Digital, as we have seen in previous issues of Son Pro and other specialist publi-cations, brings significant improvement in the ease of use, and to certain parameters of musical reproduction. However just as nothing is perfect in the real world, these new techniques generate flaws which have to be suppressed by making them in-audible. Certainly testing may show little measured differences, but when all is said and done, it is our experienced ears which show us the difference between two com-ponents that measure the same.

The musician searches for new sounds and will be careful to ensure that the 

( recording) medium used will be adequate to reproduce all the nuances and subtleties that they have created, just as the sound engineer wishes, apart from a few exceptions, to achieve the best sound possible.

If the (measured) limits of analogue had only been negligible values, digital would perhaps not have been born. 

Digital technology brings improvements in areas such as distortion, signal to noise ratio, phase integrity between two or more channels and in the high storage density of the medium.

But on the other hand the sampling frequencies are not high enough to give a rise time as quick as analogue, to guarantee the harmonic content is true to the original.

As we have already explained in a previous issue, it would be necessary to have a sampling frequency of 400kHz to recreate a "clean" 20kHz sine wave. It is difficult, if not impossible, to always work to the theoretical limits of digital without encountering all sorts of problems with distortion, "breath"????? and modulation of the "breath" by the useful signal. What’s more, the theoretical dynamic range of 6db per bit cannot be used as it should. On a 16 bit scale to resolve a signal down to -72 db ultimately means defining it in only the 4 least significant bits, to capture the small signals which give richness to the sounds and to recordings, putting a value on the subtitles of mixing.

The solution is to increase the precision by using 24 or even 32 bit quantitization which would give a much greater accuracy; where as the standard 16 bits only allows 65, 536 possible dynamic steps, 24 bits is clearly more precise because it exceeds 16 million steps. So it would no longer be too difficult to get an accurate and full digital range, and above all the small signals would be correctly recorded and reproduced, without the effects of modulation of sound by the "useful" signal when the latter doesn’t disappear completely in "the breath". The noise shaping, called dither, allows an increase in the usable signal to noise ratio but its effect remains limited. What is more, experience shows that while convertor should, in principle, be adjusted to have a linear function , without "switching noise" as the signal passes through zero. While the majority of specialist integrated circuit convertors have pins which can be used to adjust "digital offset" by using various resistances, the appliances equipped to do so are few and far between. Some will get as far as silk screen printing the printed circuit board without fitting on the components required for adjustment.

It is not sufficient today to specify an oversampling filter to recalculate the digital from 16 to 20 bit for example, it is also necessary to manually refine the work of convertors by the above mentioned adjustments. But let’s not draw such a gloomy picture and let’s set aside for the time being this potential hypothetical sampling to 400kHz on 24 or 32 bit and content ourselves with the standards in use. These are not too bad but are not too good either, since some of the faults that are detected can be greatly reduced, or even overcome entirely.  

The fact that a digital signal is sampled at 44,100Hz for example quite obviously means that the signal is not continuous. On the contrary, a digital signal, called also "modulation of code impulse", is a discontinuous structure. The "jitter margin" is a temporal variation, at high frequency, of the digital signal by a clock signal. A logical 1 is replaced by a transition/change from high to low or from low to high, which is something that happens when there is no jitter, half way between the bit and the source. An absence of transition is a logical zero. Furthermore there is change of logical level at the end of each bit, that is to say the digital words are not replaced by the levels, but by the intervals that separate the different sides (of the waveform) (climbing or descending). The digital signalhas been conceived in such a way as to carry the clock signal ( the sampling frequency), and the data. Each word is preceded by a formed preamble by fixed logical levels, not modulated, which allows convertors to be synchronised and to detect the beginning of a block (which represents 384 words).

The causes of jitter are numerous, for example, the references of unstable clocks would be shown on an oscilloscope as a kind of fluttering of the synchronisation signal vis a vis a marker or window. If a side of a signal does not appear in its 

widow at the desired moment, it is either early or late. This phenomenon is a little like a frequency modulation altogether comparable with "crying" and sparkling that you find on an analogue machine except that jitter is in the megachertz region. Sometimes the problem of drift is such that the noise superimposes on the clock frequency. With the data having a discontinuous structure, there is nothing more difficult than the optimum synchronisation of the transported data signal in order to cnvert it in the best way possible. Unfortunately these discontinuous signals, at high frequency (several megahertz) and at low voltage are corrupted by the technology which can never reach the perfection that one would wish to be able to obtain. In accordance with Murphy’s Law, in its chapter about electronics, it states that if faults are noticed on a circuit there is no doubt that they will accumulate in the most harmful way, and not cancel each other out. What is more, the most commonly used digital standards have serious limits. Only the SDIF2 professional is the most stable and reliable because it is necessary to have three cables, one for the left data, another for the right data and finally a third which carries the rest such as the clock which is going to allow the source and the receiver to be synchronised to a great level of precision. Unfortunately the other digital formats in the audio world are multiplex, that is to say that you pass the two channels and the clock on the same cable which increases the risk of error.  

Several types of instability can interfere in a digital link such as jitter from the digital data, jitter from the clock frequency, caused by the drift of the components, imprecision all relative to the base of time to quartz which synchronises roughly all the data and also the link cables: a capacitive cable and/or inducive cable will smooth out the sides of the signals which should be, in order to be recognised and converted correctly, as straight and thin as possible so as to appear in the correct order in the "window of measurements" cadenced at very high speed. The electrical magnitudes of the digital signal aconveyed/translated here into....... current and voltage to see a luminous intensity and it is there where the quality of the vector of transmission enters the game. A capacitive cable displaying a poor band width will smooth out the sides of the signal, by transforming the line into the narrowest possible sinusoidal form, Similarly, for a lowest grade optical link terminated with too narrow a pass band opto coupller. In a fibre, as in a cable, it is possible to have signal reflections, and the leading edge of the waveform can spread out and the signal receiver (the famous DIR or reception interface of digital signals) will see and pass two logical levels on both sides of the window which will pose inevitable problems of synchronisation.

The "prelude/preamble", that we have seen above not being biphased, can mislead the phase lock loop (PLL) of the DIR which will desperately try to follow these disruptive signals for its operation. In fact the locking circuit of DIR will synchronise its PLL on the signal only if this is of a sufficient quality. If not it will try vainly to lock into the "prelude" which is only a useful signal in which appears information such as the sampling frequency, the presence of emphasis etc. At worst the errors are such that audio outputs waiting for a signal that it can decode. At best it produces errors which, if they are either all or partly compensated, thanks to the parity bits will always bring a host of errors and therefore approximations in the convertor, and have a negative influence on the sound output.  

Hidden behind these three letters is the designation Digital Interface Processor. This piece of equipment from Monarchy Audio serves to virtually eliminate as far as possible this instability that we call jitter. It comes in a thick anodised aluminium casing which gives this piece of equipment a very respectable form as well as an excellent robustness. The DIP has a filter section incorporated into the IEC, high quality components and notably Texas Instruments digital processor of which the reference has been removed and which constitutes the heart of the system. This piece of equipment does not have a mains switch, so it can be left switched on all the time which guarantees good stability, an essential quality for this type of product. It is also equipped with an enormous transformer of galvanic isolation of digital input, triple plating, totaly isolating the output socket. Two standard SPDIF inputs are connected directly to the front by a switch either by electrical phono ocnnector or optically on Toslink. There are two outputs, a phono for the SPDIF (amplified by the DIP from 0.5V to 3V peak to peak at 75 ohms) and an XLR for the AES/EBU (balanced link, 5Vpeak to peak at 110 ohms). This piece of equipment could, besides greatly reducing jitter, be used as a type of connection convertor, taking the single ended phono coming out of the source and outputting an AES/EBU to a convertor, benefiting from the higher definition provided by this type of input. The DIP is designed not only for the present but also for the future (we have already mentioned the military spec. construction) since it accepts, without any degradation, digital signals from 16 to 24 bit. The way in which the DIP works is to extract the synchronising clock frequency form the digital signal, reclock the signal and output it, stabilised and amplified. The DIP, besides providing a common interface between the source and the convertor, also succeeds in strongly attenuating the high frequency instability called jitter: its work begins at 25KHz which is just above the audio band and attains the impressive value of 50dB attenuation at 1MHz, an absolutely remarkable result!

The DIP locks onto the digital signals (that is to say the biphased signals) which will be converted into sound. A small interesting detail is that the LED only lights up if a modulated digital signal is present at the input. This is very different to the majority of digital systems which permanently display the sampling frequency, having detected it even if the signal is not strong enough to decode. With jitter increasing proportionately to the length of the digital cable, Monarchy recommends placing the DIP as near to the convertor as possible. The audiophiles on the other side of the Atlantic go even as far as to link two DIPs to establish a special link between the source and the convertor in order to obtain a better rejection of exterior disturbances and greatly alleviate the undesirable effects of the cables used, by connecting the first DIP close to the source and the second in immediate prosimity to the convertor. These digital connections, being serial links, the jitter attenuations is increased. The exceptional price performance ratio of the DIP easily allows this sort of operation. Let us profit from this and bow to the wisdom of audiophiles who understand and are passionately interested in the technicalities (the true ones). Their views are not as far from our own professional views as one would perhaps believe.......

The DIP does not have optical output, because if you class the connections in increasing order of quality, this is what you obtain: SPDIF on Toslink, SPDIF on phono (or BNC), optical AT & T on ST, AES on XLR and finally SDIF2 on three BNCs.

There is another model, the superdrive DIP, newly arrived in France which has an SPDIF output at high current available on BNC in place of the XLR socket of the AES, without passing through the isolating transformer to the output (see diagram). The objective is to avoid the line losses of the digital output signal if you use very long lengths of cable, qualified here as hundreds of meters! If, taking into account the excessive length of the vector and the risks of jitter, you could not do anything else, then at least there will always be a signal getting to the other end of the cable, which would probably not have been the case without the insertion of the DIP.

When you rub shoulders each day with colleagues who work at the sister publications of Son Pro - Audiophile, La Nouvelle Revue du Son and Son Mag, with whom yours truly, also being an audiophile, collaborates regularly, one becomes particularly critical about sound reproduction in general, continually improving a certain auditory acuity, which is vital even in the professional world. Those of our readers who have evolved from the world of radio know of the pschoacoustic influences in the processing of FM sound. This research is not limited solely to the study of the change in percevied sound, caused by artifacts knowingly added to the wave form because as in the case of this present test, it is not about adding a special effect but about a piece of equipment which effects a drastic correction to jitter which is, as we have seen, a drift in the tine base of data in the digital signal.

When you listen to a system comprising, for example, a transport and digital convertor, you can, with the right material immediately apreciate the sound reproduction. But as soon as you insert the Monarchy Audio DIP in the chain, the bass becomes immediately less blurred, more defined and the mid and top frequency bands have truly gained in transparency and naturalness. The differences are quite spectacular. To demonstrate the effect even better, and this is definitely the amusing part of psychoacoustics, take out the DIP and you hear the base more imprecise, the mid and top as aggressive to the extent of shrilling and this seems to sound much worse than (the system did) before putting in the DIP. It seens therefore that the ear quickly gets used to natura sound, untainted by the distortion caused by jitter. It is quite difficult to evaluate precisely the quality of the transport when auditioning the DIP in the digital chain. The tones sound more true, giving a clear impression that the convertor finds it easier to decode the data coming from the transport without the problem of error correction circuits having to correct or compensate for, the signal values which are usually lost because of jitter, sometimes in a manner too approximate for the signal to keep its integrity. The DIP preserves two important parameters which are sense of space and the realism of sound levels and timbres.  

So here is a reasonably priced piece of equipment, capable of making great improvements in systems whether they be large studios or personal installations. It has two great advantages. Looking at the price of other processors offering SPDIF/AES/EBU, the Monarchy DIP is clearly more competitive which, alone justifies its acquisition. As for the correction to jitter, the DIP is the cheapest on the market facing off competition which reaches more than four times the price and yet which may not offer the same characteristics (50db of rejection fo jitter at 1MHz).

Let s not get hung up about the DIP? Ah it was easy. So much so that this piece of equipment could be used for example as a complement to a word clock in a digital editing studio working with more than two machines improving, the link coming on the input from the recorder.