Dynamic Noise Filter Project (Part 3)
The third instalment of our series describing the design and construction of a high quality modular effects rack system.
To continue our series of high quality audio processor projects for the Modular Effects Rack described in the December 1984 issue, Paul Williams reveals an effective, single-ended, easy to use, stereo noise reducer which will be found invaluable in any home studio set-up as a supplement to normal tape noise reduction systems or to enhance the signal-to-noise ratio of outboard effects, including those which exhibit digital quantisation noise.
The noise reduction system, be it Dolby, dbx or whatever, that you probably use on your multitrack recordings involves a double-ended, encode-decode process to reduce tape noise; and they work very well, rarely producing any side effects on the programme. However, even if your multitrack is endowed with such a system, you may not find it a complete solution, especially when you have done several track bounces. Moreover, these systems invariably have fixed characteristics which, to minimise side effects, are optimised for a particular noise level. Obviously by using a different type of tape, or by bouncing tracks, the noise levels may not be optimum. The other problem you may be faced with is what to do with noisy recordings you made before you obtained your noise-reduced equipment. Your recorder noise reduction system does nothing either for the noise generated further down the line, such as in mixing, or from effects like flangers and echo units etc. And what about the quantisation noise generated by some digital instruments and effects, all faithfully reproduced by your recorder?
Clearly what is needed is an effective single-ended noise reducer which can be used at final mixdown stage, and is capable of being adjusted to cater for a wide range of signal and noise levels. HSR's Dynamic Noise Filter 2 (DNF2) was designed to satisfy just this requirement. It will significantly reduce any level of tape or equipment noise - yes, even digital quantisation noise, and yet when correctly adjusted will have no noticeable effect on the music programme, which still retains all of its original brightness and clarity. Impossible? Read on...
The DNF is effectively a self-adjusting low-pass filter whose cut-off frequency rises when high frequency signals are present above a predetermined threshold level. The operation of the DNF relies on the well known masking property of the human ear, where high frequency sounds in the programme will disguise noise of a lower level but similar frequency band, such as tape or equipment noise. When these high frequencies are not present to mask the noise, the cut-off frequency of the DNF's filter lowers to attenuate the noise, but since no high frequencies are present in the programme at this time, the programme is not affected. The filter cut-off frequency will only rise when there is sufficient high frequency signal programme content to mask the noise.
DNF2 contains two such filters to provide two channels for stereo operation. The filters share a common sidechain circuit, along with a key input, allowing the experimenting home recordist to produce some unusual effects. The filters are of the second order 12dB/octave type, which can much more effectively separate the noise from the signal than would the shallower cut-off slope of a more simple 6dB/octave first order design. The variable threshold of DNF2 allows a wide range of signal levels and signal-to-noise ratios to be handled with optimum performance.
Figure 1 shows the dynamic response curves for DNF2, from which it can be seen that input levels which exceed the threshold level by 20dB or more will be completely unaffected, but lower levels become progressively more attenuated at higher frequencies. Note, that the signal levels shown are not absolute, but referenced to the threshold level.
An LED indication of the filtering operation is provided so that a check can be kept on how the programme is being affected.
Turning to DNF2 certainly shouldn't be considered as a last resort if all fails; it is subtle and forgiving enough to allow it to be permanently patched into the mixer output to be used on every mixdown.
As shown in Figure 2, each audio channel comprises two cascaded current controlled low-pass filters formed by IC2 and IC3, both dual Operational Transconductance Amplifiers (OTAs). Although the input offset voltages of the OTAs are nulled out using VR2 and VR3, the simple cascade connection of the filters, each using two inverted stages, further helps to minimise any tendency towards switching 'thumps' being transmitted through to the output.
For control purposes, the two input signals are mixed together by R3 and R4, which present the composite signal via the high-pass filter, C3 and R5, to IC1a. The high-pass filter allows only high frequency signals to open the filters, while the low-pass characteristic caused by C3 prevents any ultrasonic signals from unintentionally opening the filters. The gain of IC1a is adjustable by means of VR1 to allow a wide range of threshold levels to be set. C5 and C6 further promote the high-pass characteristic.
IC1b and IC1c form a precision full wave rectifier which eliminates the forward voltage of the diodes, so no temperature dependent terms are involved in determining the threshold level. IC1c also acts as a peak detector, which charges C7 to the peak level detected. R12 prevents any very short transient peaks from opening the filters. In the absence of a signal, C7 will discharge via R11 and R12, resulting in a decay time constant of about 100 milliseconds.
IC1d and TR1 convert the negative voltage on C7 into a filter control current which is injected into the OTAs via R18 and R19. When there is little HF (high frequency) input, the bias current produced by R14 prevents the cut-off frequency on the filters from falling below 500Hz. The output of IC1d, being close to 0 volts at this time, causes the LED D5 to illuminate, indicating that the programme is being filtered. Once the bandwidth has opened above 12kHz or so, the output of IC1d is negative enough to cause D5 to extinguish.
Building this project using the high quality kit should present no problems, especially since, by extensive use of PC mounting connectors, switches and potentiometer, there is no interwiring to do.
The first step in construction is to insert, solder and crop the resistors, say ten or so at a time to prevent crowding, into the appropriate positions as printed on the PCB, and according to the parts list. Solder the four links in place using resistor lead off-cuts, then the diodes D1-D4 and D6, and the transistor TR1. The IC sockets come next, making sure that they are pressed down onto the PCB whilst soldering. The ICs themselves are left out until later though.
Now insert and solder all the capacitors except C7, taking care with the polarity of the electrolytic types. Mount and solder the presets VR2 and VR3 so that they lean back slightly to ease adjustment later. The bus connector and the five jack sockets can then be soldered whilst holding them firmly down onto the PCB. A piece of foam rubber laid on the bench always comes in handy for holding connectors and the like in place on up-turned PCBs during soldering.
Holding the pot shaft in a vice, use a hacksaw to trim it within 8mm of the bush.
Failing that, a pair of bolt cutters or similar will do the job just as well, and much more quickly. Fit a PC bracket to the pot and locate it into the appropriate PCB position, but don't solder at this point.
After determining the correct orientation of the LED, bend its leads down at right angles 4mm from its body and locate into the PCB without soldering. Screw one nut onto each toggle switch then locate them into their PCB positions. Place shakeproof washers on the pot and switches, then offer the front panel up, feeding the pot and switch bushes and LED dome into the appropriate panel apertures. The panel is then fixed in place by means of the pot nut, which should be fully tightened. The toggle switch front nuts are, however, only finger tightened, after which the backing nuts are fully tightened onto the back of the panel. The pot, bracket, switches and LED can now all be soldered, after making sure that they are all positioned correctly, and that the panel is at right angles to the PCB.
Now before going any further, check the assembly very carefully, especially on the track side where dry joints and solder splashes are quite possible, no matter how experienced or how careful you are. When you are completely satisfied with the assembly, locate the ICs into their respective sockets. Note though, that IC1 is fitted the opposite way around from the others. Fit the knob, complete with its cap so that the marker line covers the entire scale evenly, then fit the switch lever covers.
If you are worried about the capacitor C7 that you still haven't fitted yet, don't be: all will become clear soon.
Slide the completed DNF2 module into the Sub-Rack, with the module to its right removed. The presets VR2 and VR3 now have to be adjusted to null out any control noise but, since the cancellation in the cascaded filter stages is so good anyway, the noise is very difficult to detect with C7 in circuit, hence its omission!
Switch the key toggle on and apply the rack power. Inject a steady tone of medium frequency to the key input from a synth or whatever is to hand, preferably with a 'raunchy' edge (a square wave, for instance). Adjust the threshold control and input level so that the LED glows at about half brightness. Monitor the left output with plenty of gain, and adjust VR2, the lower preset, to minimise the signal breaking through to the output. VR3 is then adjusted to null out the right channel breakthrough.
Now remove the module and solder capacitor C7 in place, since the preset should never again need readjustment. The module can now be fixed into its rack position using black M2.5 screws. Although not essential, you can screw the jack socket nuts onto the backplane if the module is not to be removed frequently.
DNF2 is best placed right at the end of the playback chain on final mixdown, after all submixing and processing. This way, not only is the tape noise reduced, but so is the mixing and processing equipment noise with it. It also means that no matter how many tracks you mix down, a single DNF2 will suffice, so long as the mixdown is only for two channel stereo. DNFs can, of course, be used on individual tracks and instruments so that the threshold can be optimised for each.
Your noisy old recordings can be cleaned up too, either by playing them back via DNF2 to your amplifier, or for a more permanent solution, re-record via DNF2 onto another machine (which is endowed with noise reduction, of course!). DNF2 has been optimised for the home studio operating level of -10dBV, but it has sufficient headroom to allow operation at higher professional levels.
The rack's linking system will route signals via the left channel, which is switched in/out using the 'in' toggle. The right channel which shares a common control sidechain with the left channel, is accessible via the rear jack sockets only, and is permanently 'in'. The left channel is also, of course, available at the rear jack sockets.
Correct adjustment of the threshold control will depend very much on the level and nature of the programme and on the noise level, but in any case it is vital to the performance of the unit. Adjustment is best done during periods of complete silence, or during quiet sound passages. Start with the control fully clockwise, and gradually adjust it anticlockwise until the noise just disappears. The LED will be glowing by this time, indicating that the filters are closing. See if the control can be advanced clockwise without re-introducing the noise.
Again, depending on the nature of the noise and programme, it may be necessary to trim the control a little further anticlockwise if, during moderately quiet passages, the noise starts to 'peep' through again. When the LED extinguishes completely, then the filters are open fully and the programme is being passed unaffected. You will see the LED flashing on and off continuously as the music plays, indicating that DNF2 is reducing the noise, but its effect on the music will usually be imperceptible.
The reader may be forgiven for thinking that the usefulness of a key input on a noise filter must be very limited. Well, as with most creative applications, limits are set only by the imagination. Here are a few experiments to try using the key input, with the key toggle switched on. Try passing vocals or strings through DNF2 while keying from a guitar. The 'highs' can be 'played' on the guitar strings. Or, how about adding texture to the highs by running your fingernail along a wound string on the keying guitar.
The key input will also accept a high impedance microphone input directly, so try singing, tapping or rubbing the highs! Tapping is particularly effective for bringing out percussiveness, whether it be on drums, or any other instrument in the mix. A crescendo that didn't quite work out can really be brought to life by keying the whole mix from a mic into which the crescendo is sung or blown. And, since the in/out switching is virtually silent, you can even 'punch' the left channel into service while the mix is in progress. Anything goes, as they say!
Next month: Parametric EQ.
Feature by Paul Williams