. aliasing). Some high-quality dynamic processing devices (such as Weiss) use double sampling technology, i.e. respective of the digital signal before dynamic processing. The advantages of such a technology for maximizers are undeniable, although it is used, for example, in the Plug-in Voxengo Elephant Hz (and may lead to inaccurate threshold follow). However, compressors are reported to have reported an expected improvement in sound quality.

So how do you increase the objectivity of testing? We propose to compare the sound of maximizers on real sound material, exposing the same threshold values and achieving the same aggressiveness from the testers tested. How to achieve the same aggressiveness? Only by measuring the RMS resulting signal! Let the first maximiser give the output signal with RMS, say- -9.25 dB, and the second maximizer at the same settings – -9.89 dB. Although the difference is small, it is an average difference in volume throughout the sound fragment. If we consider that the maximizer works only on a small part of the fragment (where there were exceeding thresholds), in fact the volume of really limited areas could vary very markedly. Therefore, we conclude that the aggressiveness of the comparisond maximisors in these settings is different and change the parameters (recovery time) of one of the maximizers so that RMS output signals are equal. Now the maximizers work with the same efficiency, and it is possible to make a comparison by ear.

Usually, first of all pay attention to the presence and visibility of intermodulation distortions, trembling and failures of volume (pumping), crackling and other possible additional sounds. Thus, when all maximizers work in equal conditions, giving a signal of the same volume, the best of them choose by ear.

If it is desirable to give some objective numerical data, it is possible to conduct tests of intermodulation distortions and assess the spectrum and magnitude of distortions (of course, having achieved a pre-equal aggressiveness of maximizers).

On rice. 6 compares the two maximisers with two different aggressiveness values. Within each picture, the aggressiveness of the maximizers is the same. The graphs of intermodulation distortions show that the peaks of distortions are almost the same, but the width of the spectrum of distortions varies greatly. Therefore, the visibility of distortions will be much higher for the maximizer with a wider range of distortions.

The article by Aleksey Seitsev suggests that the wide range of intermodulation distortions of the Timeworks Mastering Compressor plug-in symbolizes the work of special saturative algorithms that simulate lamp distortions. Unfortunately, this is far from reality. The wide range of intermodulation distortions shown in the graph arise from the clipping that this plug-in misses. Despite the fact that the graph shown in the article is not very noticeable, in reality, distortions in this case are not harmonics of the test signal. Each of these “harmonics” consists of several closely spaced intermodulation tones. In addition, the spectrum of lamp distortions is valued precisely because it has a small number of harmonics. In the case of maximisors, we often have a spectrum with an endless slowly fading row of harmonics, not particularly reminiscent of lamp distortions even on the spectrum, and the sound is more reminiscent of crackling (rice. 8).


In this table, we provide data on the presence of tested maximisors of certain capabilities (or lack of certain shortcomings). I tried to take into account as many objective criteria as possible, important for quality sound. You can expect maximisors with similar characteristics to sound similar. Although a detailed analysis of the sound requires a close listening of specific devices, the experienced user this table will allow to form a good idea of the sound of a particular model. The table lacks the functionality that is obvious from the documentation (for example, the presence of certain controls), the focus is on the features of the implementation of algorithms. The order of algorithms in the table is not an unambiguous rating of quality (although it can be considered as an approximate rating).

Comparison criteria:

  1. No clipping – no clipping.
  2. Full range – full use of dynamic range (the possibility of maximizing with a strict threshold of 0 dB).
  3. Look-ahead is an opportunity to look ahead.
  4. Continuous env. – No gaps in the amplitude.
  5. Smooth env. – smoothing the amplitude of the envelope, a narrow spectrum of distortions.
  6. ARC is an automatic recovery time control (auto release control).
  7. Adjustable ARC – the ability to vary aggressiveness when the ARC mode is enabled.
  8. Analog detection is the ability to detect the peaks of analog signal between digital counts.


  No clipping Full range Look ahead Continuous env. Smooth Env. ARC Adjustable ARC Analog detection
Steinberg PeakMaster + ?
Steinberg BuzMaxi 3 ? + +
TC Native L + +
Anwida L1V + + +
Digidesign Maxim + + +
iZotope Ozone 2 + + +
DSP-FX Optimizer + +/- + +
Kjaerhus Classic + +/- + +
Crysonic SpectraPhy 1.0 + + + -/+
4Front XLimiter + + + +
DB Mastering Limiter + + + +
GVST GMax + + + +
Logic Limiter + + + +
Cubase SX Dynamics + + + +
Nuendo Limiter + + + +
CEP Hard Limiting + + + +
SF Graphic Dynamics + + + +
Voxengo Elephant HQ +/- -/+ + + +
Kjaerhus GPP-1 + + + +/- -/+
Massey L2007 + + + + +
TB Barricade 2.1 + + + + +
Wave Arts FinalPlug + + + + +
G. Yohng’s W1 Limiter + + + + +
Sonalksis MaxLimit + + + + +
Slate FG-X 1.1.2 + + + + +
Sonnox Oxford Limiter + + + + +/- -/+ -/+
Waves L1+ + + + + + +
Cakewalk Boost 11 + +/- + + +/- +/-
Flux Pure Limiter + + + + + +
Waves L2 + + + + + +
PSP Xenon + + + + + + -/+
FabFilter Pro-L + + + + -/+ + + -/+
Voxengo Elephant 2.0 + +/- + + + + +
Voxengo Elephant 3.0 + +/- + + +/- + + -/+
iZotope Ozone 9 + + + + + + + +



Brief conclusions

The best quality is demonstrated by maximizers who have the technology of automatic control of recovery time. The Waves L2 Maximiser has actually become an industry standard for high-quality processing. It combines ARC technology with smoothing amplitude. The drawbacks of L2 is that for some reason the function of detecting “similar” peaks was excluded from it, although this feature was present in Waves L1, and in ARC mode it has no possibility to regulate aggressiveness. The aggressiveness of L2 is quite high (and it sounds really loud), and at noticeable levels of limitation intermodulation distortions begin to appear. You can get rid of them only by abandoning the ARC mode and manually setting a high recovery time.

Maximizers in the construction of essb signal

The Intelligent Maximizer Maximisor from i’otope Ozone 9 was released recently. It combines ARC mode with adjustable aggressiveness, smoothing amplitude and the ability to detect “similar” peaks. This maximizer has the best sound and can be seen as an extension of Waves L2 towards customizable aggressiveness and detection of analog peaks.

  • Transparent restriction with multiple IRC ™ (Intelligent Release Control) technology modes, now with several improved IRC IV modes and IRC Low Latency mode.
  • Threshold Learn: Automatically setting the threshold based on a user’s LUFS target, perfect for finding the right volume for your preferred streaming service.
  • Threshold and Ceiling Link: Reduce the threshold without raising the perceived level to hear the audible effect of your processing.
Maximizers in the construction of essb signal

I’otope Ozone 9 sounds softer than Waves L2. But I would advise putting them in the kit first Waves L2 and at the end of the i’otope Ozone 9

About The Author

Leave a reply