Rob's web

More on speech clipping

Home - Techniek - Electronica - Radiotechniek - Radio amateur bladen - QST - More on speech clipping

---

Practical design data and circuit information.

Speech clipping and filtering can greatly increase the effectiveness of a phone transmitter. It can also make that transmitter take up less room on the air than an unfiltered job that isn't properly operated. But it also requires great care and attention to design and operating details that, unfortunately, are only too frequently neglected in the average 'phone. This article emphasizes those points in addition to introducing some new circuits.

The author's first QST article on speech clipping(1) was intended primarily as an introduction to the basic principles of pre-modulation speech clipping and the considerations involved. It must be admitted that the low-level clipper-filter circuit that was shown as a prototype in order to facilitate explanation of the basic system is somewhat more elaborate than necessary for practical application to typical amateur transmitters. While it could be used to advantage in a transmitter boasting a modulator with virtually zero distortion and a modulated Class C stage with virtually perfect linearity, the residual splatter resulting from the harmonic and intermodulation distortion generated after a low-level clipper-filter is, in practical amateur transmitters, far from zero. Therefore there is no point in trying to achieve absolute perfection in the clipper-filter.

A much simpler low-level circuit which gives very good results was shown in November QST,(2) and one of still different persuasion and intermediate complexity is shown in Figs. 1 and 2. Fig. 1 shows the basic series clipper circuit, and Fig. 2 shows the complete clipper-filter incorporated in a speech amplifier suitable for amateur work.

Fig. 1. Full-wave series clipper which maintains constant load on an RC driving circuit, thus preventing axis shift. At low clipping levels a series clipper gives cleaner chopping than a shunt clipper, although the difference is not great at levels of 2 volts or more. The values of resistors R2, R3, R4, R5 and Re should be equal to each other and to twice the value of Rt. The clipping level is one-fourth the positive bias voltage.
The positive bias source should have low impedance. There should be no external path for d.c. across either the input or output terminals.

As was stated in the author's original article, there are many types of series limiters and many types of shunt limiters which can be made to work satisfactorily, but the many limiters in either class differ only as to detail. When one analyzes them carefully it becomes apparent that all shunt limiters are basically the same and all series limiters are basically the same, even though there are many ways of obtaining delay bias and feeding the signal in and out. The similarity between speech-clipper circuits and conventional receiver noise-limiter or "chopper" circuits also becomes apparent.

A practical speech-amplifier front end

Fig. 2 is intended as the front end of a speech amplifier for an amateur phone transmitter. It has sufficient gain for any of the common p.a.- type diaphragm crystal or high-impedance dynamic microphones for moderately close talking. It is not designed for the ham who likes to sit across the room from the microphone and whisper at it. The amplifier incorporates a moderate amount of bass suppression ahead of the clipperfilter, a highly desirable feature for communications work at any time and a virtual necessity when a low-level clipper-filter is employed in a transmitter using anything other than "broadcast-quality" transformers following the clipperfilter. The suppression is obtained by proportioning the RC values in the grid coupling circuits of the first two stages to give a cut-off frequency that is about as high as can be tolerated without the quality becoming quite thin and unnatural. The values shown are recommended as a good compromise between voice "naturalness" and "get-through" ability, but if the reader insists upon slightly more bass the values of C1 and C2 can be increased to 0.002 µF.

Fig. 2. Speech-amplifier front end for amateur work, incorporating bass suppression and speech clipping and filtering, utilizing the series dipper of Fig. 1.

Partlist fig. 2

C1,C2	0.001 µF mica.
C3	8 µF 450 w.v. electrolytic.
C4	25 µF 50 w.v. electrolytic.
C5	0.05 µF tubular paper, 1000 w.v., best quality.
C6	50 µF 50 w.v. electrolytic.
C7	16 µF 450 w.v. electrolytic.
C8,C12	0.008 µF mica.
C9,C11	0.007 µF mica.
C10	0.016 pµF mica.
C13	0.02 µF tubular paper, 1000 w.v., best quality.
R1	560 kΩ, ½ watt.
R2	27 kΩ, ½ watt.
R3	270 kΩ, 1 watt.
R4	500 kΩ potentiometer, a.f. taper.
R5	68 kΩ, 1 watt.
R6	1500 Ω, 14 watt.
R7	47 kΩ, 1 watt.
R8	110 kΩ, ½ watt (see text).
R9,R10,R11,R12,13	220 kΩ, 3 watt (see text).
R14	1 kΩ, 3 watt.
R15	3 kΩ potentiometer, wire-wound.
R16	470 kΩ, ½ watt.
L1,L2	125 mH powdered-iron core "r.f." choke.
B1	Bias cell, 1.1 or 1.25 volts.
J1	Microphone connector.

Mica condensers are specified at these points to avoid possible leakage and consequent short life of the bias cell, and to prevent the possibility of application of positive voltage on the grid of the next stage. When a high value of grid resistance is employed, many paper capacitors have or eventually develop enough leakage to put a volt or two of positive voltage on the grid of the next tube. There is no excuse for using anything but mica when 0.006 afd. or less is required, but when larger capacities are indicated the writer has found it wise to use only top-quality paper capacitors rated at 1000 working volts and to make the capacity as small as can be tolerated from the standpoint of frequency response or phase shift.

The resistors R8, R11, and R13 are very critical if symmetrical clipping is to be accomplished. R9, R10 and R12 are also fairly critical. The exact values are not so important as their uniformity and ratio. R9, R10, R11, R12 and R13 should all be equal in value and R8 should be equal to exactly half this value. The best way to get a perfect match is to persuade the clerk to let you get into the resistor bin with an ohmmeter and measure "220,000-ohm" resistors until five are found which are within 1 or 2 per cent, then tie two in parallel and look for a "110,000-ohm" resistor which is exactly equal to the parallel combination. When this procedure is used, the absolute accuracy of the ohmmeter is unimportant. Good quality resistors should be used at these points to ensure that they retain their characteristics with age.

A 6F5 is used in the first socket because the tube is relatively nonmicrophonic (which is more than can be said of some 6SJ7s), and because it provides good gain with low noise and hum. Also, the writer prefers a double-ended tube for the input stage in any amplifier, as it makes for less grief with hum and feed-back. R2 serves in conjunction with the input capacity of the tube to prevent excessive r.f. from getting on the first grid, a common trouble around high-power rigs when a high-impedance microphone with a long cable is employed.

It is recommended that the 6F5 be mounted so that the grid cap is close to the microphone connector, and that C1, R1, R2 and B1 be placed in a shield can which also shields the grid end of the tube and the back side of the microphone connector. Sometimes this saves trouble later on, and it is more easily done in the first place.

The filter uses two standard 125 mH. powdered-iron core "r.f." chokes, such as are made by Meissner or Bud. When a wire-wound potentiometer is used at R15 to set the output level it need not be touched after once being set, regardless of changes in the weather, provided that the modulated stage is run at the same input. A slight change in the characteristics of a composition- or carbon-type potentiometer will not be noticed when the potentiometer is used as a volume control, but even a slight change can cause trouble if it is used to set the modulation "ceiling," and initially is advanced as far as is possible without producing splatter.

The maximum peak voltage available across R15 is about 8 volts. The rest of the speech system should be so designed or altered that between 2 and 8 volts peak input produces approximately 95-per-cent modulation. If less than 2 volts is required, proper adjustment of R15 becomes difficult, because wire-wound potentiometers ordinarily are not available with tapered windings. If less than 2 volts is needed, the situation can be saved by using two resistors as a voltage divider at R16, proportioning them to give the desired voltage with R15 at about half scale. If between 10 and 25 volts is required at the next grid, the primary of a 1-to-3 ratio interstage transformer can be connected from the pot arm to "B" plus, eliminating C13 and R16 - but be sure the transformer is not of the bargain-counter variety.

The correct adjustment of R15 is the one that gives the highest percentage modulation that can be used without splatter even when screaming into the microphone with the gain control full on. If the following stages have very good low-end phase-shift characteristics, R15 can be advanced somewhat more than would otherwise be the case. Phase shift tends to cant the flat-topped waves after they leave the clipper-filter.

Actually, phase shift can be tolerated if it is linear with respect to frequency, but when phase shift is due to inadequate capacity in coupling condensers or inadequate transformer inductance, the resulting phase shift is not linear with respect to frequency and the waveform is distorted.

Modulation distortion

If a scope check should indicate that the clipped waves are substantially flat-topped into the modulator but no longer so when the rectified carrier envelope is viewed on the 'scope, the Class B modulation transformer is guilty. This condition is most common with a combination of cheap modulation transformer, high plate-to-plate load on the modulators, and Class C plate current flowing in the transformer secondary. If the condition is not cured when the d.c. is eliminated from the secondary by resorting to shunt feed, a bigger and better modulation transformer is in order.

With a low-level clipper-filter the modulator distortion either must be kept low at full modulation, or else the high-order components must be removed from the modulator output by filtering. The latter can be done fairly well, simply by shunting both ' primary and secondary of the Class B output transformer with as much capacity as can be employed without excessive attenuation at 3500 cycles. These capacitors act in conjunction with the leakage inductance of the transformer to constitute a pi-section filter. If this does not do the trick - and it may not if the transformer is of very good quality and has low leakage reactance - then the solution is to augment the leakage inductance with a filter choke designed for the purpose, such as a Thordarson "splatter choke."

Linearity of modulated amplifier

With any type of speech clipper, linearity in the modulated Class C stage is of vital importance. Distortion generated here can produce splatter just as well as anywhere else in the rig, and it is too late to filter it out. One cure for a stubborn Class C modulated amplifier is to modulate the r.f. driver stage along with the final stage. (This is not so wasteful of audio power as it might seem, particularly if a high-voltage low-current tube is used as an r.f. driver.) The reason is that excitation of a value that ordinarily would be used for c.w. operation will suffice, because on a modulation up-peak the excitation power will increase to four times the unmodulated carrier excitation if the r.f. driver is fully modulated along with the final r.f. stage. Also, on a down-peak the driver power decreases with the final plate voltage, and to avoid modulation distortion caused by a discontinuous characteristic caused by insufficient excitation - which would occur if fixed bias were used on the final amplifier - the final must have all-grid-leak bias and a very small grid by-pass or coupling condenser.

The plate voltage for the r.f. driver in such an arrangement is best obtained from a series dropping resistor, in the same manner as screen voltage commonly is obtained for modulated tetrodes and pentodes. The same precautions apply with regard to keeping the modulation voltages in phase. The plate by-pass condenser in the driver stage should be small, or detrimental phase shift will result at the higher voice frequencies. The same applies to the grid coupling or by-pass condenser in the final. Shunt plate feed to the driver permits use of a much smaller plate blocking condenser, and is recommended when the arrangement is otherwise acceptable.

Clipping symmetry

When a speech clipper is built into any speech system, all single-ended stages ahead of the clipper should have fairly low distortion even under conditions of maximum clipping. Bad overloading of any single-ended stage ahead of the clipper will cause an asymmetrical wave to be fed to the clipper because of shifting of the axis, and will prevent maximum realization of the clipper benefits. In other words, for best performance the clipping should be confined to the clipper.

The series clipper shown in Figs. 1 and 2 utilizes two diodes and three resistors whose only purpose is to maintain a constant load on the RC driving circuit over a complete cycle regardless of the amount of clipping. This provides a slight improvement in the performance of the clipper proper by keeping the axis where it belongs. Although the improvement is slight it is not negligible, and it certainly is worth the cost of one 6H6 and three carbon resistors.

High-level clipper

Illustrated in Fig. 3 is a practical high-level clipper-filter system for use at Class C inputs up to 2000 volts at 250 ma. The particular filter constants shown are for a load impedance of 8000 to 10,000 ohms, and the filter is designed to use some of the thousands of surplus 0.002-afd. 5000volt-test mica condensers that are reposing on bargain counters around the country.

Fig. 3. High-level half-wave clipper-filter system for use with 8000- to 10,000-ohm loads and plate voltages up to 2000. If the same power supply is used for both the modulator and Class C amplifier, the latter should be decoupled by means of a suitable choke and capacitor.
C1,C2,C4,C5 0.002 µF 5000 volt mica.
C3 0.004 µF 5000 volt mica.
L1,L2 Approximately 0.4 H, capable of carrying 250 mA without overheating; high-voltage insulation. See text.
T1 Class B modulation transformer.
T2 866 filament transformer, 7500-volt insulation.
NOTE: If plate blocking condenser is 0.001 µF or larger, refer to text regarding value of C5.

Because inexpensive filament transformers with high-voltage insulation are more readily available in 2.5-volt rating, an 866 is shown as a clipper tube instead of a 5R4GY or other high-vacuum. rectifier. The use of an 866 is perfectly feasible so long as it is run at reduced rating, as is done here, and careful checks indicate that the performance is as good as with a high-vacuum rectifier.

The two air-core "splatter chokes" for the high-level filter are obtained by removing the iron cores from ordinary filter chokes of suitable current rating, and then removing turns or adding a few laminations to trim up the inductance to the desired value. The nonlinear characteristic of conventional laminated iron-core chokes makes them inferior to air-core chokes for use in splatter filters which must carry lots of d.c. and handle high a.c. voltages within the pass-band, as is the case when the filter is placed after the modulator. The slightly-greater d.c. drop in an air-core choke, because of the greater number of turns required, is not of serious consequence.

Inexpensive uncased filter chokes having comparatively low-voltage insulation can be doctored up to make excellent air-core splatter chokes if the current rating is adequate. With the core removed it cannot "talk back," and it is an easy matter to insulate the whole coil from a metal chassis. The greater number of turns reduces the voltage between turns and between windings, and the only possible vulnerable point is the termination for the inner end of the winding. If the choke originally was designed for low-voltage use, this wire probably crosses over one end of the coil and is anchored to a lug or pigtail lead taped to the outside of the coil and not too well insulated from it. This is easily remedied by snipping the wire and fixing it up with its own tie-point or lug, well spaced from the outside end of the coil.

Some very useful information would be a list of all the popular filter chokes of commercial manufacture by type number, with exact inductance with the core removed. Unfortunately the author is not in a position to supply such information at the present time, but it is hoped that such information can be made available at a later date. The two chokes used in the filter shown in Fig. 3 were obtained by removing the cores from two old homemade filter chokes, vintage 1927, which were found in the junk box. The inductance with no d.c. superimposed was measured at 17 henries before removal of the core, and 0.32 henry with core removed. Just to get some idea of what could be expected of typical chokes of recent commercial manufacture, a choke rated at 8 henries at 150 ma. was measured with no d.c. in the winding and the inductance found to be 21 henries. With the core removed the inductance was measured at 0.43 henry. A 250-ma. factory-made choke rated at from 6 to 10 henries with rated d.c. in the winding probably would give just about the right inductance for the filter of Fig. 3, without removing any turns or replacing any laminations.

Naturally it is desirable to hit the inductance value on the button (or nearly so) with all of the core removed and no alteration of the winding. However, removing a few layers of the winding is not a very big job should this be necessary. If the inductance is too low, it can be raised by sticking a few of the straight laminations (not the "E" pieces) through the coil, separating them with tape and wedging them in tightly to prevent "talking" once the right number of laminations has been determined. This is a little more messy than removing turns, but still is not a formidable job. Even if enough iron is inserted to double the inductance, the total reluctance still is so high that the characteristics of the choke still will be substantially those of an air-core choke.

When checking the inductance it is best to make the measurement at approximately the cutoff frequency of the filter. Because of the comparatively high distributed capacity of an air-core choke of this type, and the tendency for it to increase the mutual inductive coupling between different portions of the coil at the higher frgquencies, the inductance as measured at or near the cut-off frequency of the filter will be appreciably higher than that measured at 60 cycles. If one does not have access to equipment suitable for direct measurement, the inductance can be determined with fair accuracy using an audio oscillator and diode peak voltmeter, making reference to a reactance-frequency chart to determine the reactance of a capacitor of known value which resonates the choke at a known frequency.

If the plate blocking condenser in the modulated stage is 0.001 pfd. or larger, the value of C5 should be reduced by approximately the same amount. Obviously the blocking condenser should never be much larger than 0.002 µF. The filter constants are not extremely critical, and some leeway can be tolerated. But for best performance the values should not deviate too much from those specified.

When using a high-level clipper-filter it is important that there not be appreciable "lopsided" overloading of any-stage ahead of the modulator at maximum clipping level. It is recommended that a Class Al push-pull a.f. driver stage be used ahead of the modulator, and that a 250,000-ohm resistor be placed in series with the grid of each driver tube, right at the grid. If the driver stage uses fixed bias, then the resistance value should be as high as is permitted for fixed-bias operation, usually 50,000 ohms. The stage ahead of the a.f. driver also should use 250,000-ohm series resistors at each grid, and preferably be push-pull, although this is not absolutely necessary if the stage is capable of delivering without serious distortion several times the peak output voltage required for 95-per-cent modulation.

It is true that the high-level clipper-filter system is more expensive than the low-level system, but it has the advantage of being self-adjusting, removes splatter components generated in the modulator stage, and renders harmless any phase shift in the modulation transformer. Also, if a husky modulator is employed in conjunction with plenty of r.f. excitation to the Class C stage and a comparatively low plate-to-plate load on the modulator, the signal will have noticeably more "punch" for the same resting carrier power.

Notes

1. W. W. Smith, "Premodulation speech clipping and filtering," QST, February, 1946.
2. J. W. Smith and N. H. Hale, "Let's not overmodulate - It isn't necessary!," QST, November, 1946.

W. W. Smith, W6BCX.