Intermodulation distortion occurs between the lowest and highest modulation frequencies and increases bandwidth of products outside the normally expected limits (guessed at from the highest modulating frequency, normally experienced as splatter. This is true even on SSB when running ESSB. For an authoritative technical discussion of this topic, visit: http://www.w8ji.com/transmitter_splatter.htm.

Another opinion advocates ESSB and has important technical information: https://www.nu9n.com/news_july_2004.html

Here is the definitive article on how to properly adjust a linear amplifer: http://www.w8ji.com/loading_amplifier.htm

This article is provided to clear some of the fog that surrounds the folklore that has been handed down over the years concerning high level splatter filters, and splatter considerations in general. Please note that all this discussion addresses transmitters of the size of the Valiant or smaller. Larger rigs employing plate voltages in excess of 1500 Volts or exceed 200 Watts involve higher impedances and voltages that require the use of spark gaps and special techniques that are beyond the scope of this discussion. The Chicago Transformer filter referenced later purported to be suitable for such applications, but I have not independently verified this, and there are contradictory discussions on the subject. References to operating practices on AM are presented for the sake of comparison, not to say something is wrong, only different from my intent.

No discussion of splatter suppression measures is complete without at least a perfunctory discussion of what constitutes an excessively wide signal, often characterized by "buckshot" on either side of the signal.

FIRST, what is an "excessively wide" signal? Probably depends on who you ask. I limit this discussion to voice or "phone" emissions. Aside from classic plate modulated rigs, there are software defined radios (which can be dialed up to nearly any bandwidth) and class E rigs (whose bandwidth is set in the filters associated with removing the switch mode transients inherent in the design).

If you ask the IARU, in its band plans for regions 1, 2, and 3 (under current revision to harmonize the various regions) a SSB signal is 2700 Hz wide and an AM signal is 6 KHz wide.

If you ask the FCC, you get no clear answer:
§97.307 Emission standards.
(a) No amateur station transmission shall occupy more bandwidth than necessary for the information rate and emission type being transmitted, in accordance with good amateur practice.
(b) Emissions resulting from modulation must be confined to the band or segment available to the control operator. Emissions outside the necessary bandwidth must not cause splatter or key click interference to operations on adjacent frequencies.
(c)All spurious emissions from a station transmitter must be reduced to the greatest extent practicable. If any spurious emission, including chassis or power line radiation, causes harmful interference to the reception of another radio station, the licensee of the interfering amateur station is required to take steps to eliminate the interference, in accordance with good engineering practice.

The implied bandwidth is based on 3 KHz maximum audio frequency in the modulating waveform (including all harmonics of the voice waveform to be transmitted). Clearly, this limit is often exceeded by many high quality amateur radio transmissions. I often observe well in excess of 5 KHz (10 KHz RF bandwidth) in signals from "broadcast like" AM operators. The FCC rules do NOT seem to prohibit this use of the ham bands. During times of low congestion on the bands, there is no harm in this practice. On 75 Meters at night, such a signal can make use of the AM window by other operators difficult at best. As noted above, your splatter should not "cause interference to operations on adjacent frequencies". Not that it will deter an inexperienced or belligerent SSB operator from snuggling up to you 2.5 KHz away and whining about your fully legal AM signal. Or trying to operate SSB on 3.880 when there are legal AM signals on 3.875 and 3.885.

Exercise caution should you wish to operate on a frequency of 7.295 MHz with a modulation audio containing in excess of 5 KHz; you risk transmitting out of band. Remember, even the best of filters have a slope that rolls off gradually, not like a theoretical ideal filter or a "brick wall" digital filter. So a 3 KHz audio bandwidth makes a lot of sense here. With an LC splatter filter having a cutoff of 3 KHz, you should be 20 dB or so down by 5 KHz. If someone on 7290 is belly aching about your splatter, you might be pushing your luck on the out of band issue at 7300. Use common sense.

Be aware that you must use a very narrow filter (such as your 400 Hz CW filter) to determine the width of a received signal. Tuning in with your trusty old SX28 is not going to give you worthwhile information. Modern software defined radios often include a spectrum analyzer and very narrow digital filters. Ask one of these operators to take a look at your signal.

The ARRL has on occasion petitioned to have its definitions of "good amateur practice" incorporated into FCC law in the form of band plans by bandwidth. This would have impact on AM signals (and in one case would have outlawed AM). Luckily, these ideas never made it into law. With changes to the IARU band plans and emission definitions, some day AM could be limited by law to 6 KHz. I hope any regulatory actions are done in a rational fashion. I plan to design to the 3 KHz or so standard in the work presented here, though the Johnson components will easily do 8 KHz audio or 16 KHz transmitted signal bandwidth.

SECOND, All this presumes that the source of the signal bandwidth is the highest frequency in the audio modulating the carrier as discussed above. This can be the cause, but also it is caused by nonlinearity in one or more of the modulating or RF systems.

Nonlinearity could happen in the case of a smaller AM transmitter like a DX60 or software defined radio attached to an over driven "linear" amplifier, which has become "nonlinear" due to incorrect adjustment.

Nonlinearity could occur in the power stage of the modulator, such as harmonic distortion or clipping caused by inadequate modulator power to achieve 100% modulation. Yes, splatter could occur at 95% modulation, when the audio power or an intermediate stage clips due to saturation. Many rigs of this era (Apache, Viking 1 and 2, Valiant) probably deliberately caused the modulator to be under powered to prevent over modulation, then "built out" the modulation transformer with capacitors to suppress the resulting splatter. It was an attempted high level clipper circuit. Often this is implemented by an impedance mismatch (Apache and Valiant have this problem) and while you may improve things, you are limited by the original design choices in the modulation transformer. I believe Peter Dahl produced a Valiant mod transformer that overcame this problem. I got considerable improvement with a total outlay of only $25. If you want to go this route, you are looking to spend serious money, and in my opinion, you should home brew the whole thing in a rack rather than trying to squeeze it in the Valiant cabinet.

Nonlinearity can occur in the driver stage (such as the single ended triode and inferior driver transformer in the stock Valiant), and then be amplified by a perfectly good audio power stage.

Nonlinearity is introduced by the clipper stage (either the stock one or as I have implemented it here). This is why Johnson used an LC filter after it to limit the distortion products above 3 KHz. Note that it does NOT eliminate in-band intermodulation distortion products. This explains why Johnson does not recommend use of the clipper in excess of 6 to 10 dB for maximum intelligibility. Let me be perfectly clear here: without the LC filter, the clipper used as a limiter to prevent modulation in excess of 100% WILL ACTUALLY CAUSE SPLATTER, even though the modulation does not begin to approach 100 percent! Properly operated, the combination clipper filter will positively prevent splatter from over modulation without introducing noticeable distortion. With 3 to 6 dB of clipping, the signal will have the punch of a complicated three diode super modulation scheme without the component stress. For good conditions, a small downward adjustment of the audio gain to zero clipping will satisfy most all of the listeners, while at the same time preventing unexpected changes in the background noise or closeness to the mike from causing over modulation. See W8JI's comments noted below on shunting one of the clipper diodes with a capacitor to get asymmetrical upward modulation.

Nonlinearity definitely occurs in the stock preamp stages. This is why I DO NOT bypass the cathode resistors. This prevents distortion inherent in the amplifying devices (12AX7) by introducing negative feedback in each of the preamp stages. Using the higher gain 12AX7 with feedback is better than using cathode bypass capacitors and a lower gain 12AU7 to compensate for excessive gain. This nonlinearity is more of the intermodulation type, which causes an unpleasant gravelly sounding signal. The preamp is unlikely to produce splatter, unless it is over driven with a "power mike" like the Turner Plus 3 or amplified D104 when they are adjusted to maximum output. You commonly hear this problem on CB.

Nonlinearity can come from a source that is the most overlooked cause in many of the modification articles. The audio system is working perfectly. However, if the RF stage (final amplifier) is incapable of following the modulating waveform. This can be evidenced by either failure to go all the way to zero percent modulation or 100 percent modulation. Nonlinearity can also occur as irregular response to the audio modulation waveform at intermediate levels, but it is not likely to result in splatter. This type of distortion shows up as curvature in the trapezoid pattern. The other distortion described shows up as a sharp change at the left zero percent cusp or the right 100 percent edge in the trapezoid pattern. On a simple RF envelope display, it shows up as flat clipped wave shape instead of a clean sine wave shape. Commonly it occurs when the RF waveform disappears at zero percent (hits the baseline). If the RF stage will not make zero percent properly, the flat topping can be seen above the baseline. As noted above, this defective wave shape can also be caused by nonlinearity in the modulator itself, especially if it is under powered, as in the case of the stock Valiant.

FINALLY, a square wave (with its resulting harmonics) is the nonlinearity that is generally believed to be the cause of the splatter. For a period of time, the handbooks, even the "west coast"handbook, included a vacuum tube diode and high voltage insulation filament transformer and often a filter of some type. W8JI debunks this notion. Simply put, the 6146 RF stage ALREADY IS A DIODE. You do not need another one. Around 1955, this erroneous series diode thinking disappeared from the handbooks. The clipped waveform causes audio frequency harmonics in the modulated RF waveform. The LC splatter filters were an attempt to eliminate those harmonics above the cutoff frequency of the filter. They worked well, and were included in broadcast transmitters as well. They were often blamed for modulation transformer failure. Surely the diode could cause this problem by operating the modulator at maximum output with no load connected (because the diode became an open circuit). Excessive circuit Q could cause the tuned circuit of the LC filter to develop damaging peak voltages. Properly designed pi section filters positively will not cause modulation transformer failure. Chicago Transformer made a splatter filter kit, the SR500 or the SR300, rated for 500 or 300 mA plate current. Use the data sheet for that device as a guide for your design. Addition of negative cycle loading as described elsewhere on this web page will prevent excessive voltages on the modulation transformer by keeping the rated load in circuit when the RF stage "diode" is "open". A diode is used in this scheme to connect a resistor across the modulation transformer when the plate voltage goes below zero on the RF stage. Please note that all the circuits described here are NOT the 3 diode super modulation circuit or a keep alive diode as described on the AM window or elsewhere.

Please reference the Chicago Transformer SR300 and SR500 data sheet: http://www.813am.qsl.br/artigos/teoria/Splatter_Choque_Chicago.pdf

For a 2K ohm modulating impedance such as the Valiant, use a paralleled 0.2uF in parallel with a 0.2 Henry choke and 0.157uF to ground on both sides, according to the data sheet, for a 3 KHz cutoff. This is an M derived filter rather than a simple pi filter. As designed, the M derived filter exhibits constant impedance characteristics, which protects the modulation transformer. The data sheet shows that 5.5 KHz is 20 dB down. Capacitors and inductor should be rated for a minimum of 3 times the RF plate B+ supply. You are highly unlikely to acquire the original Chicago Transformer components. So I plan to improvise and describe the process if I ever use it.

The site shown above also offers a low level filter by Chicago similar to the Johnson filter and the one shown in my schematics.

VERY IMPORTANT SAFETY NOTE: Many modifications increase the efficiency of the modulator stage so that it will develop in excess of 100% modulation (or worse yet, go below zero percent). IF YOU PERFORM THESE MODIFICATIONS WITHOUT INCORPORATING THE NEGATIVE CYCLE LOADING YOU RISK DAMAGING THE MODULATION TRANSFORMER! DO NOT DO ANY MODIFICATIONS WITHOUT INCLUDING THIS SIMPLE PROTECTIVE CIRCUIT!

These notes are just a collection of unfinished thoughts and research as the project unfolded. W8JI states that a diode in series with the modulated plate voltage feed to the final does not stop splatter. This diode is featured in some handbooks from the 50s. Just because its in print doesn't mean its right. By 1955, the diode disappears from the handbook splatter filter discussions. (I home brewed a phasing rig using handbook designs that did not tell you to use matched resistors in the op amp voltage amplifiers to maintain equal levels. To achieve unwanted sideband suppression, you must have 180 degree phase shift and EQUAL AMPLITUDE, something the circuit shown in the handbook did not provide for. The error was reproduced for more than a decade in subsequent handbooks.) My Valiant contained a series diode when I received it, as well as a large choke in series with the screen (which was not switched out in CW). I bet the clicks and waveform on CW were really ratty. I disposed of all this crap in short order, putting the rig back to fully stock to get it running right before modifications. (Someone had also used a 50K pot for the audio gain, explaining why the rig sounded like "space shuttle" audio. Check everything against the blueprint first before diving into modifications.) A series choke in the screen lead is normally not necessary when the screen voltage is derived from the modulated plate voltage. Normally it is used when the screen voltage is obtained from a separate low voltage supply to make it "self modulating" due to the variations in screen current with modulation of the plate current. Just because you saw it somewhere in a handbook, does not mean that it is appropriate for your radio.

Please note that a diode with a power supply in a "keep alive" configuration is different and may help, but that is more like the multiple diode super modulation circuit described below. However – and this is VERY important – if you put a simple series diode in the circuit – and it disconnects the RF amp – which is a load across the secondary of the mod transformer – very high voltage across the secondary of the mod transformer may destroy it instantly. Bottom line: just because you saw it in a handbook somewhere don't always make it true.

I used a home brew pi section splatter filter WITHOUT THE DIODE in a Viking 2 and it worked fine. It would be important to have the negative cycle loading too. There are also tuning effects from the pi network that increase voltage if you choose unwise amounts of Q. These could damage your mod transformer. You need a choke that is in the range of 0.05 to 2 Henries. I salvaged one from an ancient transistor equipment power supply and mounted it on ceramic standoffs so that it had enough insulation for the core. You can also dismantle a tube power supply filter choke and rewind it with less turns. You do need to do calculations and tests to ensure that the core (with fewer turns) will tolerate the magnetic field generated by the plate current of the modulated RF stage. Better yet, if it has an air gap, you can simply increase the air gap with insulating shims on a 1 Henry transformer until you obtain the correct inductance. You will also have to insulate the core from the windings or float the whole thing on standoffs. You also need capacitors that are rated for at least 3 or 4 times the modulated stage plate voltage and the circulating currents that flow; again, if you design it with higher than necessary Q, the extreme voltages will damage things in an instant. All very tricky stuff, and all sorts of safety issues. IF YOU USE LOW LEVEL CLIPPING, NONE OF THIS IS ABSOLUTELY NECESSARY. Never get near the floated choke with your body or anything that may arc to it.

I plan to include a scheme for testing the modulation transformer with components to "build it out" with shunt capacitors. This technique also will work fine to confirm design of a high level pi or M derived filter. Most importantly, it is all done without high voltages on the transformer which could damage it during testing. This comes from an old "west coast" handbook. All testing is done with the transformer isolated from all circuitry, using a low level audio oscillator, voltmeter, and scope. Basically, drive the modulator plate leads (using resistors to simulate the plate resistance of the modulator tubes) and watch the frequency response at the plate of the RF stage.

Splatter and key clicks are caused by sharp transitions (square waves) in the output. These occur when the RF waveform tries to go below 0% modulation. Sticking a diode in series with the plate feed to the RF tube does not make the high frequency component go away, because the RF stage already turns into a diode when the plate voltage goes below zero. A high level LC splatter filter fixes it because it rounds off the sharp edges by reducing the high frequency content. W8JI has a good explanation of this on his page. This all comes from higher level math (Google Fourier and Laplace) that shows that any complex waveform can be constructed from a fundamental sine wave as well as additional sine waves in varying amplitudes and harmonics.

W8JI comments on the multiple diode super modulation scheme and dismisses it probably due to complexity. Some people like non symmetrical modulation, but vintage receivers only tolerate a little bit. Something like this with a high level splatter filter could be very effective, but would require heavy duty components long out of production. If you were lucky enough to possess parts scrounged from tube type AM broadcast transmitters, you could choose this path. I chose not to pursue it, because I retained the low level clipper circuit and its associated low pass filter. You might be able to mount the external high level clipper on the back of the cabinet and attach it via the J8 connector. W8JI suggests that you disable one of the LOW level clipper diodes with a shunt capacitor if you want more upward modulation. I did not try this technique yet. Doing the non symmetrical modulation at lower levels certainly makes things easier, in my thinking.

The FCC DENIED this Petition for Rulemaking, RM-10740, which sought to severely limit amateur radio experimentation with AM and SSB modes and impose unrealistic standards for band width: https://docs.fcc.gov/public/attachments/DA-04-3661A1.pdf

On the other hand, the FCC did issue advisory notices about ESSB band widths, from Riley Hollingsworth himself to a number of operators: https://www.eham.net/article/5273

NU9N weighs in, in favor of ESSB:https://www.nu9n.com/apologetics_2.html

Similar problems occur on CW, called clicks. They are caused by rise and fall times that are not controlled properly in the transmitter.
https://www.w8ji.com/what_causes_clicks.htm