Sound

Digital Delay For Application in Surround Sound

5.0 Discussion

5.1 DESIGN

5.1.1 Frequency Response

The processor was designed to have a full audio bandwidth, i.e. the lower -3 dB point at 20 Hz (or below), and an upper -3 dB point at 20kHz. FIG 4.0 shows that at 20kHz the output was -4 dB down. This is acceptable, as the slight deviation can be attributed to component tolerances. The lower -3 dB point is not shown in FIG 4.0, but was measured using an oscilloscope and signal generator and was found to be 10 Hz. The limiting factor for the lower frequency cut-off was the value of the DC blocks used in the analogue circuit.

5.1.2 Signal to Noise Ratio

The main problem was initially that of the poor signal to noise ratio, typically 59 dB.
The majority of the noise lay in the very highest part of the audio spectrum, as could be seen by using a real time analyser, but rolled off into the mid-upper part of the audio spectrum - this is where it was most audibly noticeable.

At first there was no definite conclusion as to what was causing the noise. The display which is clocked internally was eliminated as the source (by simply removing the display I.C.).
From further investigation it was concluded that the noise must be being generated from the digital side, i.e. within the digital domain.

It was discovered however that by running the MASTER CLOCK at a higher frequency (twice that of 384kHz), then the signal to noise level was improved.
Although this helped eliminate some of the noise (by shifting the spectral content above the audio range), it would probably not be advantageous to do this permanently, as the ADC is specified to run within a specific sampling frequency range, and doubling the master clock frequency would cause it to run beyond that.

It was eventually discovered that the noise was being caused by the fact that no external off-sets were configured to the ADC or the DAC. They had been omitted in favour of leaving the I.C.'s at their factory preset values.
In the case of the ADC there were two off-sets which could be fitted. One was to trim the input to exactly 0V when no input signal was applied, and the other to trim the gain to +10V.
Only the 0V off-set was added for the ADC, and this helped eliminate the noise, reducing it by 18.8 dB.

For the DAC one off-set could be applied to help improve on the bipolar zero error - the deviation form 0V out (factory trimmed to typically ±10 mV), and the differential linearity error (DLE) - the deviation from an ideal 1 LSB change from one adjacent output state to the next. DLE may result in audible crossover distortion for low level output signals (factory trimmed to typically ±0.001% of FSR).
This off-set was added, but did not help improve on the S/N further.

5.1.3 Phase Characteristic and Delay Time

The overall phase characteristic of the processor was poor, particularly at the higher frequencies. This may not unduly affect the quality of the sound, but transient responses may be affected. This can be improved upon by using OVERSAMPLING techniques (see FURTHER WORK section).
The delay times tested gave accurate and precise results.

IMPROVEMENTS
In general, the building of the processor in future would benefit from a number of changes. Although these factors were known about at the time of design of the prototype, they were not be implemented because of time and financial constraints.

They are:

1) Instead of using veroboard, printed circuit boards (PCBs) could be used. This would have a number of advantages;
(a) The number of tracks which have to cross directly over the paths of others could be minimised.
(b) The effective area taken up by circuit boards could be reduced.
(c) Tracks could be made to turn through angles greater than 90° (better electrical transmission).
(d) Track widths could be varied, so that ground rails could be made as large as possible, and ground planes in general introduced throughout the PCB's.

2) Separate analogue and digital power supplies could be used, reducing the possibility of noise travelling from the digital circuits to the analogue circuits.

If the processor was to be built again, then the conversion board need not have the INVERTERS prior to the DAC included. Instead a single inverter on the MSB (prior to the DAC) would suffice. The resulting effect of this change is that the output signal from the digital circuitry would be 180° out of phase with respect to the input. This would not create any problems, as the surround sound will benefit from having the surround speakers wired in anti-phase with respect to each other.

No tests were done on electro-magnetic interference which has gained prominence particularly due to recent legislation. If such a device was to be built on a commercial basis, then tests of this kind would be necessary.

5.2 PSYCHOACOUSTIC TESTING

Ideally more candidates for the subjective testing would have been better, and so the fourteen who were tested represent only a relatively small test group. It would also have been better to allow each subject more time to listen to each piece of music.

From the results, it would appear that the most pleasing effect is generated through the playing of electronic music. The words ticked to describe the effect in the majority of cases imply that people find the effect to be one of involvement with the music, i.e. to be absorbed in it; a more subtle effect (more spacial, engulfing). This is as opposed to an effect which has sounds flying around the room from speaker to speaker (dramatic), or one in which they feel as though are actually in a concert hall (concert hall).

Rock music may also tend to benefit from the processor, although there is a greater risk that it may "destroy" the music on some pieces.

Classical music may or may not benefit from the effect created by the processor, and on the strength of just two subjects it is very difficult to draw any definite conclusions.

Had time and circumstances allowed, it would have been interesting to conduct listening tests on surround sound encoded video films.

It may be unnecessary to have such a wide audio bandwidth for the surround sound, due to the localisation effect of high frequencies which may be displeasing to the listener. This would confirm with the research carried out by Michael Gerzon [9], instead bandlimiting the signal may be advantageous.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -