negative peak limiting

[ab3al]

Ok I have been slowly upgrading the system here,  flex 1k is much better than the icom  crapola. henry 2k classic x.. legal limit plus never hurts. just hung a full wave loop in place of a short 80 diaperpole. behringer vx2000 and 9024 poor mans optimod.  Mics range from nuemans re20-27 marshal 2006 v67 mxl900 and some other crap. 

What about negative peak limiting.  all of the circuits i have found deal with this in the modulators of pm rigs. I know It could be done in the software for flex, but im not a programmer.  Is there a way to accomplish the same thing in the audio chain befor the rig

73
cua maaaan

[Bacon, WA3WDR]

I got pretty good results with a very very fast peak limiter that would follow the waveform.  It acted as a clipper and it worked outboard.

You need some sort of dynamic DC correction if there will be AC coupling after the negative clipper, because the clipping will produce a DC offset.  But if you use a servo-clipper (a peak limiter with no filtering of the gain-control signal) and it has matching AC coupling to its level detector, this is automatic.  You can configure the servo-clipper to clip only one side of the waveform, or to clip asymmetrically, etc.

The clipped signal can then be low-pass filtered with something that does not produce ringing spikes that exceed the peak level, and you have a nice splatter filter.

   Bacon, WA3WDR

[Todd, KA1KAQ]

NU9N has some good info here:

http://www.icycolors.com/nu9n/am.html


I recall looking into this or something similar years ago. Maybe it was Ultra modulation or ? I also recall a problem of distoriton being introduced, modulator being left unloaded, stuff along those lines that seemed to negate any benefit. But it was long ago, and I knew more then.

Looking forward to hearing what those in the know have to say.

[Steve - WB3HUZ]

There are lots of ways to do it. But, it's probably more bother than it is worth. Just make sure your have things phased or polarized such that the highest peaks are in the positive modulation direction and then use the processor(s) you have to get the average level up to where you need it without over modulating in the negative direction. Another dB or so provided by agressive negative peak modulation isn't even going to be needed or noticed under most band conditions.

Less is more.

[ab3al]

this is what i love about this forum// quick answers and no flaming.. thanks guys.. I actually found a circuit on qix e page. Its in his pdm generator right after the mic pre.  If it helps anyones pondering and thought.  the entire chain is balanced in and out  The mike input to the sound card is balanced as well.. I understand negative limiting may or may not help but what the heck batman its ham radio.  Ill try it at least once to learn sumfin

[flintstone mop]

I agree with the HUZMAN!! The days of 150% pos peaks was a waste of time. Seems like anything beyond 120% pos. overloads the detector in the receiver. A fully modulated AM carrier with nice balanced, not over bassed, audio will be music to our ears. Sounds like you have a recipe for a nice station that will quiet the ether.
Fred

[k4kyv]

I agree with the HUZMAN!! The days of 150% pos peaks was a waste of time. Seems like anything beyond 120% pos. overloads the detector in the receiver. A fully modulated AM carrier with nice balanced, not over bassed, audio will be music to our ears. Sounds like you have a recipe for a nice station that will quiet the ether.

Some time back in the 60's I tried ultramodulation with the three 866-A's and the big resistor.  On the scope the waveform looked just as it was supposed to look.  But the signal reports I got over the air were that there was no significant loudness of the audio, but it sounded more distorted.  After some experimenting with it, I finally abandoned the idea.  Besides, audio power is too precious a commodity, too expensive to generate, to waste a significant part of it heating up a resistor.

The next method I tried actually worked fairly well; high level symmetrical clipping in the modulator stage with a splatter filter following the mod transformer.  The clipping action was accomplished by using a variac in the modulator power supply, driving the hell out of the modulator stage, to the point that it flat-topped on both positive and negative peaks, and adjusting the variac until the negative peaks just barely missed 100%.  There was some tilting of the clipped waveform, but careful design of the splatter filter, using an air-core splatter choke, kept the tilting to a minimum.  It was very effective in slopbucket battles and poor band conditions, and sounded pretty good if the clipping level was kept at 6 dB or so.  Under battlefield condx, the audio was readable, with approximately slopbucket quality, up to 20 dB of clipping.

But there was one weak  link in the chain - the modulation transformer.  It did not like looking into the highly reactive load presented by the low-pass splatter filter at the knee of the curve right at frequencies near cutoff.  I blew a total of six kilowatt size modulation transformers, including a very nice VM-5 and another very nice one that I acquired new in the box, rated at 2 kw of audio, while experimenting with high level clipping.  I finally took the splatter filter out and cut back on the audio drive, and never blew another mod xfmr.  This is the same weak link that causes KW-1's to blow the mod xfmr, and why the splatter filter in those rigs should be disconnected.

Now, I use the natural asymmetry of the human voice to achieve extended positeve peaks.  Just a good quality mic, good audio iron from mic input to mod xfmr, and flat audio response at least one octave above and  below the targeted frequency response of the audio chain, and phase the audio so that the peak with the greater amplitude is in the positive direction.  Mine averages about 135% in the positive direction.

The advantage of extended positive peaks is to allow the negative peaks to  go closer to 100% before the positive peaks flat-top, thus increasing the effective percentage of modulation.  You don't gain any extra audio (and thus sideband) power, since the system is a.c. coupled, and by the laws of physics, the area under the positive half of the curve must be equal to the area under the negative half.  In other words, the extended positive peaks are greater in amplitude, but of proportionally shorter duration, while the lesser amplitude negative peaks are of longer duration - so that the energy under the baseline is exactly the same as the energy above the baseline.

A more efficient method, giving greater sideband density, would be to use a type of  speech processor that some broadcast stations use, which shifts the phase in such a way that the voice waveform becomes symmetrical.  Supposedly the change in voice quality is imperceptible. I'm not sure what this unit is called, but I'm sure someone here will be familiar with it; they have been in use in AM broadcast stations for a long time.

A major drawback to asymmetrical modulation is the peak power capability required of the modulator.  Audio power increases as the square of the amplitude of the signal and thus percentage of modulation.  50% modulation requires only 25% as much power as 100% modulation, and 140% modulation requires approximately twice as much power.  200% modulation requires 4 times as much power as 100%.  So in order to significantly increase the percentage beyond 100%, a large modulator is required.  Otherwise it will self-clip the positive peaks and produce the same type of splatter and distortion as overmodulation in the negative direction.

The easiest way to achieve this positive peak capability is to increase the voltage on the modulator tubes (while maintaining appropriate grid bias), and have plenty of undistorted reserve power in the driver stage.  This becomes a rather extreme requirement very rapidly as you try to extend positive peaks, regardless of method used, whether negative peak clipping using diodes, or simply using natural voice asymmetry.

The once popular type of negative peak clipping that uses a single diode in series with the bottom side of the secondary of the modulation transformer, plus a splatter filter, actually causes more splatter than the same modulator would cause, running at the same level, with an identical circuit except that the diode is removed from the circuit.  The explanation for this is rather involved, beyond the scope of this posting, but the technical reasons are clear and demonstratable, involving the switching action of the diode combined the decay rate of the charge on the capacitors in the splatter filter.

[W1DAN]

Hi:

The 9024 has some poor dynamics on the final limiter. When I had one I did not like the final limiter. A simple diode (or series group of them) across your audio going into the SDR might do it...simple, but you would have to play with levels to set it well. I'd set it to shave off any overshoots, but not be active very long. This would obviously create harmonics when clipping occurrs.

If you wanna protect bandwidth a linear phase low pass filter after the clipper will help prevent splattering the nearby frequencies. A good starting choice is 3-6kc. Maybe the SDR does this already?

WA1QIX's circuit is good to use.

Also this can be done in software with simple code that asks if a sample is greater than or equal to XX amplitude and then outputs a fixed result.

Let us know what you do.

73,
Dan
w1DAN

[K9ACT]

Don's comments saved me an exercise in futility.  I have been pouring over old articles and contemplating high level limiting/cliiping but wondering why no one ever mentions it any more.

I went over NU9N's audio page and found it very informative.

Before I go plop done money for an Inovonics 222 does anyone have any other inputs on it? 

Don has heard my current audio and perhaps he could comment on whether this would be something useful for my station.

js

[K3ZS]

I am using the high level negative limiter as shown elsewhere in this forum.   However the main reason is to protect the hard-to-get modulation transformer.   I set the modulation so that peak limiting normally doesn't occur as shown on the scope.

[Steve - WB3HUZ]

Before I go plop done money for an Inovonics 222 does anyone have any other inputs on it?

Nice unit but probably over kill. Plenty of guys on the air that sound fantastic and they aren't using 222's.

[w5omr]

Before I go plop done money for an Inovonics 222 does anyone have any other inputs on it?

Nice unit but probably over kill. Plenty of guys on the air that sound fantastic and they aren't using 222's.

Besides, doesn't negative peak limiting/clipping really just add distortion, if the audio is up too far, anyway?

[Ian VK3KRI]

Besides, doesn't negative peak limiting/clipping really just add distortion, if the audio is up too far, anyway?

The way I see it, the negative peak limiter should be preventing inadvertant splatter due to over modulation. Of course you need filtering AFTER the neg peak limiter to stop the harmonics from the clipping.

Just remember if you have 6dB of peak clipping, then your audio is 6dB louder than if you wound back the audio to just before 100% neg modulation. Its all a trade off.

Also it pays to match the cutoff freqs of various bits of the audio chain. If you put 4 khz filtered audio into a clipper and then follow that with say an 8khz LPF , the distorion products are MUCH more noticable than if if you put 8Khz audio into the clipper and then followed with another 8khz filter.

I was amazed when I first heard this, but its true, with 4Khz audio you can clearly hear distortion due to the clipping, as you'd expect, but when you feed in 8khz audio at the same level, the distorion products are masked , at least some of the time, by higher frequency components of the original. It really sounds like the distortion products dissappear.
Psychoacoustics is an amazing thing...
                                                                          Ian VK3KRI


 

 

[Bacon, WA3WDR]

Clipping can be done at low-level, but you either need DC coupling from there on to the modulator, or you need some dynamic adjustment of the clipping level to compensate for AC coupling.

I accidentally created exactly this kind of compensation in my speech processor.  Basically it has an AC-coupled side chain that resembles a typical analog modulator, and the time constants happened to match pretty well, result was that low-level clippping was still viable after AC coupling and a modulation transformer coupled modulator (no driver transformer, though).

I was yacking with HLR Tim, and I was changing the value of the capacitor in the gain control loop, and just then he said something and turned it over to me.  So I gulped and answered, not having any idea what would happen if I transmitted with no capacitor to ground in the gain control loop of the peak-limiting compressor - and I reached over and turned down the TX gain, and on the scope I saw perfect flat-topped and bottomed clipped audio waveform modulating the carrier wave.  Huh???  That couldn't be... could it???  It was a total coincidence that it happened, and that I noticed it.  I had to think about it for a while before I understood what had happened.

The basic design is a peak-limiting compressor, but it permits an adjustable degree of super-fast peak detection that actually causes the waveform to be peak-limited, along with the envelope of the waveform as in a normal peak limiter.  This winds up peak-limiting and slightly clipping the audio waveform, and the tilt that you would normally expect from the clipping is reversed.  This compensates for the tilt in the AC coupled and transformer coupled modulator, and the result is that the low-level clipping remains effective.  This is a crude design that I did back in the late 80s, and I still use it, except I zapped the dual-gate mosfet and replaced it with an MPF102 and adjusted the bias on the fet a little bit.  I'll have to update this design and publish it; it works pretty well.  It has some capacitive filtering and some partly filtered control as well, so it acts like a peak limiter and an average compressor and an adjustable clipper.  Also it is asymmetrical, and asymmetry is adjustable, so it can clip at 100% negative and 200% positive if desired.

The trick is the ac coupled and transformer coupled side-chain that drives the level detector.  The ac-coupling and transformer coupling there mimics the ac coupling and transformer coupling in the modulator, and the resulting processing comes into focus at the modulated B+ going to the modulated stage.

I added a second-order Sallen-Key low-pass filter after the processor, which is not shown.  I fiddled with the values to adjust the Q to control the ringing to avoid overmodulation.  I'll have to publish this, too.

http://www.amfone.net/AMPX/issue_number_79.htm