Noise Reduction for High Quality Reproducing Systems
HIGHLAND MANOR
164 SOUTH MAIN ST.
CARBONDALE PA, 18407-2655 USA
kendricksellen@hotmail.com
                                               Noise Reduction for High Quality Reproducing Systems

   The fundamental conception of a dynamically operated filter, one which whould hold down the higher frequencies of both the noise and the music when there were no highs in the music and pass all the highs when the high frequency content of the signal was strong enough to mask out the noise was a real contribution to the audio art. However, the technology for achieving this fundamental goal and the determination of the precise
electrical characteristic which would do the most good with the least harm turned out to be an extremely difficult design problem. 
   The noise suppressor about to be described fully for the first time in ant webpage is the result of several years of research and development. None of the retarding features of prior units are to be found in this suppressor, which is simple to construct and so smooth in operation that it has been used on live FM program material without detracting in the least from the over-all quality. The most elaborate model of the suppressor may be built from new parts for less than hundred fifty dollars.
   The operation of the suppressor is truly dynamic. The actual suppressor is divided up into two sections; a gate section which filters out all highs when the gate is closed and passes them when the gate is open. And a “swinging” section which actually swings the gate open or closed depending on the nature of the 
signal passing through the suppressor at a given instant.
   Regulation of the filter action of the gate is accomplished by a panel control which varies its characteristic from one with a very gradual roll-off, starting its drop at several thousand Hz. To a very steep roll-off, starting at about one thousand Hz. The static characteristic of this filter is, therefore, both horizontal and vertical. The dynamic characteristic, as the gate swings, is primarily vertical as contrasted with the strictly horizontal action of most previous suppressors. Indeed, the difference between horizontal and vertical suppression can only be fully appreciated by a listening, or A-B test; making a direct comparision between the two types of filter. This comparison is one of the most striking examples of the necessity for listening to audio equipment. The result is opposite to the one expected from a study of the performance curves.
   Similarly the frequency-amplitude characteristic of the swinging section is adjusted by
a panel control, R25 (Gig. 3), which regulates the “masking threshold” of the suppressor both horizontally and vertically in a carefully designed manner to provide correct behavior with various record surfaces The screwdriver adjusted control R20 limits the gain sensitivity of the swinging section so that the frequency-amplitude characteristic of  R25 operates over the desired range. Quiet records will have objectionable noise at higher frequencys than those with gritty surfaces. This is why both the filter and the swinging sensitivity characteristic must vary with frequency as they take bigger bites out of the noise. 
   Actual operation of this suppressor is extremely simple. The gate control is first advanced until all record noise is satisfactorily removed. Next, the swinging control is advanced to the point where the gate swings open or surface noise; it is then backed down to the point where the noise again disappears. (This point will vary with various records used.) Set this way the suppressor will let even the very softest high frequency passages be heard. A signal only slightly louder than the record noise that is one at the masking threshold, will devlop sufficient voltage in the swinging section to bias the gate filter to a completely inactive state. The sensitivity of the swinging section rises very rapidly at the masking threshold. For most program material the gate will be either entirely open or entirely closed at any given instant. This is due not only to the sharply rizing sensitivity of the swinging section, but also to the action of the “Automatically Variable Time Constant” incorporated in its circuit.
   Is the gate section two tubes are employed, a 12AU7 twin triode and a 6BA6-6BJ6 remote cut-off pentode. The potentiometer network at the grid of the input 12AU7 (R33, C25, R34) is for limiting the input to this tube to prevent overload. Where signal sources of less than one volt output are to be used, it may be replaced by a single fixed resistor.
The output of the voltage amplifying half of the 12AU7 is fed through a very small coupling capacitor, which acts as a high pass filter, to the grid of the 6BA6. The audio output of the 6BA6 is fed back through another filter network, which has been deliberately made variable as to frequency and amplitude, to the other section of the 12AU7, which is a cathode coupling element. The combined effect of the two high-pass filters, canceling high frequencys by selective inverse feedback, results in a maximum possible roll-off approaching 12 db. Per octave, or expressing it another way, the response is over 20db. Down at 10 kHz. This is more roll-off than is desirable; however, because the degree of roll-off is completely variable, there is no harm in having this maximum reserve available. When the gate filter, set at maximum, is used as a fixed filter swinging section “off”) the tonal quality of reproduced music is so muddy as to be unpleasant. Here it is the action of the swinging section which restores the full musical brilliance without allowing noise to intrude. 
   This swinging section is the novel part of this circuit. It features an “Automatically Variable Time Constant” (a.v.t.c.) which facilitates the rapid swinging of the gate.
   The same audio signal which feeds the voltage amplifier also enters a separate amplifier in the swinging sectionwhere it is amplified, coupled through a high pass filter system which has also been made deliberately variable as to frequency and amplitude, again amplified, and finally rectified and filtered. The result is a d.c. voltage which is proportional to the high frequency content of the record. This voltage could be made to bias the gate tube without any further modification. However, it has proven desirable to incorporate the additional three triode stages which make up the (a.v.t.c.) and d.c. threshold control portion of the swinging section. The complex behavior of this part of the circuit is required to make the action of the gate so rapid as to be unobtrusive. This may readily be demonstrated by removing the a.v.t.c. tube from its socket, articularly at a time when it is operating on such percussive modulation as pinao or vioin music or hot 
jazz. 
   The function of the a.v.t.c. circuit is to convert the conventional fixed time constant filter circuit that follows the rectifier into a circuit which has a time constant automatically varied over a range which is nearly instantaneous at one extreme and 0.1 second at the other. This permits a much shorter effective time constant than that afforded by a conventional RC time constant. The usual RC circuit, with a time constant short enough to act without inertia whould introduce serious transient distortion. 
   The first half of the a.v.t.c. tube acts as a shunt resistor across R29. The time required to charge capacitor C21 under rising signal conditions is controlled by the dynamically varied plate resistance of this tube. Conversely, the time required to discharge capacitor C21 is dynamically controlled in accordance with the decaying signal conditions and their relation to the average signal level. The a.v.t.c. action is further modified by capacitors C22, C23, C24, and resistor R31. The precise manner in which the build up and decay is shaped can only be illustrated graphically and by lengthy description.
   The threshold control circuit is designed to keep the gate from opening or closing on noise signals in the absence of a desired control signal, or when noise greatly overrides low level signals. Under no-signal condition the cathode resistor develops about 4.6 volts bias across the 12AT7 cathodes. The 12AT7 normally requires about 1 volt bias; with 4.6 volts this stage operates nearly, but not quite, at cut-off. It is effectively paralyzed at low signal levels. A signal just over the level of the noise will reduce this cut-off bias so that a positive feedback of d.c. voltage will restore normal 12AT7 bias very shortly after the threshold level of noise is exceeded. It can be seen that because of this self-resensitizing action of the swinging section amplifier stage, a very sharply defined threshold point will result due to the action of this d.c. threshold control circuit. It should be further pointed out that due to Miller Effects in the various stages of this circuit it is insensitive to extrems high frequencys and will not respond to most noise.
   The diagrams of Fig. 2 and 3 illustrate two versions of the suppressor. 
   Fig. 2 is actually patterned after an early development model which did not employ a.v.t.c. or the cathode coupled type of feedback circuit. It will not permit quite as much noise suppression as the more elaborate versions; the gate action is perceptible, but not to such an objectionable degree as to be without merit. 
   Fig. 3 illustrate the model shown in the photographs. It was built by the author and incorporates, in addition to the latest elaborate version of the basic suppressor, a power supply and preamplifier for a magnetic pick up as well as a compleat tone control preamplifier based on design data. A cathode follower in the output of this unit permits connection through a long shielded cable to a fixed gain power amplifier. It provides straight ilne frequency response over the entire range of 20-20kHz as measured from the input of the suppressor to the speaker voice coil. 
   Nevertheless, actual construction of either basic suppressor is not critical. The audio signal itself is channeled through only two tubes and if normal care is excrcised in keeping leads reasonably short and the components are not crowed to the point where they interact, no difficulty need be expected. The swinging section is even less critical.
Remember only that it has fairly high gain at the high frequencies and very little stray coupling capacity is required to cause interaction between it and the gate section. A layout similar to that shown in the photographs is easy to follow and applies equally to both models shown schematically. The simpler model merly requires less chassis space. If desired, the unit built by the auther may be duplicated exactly, or the phono-preamp and tone control stage may be left out if not required. The author’s unit is built on a 7x17x3 inch aluminum chassis which was brushed with a rotating wire brush to give it a velvet grined appearance. The controls are located on two-inch centers on a straight line running through the center of the front apron. From left to right they are volume, on-off and bass frequency, bass (amplitude). Treble, swinging control, gate control, and radio-phono-microphone selector, There are three phono positions on this second switch providing low frequency turnover points of 300, 500, and 800 kHz that recordings of the varions companies may be properly equalized. The constants of the treble control afford proper high frequency equalization to accommodate any pre-emphasis recorded on the disk 
   The tube sockets are centered on 2 1/8” from each other in the longest direction the 6AU6A/6CB6A 1 3/8” from the right edge og the chassis. The front row centers on 1 5/8” from the front edge, the middle row centers on 3 ¾” from the front edge and the back row of tubes centers on 5 ¾” from the same  edge. With these figures as an accurate guide to location the other parts may be readily positioned. An a.c. receptacle on the rear apron is wired through the power switch and permits preamp. Power amplifier, and tuner to be turned on or off together.
   The wiring is straightforward; filament wiring is twisted and runs from socket to socket. One leg of the filament wiring is grounded. The ground side of the filament string is directly wired from tube to tube. Using the chassis as a conductor for part of the filament circuit is a dangerous procedure in any audio work. The connection between all wiring grounds and chassis is accomplished at only one point, the phono input connector. All other wiring ground points are grouped by stages and electrically isolated from the chassis. The electrolytic capacitor is mounted on a bakelite mounting plate in a manner which insulates its can from the chassis. To the uninitiated this may seem like a lot of trouble to keep two parts, which are schematically tied together, from connecting. The reader may be assured that employment of this technique goes further towards eliminating hum and oscillations than anything else he  might do in the actual construction of an audio component. Remember schematics are only a guide, they can never show the whole picture. Ground loops are almost always the real cause of trouble usually attributed to the use of a.c. heater supplies. The author has never had to resort to d.c. on high gain heater circuits to eliminate hum. Ground loops also cause those unexplainable” squeals many p.a. men have encountered in complex installations. 
   Shielded wire has not been used in the construction of this unit, except in the a.c. switch leads. The only point with enough gain ahead of it to cause trouble is the input lead from the mike jack. Running this lead along the corner of the chassis will keep it away from other components and out of trouble. By not using shielded wire in any audio circuits, the high frequency response is kept flat far beyond the 20 kHz goal. It is a simpler task to keep the input wiring of the early stages away from filament circuits than to use shieided wire. Since the power supply is located at the far end of the chassis away from low level circuits it cannot cause trouble. 
   Construction of either of the models of noise suppressors discussed here is a simple matter. The use of one of these suppressors to remove high frequency noise from any signal source, including the hiss sometimes encountered in FM reception of weak stations, will bring gratifying results. The manner in which these units behave, without detracting from program quality in any way, is astounding. Certainly it must be heard to be appreciated.