LINE AMPLIFIER

Controlling the amplitude of the incoming signal is crucial. No matter how great a digital recorder may be, if the amplitude level of the incoming signal is wrong, it will be recorded just as wrong. Whether you are using a physical mixer, or software based attenuation, if the level is too low, fewer bits are used which results in grainy low bit samples. If the level is to high the signal turns to digital mush. There is no 'fixing in the mix' .(garbage in ....garbage out).

So proper amplitude control and a clean mixer with great dynamic range and frequency response, (a noisy mixer may be covered by the inherent noise of analog recording, but it will stick way out with digital recorders), can make or break a recording .
 
 

DITHER GENERATOR

Calibrated random noise used to linearize low level information.

As we learned in the previous chapter, a little added noise is better than the distortion caused by quantize error. The dither noise is concealed by the audio signal. (As we will see, to hear dither we have to push the mixer output to an extreme level.)

The quantize error (distortion) is not eliminated, but turned into broadband noise ,which is more acceptable.

As we also know from the last chapter, the extra dB actually adds to the dynamic range of the system. A 16 bit CD can sound 20 bit. Dither adds a small amount of random hiss at around -96 dB. Sound below -96 dB is pushed up into audible levels . (Although, it also may mask the very sound that is pushed up).

By pushing up the dB , quantization error is minimized since the signal is rounded up to higher values. The chance of the voltage hitting the 1/2 point of the bit is also reduced.  Without dither, the LSB is toggled on and off (like being able to see the flicker of the shutter in a film). With the extra noise of dither, the sound is pushed up and the LSB is toggled by noise below the wanted sound instead.

Three types of dither are available.

1. Gaussian - 6 dB boost - easiest to implement in digital domain.
2. Rectangular - 3 dB boost -constant and precise amplitude.
3. Triangular - 4.77 dB boost - costly but optimal dither.

Gaussian is the most popular since a simple diode can be used as the noise source.

Rectangular and triangular dither are best used during playback (D/A).

Dither signals have a white spectrum, also known as white noise. White noise is used because it has a constant power unit bandwidth over the audible frequency range,

Dither noise must also vary between + and - in each sample period.

Dither is introduced before the LPF.

We will listen to dither itself, as well as music recorded with and without dither.

LOW PASS FILTER ( LPF)
(Also anti-alias or anti-image filter)

We have already seen that the LPF filters out frequencies that surpass the Nyquist level of the sample rate.

There are various LPFs available with varying levels of success. Ideally an LPF stops the unwanted frequencies cold right at the Nyquist level.

One would think that a 'brick wall' cutoff would be ideal. In reality, the vertical slope causes phase problems. Resonance is created at the cut off point causing ringing. Super sharp filters also require lots of capacitors, resistors and amp stages which contributes noise and distortion.

Therefore a sloped cutoff reduces phase problems, yet has it's own set of problems. The slope causes time domain response variation. A 2 kHz guardband is used so that the angle of slope reaches the nyquist line in time.

One problem with low pass filters is an anomaly called ripple. Ripple is frequency irregularity. As mentioned earlier, the LPF guard band actually extends below the Nyquist frequency level. Symptoms of ripple include ringing and phase distortion. Fortunately, these frequency anomalies happen at the top of our threshold of hearing so they usually are not noticeable. On the other hand , if enough anomalies pile up there may be audible detection.
 

SAMPLE AND HOLD (S / H)

After passing through the LPF, the audio signal reaches a very critical circuit, the sample and hold circuit. The S/H circuit performs two functions.

1. Periodically samples the analog signal.
2. Holds the analog value of the sample while the analog to digital converter analyzes and outputs digital words.

A potential error may occur as the signal is being held. The voltage current may leak through the capacitor. This is referred to as droop. The leak causes the held voltage to drop which can result in a wrong calculation by the A/D converter. Better circuit designs and components minimize erroneous droop.
 
 

ANALOG TO DIGITAL CONVERTER ( A / D)

The A/D converter is the most critical circuit of the digital recording process. It is the heart of the process of encoding digital audio. It is at this stage that the analog signal is actually converted to digital information. The S/H circuit and A/D circuit are intimately linked. The two circuits must be perfectly in sync to work properly. The S/H takes a sample and holds it. The A/D analyzes the held signal and converts the signal to a digital word. There is no room for error.

In a 16 bit system with a 44.1 sample rate, the A/D must do 16 approximations (one for each bit) for each held sample. That is 16 calculations every 44,100th of a second. That is 705,600 calculations per second.

If the timing varies, phase jitter is introduced. Jitter adds noise to the sample, especially in high frequency signals. The added noise results in miscalculation of the digital word assigned to the sample. The audible symptom is clicks or ticks during playback. Therefore, the internal clock of the device must be controlled by a highly accurate crystal quartz oscillator.

The A/D converter analyzes the captured held signal one of three ways.

1. Successive approximation register (SAR). The analog voltage is compared to a reference voltage within a feedback loop to determine the word. In a 16 bit system, the voltage is compared to 16 reference voltages. If the voltage falls below the 1st reference voltage, it is assigned a 1. If it is above the reference voltage, it is assigned a 0. This happens 16 times in a 16 bit system resulting in a 16 bit word.
2. Integration approximation. The analog voltage is allowed to decrease. The time it takes to reach zero determines the digital word.
3. Oversampling or low bit conversion. A higher sample frequency is used to dispose the low pass filter, then reduces the signal to a low sample frequency digitally removing any components that alias; low bit conversion provides a high resolution.

With an internal clock sync, the A/D is controlled by a master crystal oscillator.

With an external clock, the A/D's clock is driven by a phase lock loop (PLL) which tends to be more jittery than using a crystal clock. Therefore , if possible the internal crystal clock should be used.

Exceptions would be:

1. Synching an A/D to video.
2. Synching an A/D to another A/D.
3. Some computer clocks are prone to jitter, therefore you may choose an external clock source.

1. describe using SMPTE slave driver for pro tools. (better crystal- computer can be overtaxed.)
2. if possible show chips on mainboard with each component.
3. discuss clock speeds of computers.
4. discuss how digital devices use a clock to keep hours used.
5. discuss different a/d converters pertaining to gear.
6. Listen to dither and examples with and without. 

   ---


MRT 374 | (be sure to go to Oversampling or low bit conversion and| jitter ).