The AM Forum

I recently acquired a Viking Valiant with the stock clipper circuit in place and I've noticed that there is no shortage of posts urging the removal (or bypassing) of that circuit to "improve frequency response and reduce distortion". This makes sense since even a properly designed clipper creates noticeable distortion when over-driven and a low pass filter is required to reduce harmonic content generated by clipping.

However, every single professional broadcast audio processor includes a clipping and LPF circuit as the last stage. Legally this is a must to ensure that peak energy is tightly controlled within -99% to +125%, an FCC rule. As a safety feature, clippers protect expensive modulation transformers from spikes that can make it through compressors. Competatively, properly set clippers produce noticably louder audio than limiters, a must in commercial broadcasting.

So why are clippers scorned in the amateur world?

There's a fine line between loud and distorted when it comes to clippers and properly setting the amount of drive is critical. In broadcast applications, clippers follow compressors (or limiters) so the dynamic range is already tight. Setting a clipper to handle a consistent 1 or 2 dB is considered max, but it makes a big difference in perceived loudness to listeners.

Unfortunately, clippers in most amateur circuits aren't preceded by anything so they become driven by 8-20 dB of dynamic range as if they were compressors and sound like hell. From the Valiant manual: "12 dB peak clipping - not at all objectionable, on the contrary, speech sounds as though the speaker is enunciating with special care".  Hmm.

By the way, the less clipping is needed the less low pass filtering is needed. That means that the LPFs that follow many amateur clipper circuits are designed to manage copious amounts of harmonic content. A clipper used with discretion would require a far more gentle rolloff than is usually employed.

So, my plan is to flatten out the frequency response of my Valiant, use a microphone processor to feed the Valiant with a consistent audio level, and set the 6AL5 circuit to catch the peaks that get through. The LPF can easily be redesigned with some minor component value changes.

I'd be happy to share these circuit suggestions here if anyone's interested...

The usual recipe is to use the 3 diode high level negative limiter circuit in the modulator, and let the positive peaks drive as high as they happen to go.

My Valiant II does about 115% positive according to WA1QIX's monitoring. It has sweep tubes for modulators.

There's absolutely no reason to limit positive peaks to 100%.
There is a good reason to limit negative peaks to no more than 99%.

The clipper circuit in the Valiant is not a good idea, generally speaking.
It came at a time when "talk power" to compete with the coming SSB "revolution" was important.

A soft knee limiter is a much better plan, if you want "dense" audio. Imo.

                        _-_-

I'm interested in your circuit suggestions.

I have a Heathkit Apache, that, in its stock configuration, included a 6AL5 clipper.  When adjusted properly, the result was good communications grade audio.

73,
Brad

Everyone understands the importance of keeping negative peaks from cutting off the carrier but the benefit of positive peaks is widely misunderstood.

The problem, particularly for older rigs, is subjecting mod transformers to the higher voltages as a result of >100% positive modulation. Additionally, the audio energy contained in those peaks is miniscule compared to the average modulation level although peaks look cool on mod monitors and they give operators all kinds of bragging material. High risk, low benefit. Besides, perceived loudness is a result of average, not peak, levels.

Properly clipped and symmetric modulation will be much louder and cleaner to the listener than allowing unclipped peaks to reach 200%, and why risk the iron? Motorola released an extensive report in the 1980s outlining the importance of symmetric modulation to reduce distortion generated within receiver detector circuits. Asymmetric modulation also suffers from greater distortion due to propogation.

BTW, the 6AL5 is a soft clipper...

Interesting comment, Clark.  I'm interested especially because my Collins 32V1 is capable of greater than 125% positive peaks, occasionally going as high at 140%.  So far, so good..

BTW, what operating times are you doing these days?  I confess that my early evening AM operations have suffered a bit lately.

Al

Hi:

The Valiant (and Apache, and others) clipper preceded audio stages with substantial low frequency roll-off (and often a crystal mic). This reduced the IM distortion created in the clipper. Also remember these manufacturers wanted to create a product with the fewest parts, so that meant no compressor. I agree that carefully controlling the level going to the clipper is a key to it's not getting in the way of the sound. I agree that negative clipping at the modulation secondary is the best place to clip, as there is often less or no phase change after this point, where the low level clipper square waves will be rounded and tilted when going through the driver and modulation transformers.

Less low pass filtering with less clipping is not really true. While a hard clipper will create harmonics, your natural voice goes beyond 2KHz, the desired bandwidth of SSB stations. If you want to stay within a certain bandwidth, the only way to do that is with a properly designed LPF. Note that any LPF will also tilt a square wave that results from clipping, and your ceiling is no longer as defined. In the end you need to look at occupied bandwidth with whatever you do in the transmitter, compare it to desired bandwidth, and add some form of LPF. My vocal "esses" go out to 6Kc (12Kc occupied bandwidth) before filtering. For broadcasting, AM stations need to stay within their channel, and must have tight filtering to protect the adjacent channel. The designers of ham tube transmitters in the 1950s had limited occupied bandwidth and communications quality audio as main goals. So they had steep high pass and low pass filters. Today, we often desire "hi fi" audio on the ham bands. I clip a little and limit my bandwidth.

73,
Dan

Not sure anyone was suggesting 200% peaks.

What Motorola was trying to do in the 1980s differs sufficiently, imo, from what is generally aimed or in today's AM transmitters and operations as to be not terribly relevant.

If you want very symmetric audio, run it through a few sections of all pass filter and ur done. That's pretty much what some broadcast "processors" did or do.

I'm not sure I agree entirely with the perceived loudness being solely a function of average level. Certainly a full modulation is more audible than less than full modulation, under conditions that are less than good. Otoh, I have heard stations that run more than 100% modulation and their audio is both cleaner and appears to have a greater ability to get through marginal conditions.

High average modulation using compressors and peak limiters and the like generally sounds awful except for "battle conditions" that I have heard.

I guess one could make the case for higher peak modulation and less carrier the result eventually being DSB! :D

But there are all sorts of ways to do things, and everyone has their preferences and ideas. Nothing wrong with limiting the rig to 100% positive peaks...

                                   

I don't recall ever seeing anything that approaches 200% pos peaks.  I have, however, had levels approaching 175% with my class E transmitter, which is very linear.

I use enhanced density from my dBx 285s but do not think I sound particularly distorted -  badly adjusted transmitters may result in a distorted sound but not properly designed / adjusted ones.  

I see no need to make the AM modulation component symmetrical.  Just design / adjust one's station according to good engineering practice.  Having said that, one should take measures to detect the incoming AM signal with adequate linearity so that transmitting stations that are properly producing the  asymmetrical human voice are accurately reproduced. There are several approaches that achieve that goal, including enhanced carrier detection, precision detectors and Sync AM detection. Several of our folks have designed these type of detectors and have published the results.

Al

For great and loud audio, referring to AM, the most important thing is to set up and/or design your transmitter to be as flat in audio response as possible, as low in distortion as possible, and able to modulate any waveform applied to it as accurately as possible.  ALL audio processing should be done BEFORE applying it to the transmitter.  That's the way the broadcasters do it.  ALL phase rotation, AGC, limiting, clipping, filtering, should be done BEFORE sending it to the transmitter.

I have spent my entire career in broadcast radio, most of it was AM with a number of AM transmitters in the 50 kilowatt range.  I also design the parameters of some of the AM (and FM as well) Omnia audio processors for broadcast.  And that is how we do it.

73
Ted  W8IXY

Here's a link to what I consider THE Bible of audio processing as it was co-written by Bob Orban (Orban) and Frank Foti (Omnia), designers of the two most popular broadcast processors in the world:
 
http://www.orban.com/support/orban/techtopics/Appdx_Radio_Ready_The_Truth_1.3.pdf
 
It would be amazing if more people understood and embraced what is being said here.

A valueable reference, but we are not working with the same considerations or constraints as are AM broadcast stations. Nor do we have the same goals or aims.

Old school clippers should be avoided for two reasons.

1. They create distortion than cannot be filter out with a lowpass filter.

2. The lowpass filter designs found in most 50/60s transmitters have horrible impulse response resulting in overshoot. The overshoot results in peaks, the very thing the clipper removed. So, the circuit is essentially fighting against itself.


There are many low cost processors available now or simple level control circuits that can be easily built.  These processors and circuits will control modulation levels better than clippers and will not create distortion. I don't see messing with clippers worthwhile.

Steve, do you have an example diagram of an old school clipper? I want to understand well.

Check the schematics for say, a Valiant or Apache or Eico 730 on BAMA.

I don't think I ever heard anyone using a clipper that sounded good, not that I remember anyway.
And I can never hear much difference between 120% positive and 150%.

I do think a lot of people could use some compression and agc on the audio, unless you watch it all the time, a mic into a rig does not work very well, low audio or closing off the carrier or both just by moving around or talking a bit louder.

I always thought the plan on "modernizing" these table -top transmitters was to eliminate the 6AL5 all together and update the mic circuits for wider range audio, so they would not be so screetchy. And add some negative audio feedback from the modulator to clean up any distortion.
The 3 diode set-up I tried on a DX100, from long ago, created more problems than I could ever imagine.
A famous AMer here, WA1HLR, Timtron, has some interesting approaches to firing a table-top transmitter using a power transformer as a modulator transformer.
I used one on an Elmac TX with a 6146 final and it made big  audio using 6550's as modulators!!
And 120% or around that number is about max that any AM radio can detect without distortion.
A couple others mentioned here that the audio processing, compressors/limiters, preceding the transmitter input is the real answer.
Fred

I'm proud to say that I own a Collins 32V1 that Timmy HLR did a full up mod on.  I've gotten excellent reports on it.  In fact, I'm told it is hard to tell it from the class E transmitter. It's a pleasure to operate.

Attached photo shows the 32V1, lower Collins unit.  The 75a2 is basically unmodified.  The only processing is the aforementioned dBx 286s

Al

Look at the page on an eico 730 as attached

PS: Sorry about the previous mixup on the graphic

Built-in ham transmitter clippers have a lot of problems and sound pretty bad.

But a small amount of clipping following other effective processing can help keep modulation levels high and still sound good.

Here are the keys, in my opinion, to making clipping most effective and clean-sounding:

1. The frequency response of all audio circuits past the clipper should be dead-flat and wider than the actual audio you want to transmit. Particularly, the post-clipper response should extend well into the subsonic range, or the flat tops of the clipped waves will become tilted and overshoot their target levels, eliminating the advantages of clipping. DC coupling is best, flat amplitude and phase response down to 10 Hz or so is almost as good. Hard to do with transformer coupling! (The homemade diode high-level clippers in plate-modulated rigs get around this by putting the clipper past the modulation transformer. And Class E and other pulse-modulated rigs are perfect in this regard -- that's one reason why those running them can use quite a lot of clipping and sound loud without sounding very dirty at all.)

2. The source audio and the following audio stages should also be extremely clean. (Another area where older 50s/60s commercial rigs and their clippers fail badly.) The processed audio waveform must be reproduced near-perfectly, and typical commercial ham AM transmitters don't come anywhere near doing that.

3. If low-pass filters are used, they must be of the overshoot-compensated type, or they will impair the level-control advantages of clipping.

4. Bright audio with an extended high end, with lots of pre-clipping treble density, sounds much better when clipped than dull audio. I believe this is because the natural high frequencies mask the IM and harmonics created by the clipper.

One cheap processor that uses clipping well, in my opinion, is the Aphex Dominator. Many famous and good-sounding CDs have been made with them!

Very interesting discussion.

73,

Kevin, WB4AIO.

I use negative peak clipping on 100% of my transmitters.  Would not operate without it.  If I modify an existing transmitter, I put a negative peak clipper into the circuit.

With this, you can *easily* get 6dB more audio (assuming your modulator is capable of producing the extra audio, that is) and *no one* notices the clipping on the air.

Of course as many others have pointed out here, the other half of the equation is audio processing and having a clean, flat modulator to start with.

No real need for positive peak clipping here on ham radio, since we are not limited as to positive modulation percentage.

By the way, I usually set my negative peak limiters to around 95% negative.

Under quiet band conditions and good signal to noise ratio operating, I back everything off a bit for full fidelity, largely unclipped operation. The clipper at that point is just a watchdog and hardly ever is really doing much of anything.

Yes, an AM transmitter MUST be very clean to start with.  Otherwise, by simply increasing audio power, the modulator will generally become dirtier on its own, regardless of clipping.

I have run various transmitter IMD tests using the excellent QIX clipper design taken from the class E PDM generator.  It has an audio aliasing LP filter after the clipper.   I find that the overall transmitter RF IMD will degrade very slightly from excellent to good as the NPL is dialed in aggressively. (About 5dB degradation in 3rd order IMD) There are no free rides, but it is acceptable.   These tests were run with both a class E MOSFET PDM rig and my new dual quads tube PDM rig.

I use the NPL clipper as a safety net  for occasional over-modulated negative peaks that would either cause some splatter or do damage to the mod transformer of my 4X1 plate modulated rig.   Certainly, we need a method of stopping the carrier from being totally cut off during negative peaks, whether using a negative peak clipper, limiter, or whatever.

To lean hard against a clipper all the time can invite splatter if the rig is not already very clean, so it pays to either use it as an occasional  safety net OR work on the transmitter to make it near-perfect first.

The point is, with the stock Valiant type rigs, positive peak IMD is already so bad when the audio is hit hard, that the neg peak clipping will not help - and even add some slight crud to the negative peaks.   A neg peak clipper will not change the waveform of already distorted positive peaks.

If you can build or modify a super clean AM transmitter, the whirl is your oyster. Talk anywhere and never have to worry about splatter reports.   At least -30 dB THD and -30dB 3rd order IMD is required for entry into this exclusive club.  ;)


Additional notes:

1) THD and IMD tests can be performed with simple audio tones using most any SDR pan scope.

2) On AM, there is a difference between "splatter" and normal side channel products. Splatter is the nasty, spitting, tearing sound of distorted audio up the band that extends higher than the fundamental audio freqs.  Clean side products produce a nice wispy, diminishing smooth S sound as tuned higher.

3) The comments above pertain to using a low level negative peak limiter.

T

On point "2)"  An observation not to be overlooked by the careful operator of an AM station.

All observations on this post are excellent in my view

Al

"...submitted for your approval..." - Rod Serling


Genuine old school clipper.

It's important to point out that a clipper, like any tool, has to be used properly and those early ham transmitter circuits certainly were not good examples of proper use.

Clippers should be used in conjunction with processing that will limit what the clipper stage to just peaks. Some broadcast processors even monitor the current flowing through clipper diodes and use that to adjust the gain of the preceeding limiter in order to prevent overdriving.

Music store processors are a great value for increasing average levels of modulation, but very few have the ability to catch peaks that would show up on mod monitors, even if they claim to be limiters. Also, decreasing attack times on compressors to reduce peaks will only increase gain reduction and lead to "pumping".

The best way to view a clipper is as a means to eliminate peaks that get past compressors so that modulation level can be increased. It's like trimming hedges. That's how they best increase loudness. Those same peaks can punch holes in mod iron insulation... another benefit of clipping.

A simple but effective processing combination would be to use compression with modest attack times followed by a clipper to take out the peaks. The compressor bring up the average levels and increases perceived loudness and the clipper eliminates the peaks that do nothing for loudness and cause mod levels to have to be reduced.

A lot of good information here on clippers. Just my 2 cents on this.

Yes, clippers can create nasty products up the down the band if they are conducting and the operator is leaning into them. But in the real world of operating AM under tough conditions from my observations, the stations that are also running 20-25kc of audio bandwidth using so called "good" engineering practices and that are "clean" with low IMD, can cause the just as much interference problems to co-channel QSOs as poorly used/designed clipping circuits.

Good engineering practice but very poor operating practice.

Al VE3AJM

   This is a good discussion. A few years back I bought a kit from K7DYY. This was a sophisticated mic compressor that is intended to go into the base of a D-104 microphone, and to set the speech level to drive one of the K7DYY Class D AM power transmitters. There are two designs, and the latest has just one pot. That is what I got. I was astounded to see how the thing would saturate (clip) on voice peaks during the processor "attack time". Apparently the "Super Senior" must have a clipper inside to minimize this effect. I was using this D-104 with my Gonset G-76 and Central Electronics 20A. I ended up adding a clipper to the circuit which only gets hit (limits) on the first voice peak after a long pause. After doing that, I absolutely love the thing.
   
  The following article from old man James Tonne, W4ENE covers this topic of clipping and post clipping processing very thoroughly. I suggest you print this out and read several times since it is choc full of good information. He had a QST article back in 1956 called "Compression and Clipping". His more recent work adds to that earlier article after nearly 60 years have passed by.

http://tonnesoftware.com/appnotes/speech/speechamp.html

Jim
Wd5JKO

Yes, the attack time on many music store processors - AND even on some well known AM processor/ limiters can be very slow.     I had to get rid of several units because they would produce a square wave on the first hard syllable.  Try a "tock" sound with your tongue and see what it looks like on the RF scope or output of the processor.  You may be surprised.   A pulse / square wave into a plate modulated system can cause havoc on parts and produce splatter.  

Of course, the newer software and digital boxes have "look ahead" methods that will limit leading edges well. Just be sure to test your complete system together with the fast rising "tock" of the tongue. You never know if some stage in the audio chain is not acting well.

I finally settled on an older CRL  PMC-300A peak modulation processor. The attack time was fast enuff to handle anything. This is used in conjunction with the QIX neg peak clipper circuit.    

I've used another unit called the Aphex Dominator II that is quite fast. It is a later model and often used as a failsafe to limit rogue audio spikes - to save musicians' eardrums when using headphones.

T

A few thoughts.

I'm unclear why a typical ham would need or want compression.
Seems to me that a soft knee limiter is all that is required.

The chip in the K7DYY mike bottom is a fairly sophisticated unit that has several functions built in, if I am recalling properly it can be set up to do any or all of them with various different parameters. Of course that still doesn't mean that it couldn't simply not be able miss that first attack. Imho, the earlier chip looked better as far as specs on paper compared to the later one. I have not tried them, but I looked very hard at them with an eye to using them in a multi band processor set up. Never did it.

                    _-_-

A question:

I have reasonably normal hearing.

When I listen to people talking (not over the radio, but when they are standing near me in a quiet room), they all sound different.

They don't sound distorted, or "unnatural" in any way... just different.

If I understand, correctly, the very elementary physiological aspects of how humans produce speech, and how hearing works ... there are a combination of linear and non-linear processes taking place that make one "natural" sounding individual's voice sound different from another "natural" sounding individual's voice.

Likewise, voice processing can make someone sound different, but sill sound "natural"

What types of linear and non-linear processes can be applied, between the input of a microphone and the output of a speaker, that will only change the way someone sounds; but, will not be perceived as producing an "unnatural" sounding result?

What types of linear and non-linear processing effects are most objectionable, in terms of making natural sounding voices sound unnatural?

Stu

Hi Stu,

First, I think all forms of processing, EQing, limiting and hard clipping are non-linear. Put a sinewave thru and look at the departure from the perfect sine.  Hard clipping is probably the harshest, with softer audio limiting cleaner.  EQing produces phase shift problems too. They are all  forms of distortion.

If the audio gear is set up properly, then I think generating an unnatural sound is a matter of degree. Most techniques can sound good if used sparely.  But, almost any processing technique, when pushed past a certain point will make the sound unnatural, thus the added "distortion"  sounds poor. And, it's a matter of opinion where this point is.

For example, 1 dB of compression will not be noticeable.  20dB of compression will make most people cringe.

Positive peaks:  If we use a negative peak clipper and run 110% positive and 95% negative, no one will notice. But run 200% pos by slamming the clipper hard, then many people will notice and some will object. Certainly some diode detectors will complain.

Add 1 dB of bass boast and no one will notice.  Add 15dB with a Big Bottom box and they will beat ya up.

This is why it can be an endless quest to make everyone happy. There will always be a few who will not like the sound.  It's a compromise. If we are lucky, eventually our buddies will like what they hear, we get a few good unsolicited audio reports - and we like what we hear of ourselves on the playback recording.  

T

I've noticed several versions of the qix 3 diode clipper. Which one is preferred?

Tnx

Philip

Hi Philip -

I'm talking about a low level circuit - that is a high level one.

The 3-diode clipper is a high level circuit that goes between the modulation transformer and before the final.  Personally, I have tried that circuit on at least five plated modulated rigs and later removed them due to IMD problems when hit hard.  They are probably OK as an occasional safety net, however.  

For a NPL clipper, I prefer the one diode low level circuit as found in QIX's PDM generator circuit. This can be found somewhere on his website.  There is also a thread on this BB about it with a circuit too.

If you were to order one of his kit PDM gen boards, it would be contained within it  - and also uses a 6 KHz low pass filter after the clipper stage.   I use one on the output of my AM audio chain (after the audio soft limiting) to handle neg peaks on the various PDM and plate modulated rigs.

I use a touch of two techniques together - the PMC-300A peak limiter that limits both pos and neg peaks  -   and the QIX single diode neg peak clipper with 6Khz LP filter.

T

Something often sounds better then nothing.
Some guys who run a standard old rig, with mods or without, with just a microphone into the rig often have very low weak audio.
Back away from the mike and you get 50% or less modulation.
There is someone I talk to from time to time who sort of drifts down in level, down to 20% modulation, I will say something, and its back up for a short while, then back down.
Your standalone rig has only modulator current to tell about the modulation level, and some set and forget compressor can help a lot. It gives lights to look at, and some compensation for the audio level.
A 50 to 100 watt rig needs to have a good amount of audio on it, a weak signal at 50% mod is just annoying.

On lower power rigs, I always used to set the audio to overmodulate a bit.
If a little is good, more is better!
 

For those wanting a 'look ahead' without nasty digital stuff, a 50FT coil of garden hose put neatly inside a box makes a decent delay. It is not flat or perfect and I have not done this on the air, it was before I had a license or radio transmitters. The delay reminded me of it.

A small 1-2" speaker at one end is the driver and any sort of transducer can work at the other end. A crystal earphone worked as a pickup in my experiment, based on an old popular electronics project for making 'delayed' rear channels for a hi-fi system. The article didn't complain about reverberation but it was  a slight issue I overcame mostly by putting some thick felt at the pickup end of the hose, around the ear-canal part of the earphone. It did ok for the intended purpose and YMMV.

50FT is long, 50ms delay, and shorter could be used to get past that first 100-1000us I guess. I didn't try that but thought about resonances at such a short length.

There is an audio recording from someone here, unfortunately for all their effort they allowed distortion that should not have been in there. duh. One channel is 50FT and the other 15 FT. 15FT seems much better. That is what, 15ms? enough.
http://theb-roll.com/garden-hose-stereo-delay/

Anyway there's an analog delay that could be made to work but with a lot of effort.  There is always tape as a delay, but that has moving parts.

Well, this has turned into a very interesting and informative posting...thank you all!

I mentioned earlier that I would report on the "final product" for using the Valiant with the original clipper circuit. We all seem to agree that it's undesireable to drive any clipper with raw audio even though that's exactly what the Valiant does. I imagine the EF Johnson designers decided to leave it at that in the interest of simplicity and pricing.

IF the clipper is to stay, something needs to limit the audio before the 6AL5 sees it, and I found a curious little compressor made by Waters that does just that:

(http://www.internetwork.com/radio/images/waters359a.jpg) (http://www.internetwork.com/radio/images/waters359b.jpg) (http://www.internetwork.com/radio/images/waters359c.jpg)

Delightfully, the 359 features a choice of 1/4" three conductor jacks or screw terminals (the tip carries the PTT closure) and takes a microphone level input and reduces the dynamic range.

For more flexilbility, I ended up using a combined pre-emphasis circuit & compressor with microphone level output to feed my Valiant, but I wanted to showcase the 359 because it's a neat find.

Early in my AM operating, I used some sort of signal corps tube type compressor.
Transformer input and output, and it had a round meter on the front marked gain reduction.
At the time, I thought it worked quite well, it was rack mount and cool looking.
The next thing I tried was a solid state DBX unit that could do expansion or compression, and even had a threshold setting.
I still have it, its pre IC chip stuff.
You were supposed to compress the audio before recording it on tape, then expand it on playback to drop the noise floor.
For keeping the audio in a general range, both worked, but likely with a bit of overshoot.
Back then, I think most guys were crappy sounding and wide, along with other issues, at least by todays standards...


 

Hi

I love this topic.

i would like to have a kind of "confirmation" regarding the allpass filter or sometimes said as a phase scrambler. But still related to the main topic.

As the description of all pass filter which is delaying audio starting from 300hz and more delay on higher audio freq. And my perception when receiving the transmission from AMer who use all pass filter, it seems that the talk punch is increasing. Am i right at this point? my knowledge said that using all pass filter, amplitude are spreading all over the audio, so the waveforms will become "fat" with less "thin" peaks.

So, when the allpass filter is installed do you think it is not necessarily to clip the audio after allpass in order to increase talk power (loudness).

Thank you for replying my email.

Best Regards

Agus



   

Hi Agus!  Phase rotators (or All-Pass filters) almost deserve their own thread. Basically, they were developed for broadcasters as a way to reduce the amount of gain reduction needed for a given output level by reducing the amount if asymmetry in program content, particularly voice.

Since the DC voltage used for gain control in compressors comes from the audio signal, higher peaks in one polarity cause greater gain reduction than is needed. The use of a phase rotator gives compressors that follow a more consistent waveform to sample.

They do this by gradually rotating the phase throughout the audio spectrum so that asymmetry in the different frequency ranges tend to cancel each other out. Think of it as zero phase shift at 100 Hz, 180 degrees at 3 kHz, and a full 360 at 10 kHz. That's just an example...

You can only hear a difference if you wear headphones since the audio mixes with the sound of your voice conducted through your jaw, but to anyone else listening, they won't hear any phase or equalization effects. You will, however, be louder since your compressor will not be reducing gain as much as without a rotator.

You are mostly correct. The allpass filter will reduce the asymmetry of the waveform. So some peaks will be reduced.