Fullscreen HTML5Med Res (??MB)HTML4

I _, L U _1 I_ILI IJ ,I.I ,. I LI I_ I I_I I I I An error analyser for CD players Here's your chance to put an end to the myths of the benefits of green pens, Sorbothane feet, rotational stabilisers and other gimmicks by building your own CD error analyser which connects to your CD player. This first article gives the digital background to the project. By STEPHEN McBRIDE* Since Philips introduced the first domestic CD player, the CD100, in the early BO's (the first in Australia was the Sony CDP101), many companies have been marketing CD accessory products with claims varying from intriguing to downright ridiculous. To 36 SILICON CHIP make matters worse, the "Golden Ear" brigade has brought new meaning to the use of hyperbole. Many of the big brand names are also just as guilty. In hindsight, I'm sure Philips had no idea just how much their claim that CDs offered "pure, perfect sound forever" would be twisted and used out of context as much as it has been over the past few years. Unfortunately, the unsuspecting public has been the victim of unscnipulous dealers out to make a quick buck from those left starry-eyed at the wondrous new technology of the CD system. There is no doubt Philips, and to some extent Sony, deserve to take a bow for bringing true state-ofthe-art recording technology into the home at a realistic price, despite the rubbishing it got from the analog fans in its infancy. Nor is there any doubt that the technology has improved since its inception. Yet despite the fact it's been in the marketplace for eight years, most people are totally ignorant of how the CD system works. Even if you have no desire to build the unit, this short series of articles will give you a good insight into the principles and operations of the CD system. Why have an analyser? As might have been expected, the article on CD green pens in the December 1990 issue of SILICON CHIP was met with a surge of mail. Testing fancy speaker cables is fairly easy but CD performance enhancers present a pile of obstacles, mainly because they (allegedly) work on the disc itself which operates in the digital domain. Proper CD analysers are complex (read expensive) and it would be out of the question to have one in your house. Well, now there is an easier, more economical alternative with this CD Error Analyser (CDEA) . This project will let you have an "analyser" at a fraction of the cost of the real thing. I say "analyser" because strictly speaking it doesn't give all the information a truP. analyser will, but it is as close as you are going to get without breaking the bank. In any case, the features it lacks are of no practical use to anyone except factory QC personnel. The big bonus is that the project has deliberately been designed to allow connection to most CD players on the market, from the el-cheapos to the big-buck units. This has been accomplished by its ability to be configured to suit the machine it's being connected to. Because there are so many chipsets on the market from various manufacturers, I decided to use one of the most popular sets, the Philips CDx series, comprising the CDl , CDZ, CD3, CD3a and CD'-1 units. (A CD chipset is a group of dedicated ICs intended solely for use in CD players). But it should work just as well on devices produced by other manufacturers. Philips chipsets appear in a myriad ofbrands including Philips (of course), Marantz, Aiwa, Micro-Seiki, Cambridge, Mission, Meridian, Nakamichi, Acoustic Research , Revox-Studer, Ratel, Sonographe and Toshiba. My machine is a 1985 Philips CD304 which uses the CDl chipset and this was used for testing the prototype. The machine shown in the photographs accompanying these articles is a CD104, which is virtually identical to a CD304. Incidentally, these players, as old The CD Analyser is easy to build with most of the parts mounted on two PC boards which are soldered together at right angles. We'll show you how to build it & connect it to your CD player in a future issue. as they are, can still outtrack the majority of current model players on damaged CDs and - surprise, surprise they use a single spot laser, not a 3beamer. In fact, Philips, who invented the CD system, has never used 3-beam footprints in any of their players even though they do in the VLP (Video Laserdisc) system, which is analog. Important points Before beginning assembly, a few pertinent points need to be considered. If you fit the CDEA to your CD player, you will most definitely violate the manufacturer's warranty conditions, hence the player will no longer be covered by warranty protection. So please think carefully about the implications offitting the CDEA if your player is still in its warranty period. Also, despite its ability to be adapted to various chipsets, the CDEA can't be connected to all machines. This is simply because some chipsets don't provide external connection points to the relevant information. For example, the Sony CDXl 125 decoder chip has no external access to the error flags so there is nothing you can do if your machine has this chip fitted. Likewise the Yamaha YM3817 uses multiplexed flags which require precisely, dedicated timebase decoding which kills any hope of having universal connection capabilities. Details of how to determine whether or not your machine can be fitted with the CDEA are given in future articles, so check that you can connect it up to your machine before rushing out to gather the parts. Definitions The following are terms which will be used throughout the following text and which may be new to you. Some are terms used only by Philips while others are common to all manufacturers. Note also that this article isn't intended to be one big plug for Philips, even though at times it may appear that way. But as Philips devices are so widely used and because Philips data is readily available, the choice was obvious. HF: High Frequency signal generated by the photosensitive diodes in the laser pickup assembly. This sig~ nal is a representation of the pattern impressed on the disc's information ]ULY 1991 37 Error analyser for CD players ... layer. A hole or "pit" in the surface scatters light from a semiconductor laser by causing a cancellation effect due to the depth of the hole being 1/4 the wavelength of the laser's radiation frequency. By the time the beam travels to the bottom of the pit and reflects back to the top, it is 180° out-of-phase with the light reflected off the area around the pit, which causes a cancellation effect. The flat area between the pits, called "lands", reflects a much larger percentage of light back to the photodiodes. A binary "1" is represented by the entry or exit edge of a pit while a binary "0" is represented by the flat area of the lands or the bottoms of the pits; ie, an NRZ (Non Return to Zero) code is used. HFD: HF Detector; a circuit which monitors the HF signal level and produces a signal, usually binary, flagging whether or not the HF is of sufficient amplitude for the decoder to work. A dropout will cause a temporary loss of HF level, causing the HFD to flag the event. HFD is also referred to as HFL. DROPOUT: A temporary, unwanted interruption to the HF signal caused by foreign matter on the disc blocking the light path (ie, dust or fingerprint), or by a physical defect such as a scratch on the disc surface or a hole in the reflective layer. It can also be caused by a loss of tracking as occurs while searching for a specific track (ie, in cue/review mode). The Philips specifications state that the CDs CIRC standard error correction can completely correct for loss of up to 4000 consecutive data bits (2.5mm on disc) and cope with the loss of up to 12,300 consecutive data bits (7.7mm of disc track length) by the use of linear interpolation. INTERPOLATION: The process of making a mathematical guess for the value of a missing or corrupted piece of data by averaging the values immediately before and after the bad one. For example, for the data stream containing values 1,2,?,4,5, the average of 2 and 4 is 3 and this is used to replace the"?" value. Since audio waveforms are sinusoidal in nature, interpolation provides a close approximation of the original value and hence a small, a 38 SILICON CHIP temporary increase in harmonic distortion. However, this is much better thaQ no value at all which would produce an annoying click. PCM: Pulse Coded Modulation. A system of representing analog signals with a binary weighted digital equivalent. Virtually, all A/D converters use some form of PCM as the digital output, either serial or parallel. The system was first proposed by Nippon Columbia, Japan (Denon). For the CD system, the code is in two's complement form. SAMPLE: A sample is two 16-bit PCM codes, one for left and one for right, representing the value of the original audio signal that was present at the input to the AID converters at the time of sampling. In the CD system, the AID samples are taken every 22.68µs, giving 44,100 samples per second, per channel. This value was mathematically chosen to satisfy several critical requirements. SYMBOL: Eight bits of audio data. Each sample of audio is represented by a 16-bit PCM signal which is split into upper and lower 8-bit halves, known as "audio symbols" for ease of handling. One sample period produces four symbols, two left and two right. EFM: Eight-to-Fourteen bit Modulation. Converts 8-bit symbols into a uniquely mapped 14-bit equivalent, specially chosen to make the decoding circuit as simple as possible for its 256 possible input combinations. EFM ensures there is always at least two '0's between consecutive 'l's but no more than 10 consecutive '0's in a 14bit EFM word; ie, under EFM, 00000000 becomes 01001000100000, 10001000 becomes 01001001000001 and 11111111 becomes 00100000010010. EFM prevents the data stream from containing low frequency components which could interfere with the focus, disc rotation or linear tracking servo systems. Also the code 100000000001000000000010 is the frame sync signal which contains 10 consecutive 0's so EFM prevents false sync triggering. The servos operate in the range 020kHz and the information transmission is in the range 20kHz-l.5MHz. It can also be shown that EFM enhances tracking performance over fingerprints, etc. MERGING BITS: A 3-bit block inserted between two adjacent EFM data words to ensure the 2-10 0's rule isn't violated at the block boundaries. A transition can be inserted if required to control the DC content of the HF signal. The merging bits contain no audio information and so are discarded by DEMOD during decoding. CHANNEL BIT: is a binary digit which has undergone sufficient processing to enable it to be ready for recording on the disc. C&D: Control & Data bits. Used to provide the servo and user-interface microprocessors with information about the current track number and title, the total and elapsed playing time, artist's name, album title, etc. The output is referred to as "Subcode Data" and is in eight serial bits, titled p, q, r, s, t, u, v & w. Each of the eight bits then forms part of an 8-bit parallel data stream by using a serial to parallel shift register. The eight bits clocked in serially are clocked out simultaneously (ie, in parallel) into eight separate serial shift registers, one for each bit, as each new 8-bit C&D symbol arrives. Once a sufficient number have been obtained, they are clocked out as eight individual serial channels titled p-w. Hence, the eight individual serial channels form one multiplexed serial channel during encoding, and decoding produces eight separate serial data channels. The "p" or "pause bit" is reserved to mark the silent periods between tracks, while the q or subcoding bit is used to form a serial information stream for the control microprocessor. Most machines ignore the r, s, t, u, v & w bits as the "q" bit contains the most commonly used data, such as time, track number etc. FRAME: A frame contains six sampling periods; ie, six left and six right 16-bit audio samples. This gives a 32bits per sample period or four symbols of 8-bit audio data. Hence, each frame contains 24 audio sy~bols, 12 left and 12 right, or 192 bits. Each symbol is then interleaved by the rules of CIRC. To this, one C&D symbol (eight bits) and eight parity symbols are added, producing 33 data symbols, or 264 bits. Each symbol undergoes EFM conversion and has three merging bits HIGH FREQUENCY LEVEL DETECTOR DATA SLICER A DEMODULATOR. SAA7010 ~ B : ERROR CORRECTOR . SAA7020 SAA/210. SAA7310 C CONCEALMENT. SAA7000 D : ENHANCED INTERPOLATION. SAA7210 E: TRANSVERSAL DIGITAL FILTER. SAA7030. SAA7120 F: 14-BIT DAC . TOA1540. 16-BIT DAC. TOA1541 G : BIT STREAM MODULATION DAC WITH 156x OVERSAMPLIN G. SAA7 320 H ACTIVE FILTER. TOA1542 OR DISCRETE PAR TS * DEC ODER LO GIC MASTER CLOCK ( 4.2336 DR 11.28~6MHz) CAN SE TIMING AND CONTROL GENERATED IN B. C. E OR G AND OTHERS SLAVED FROM IT AUDIO OUTPUT 8-14 BIT MODULATION AUDIO OUTPU T J~ ___:[___ -- --~L ---"'[__ :!l~~I ___ OUTPUT FLAG SHIFT REGISTER PROCESSOR SU BC ODE DATA 0 BIT P BI T - - - - - - - - - - - + - - - - - - - t - - - - , SERIAL TO PARALLEL SHIFT REGISTER DE-EMPHASI S 14 OR 16-BIT DAC TIMING AND CONTROL MUTE FROM SERVO SYSTEM CLOCK DATA DESCRAMBLER BIT STR EAM DAC 4-SAMPLE SHIFT REGISTER TIMING AND CONTROL COEFFICIENT GENERATOR RIGHT SHIFT REGISTER " EFAB " DIGITAL MULTI PLIER TIMING AND CONTROL SYSTEM CLOCK MUTE - oodB.- ...._,,_ ATTENUATE - 12aB Fig.1: block diagram of a CD player which uses the Philips CDl chipset. Most of the circuit functions & processes are described in the text. Philips chipsets are used in many brands (see text). ] ULY 1991 39 Error analyser for CD players ... · added, giving 17 channel bits per symbol and thus a total of 561 channel bits. 27 sync bits are then added, producing a total of 588 channel bits. This final result is the "channel bit stream" which is recorded on the disc surface such that a "1" is represented by a pit edge . So, our original 192 audio bits ends up as 588 channel bits on the disc; ie, for every 588 channel bits read from the disc, only 192 are PCM audio code. INTERLEAVING: Prior to EFM conversion, the encoding process interleaves (ie, jumbles up) the order of the data symbo.Is relative to time. To illustrate, consider a timeframe containing seven sample periods: 1L, lR, 2L, 2R, 3L, 3R, 4L, 4R, 5L, 5R, 6L, 6R, 7L & 7R. If this data was recorded serially in this order, a fault causing a dropout of, for example, six consecutive symbols would leave us with 1L, lR, 2L, 2R, 3L, 3R, 4L, 4R, ?, ?, ?, ?, ?, ess. In the CD player, a reverse CIRC operation takes place; ie, de-interleaving. CIRC: Cross Interleaved Reed-Solomon Code. Uses interleaving and two decoders , Cl and C2, to correct errors. The de-interleaving process is performed while the symbols are in RAM. When the data is recorded onto a disc, an encoder inserts parity check blocks so that the Cl and C2 decoders can determine whether or not the deinterleaved data is valid. The interleaving process only serves to scatter any data losses; it can't detect or correct for them. That's where Cl and C2 are used. First, the 32symbol frame is de-interleaved and moved into the Cl decoder where four parity symbols are stripped off and used to generate four error syndromes on the remaining 28 symbols. If there are no errors in C1, the symbols are written back to RAM. If one error is detected, it is corrected, then output to RAM. If two or more errors are detected, Cl flags the incorrect symbols as being unreliable and writes them to RAM for further processing. The 28 data symbols output from Cl are further de-interleaved then clocked into the C2 decoder. C2 strips off four parity symbols to create the error syndromes. If there are no errors, the remaining 24 symbols are written back to RAM and the four symbols used for the parity check matrix are discarded. If there is one error, C2 handles it the same way as Cl. If there are two errors, the flags set by Cl mark the errors and C2 uses these flags and its own error syndromes to create erasure positions to enable correction of two errors, then all 24 correct symbols are rewritten back to RAM. In the case of more than two erroneous symbols in C2 , all 24 are rewritten back to the RAM unchanged and a C2 flag is set to mark these 24 symbols as being unreliable. The data, regardless of whether or not it is correct, is then kept in RAM for a 5-frame duration then clocked out to Virtually, all AID converters use some form of PCM as the digital output, either serial or parallel. ? . This poses the problem of how can we guess what the missing values were. How ever, if the original serial stream is interleaved in the form: 1L, 1R, 3L, 3R. 5L, 5R, 7L, 7R, 2L, 2R. 4L, 4R, 6L, 6R, the loss of six consecutive symbols would give 1L, lR, 3L, 3R, 5L, 5R, 7L, 7R, ?, ?, ?, ?, ?, ?. This-may appear to be no different than before, but when we rearrange the order in the original way we have 1L, lR, ?, ?, 3L, 3R, ?, ?, 5L, 5R, ?, ?, 7L, 7R. This is much better than before and if we now split the left and right symbols apart, we end up with 1L, ?, 3L, ?, 5L, ?, 7L and 1R, ?, 3R, ?, 5R, ?, 7R. We can now use the error corrector's parity matrix to determine the missing values since each missing value is surround ed by valid values. Any values which can't be corrected reliably can still undergo linear interpolation to approximate the missing values. Neither function could be performed without the interleaving proc40 SILICON CHIP the next step in the chain, which in most cases is the CIM. Since CIRC spreads out any errors, most error bursts are fully corrected. DEMOD: DEMODulator. The circuit which takes the HF signal (4.3218 MBits/sec) from the laser assembly's photodiodes as its input and extracts and/or recovers frame and block sync signals, the data and parity symbols, the original bit clock rate and subcoding information for control and data displays. HFD notifies DEMOD of dropouts to ensure stability during periods where there is no HF signal. DEMOD also converts the 14-bit EFM words into 8-bit data symbols. ERCO: ERror COrrector. The circuit which de-interleaves the demodulated data from DEMOD, detects and, if necessary, corrects errors in the audio data stream. If ERCO is unable to correct an error, it outputs a flag, UNEC, to notify any following signal processing devices that the data is unreliable. It also removes any speed instability (ie, wow and flutter) by using RAM to buffer the incoming data rate from the outgoing data rate, thus allowing the output data to be re-synchronised to a clock signal derived from a quartz crystal oscillator. Hence, wow and flutter is eliminated. In most cases, ERCO also generates a motor speed control line; ie, MCES (Motor Control from Erco to Servo). The output from ERCO is the left and right audio data in 16-bit serial fashion, and any error flags resulting from the CIRC process not being able to fully correct any errors. UNEC: UNreliable data from Erco to Cim. A flag which signals that the current audio symbol has failed to pass the full requirements of the CIRC error corrector and thus needs further processing before being passed to the DI A converters. In some implementations, UNEC can be used to notify CIM that there will be a symbol arriving five frames later which is corrupted, hence giving advance warning that action will need to be taken. More about this in CIM (below). CIM: Concealment by Interpolation and Muting. If ERCO can completely correct all audio data, CIM is transparent to the data stream. If ERCO can't correct a single symbol (ie, a bad symbol has a good one either side), CIM will use the principle of linear interpolation to replace the error. If the error exists in two or more The CD Analyser is connected to the CD player via this small interface board. Note, however, that the unit can not be connected to all machines. consecutive symbols, the device mutes them until valid data is again available. UNEC gives CIM five frames advance warning when a mute is needed. When CIM receives a MUTE command, either via ERCO 's UNEC or from the control microprocessor, it immediate! y starts a digital attenuation process whereby the audio data values are brought to zero by following a cosine curve (0 - re) over a 30-symbol period and holds the audio values at zero so long as MUTE is active. When MUTE is released, the audio symbols are returned to their normal values (ie, the attenuation is removed) over a 30 symbol period, following a cosine curve (re - 2TC). This action occurs when track jumping is expected by the control microprocessor or when ERCO is unable to correct a burst of errors, there being too many for the interpolator to handle. This smooth muting action prevents annoying and potentially dangerous (to speakers) transients from occurring. Because it takes 30 symbols to reach the zero point, ERCO keeps all audio symbols in a 5-frame (ie, 30 symbol) delay before sending them to CIM. If ERCO decides muting action is needed, the delay line allows ERCO to give CIM a 5-frame advance warning so that CIM will have time to acli vale the digital attenuator down to zero by the time the erroneous data reaches CIM's input. Unfortunately, not all manufacturers' devices offer this critically acclaimed feature. EFAB: Error Flag from A-chip (SAA7210, SAA7310) to B-chip (SAA7220). EFAB is almost the same as UNEC except it only flags current symbols as being unreliable. It doesn't need a 5-frame advance warning because the SAA7220 (B-chip) has internal delay facilities to cope with large error bursts. FCO: Focus Control 0, one of three focus states for the servo microprocessor. FCl and FC2 are used to move the focus lens up and down during the startup procedure by the servo microprocessor. When FC0 is high, the focus and disc rotation servo circuits are overridden and forced off, making the focus lens sit in its home position and leaving the disc stationary. Upon receiving a startup corn- mand from the control microprocessor, the servo microprocessor uses FCl and FC2 to determine if focus can be achieved, indicating that a disc is on the turntable and is the right way up. Once satisfied all is well, the servo microprocessor releases FC0 (ie, it goes from a high to a low), allowing the focus and disc rotation servos to do their thing. It also gives the disc motor an initial shove to help get things up and running as quickly as possible. FC0 only creates a falling edge at startup; ie, an ideal time to reset the counters. RD: Ready Signal. A flag set by the TDA5 708 photo diode processor to signal to the control microprocessor that the laser startup routine has successfully been completed. We can use it for the same purpose as FC0. On the TDA8808, the RD pin is combined with the Si (attempt Startup routine instruction) input. This input is a 2way communications line between the control microprocessor and the TDA8808 and serves fine for use as a counter reset command line (ie, the same as FC0), even though it performs additional tasks. On the TDA8808, it's called Si/RD. 'MUTE: As the name implies, a signal from the control microprocessor (to the CIM) requesting that the audio output signal be muted. This occurs when the control microprocessor expects corrupted audio information to be received by the decoder, such as happens when in the pause mode or others such as skip, etc. The SAA7220 has a similar input: MUSB (MUte from Servo to B-chip). See also ATSB. ATSB: ATtenuate command from Servo to B-chip. This line is used by the control microprocessor to force the SAA7220 to digitally attenuate the audio signal by -12dB. This is so you can hear where the laser is in the high speed searching mode, enabling easier location of the desired portion of the disc. The SAA70x0 (CDl) devices don't offer this facility. In this project, if the SAA7220 is fitted, it is necessary to tap into both MUSB and ATSB, ·as errors are produced when either line is pulled low. In the next article, we will go on to describe the CD chipsets and present the circuit of the CD Error Analyser. * Stud. I.E. Aust; Dept. of Electronic and Computer Engineering, James Cook University, Townsville. sc JULY 1991 41