The AM Forum

Hi, I am now considering this kind of circuits:

http://www.amwindow.org/tech/htm/3diode.htm

Just for curiosity I have simulated a circuit that includes a 8 kohm resistance and a diode (cathode to ground) to emulate the load of an RF tube. The "tube" is powered by DC voltage in series with a modulation transformer 1:1 - 10H (coupling k=0.98). The audio amplifier, excited with 1 kHz sine wave, is ideal but has an internal resistance of 1K.
If modulation amplitude is below clipping, the audio amplifier "sees" a nearly resistive load, "voltage and current are sine waves", this is good.
When the anode voltage of the emulated RF tube goes below zero the loading circuit is opened by the diode (it emulates the fact that our RF tube does not take any current with negative plate voltage - it is like an open diode) and the amplifier current (ideally sine wave) has a sort of "shortcut" at the corresponding peak. This is not so dramatic, but various spurious oscillations appear, that are caused by the transformer leakage inductance and distributed capacitances. I suppose that those  oscillations (10 kHz - 200 kHz range)are the origin of most "splatters".
Overmodulation zero clipping cannot be avoided at this stage, on the other hand the perfect sine shape of the current in the audio transformer can be exactly restored and spurious oscillation removed by loading the modulation circuit, in parallel to the RF tube, with a series of a diode and a resistor.
The resistor in my simulation is obviously 8 kohm and the diode, anode to ground, must be capable to sustain the full peak anode voltage.
This is NOT a clipper or a limiter, it is a load impedance "linearizer". Exactly what is needed to make the audio amplifier and the modulation transformer work the best they can  under clipping conditions. Obviously, clipping conditions are those of the modulated RF stage, it is supposed that the audio amplifier never clips and never overloads.
The power dissipation of the linearizing load resistor can be about 1/5 of the power of the audio amplifier.

I shall include this circuit in my future AM tube transmitter...
73 Giorgio

Now I see that this circuit: http://amwindow.org/tech/htm/3diodeka.htm is quite elaborate and D1-R1 perform the function of proper audio amplifier loading.
Very good circuit. It takes into account what I think is the main issue.
See also page 25 of http://www.arizona-am.net/PHOENIX/NA7RH/6DQ5_Charlie_Pics_rev3.pdf

You don't need three diodes to make a neg peak limiter.

You only need a single diode and a keep alive voltage of about 10% of your plate voltage.  The keep alive PS should be built with poor regulation.  The poor regulation allows the voltage to sag as current is pulled from it.  This sag in voltage rounds the corners of the clipping.  The sudden brick-wall clip is where the distortion comes from.

You need a high speed diode, ordinary PS diodes are usually not fast enough.  The PIV of the diode should be 4-5 times your plate voltage.  So, the diode will be a string of diodes to get the needed PIV.  I use a .5meg resistor across each diode (no caps).

The cathode end of the diode string gets connected directly to the RF end of the mod xfmr.  I put a small fuse (about 100ma, more for big rigs) between the PS and the anode of the first diode.

Any load that the mod xfmr needs to see is presented by the internal impedance of the PS.

For 50-100 watt rigs, your can use a backwards connected filament xfmr.  Use a RCRC filter, then add another resistor (about 2-3K 2-5W) and maybe about a .1-.5uf cap to ground.  So you end up with a RCRCRC filter.  I also use a LED is series with the PS (after filter before the fuse).

Circuit works FB. I've been using this set-up for years.

You can experiment somewhat with your PS, but remember it should allow for voltage sag.

The other circuit you posted seems way over engineered.  IMO, my circuit is simple and it works fine.

Fred

I suppose this is very relevant to transmitters having no proper audio processing ahead of the modulator input.

Mic to amp to modulator, barefoot, peaks galore.

Those diode jobs sound a lot like a clipper, and clippers do add loudness at the expense of audio crud & distortion.  Even one that 'rounds the corners' a bit.

For higher 'FI', you wouldn't see any such thing in a broadcast rig, nor should one be added to a converted BC transmitter for ham use.

It is all taken care of in a proper audio AGC/compressor/limiter after the mic preamp and before the transmitter, like a DAP310, Optimod, or Volumax/Audimax.  Correctly phased for symmetry advantage.

Yeah, I know, that is a LOT more complex and way overboard for something like an Apache.

But, as we all find out, the 'next step' in quality of anything is at least 10X in effort/money/complexity. 

i.e., You were happy with a bunch of good Canon camera stuff till you played with a pro Hasselblad & Zeiss combo.

Off the soapbox for now, and 73. :D

DG

DG,

My circuit is designed for us poorer hams who don't have any of that fancy high priced audio gear. :)

Fred

Chops sufficiently busted.

If I had any $, I'd buy you lunch! :D

I do have plenty of ideas, but those are cheap.... :P

73DG

Eh, you can buy most 70's and 80's vintage bc processors for $50-100. Or get a newer vintage Behringer 2496 for about $100 (it has six band processing).

Fred spent more than that on the knobs for his 51 watt rig.  ;)

I believe that when Steve, WA1QIX worked at the broadcast station he installed the 3-diode ultra-modulation circuit with an added keep-alive feature in the transmitter.  Said it worked well.  Maybe he could furnish more details.

I prefer to rely on the natural asymmetry of the human voice to achieve extended positive peaks.  The additional apparent loudness is NOT due to the added sideband power from the positive peaks, but due to the overall increase in average percentage of modulation, since the additional positive peak headroom allows the negative peaks to hit closer to 100% without flat-topping in the positive direction. The asymmetrical modulation waveform (phased in the proper direction) does generate slightly more average sideband power compared to a symmetrical one, but that additional SB power is too insignificant to contribute much to the apparent loudness of the modulated signal.

I would like to get my hands on one of those Kahn gadgets that uses a passive R-L-C circuit in the audio line to shift the phases in a manner that renders the voice waveform completely symmetrical, supposedly without introducing audible distortion. Theoretically, that would allow full 100% modulation in the negative direction without positive peak clipping, even using a transmitter just barely capable of achieving 100% positive.

Well, Don, the next Kahn Symmetra-Peak I come across will go your way. :D

They weren't very successful in the BC biz.  In theory it was supposed to do just what you said.  The last one I saw working in a station was about 30 years ago.

I've passed on several over the years lurking in storage rooms.

In practice they had problems, like audible phase flipping artifacts. Not a good listen when blowing music through them.

One topology popular was to have two audio chains, one for the jock, the other for everything else.  They summed together at the TX input.

We don't usually play music (damn), so one of Leonard's boxes might just be the ticket for PTT voice.

73DG

You can build one with nothing more than a few op-amps, resistors and capacitors. It's just an allpass filter.

W3AM has a schematic on line.

http://www.w3am.com/8poleapf.html

http://www.ebay.com/itm/Audio-All-Pass-Filter-Phase-Rotator-Module-Based-OPAMP-NJM072D-/130604742281?pt=LH_DefaultDomain_0&hash=item1e68a63689

these are some nice people. they also build a little bbe sonic maximizer.

i sent them an email a while back asking if they woudl build a few and it seems like they have sold a few of them.

all you have to do is build a power supply for it.

K5IIA

  Hey Thanks for the link . Those guys on Ebay have some very cool stuff for DYI's.

I bought one of the Phase Rotator's today as I was in the market for something just like that. I will let you know how it works . I am trying to reverse my phase here as I am getting about 10% higher neg peaking . If I can reverse that, SWEET!

I also bought a Behringer  tube pre-amp with phase rotation as well . Should be interesting to see which works better.

I've been slowly delving into the audiofile syndrome and came up with this interesting article which delves into the history of audio processing and where it stands now . Found it researching Kahn who Don mentioned in a previous post in this thread.

All said and done trying to achieve perfect audio is a slippery slope to nowhere that could be a lot of fun.

http://www.broadcasttechnical.com/articles/fourtyfive.pdf

Mike Dorrough KO6NM is one of the best in the field.

Go forth and Google.

73DG

The symmetra-peak patents and info is online. It ought not be hard to make one, a very simple passive circuit for those who don't want op-amps in the path.

Please do keep a lookout for one for me.  Thanks.

Anyone looking for the Behringer 5 band compressor should keep in mind a few things:

It is a compressor only; not a peak limiter. 
It is extremely difficult to use (or it was for me at least, YMMV)
From what I have been told, if the lithium battery in one goes kaput, you loose everything, including the program or programs in the chip sets that make the whole thing function. 
The Kahn Symmetra-Peak was not popular because all it did was eliminate the asymmetry in the program material and broadcasters wanted to exploit the asymmetry by keeping the high peaks on the upside.   The Kahn basically reduced "loudness" instead of improving it.   With a big carrier on a vintage rig it might work well.   I am not a fan of super high positive peaks--vintage envelope detectors don't like them--also the occasional ham with super modulation is a PITA in a group because when his turn comes my speaker cones blow if I fail to get to the gain pot in time.   

Some AGC could be applied to the listening/audio side of things. Even a simple photocell across the amp's input and a #47 or lower voltage lamp across the speaker will do pretty well for the clip lead crowd.

no you just loose the settings that you had. the rest is safe.

I see...

The problem is that the envelope cannot go "below zero". Obviously if we want to stay near 100%, it may easily happen that the envelope goes "below zero"
Now, what can we do to avoid these events?
I think that audio processing certainly works but it must be carefully tuned after the full RF tuning of the transmitter. It may not be practical for ham radio operations.
Negative peak limiters may save the life of modulation transformer, but still produce distortion and splatter.
I think that the best approach to the problem is to dected when the modulated RF output of the class C power device goes below say 5% of unmodulated supply voltage. This "detected" signal could control the gain of an audio compressor located after the mike gain control. This is a feedback ALC that will work fine independently of the loading conditions of the power RF tube. Alternatively a special audio compressor controlled by the RF signal produced by the AM transmitter could be used.
Now the question is: are there examples around?

Giorgio

This may be the solution:

1)Simple
2)Full tube circuit.

First) change the first audio tube (mike preamp) with a variable mu pentode like 6BA6.

Second) install a negative voltage loading circuit at the output of the modulation transformer. Diodes towards the transformer and resistor (=RF tube plate input resistance) to ground. A negative voltage will develop on the resistor when overmodulating. In fact the modulation transformer will see the anode resistance of the RF tube (say 8kohm) if the voltage is positive respect to ground, alternatively if the voltage is negative respect to ground it will see the loading resistor (say 8kohm). With the loading circuit the modulation transformer will always see 8kohm (in this example).

Third) RC filter and "voltage divide" with a pot this negative voltage and give bias to the 6BA6. High pass filter (300Hz) at the output of the 6BA6. This is the ALC that I propose to keep modulation limited to 100% anytime.

Alternatively) because the power available on the resistor of the negative voltage loading circuit is high, a simple RC and LED coupled to a photoresistor shunting the audio signal could easily do the ALC task.  

What do you think?

Giorgio

DO NOT USE THE *First* CIRCUIT linked from the first post!  The 2nd one is fine.  The first one works by dissipating more power in the modulator tubes, changes the load on the modulator, and will shorten the life or damage any GOOD modulator.

Why?  Perfect modulators should function as ideal voltage sources - meaning, the modulator should be able to supply whatever output voltage is required, regardless of the load.  Albeit, there are no perfect modulators, but you can get close by using negative feedback and a good design.

So, the very last thing you want to do is to put something across such a modulator to attempt to reduce its voltage output, since the modulator will then work very hard to maintain whatever output voltage it is trying to deliver, and a whole lot of heat will be produced, and stress placed on the modulator components.

Use the 3 diode circuit shown below.  This one keeps the load on the modulator constant, and simply "switches" between the modulator output and the "keepalive" voltage, depending on the modulation level.  I have used this circuit with 5kW broadcast transmitters.

I have a variant which does not require a separate keep alive supply, although having the adjustment available is nice.

(http://www.classeradio.com/3-diode.jpg)

I'm not sure about the BehRINGer 2496 but the DSP-9024 features 6 bands.
The bands can be frequency taylored by the user. It also has a limiter which works well.

Being able to set band edges can make these inexpensive boxes more useful than broadcast surplus units which are pre-set.

Yes, they are not easy to set up. IMO the work is well worth the effort.

The 9024 can be had on ebay and elsewhere for under 100 bux used. Replace the supply electrolytics. They are a high failure item. Replace the backup battery as the unit will dump your settings. Look for a UPS to run the thing on because fast cycling of the mains power will make the 9024 dump the settings also.

My DSP-9024 dumped all the settings 3 weeks ago due to the power company doing some scheduled maintenance. I forgot to put it back on the UPS which was disconnected during the recent extended power outage from wind storm. So much for K.I.S.S.

Here it is in use. Not a direct connection IIRC a camcorder mic. Thanks to G3YPZ for putting the audio up:

http://www.traditional-jazz.com/mainpages/w2vw1.mp3

That is a very sane starting point.

It works:


I just installed a negative voltage loading circuit at the output of the modulation transformer. Diodes towards the transformer and resistor (=RF tube plate input resistance) to ground. A negative voltage will develop on the resistor when overmodulating. In fact the modulation transformer will see the anode resistance of the RF tube (say 8kohm) if the voltage is positive respect to ground, alternatively if the voltage is negative respect to ground it will see the loading resistor (say 8kohm). With the loading circuit the modulation transformer will always see 8kohm (in this example).

The negative voltage is filtered by a low pass RC (22 kohm and 47 uF ---- in place of the 47 uF capacitor it is possible to use a series of a 200 ohm trimmer and 330uF capacitor that is a pole-zero compensation) and sent to a high efficiency red LED coupled to a small photoresistor shunting the audio signal. The voltage divider is a 47kohm resistor in series and the photoresistor to ground. This is located after the mike gain pot. The 200 ohm trimmer must be tuned for best transient response of the control loop.


Giorgio

Great! Its always good to read about the results and success. The mike AGC is very interesting, good job on that!

Giorgio,

   This is good stuff! You will find that depending on your R-C time constant, that the first syllable or two will squeak through before the gain reduction takes place. It's hard to get around this without a lot more circuitry.

   I do something similar with my Retro-75 AM QRP rig where I threshold the positive modulation peaks at about 150%. Here I RC filter the pulses that try to cross the threshold to reduce the gain at a lower level. My problem was the initial R-C lag, and when trying to minimize it, I could see the ripple across the C varied at the audio rate when the frequency was < 500 hz. This caused a noticeable audio distortion. So my circuit got a little more complicated.

   You might add another high efficiency LED in series with the one there already and use it to monitor the circuit behavior. Anotherwords, put it on the front panel.

  With your circuit, also try swapping the audio phase polarity. For many of us guys, our voices are highly asymmetric, so polarity chosen makes a big difference.

Jim
WD5JKO
   

The IEX limiter (let's say) is very similar to the ALC of SSB tube transmitters.
For SSB linear amplification by class AB1 output stages, the control grid current is monitored. If grid current is zero, everything is fine. If there is current, it is converted to voltage, amplified, filtered and used to control the gain of an IF tx stage by a variable transconductance tube.
My approach is conceptually the same, but used for plate and G2 modulated AM transmitters. If there is no voltage on the loading resistor everything is fine. If there is voltage, it is filtered and used to control the gain of a modulator gain stage by the LED - photoresistor method. Simple, inexpensive, and sounds good for voice.
Case closed. ... for today....

As with any wideband feedback control system, there will be problems with rise time and transients. There is no way around this. Much of the splatter your hear from SSB rigs is due to the ALC. There is a reason why broadcasters went to multichannel processing over three decades ago.

While Steve is correct an old trick is to apply control voltage to a push-pull stage. This will help with transients.

How about delaying the audio using a bucket brigade chip, then processing the audio before it gets to the modulator.

I am experimenting and it sounds good.
I used a 5mm diameter photoresistor in front of a 5mm LED. Both are inside a black plastic tube.
I think of this limiter as a protection accessory. Protection for the modulation transformer, for the audio amplifier and for the RF spectrum.
It is hard-wired in my TX and cannot be disabled.
Audio processing will be added in the future, without removing this circuit.
The measured LDR resistance vs LED current is attached. I am not certain of the first point, maybe there was some leakage of ambient light.

Giorgio,

   I remember an old guitar amplifier that used a #49 lamp coupled to an LDR with a cardboard sleeve. It was used in a tremelo circuit that worked very well. There is a device already made that is more linear for low level AC signals, and this is an opto-fet in a 6 pin dip package:

http://www.fairchildsemi.com/ds/H1/H11F1M.pdf

These are getting scarce these days, but if you can get some, they are very interesting to experiment with.

Jim
WD5JKO

This is really interesting. The LDR is slow, it has 30ms response time.
Maybe I can put a pole-zero compensation to compensate for the slow response of the LDR: previous post updated.
Thank you.

LDRs are really slow.  One type of "gain cell" you could use is a 4-quadrant multiplier.  I have some here, and they work very well as a gain cell - and they are quite fast.

THAT corporation (that is really the name of the company) also sells a whole line of ICs, including some very linear gain cells.

We like Tubes in this context

LED: hmm.....a strange lamp, better not to investigate...

chips: oh no !!

LDR are ok because they are older than me.

Anyway the audio limiter works in the band DC - 10 Hz so 30ms is ok (faster will make no difference), then the pole-zero compensation will remove this time constant from the transfer function and the dominant pole will be the RC (22k and 330uF) = 7.26 seconds. Then this time must be divided by the open loop gain of the control loop and should fall in the 100ms region....
Because of the LED (that emits as a function of the current) R (22k) will affect the dominant pole RC and the open loop gain. C (330uF) will only affect the dominant pole RC.

Yesterday I played with the circuit and tuned the 200 ohm trimmer. I started with zero resistance. With audio bursts over 100% I got about 100ms of overmodulation, then 100% (330uF capacitor with "equilibrium" voltage on it). I increased the resistance until the response of the control system was smooth, never overmodulating.
Resistance was about 30 ohm. When this happens the time constant of the LDR is compensated, the control loop is now first order (like an OPAMP with dominant pole compensation (741), and has no overshoot.
Further increase of the resistance of the trimmer makes the control loop more sensitive to audio frequencies.

Giorgio

Here we have the suggested schematic:

http://www.ing.unitn.it/~fontana/AM%20ALC.pdf

It should work in any transformer plate modulated transmitter. Put enough diodes according to maximum anode voltage.

Please report success or failure ....

Giorgio

Hi

A low level negative peak limiter circuit from an old hand book,maybe looking at the basic circuit we can modify our audio circuit .

Gito.N

After extended experimentation and after reading this article:

http://www.w8ji.com/amplitude_modulation.htm

The circuit has been improved, while keeping its simplicity.

http://www.ing.unitn.it/~fontana/AM%20ALC.pdf

As you can see with the two additional resistors and the 250V supply, the diode network will conduce "before" the anode voltage of the RF tube reaches zero volt. You can decide at which voltage. In the schematic with 250V, 3.3k and 14k we have a threshold voltage of +47.6 - about 1.2V = +46.4V. Approximate solution because of RL. When the anode voltage will go below  +46.4V the LED (it takes about 1.2V) will start to emit light and the low level limiter will be activated.
Due to the finite gain of the control loop the voltage might drop below 46.4V, but not too much. Trim here to have 99.9% with your TX !

With this low level (high level activated) limiter/compressor most modulation enhancers can be used in combination. The gain will be automatically adjusted to avoid overmodulation, as defined for this class of modulators and measured exactly where it manifests itself.

Audio quality is top.
Modulation % is kept always slightly below 100% under any tuning condition.
Threshold voltage of loading resistor and audio limiting circuit can be adapted to any RF tetrode.

No transistor or IC......could be made in 1960 with vacuum or selenium diodes and a filament lamp.

You are a lucky guy DAVE.......
Dave Said::::
The 9024 can be had on ebay and elsewhere for under 100 bux used. Replace the supply electrolytics. They are a high failure item. Replace the backup battery as the unit will dump your settings. Look for a UPS to run the thing on because fast cycling of the mains power will make the 9024 dump the settings also.

I SEE:
I see DSP 9024's going for $325 on the BAY....nice box

Not to disrupt this fB thread but:

http://www.ebay.com/csc/i.html?_nkw=9024%20behringer&_clu=2&_dlg=1&LH_PrefLoc=0&_fsct=&LH_Complete=1

Also not to disrupt Dave's FB used Behringer price list.. ;)
 
As mentioned, asymmetrical peak limiting is a different animal from any of these low priced boxes. Tight multiband compression control they do well, and when set up right can make you sound louder.. Seems like Behringer still might have power supply issues in even their newest stuff..

Interesting video on NEW DEQ2496's "do it all processing"
 (2 out of 3 isn't THAT bad)


http://www.youtube.com/watch?v=H-l_P8hROLY

OK OK you gots me!!!

I'm in the process of trying to eventually become Behringer-free.   Their gear is either too complex (or else the manuals are too vague) or if nothing else the power supplies are apt to blow out all of a sudden.  I use the ultracurve-pro box but once I no longer need a noise gate (it has a great one in it I have to admit) I'll just use an old analog graphic eq.:  No memory, no batteries, no delicate microchips to get zapped, no settings that can suddenly get lost.   I had one of those 9024 compressors and gave it to Derb.   It was too much of a hassle to figure out and I thought he might have better luck with it on Ray.   :-[   Anything with "menus" and nothing but push buttons...forget it.  

"I'll just use an old analog graphic eq.:  No memory, no batteries, no delicate microchips to get zapped, no settings that can suddenly get lost."

Darn tootin', Analog knobs are real knobs and stay where they are put. dang it.

After 3 years. I am Behringer free here.   :-* Some of the B gear is ok. But most of it, failed or was complex.

Now you can get a real voice channel at a good price. I first saw this unit in the BHW catalog.  Now that I have one, I wont go back. The entire focus of this unit is for exactly what we are doing, Speaking into the mic. The EQ is perfect for voice and there was no reason to keep the 32 band I had so I ditched it.  The USB connectivity is a very nice feature also. It plugs right into your computer.  I would rather have one complete unit rather then a bunch of crap daisy chained together.

http://youtu.be/XOdi50HRqAk

Sorry for the OT

One possible solution, it seems to me, might  be to take the incoming audio, run it through a "precision rectifier" (a diode wrapped around an opamp), one in the + and the other in the - , then run each half through one of those nice IC chimps that do compression/limiting, then recombine at the output end...

The idea of the precision rectifier is to avoid the "dead" zone around zero crossing...

with a set up like that, and knowing which phase modulated upward in ur shack you could limit the negative going to 99% and make the upward direction whatever ur rig wants or can stand...

I have no idea if this has been done or not...

                       _-_-bear

My $0.02 from a long time (and still at it) AM broadcaster.   "I know its been said, many times, many ways...."   Negative AND positive peak control should be done before the audio is ever supplied to the transmitter.   The modulator should be able to reproduce and impress on the carrier the EXACT audio waveform fed to the input of the transmitter.  Attempting to modify the audio waveform AFTER the input stage of the transmitter is a compromise at best, and can be a disaster at worst.

Make the transmitter as linear as possible (not to be confused with a linear amplifier) from audio input to modulated output.  Treat the audio BEFORE not during the transmitter's modulation process.

73
Ted  W8IXY

AGREED and passed......As Steve (HX) said .............about $150 and you're there.
The negative loading possibly strains the modulator in those desk top TX's

fred

Hi

What about this high level negative peak limiter ,that used a diode and a low pass audio filter,design with an input impedance = impedance of the transmitter.

Gito.N

A sure way to destroy your modulation transformer.

Like the old spatter filters but faster!

Those wanting to avoid solid state could try mercury vapor tubes. There was an article using 866's it posted here in the past, and someone had found an adapter and shown pictures of it. But I agree that peak limiting should be done before the modulator.

It looks like negative peak limiters/transformer loaders/keep alives should be intended as a safety valve, not a controller.

A problem I had with keep-alive experiment was that when the modulation ws forced negative enough to activate it, it created some splatter as the modulated B+ transitioned from the downward-slope of the modulating wave to hit the hard floor of the keep-alive voltage. The only way I can describe it is that when the switching point was crossed, that little 'corner' in the waveform was full of high frequencies. Maybe I did it wrong but I couldn't avoid it.

Hi


Yes,when I read this article ,i have the same opinion,but when i reread it  
,with no load the modulation transformer has a high peak voltage developed,/transient voltage

but does not the audio choke inductance  ,"makes" a load /blocks  the transient voltage ,since the choke is connected between the Mod transformer and the RF transmitter tube.


Gito.n

If a filter is used, yes the transients will be prevented from reaching the RF amplifier.

In my experiments I did not use a low pass filter because of opinions here that the transient voltages can possibly damage the modulation transformer.

Yes ,You are right,but is it possible that ,this audio low pass filter (limiting the audio to 6 khz maybe)also absorb this transient voltage .
limiting the transient voltage develop across the modulation transformer,since the L blocks the high frequency(high frequency load)  and the C bypass the high frequency to ground.

Gito

Patrick,

   You might remember when you and I were doing modeling of the three diode circuit, and variations thereof. I was using a LPF for the reasons you and Gito have mentioned. When using SS diodes, the reverse recovery time is a consideration as well. When I actually ran a multi-diode circuit years ago, I did have a LPF, and splatter was not an issue even at 200% positive peaks. I did use HV fast diodes surplus from HV switching power supplies.

  More recently, I developed a Ultra-Mod circuit for the SS Retro-75 XCVR. Here I was able to get by with only one diode, and still had a keep alive supply which acted somewhat as a LPF as well. This arrangement with a larger Mod transformer, allowed clean peaks to 150% while limiting the negative peaks to just under 100%. I attach that circuit here. I wonder if this idea would be workable when scaled to high impedance tube circuitry.

  Also, the idea to process the audio prior to going in the rig might work when the modulator is transparent input to output. Broadcast transmitters might do this, as well as some of the Class E setups. On the other hand, take the average Ham transmitter like a Viking II, DX-100, etc. where we have multiple transformers in the audio chain, along with frequency limiting (hi and lo) R-C circuits. Here we simply cannot gain much from upstream processing, especially if we wish to enhance the positive peaks by embracing asymmetrical modulation. Therefore we must 'process' the audio at a point where there will be little to no phase shift, or level shift following the processing point in the audio chain. With amateur transmitters like the ones I mentioned, the only place to minimize the phase shift, and level shift problems is AFTER the modulation transformer.

Jim
WD5JKO

Hi

I am thinking about the transient voltage that is developed across the mod.Transformer,Yes there must be a transient voltage if the peak modulation goes below zero.

But there's a difference with loaded and unloaded Mod.Transformer.
When a transient voltage appear (positive pulse) it  is absorb/loaded by the RF tube itself ,like a " B+" supply as long there's a grid drive in this RF tube
So the transient voltage may not be too high.
It;s just my thinking.

the high level clipper is used in the military transmitter.
here's a PDF file ,about this High level Clipper.

Gito

Measure the impulse response of that filter circuit. I probably creates some nasty transients vice reducing them.