A Binaural Processor for Any Rig
|By Joe Street, VE3VXO
Special to The ARS Sojourner
|A short time ago, I became aware of the ongoing research of Rick Campbell, KK7B, into the area of binaural presentation of audio signals to a radio listener. For those who aren't aware, of what this means, briefly it involves sending slightly different audio signals to each ear in the same sense that a stereo viewer or a pair of binoculars sends a slightly different view to each eye resulting in a 3D view with depth perception. In this case a 3 dimensional soundscape is created in which sounds of different pitch are perceived in different positions in space within the listening environment of a pair of stereo headphones. As recently as March 1999, QST ran an article by Rick detailing the design of a new receiver descending from the design of the famous R2, and intended specifically to take advantage of this effect.
The article makes reference to experimental literature which indicates that binaural presentation of tones appears to offer a 3 dB advantage with respect to uncorrelated noise. In other words the noise is spread out over the entire soundscape whereas the signal of interest occupies a finite position within that space. The ability of the mind to concentrate in this way is a difficult thing to quantify, and I haven't checked the references to see just how the figure of 3dB that was quoted was arrived at, but it sounded so interesting that I just had to give it a try. For a more in depth understanding of the effect, I highly recommend reading the original article. One thing that caught my attention in that article which was inherent in the design of the receiver was the following. The receiver front end is a classic twin mixer quadrature phasing design. Normally this is followed by two precisely balanced IF (or audio in this case) phasing networks , the output of which are summed and magically, the undesired sideband is cancelled through the joys of trigonometric identities.
In this KK7B's design no summation of the I and Q channels is done. They are simply fed straight into two separate identical audio filters and amplifiers and then on to the headphones. This makes use of the inherent phase difference in the two channels to create a spreading effect in the listener's environment. One drawback of this approach is that both the desired and the image sidebands are present in this system. This means that there are twice as many signals, and twice as much noise, as is the case in any DC receiver, so the wonderful 3dB gain in apparent signal to noise ratio is nullified, but there is still the spreading effect between signals of a different pitch, which has its own advantage as well.
Immediately I went straight to work learning all about binaural presentation of audio signals. I don't have a phasing rig yet to use as a test bed, so I started off by building a circuit to simulate an I-Q audio chain. I used an all pass filter to obtain a unity gain but frequency dependent phase shift of the audio signal I wished to use. I split the signal from an audio signal generator into two channels and fed one into the phasing network, and the other not. A couple of LM386's served as a cheap and dirty headphone driver. I also tried a phase lead filter on one side and a phase lag on the other. Then I did a lot of frequency sweeping and a lot of listening.
While I did definitely perceive a change in spacial position with a change in frequency due solely to a change in phase of the two signals which were going to either side of my headphones, I was not satisfied. The spreading had a limited scope, and an uncomfortable effect which cannot be described except to say that it was somewhat stress inducing. At some point I realized it was due to another parameter that the brain pays attention to, which is the relative amplitude of the signals presented to each ear. If you hear a sound coming from directly in front of you, the soundwaves reaching each of your ears will be equal in amplitude and phase. When the sound moves off to the side, say the left, then the two ears hear the sound differently. Since it is somewhat closer, the left ear will hear the sound slightly louder and the soundwaves will arrive at the left ear slightly before the right. This means that the soundwaves reaching the right ear will be slightly attenuated and lagging in phase. This is the set of conditions which the brain perceives as a sound coming from the left.
There are other effects such as the millisecond delays between the arrival of the incident soundwaves and the arrival of reflections from nearby objects and walls which are substantially attenuated but give clues to the brain as to the precise position of the sound in 3 dimensions. Trying to simulate all of these effects would be exceedingly difficult, but I was sufficiently fired up at this point to proceed further.
I then tried a different approach. I built two first order filters both with a cutoff of 750 Hz. One was a highpass and the other was a lowpass. I tried one of these in either side of the audio chain. This had a profound effect. Low frequencies would be loud on the left but as the frequency approached 750 Hz and both filters were passing the signal equally, the loudness was dead center, and then higher frequencies shifted to the right.
However the sound was still not true to life and this was due to improper phase shifts. A low pass filter retards the signal as it attenuates it, which is what we want, but the highpass filter advances the signal in the stopband and then it approaches zero phase shift in the passband. After a little pondering and playing I realized that what is really needed is an amplitude, and phase relationship, which is controlled so that at the design frequency, the signals to both ears are matched in frequency and phase. As the frequency varies, one side or the other must be attenuated but also must be retarded in phase. This is exactly what happens in nature and this is what I wanted to reproduce.
After much fiddling and exhaustive checking with a dual trace oscilloscope it was finally accomplished with the use of two filters and two phase compensation networks. I am very pleased with the results. The combination of attenuation and phase lag is interpreted in the brain as a more realistic change in spacial position. I connected it to my superhet which has a 500 Hz IF and gave it a whirl. Amazing! I tune the desired signal till it is center stage and anyone else who is not exactly zero beat is off to one side or the other. It was tough to get the right combination of circuitry to end up with the desired transfer function but it really makes a huge difference. In the end the circuit is not complicated or critical in any way and it is well within the capability of even a fairly inexperienced homebrewer to get it going.
At a first glance this would appear to be an imbalance in the two sides of the circuit. The reason for this becomes apparent when you consider the attenuation of signals an equal offset from the center frequency on each side. Suppose your receiver has a passband of 500 Hz with a center of 800 Hz. The cutoff point of the passband is 250 Hz on either side, giving a lower point of 550 Hz and an upper cutoff of 1050 Hz. A first order filter attenuates 6 dB per octave. An octave below 800 Hz is 400 Hz, and an octave above is 1600 Hz. So the 250 Hz on the low side represents 5/8 of an octave resulting in 3.75 dB attenuation of the signals through the first order highpass filter.
If a first order filter were used on the lowpass side it would give 250/1600 = 5/16 of an octave or 1.875 dB attenuation of the higher frequency signals to the earphone on that side. In order to obtain equal attenuation of signals an equal offset on either side of the center, the lowpass filter must be twice the order of the highpass.
Following the filters I used a pair of audio drivers which are built around an op amp with class AB push-pull complementary pair transistors for power gain as part of a feedback loop along with the RC components necessary to get some extra filtering at the same time. The bias was chosen so that the supply voltage could fall to 8 volts before the onset of significant crossover distortion. By using TL064 type micropower op amps throughout, the overall current draw is only about 3 mA, yet there is plenty of audio even for my big audiophile headphones which work great with this circuit by the way.
Receivers which have a more open frequency response such as the R2 with its 1 KHz lowpass together with good headphones make the frequency spreading effect of this circuit more dramatic. Narrower IF responses result in signals simply disappearing as they get farther from the center rather than continuing to shift to the side. Ihave made some recordings of signals on the 20 m band using my QRP rig which has a 500 Hz IF, which are available on my website at http://microwatter.homestead.com/firstpage.html
In addition, I tried out a new wrinkle in prototyping techniques as evidenced in the photo. I call it notsougly construction, which will be the subject of another short article to follow. The basic idea was to use a board layout program to do the artwork for a board which looks like a single sided surface mount type of construction. What makes this different though is that I used regular sized components. Of course any combination of regular and surface mount devices could be used. The advantage of this technique over a normal board is that there is no drilling holes which saves time, and all circuit traces are on the same side as the components making circuit tracing very easy. This also means that components can be added or changed very easily and quickly. Even a mistake in the layout (which I confess crept in, careful as I was) is easily corrected with traditional ugly techniques. The components were all mounted within 2 hours of etching the board and aside from the error which was quickly found and fixed, worked very well first time.
Joe Street, VE3VXO, is one of the more creative members of the low power community. His articles are always stimulating.