Clever Geek Handbook
📜 ⬆️ ⬇️

RIAA curve

The frequency response of the recording (“anti-RIAA function”, the red curve) and the frequency response of the gramophone (“RIAA function”, the blue curve) normalized to 1 kHz A. The dashed curve is the frequency response of the IEC 1976 edition. Numbering of cutoff frequencies in chronological order of implementation

The RIAA curve is the standard amplitude-frequency characteristic (AFC) of the recorders of a long-playing gramophone record and the inverse of its amplitude-frequency characteristic of the preamps-correctors , restoring the original signal spectrum during playback. When writing the original program to a varnish disc signal is processed by a predistortion circuit with time constants 3180, 318, and 75 μs , which corresponds to frequency bending frequencies of the frequency response of 50.05, 500.5, and 2122.1 Hz [comm. one] . When playing a plate with an electromagnetic pickup, the original signal spectrum is restored by the reverse circuit with the same time constants . The complex shape of the RIAA curve is a compromise that has emerged from the need to get the best playback quality from technically imperfect mechanical recording devices .

The first serial records recorded using this frequency predistortion scheme were released by RCA Victor in August 1952. . In June 1953, [1] the RCA was approved by the (NARTB) as a national standard; NARTB's choice has been supported by other industry institutions, including the American Recording Association (RIAA) . By 1956, a new standard, known as the RIAA Curve, supplanted competing formats and captured the markets of the United States and Western Europe. In 1959, the RIAA curve was approved, and in 1964 standardized by the International Electrotechnical Commission . In 1972, the standard as amended by the IEC was adopted in the USSR. In 1976, the IEC modified the standard RIAA reproduction curve in the low-frequency region; the innovation met with fierce criticism and was not accepted by industry . In the XXI century, the vast majority of manufacturers of preamps-correctors follow the original RIAA curve standard without changes introduced by the IEC in 1976 [2] .

Mathematical Description

AFC Records

The standard amplitude-frequency characteristic of the recording channel of long-playing records (the “anti-RIAA function” [3] ) is described by the formula for sequentially connecting three frequency-dependent first-order links - two differentiators (numerator) and one high-pass filter (denominator) [4] :

Vx(ω)∝one+(ωT2)2one+(ωT3)2one+(ωTone)2{\ displaystyle V_ {x} (\ omega) ~ \ propto ~ {\ frac {{\ sqrt {1 + (\ omega T_ {2}) ^ {2}}} {\ sqrt {1 + (\ omega T_ { 3}) ^ {2}}}} {\ sqrt {1 + (\ omega T_ {1}) ^ {2}}}}}   [5]

or

Vx(f)∝one+(f/f2)2one+(f/f3)2one+(f/fone)2{\ displaystyle V_ {x} (f) ~ \ propto ~ {\ frac {{\ sqrt {1+ (f / f_ {2}) ^ {2}}} {\ sqrt {1+ (f / f_ {3 }) ^ {2}}}} {\ sqrt {1+ (f / f_ {1}) ^ {2}}}}}   ,

WhereVx {\ displaystyle V_ {x}}   - vibrational displacement speed of the grooves,f {\ displaystyle f}   andω {\ displaystyle \ omega}   - frequency and angular frequency of the signal, andTone {\ displaystyle T_ {1}}   ,T2 {\ displaystyle T_ {2}}   andT3 {\ displaystyle T_ {3}}   - RIAA-specific time constants that determine the cutoff frequenciesfone {\ displaystyle f_ {1}}   ,f2 {\ displaystyle f_ {2}}   ,f3 {\ displaystyle f_ {3}}   . Different methods of numbering these frequencies and time constants are used in the literature; in the above formulas they are numbered in chronological order of their introduction into production (fone {\ displaystyle f_ {1}}   - 1926 [6]f2 {\ displaystyle f_ {2}}   - 1938 [7] ,f3 {\ displaystyle f_ {3}}   - 1948 [8] ):

  • Tone{\ displaystyle T_ {1}}   = 318 μs sets the crossover frequency of the low-frequency ( constant-amplitude displacement amplitude mode ) and medium-frequency ( constant -frequency mode of vibrational amplitude amplitude ) regions,fone {\ displaystyle f_ {1}}   = 500.5 Hz;
  • T2{\ displaystyle T_ {2}}   = 75 μs sets the crossover frequency of the mid-frequency (constant-amplitude mode of vibrational velocity) and high-frequency (constant-amplitude shift mode) regions,f2 {\ displaystyle f_ {2}}   = 2122.1 Hz. Interval betweenTone {\ displaystyle T_ {1}}   andT2 {\ displaystyle T_ {2}}   is only two octaves , so the “kinks” of the idealized piecewise linear frequency response are actually smooth bends;
  • T3{\ displaystyle T_ {3}}   = 3180 μs sets the frequency of the rise of low frequencies during recording (f3 {\ displaystyle f_ {3}}   = 50.05 Hz) - in order to reduce the relative level of rumble and low-frequency noise during subsequent playback [5] .
 

The frequency response of the recording (“anti-RIAA function”), defined in terms of the vibrational velocity of the groove, is measured in practice in the through path from the linear output of the source of the recorded signal to the output terminals of the reference electromagnetic pickup [7] and characterizes not production equipment, but its end product - phonograph record. The deviation of the real frequency response of the recording from the given formula, according to the publication of IEC-98, should not exceed 2 dB [9] .

Frequency Response

The inverse conversion of the voltage at the output of the electromagnetic pickup, which is proportional to the vibrational velocity, into the output voltage of the preamp-correctorU {\ displaystyle U}   performed by the “RIAA function”. The standard RIAA filter is equivalent to the serial connection of two first-order low-pass filters (denominator) and one differentiator (numerator) [10] :

U(ω)∝one+(ωTone)2one+(ωT2 ) 2 one + ( ω T 3 ) 2{\ displaystyle U (\ omega) ~ \ propto ~ {\ frac {\ sqrt {1 + (\ omega T_ {1}) ^ {2}}} {{\ sqrt {1 + (\ omega T_ {2}) ^ {2}}} {\ sqrt {1 + (\ omega T_ {3}) ^ {2}}}}}}   [5]

or

Vx(f)∝one+(f/fone)2one+(f/f2)2one+(f/f3)2{\ displaystyle V_ {x} (f) ~ \ propto ~ {\ frac {\ sqrt {1+ (f / f_ {1}) ^ {2}}} {{\ sqrt {1+ (f / f_ {2 }) ^ {2}}} {\ sqrt {1+ (f / f_ {3}) ^ {2}}}}}}   ,

with the same as in the frequency response of the record, the values ​​of time constants and frequencies. The deviation of the frequency response of real devices from the standard is not normalized based on the assumption that such a deviation can be corrected by the timbral block of the amplifier [9] . The target value of the maximum deviation of the frequency response from the standard, adopted during the development of high-quality preamps-corrector, is ± 0.1 dB [11] .

The frequency response of the playback channel (“RIAA function”) is always concentrated in the preamp-equalizer. These preamps are practically unsuitable for reproducing the absolute majority of “gramophone” records at 78 rpm due to the decrease in frequency response at medium and high frequencies [12] . The sound of such records turns dull, devoid of overtones [12] . When playing records recorded with first-generation electric recorders with particularly lowfone {\ displaystyle f_ {1}}   , this effect is exacerbated by an additional increase in low frequencies [12] .

Scope and rationing

Both formulas are defined in the frequency range from 20 Hz to 20 kHz; beyond its frequency response is not regulated [10] . Formal extrapolation outside the audio range shows that with decreasing frequency below 20 Hz, the recording frequency response module asymptotically approaches unity, and with increasing frequency above 20 kHz it grows infinitely, in direct proportion to the frequency. In real recorders, in addition to RIAA recording filters, inevitably there are filters that are not provided for by the standard, which block the passage of direct current, infrasound , ultrasonic and radio frequencies to the cutter drives and do not affect the transmission of sound frequencies [13] . For example, in the most common [14] recording amplifier SAL 74B, high-frequency noise is cut off by a second-order Butterworth filter with a cutoff frequency of 49.9 kHz [13] . The sound attenuation introduced by him, less than 0.1 dB at 20 kHz, is not audible and does not require any compensation in the playback channel [13] .

In practice, both formulas are always calculated in decibels and normalized to a frequency of 1 kHz. At this frequency, the normalized values ​​of the frequency response of both recording and playback are 0 dB [10] ; the normalized frequency response of reproduction at a frequency of 20 Hz is +19.274 dB (gain of 9.198 times relative to the level at 1 kHz), and at a frequency of 20 kHz it drops to −19.62 dB (attenuation of 9.572 times) [15] . Thus, the gain of the RIAA preamplifier at frequencies of 20 Hz and 20 kHz differ by 39 dB, or 88 times. A common statement that at frequenciesfone {\ displaystyle f_ {1}}   andf2 {\ displaystyle f_ {2}}   normalized frequency response of the reproduction takes values ​​of 3 dB and −3 dB, not true [16] . It is valid for single filters of the first order, but not for a chain of series-connected filters with sufficiently close cutoff frequencies. Exact RIAA function values ​​onfone {\ displaystyle f_ {1}}   andf2 {\ displaystyle f_ {2}}   equal to +2.648 dB and −2.866 dB, respectively [17] [16] .

Frequency correction purpose

Features of long-playing sound recording

 
Recorder and System Copper Disc OriginalDMM A nozzle for helium supply is visible in the upper corner of the front triangular panel [18]

The classic production cycle of stereo records begins with cutting the original record in thin [comm. 2] a nitrocellulose layer [comm. 3] varnish applied to an aluminum disk [21] . Triangular in plan [comm. 4] , forcedly heated to 200-300 ° С [23], the sapphire cutter mounted on the massive tangential “tonearm” of the recorder is controlled by two light but powerful electromagnetic drives cooled by air or helium jets [21] [comm. 5] . Frequency distortion, intrinsic resonance, and nonlinearity of the recorder’s mobile system are effectively suppressed by the electromechanical feedback circuit developed in the late 1930s and became the de facto industry standard by the mid-1960s [27] [28] [29] . The cutter moves from the edge to the center of the disk strictly along its radius, and the axis of symmetry of the cutter is always directed tangentially to the cut groove [21] .

The signals of both stereo channels are encoded by the transverse (horizontal) displacement of the cutter [30] . The displacement of the outer side of the groove closest to the edge of the plate corresponds to the right channel, the inner side to the left [30] . Cutter drives are oriented at angles of + 45 ° and -45 ° to the axis of the cutter, and the signals supplied to them are switched so that when recording a monophonic (in-phase) signal, only the transverse displacement of the groove changes; its width and depth remain unchanged. The displacement of the cutter in the depth of the lacquer layer and vice versa corresponds to the difference of the signals of the left and right channels. During phonogram mixing, the amplitude of vertical movement is limited in order to avoid needle jumps [31] [32] [33] . This stereo recording system, known as the “45/45 system”, became an uncontested world standard in 1958 [34] .

The distance between the grooves varies from 200 to 65 μm (130-390 grooves per inch) [21] , which at a speed of 33⅓ rpm provides a playback time of one side of the plate from 13 to 40 minutes [comm. 6] . The maximum lateral displacement of the groove in the 1950s was limited to 25 microns; as the pickups improved, it gradually increased [36] . In the 1972 USSR standard, the maximum horizontal displacement of the groove was 40 μm, and the maximum vertical displacement was not more than 20 μm [37] ; by 1978, the permissible lateral displacement increased to 50 μm [36] . In the 21st century, the width of an unmodulated groove almost never drops below 50 microns; on loud fragments, the groove expands to 80–90 microns, and when recording singles by 45 rpm , the groove width can reach 125 μm [38] .

The upper cutoff frequency is determined by the high-frequency resonance of the cutter and does not exceed 25 kHz [39] . At frequencies above this boundary, the amplitude of the recorded oscillations decreases so quickly that it can be assumed that the recorded signal does not contain useful ultrasonic components [40] . An exception is the quadrophonic plates of the CD-4 system, in which the spectrum of the useful signal extends to 45 kHz [41] . The lacquered originals of these plates were cut with ordinary incisors at a half-speed of disk rotation with a half-speed magnetic phonogram. The maximum recording frequency was 22.5 kHz, but when reproduced at standard speed, it was converted to 45 kHz [41] .

Geometric recording restrictions

 
Limit recording levels for speed 33⅓ rpm (USA) [42] . The high-frequency region of the graph describes the worst case — reproduction with a standard round needle with a tip radius of 18 microns. Dotted line - absolutely record levels recorded by Shure specialists [43]

The movement of the cutter when cutting the groove should fit into three restrictions - the limiting amplitude of the displacement of the groove, its limiting vibrational velocity and limiting acceleration [44] . The first of them acts equally on the entire area of ​​the record allotted for recording. Speed ​​and acceleration limits are set for the worst case — grooves closest to the center of the plate [45] . The closer the groove to the center, the higher the likelihood of overload and distortion, and vice versa: the farther the groove from the center, the lower the recording density of the vibrations, which makes it possible to carefully calculate the excess of the speed and acceleration limits [36] .

The meaning of limiting the displacement amplitude is obvious: even a slight excess of this limit, which does not lead to the destruction of the wall between the grooves, can deform this wall and give rise to a clearly audible copy-effect [44] . Recording a signal with a maximum bias amplitude provides the best signal-to-noise ratio [46] , but it is technically possible only at low frequencies. At the turn of no more than 1 kHz, another restriction comes into effect - by the limiting speed of the groove displacement. Failure to observe this limit during recording leads to the fact that the rear edges of the cutter damage the walls of the groove cut by its front edges [37] [32] . When reproducing a groove recorded in excess of speed, its effective width narrows, the effect of extruding the needle from the groove (pinch effect) and, as a result, nonlinear distortions arise [37] . Therefore, the maximum speed of the groove displacement is always limited: in the Soviet GOST 7893–72 level of 10 cm / s for monophonic and 7 cm / s for stereo recordings [37] ; by 1978, the limit was increased to 14 cm / s [36] . The nominal recording level (“0 dB”), relative to which the gain of the reproducing path is normalized, corresponds to a peak speed of 8 cm / s; in practice, it is often equated with an rms speed of 5 cm / s [47] . In world practice, there were records with a five-fold excess of this threshold - 38 cm / s (+14 dB) at a frequency of 2 kHz, which corresponds to an acceleration of the pickup needle of 487 G [43] .

At high frequencies, the third limiting factor comes into effect, which is associated with acceleration - the maximum curvature of the groove. In order for the pickup needle to track the high-frequency displacement of the groove, the radius of this displacement must be no less than the radius of the needle tip. If this limitation is not taken into account during recording, the needle will slip past the high-frequency depressions and ridges of the groove and irreversibly damage them [48] [37] [49] . For standard round needles with a tip radius of 18 μm, this effect (“non-flexing error” [46] , English tracing error [comm. 7] ) can already appear at 2 kHz, for needles with a narrow elliptical tip - at 8 kHz [32] . The acceleration limit normalized in the USSR was initially 25 • 10 4 cm / s 2 (255 G), and by 1978 it had grown to 41 • 10 4 cm / s 2 (418 G) [36] .

The principle of predistortion

 
Communication of the frequency response of the predistortion filter of the recording channel with the limitations of displacement, speed and acceleration The ordinate axis is normalized relative to the level at a frequency of 1 kHz. Frequency response is shown schematically; in real filters, its “fractures” are smooth bends

There are two main modes of recording a harmonic signal on a varnish disc. In the mode of constant displacement amplitudes [46], the amplitude of the groove displacement depends only on the amplitude of the recorded electric signal and does not depend on its frequency. In this case, the rate of change in the bias increases in direct proportion to the frequency of the signal and sooner or later reaches unacceptably high values. In the mode of constant amplitudes of the vibrational velocity [46], the amplitude of the rate of change of the groove displacement does not depend on the frequency, and the amplitude of the displacement is inversely proportional to the signal frequency. The most common electromagnetic pickups are sensitive specifically to vibrational speed, so playing back records recorded in this mode does not require any frequency correction. However, such recordings are characterized by unacceptably high relative noise levels at medium and especially high frequencies [46] . Due to these shortcomings, none of the two modes is applicable in its purest form. All [51] practical sound recording systems combine sections of both modes: at low frequencies, the recorder operates in a mode of constant displacement amplitudes, and at medium frequencies, in a mode of constant vibrational velocity. The transition from one mode to another takes place in a special predistortion filter , and the section frequency is selected so as to fit the maximum of the useful signal into the limits set by the technology.

An ideal solution to the problem does not exist, since any musical or speech program has its own unique spectral distribution of energy and peak signal amplitudes [52] . There is no standard of such a distribution with which it would be possible to evaluate the effectiveness of a particular filter setting [32] [comm. 8] . In practice, the simplest spectrum model is used, in which the peak amplitudes are constant in the range of 20 Hz ... 1 kHz, and in the range of 1 ... 20 kHz they decrease at a speed of about 10 dB per octave [32] [comm. 9] . The fraction of high-frequency components in this model is so small that limiting the maximum acceleration loses its meaning. On the contrary, from the point of view of the best signal-to-noise ratio, it is advisable to increase the level of the high-frequency signal in order to make maximum use of the dynamic range of the recording [37] [32] [54] . The frequency response slope of 10 dB per octave with simple filters cannot be reproduced; in practice, only combinations of first-order filters are used, each of which implements a slope of 6 dB per octave [55] . What matters is not the accuracy of “fitting” the conditional spectrum model into the conditional model of the plate, but the exact, mirror-like correspondence of the frequency response of the recording and reproducing channels [55] .

For the same reason - the need to suppress low-frequency interference of playback - the recording level at the lowest frequencies (20 ... 50 Hz in the RIAA standard) also rises [9] . Thus, the optimal frequency response of the predistortion filter of a long-playing recording has three inflection points in the sound region: two in the mid-frequency region and one low-frequency [5] .

Historical Review

Frequency correction before switching to a long-playing recording

 
Typical European (black dotted) and American (light red stripe) predistortion patterns of the 1930s and 1940s. The upper limit of the frequency range of serial plates for this period increased from 5 ... 6 kHz in the Maxfield-Harrison system (1926) to 14 kHz in the Decca ffrr system (1944) [56]

Absolutely all records in history were recorded with distortions of the spectrum of the original signal [51] . At first, these were natural, inevitable, and irreparable frequency distortions of purely mechanical recorders [51] . This stage of technology development reached its peak in the mid-1920s [57] ; at the same time, the transition began from direct recording of acoustic vibrations to the electric amplification of the recorded signal [58] . The developers of the first Bell Labs electric recorder, Joseph Maxfield and Henry Harrison , who understood the impossibility of using the modes of constant amplitude and constant vibrational velocity in their pure form, introduced a predistortion filter with a low-frequency and mid-frequency section in the circuit (fone {\ displaystyle f_ {1}}   ) 200 Hz [6] . For frequencies above 4 kHz, they recommended the transition to a constant acceleration mode, but in imperfect equipment of the 1920s it was not in demand [6] . Not immediately, gradually, the need for deliberate spectrum distortions was realized by other designers and sound engineers [51] .

In the 1930s, most manufacturers used at least two-link frequency correction, similar to the Maxfield and Harrison scheme, and standard condenser microphones designed by [57] provided an additional rise in frequency response at high frequencies. US Market Captured by Western Electric Proprietary Recording System [58] [comm. 10] ; British EMI , followed by most European manufacturers, adopted the Blumlane 250 scheme [comm. 11] ( Eng. Blumlein 250Hz ) with a crossover frequency of 250 ... 300 Hz [58] [61] .

Until the end of World War II, Europeans focused on the mechanical reproduction of records by gramophones and therefore gravitated to a regime of constant velocity amplitudes; the constancy of bias amplitudes was applied only involuntarily, at the lowest frequencies [62] . In the richer United States, where buyers could afford electrophones and radiols , the regime of constant bias amplitudes was applied over a much wider band, up to 1 kHz [62] [63] . In the mid-1930s, American studios replaced the old, “voiced” condenser microphones with the latest, relatively neutral ribbon microphones. Since the timbre of such recordings seemed dull, depleted in comparison with the old records, in order to “compensate for the losses” the studios began to raise the level of high frequencies with filters built into the microphone preamps [7] . Other technical problems when recording high frequencies are the decline in frequency response due to imperfections in the 1930s incisors [comm. 12] and the growth of nonlinear distortions with decreasing groove radius during playback - also corrected by the rise of high frequencies [8] .

In 1938, RCA Victor was the first to transfer this function from the microphone preamplifier to the recorder amplifier: this is how the first frequency correction scheme with two inflection frequency response appeared [7] [62] . According to a RCA representative, the second inflection frequency (f2 {\ displaystyle f_ {2}}   ) was 2500 Hz; according to , curator of the sound archive of the British Library , the “sonority” of real RCA Victor records of that period was generated not by high-frequency correction, but by distortions during signal compression [64] . In the industry as a whole, no “standard” predistortion schemes existed. IN USAfone {\ displaystyle f_ {1}}   ranged from 200 Hz to 1 kHz, andf2 {\ displaystyle f_ {2}}   ( if used) - from 2 to 3 kHz [63] . The selected correction scheme on the plate was indicated rarely and far from always correctly. As a result, high-quality electrophones of those years were necessarily equipped with timbroblocks (and essentially parametric equalizers ) with varying inflection frequencies to select the optimal timbre by ear [63] .

First LPs

In December 1933, Alan Blumlein recorded the first stereo record using the 45/45 system. The invention was a quarter of a century ahead of its time and was literally “put on the back burner” in EMI vaults [58] [comm. 13] . The main goal of the designers and technologists of the 1930s was not stereo recording, but the replacement of an aging shellac plate at 78 rpm with a long-playing plate [58] . Before the start of its serial production, it was necessary to solve many technical problems, and then select the frequency correction curve that is optimal for the new technology [58] . The first to reach the goal was American Columbia Records , which released the first full-fledged long-playing records in 1948 [66] .

The company, which worked on the new product from the 1930s, seriously hoped to become the author and owner of the new world standard [66] . She really managed to make the standard disk rotation speed (33⅓ revolutions per minute), the geometric specification of the grooves, she invented and introduced the notation LP itself into circulation [66] . Columbia chose the frequency correction scheme for LPs on the recommendation of her old partner, the [67] . An exact technical description of this scheme has never been published; from the published graphs it follows that NAB used the frequency response with kinks of 1590 μs (100 Hz), 350 ... 400 μs (400 ... 450 Hz) and 100 μs (1600 Hz) [68] . From an engineering point of view, this was a successful compromise solution, very close to the future RIAA standard and almost indistinguishable from it by ear [68] .

By 1952, the brand name of the Columbia curve ( English LP Curve ) became a household name in the United States [66] . Industry experts were sure that this particular scheme would become the industry standard, but Columbia lost the format war [66] . The main drawback of her scheme was that it was optimized for plates with a diameter 406 mm that have not been accepted by the market. For conquered market plates with a diameter 305 mm , more sensitive to overloads at high frequencies, the Columbia circuit fit worse [12] . Company Selected Valuef2 {\ displaystyle f_ {2}}   (1600 Hz) was too low, which only exacerbated these distortions [12] .

Formats War

 
The main schemes of frequency correction of the recording channel used in serial production before the transition to the RIAA standard. The 1953 RCA circuit (red broken line), lying approximately in the middle of the “corridor”, became the basis of the RIAA standard

Following Columbia, competitors entered the long-playing record market using alternative frequency correction schemes. About these short-lived technical solutions that have never been published in the form of full technical descriptions, only fragmentary, inaccurate and often incorrect information has been preserved. The labeling of the plates of this period is confused or completely unreliable [comm. 14] ; the actual frequency response of the predistortions used when recording them can only be estimated by ear. For example, Decca , in 1950, began selling a long-playing version of its proprietary system. ffrr , for three years published four different frequency response graphs [69] . However, according to Copland, in reality, before switching to the RIAA standard, Decca applied only two schemes - “Blumlane 500” and its variant with the increase of high frequencies above 3.18 kHz [70] . In total, at least nine different systems claimed the standard status in the postwar decade [71] . The interface between the low-frequency and mid-frequency ranges varied from 250 to 800 Hz, the high-frequency rise ranged from 8 to 16 dB at 10 kHz [1] . In addition, there were “company standards” not intended for replication of large radio stations, archives, and libraries - for example, various BBC services used three different predistortion schemes until 1963 [71] . Industry ( , 1950 [72] ) and international ( CCIR , 1953 [73] ) organizations, as they could, "controlled the process", offering their own solutions. The last of these failed standards, the German DIN 45533 , was approved in July 1957 and never reached serial production [74] .

Many incompatible formats were only to the benefit of equipment manufacturers, who offered their listeners complex timbral blocks for correcting frequency distortions. Record manufacturers, on the contrary, were interested in the speedy standardization of frequency correction. In 1953, when it became obvious that the industry was not going to adopt the NAB and Columbia correction scheme, the NARTB] conducted a comparative analysis of the frequency correction schemes used in the USA, and based on them the ideal “average” frequency response record and playback [1] . Of all the actually used circuits, the frequency response of RCA Victor was best suited to it, which was introduced into production in August 1952 under the brand name New Orthophonic [72] [1] . Its deviation from the average ideal in the entire sound range did not exceed ± 1.5 dB [1] . RCA Victor, like Columbia, used a three-bend recording curve, but optimized for 33 ⅓ rpm. It is the RCA Victor circuit, with the rise of low frequencies atf3 {\ displaystyle f_ {3}}   = 50.05 Hz, and was chosen as the national standard of the United States [1] .

Deployment

 
1956 RCA Victor label. This is a 78-rpm gramophone single, but the New Orthophonic brand name (to the right of the center hole) indicates a frequency correction according to the RIAA standard

In 1953-1954, the solution proposed by NARTB was consistently recognized by the American Association of Television and Radio Equipment Manufacturers (RETMA) and the (AES). After the American Recording Industry Association (RIAA) approved it as the US national industry standard in May 1954, it was given the name “RIAA curve” or “RIAA frequency correction” ( RIAA curve, RIAA equalization ). In 1955, the RIAA curve became the UK national standard and received preliminary approval from the International Electrotechnical Commission [1] [75] ; three years later, the IEC officially recognized the RIAA curve as a standard (Publication IEC-98-1958, now IEC 60098).

The transition of the US industry to the RIAA curve was rapid, at least in words [76] . Understanding that it would be very difficult to sell stocks of old, non-standard records in the new conditions, manufacturers hastened to declare compliance with the new standard [76] . In fact, the transition dragged on for several years, during which the companies sold out old stocks and printed out new circulations of old records [76] . It is impossible to indicate the exact date of the complete transition of a company to the RIAA curve; it can only be argued that since 1956 it was used to record almost all varnish originals of long-playing phonograms [77] in the USA and Western Europe. The only exception was Germany, where manufacturers and industry regulators experimented for several years with their own national standard, which differed from the RIAA curve byf3 {\ displaystyle f_ {3}}   [78] .

Despite the development of studio equipment and a culture of recording production, the high-quality playback capabilities in the standard did not immediately reach the mass consumer [79] . High-quality, pre-correcting preamps-correctors in household appliances of the 1950s and 1960s were rare; usually, designers used cheap, inaccurate, poor-sounding preamplifier cascades [79] . The main reason for this attitude was the low quality of the chassis and tonearms of household players, which made no improvement in the electronic path meaning [79] [comm. 15] . Even in the best correctors of that time, the frequency response deviation from the standard was significant, for example, in the Dinsdale two-transistor circuit (1965), with the exact selection of components, it was +1.6 dB at 20 Hz and +0.7 dB at 20 kHz [80] . The best circuits on discrete transistors of the 1970s deviated from the standard by a fraction of a percent, for example, the classic Technics SU9600 circuit - no more than ± 0.3% [81] (at the cost of increasing the supply voltage of the transistor circuit to 136 V [82] ). Then, in the 1970s, with the transition from discrete transistors to integrated circuits, designers switched to a relatively high-quality, easily reproduced in serial production, corrector circuit on an operational amplifier . Initially, under the influence of the authority of , a relatively noisy OA circuit in the inverting inclusion dominated; after Walker’s work came out in 1972, a low-noise, but less flexible and more complicated calculation and adjustment circuit for an op-amp in a non-inverting switch came to the fore [83] . The accuracy of reproducing the standard frequency response still remained unsatisfactory until the publication in 1979 of the fundamental work of Stanley Lipschitz , who developed a simple and reliable mathematical apparatus for calculating predistortion filters [84] .

IEC Amendment

In September 1976, the International Electrotechnical Commission approved a new edition of the IEC-98 Publication. The frequency response of the recording in the new standard has not changed, but the fourth time constant, 7950 μs, corresponding to a high-pass filter with a cutoff frequency of 20.02 Hz [85] [16] has appeared in the frequency response of playback. According to the idea of ​​the developers of the standard, the new filter was supposed to suppress the passage of infrasonic vibrations during the reproduction of warped records [85] [16] . IEC motives remained a mystery: neither ordinary listeners, nor representatives of the recording and electronic industries have ever demanded such changes [85] . Both those and others met innovation with hostility. Some manufacturers of consumer electronics refused to introduce a new filter into their amplifiers, while others made it disabled [13] . In the XXI century, the vast majority of manufacturers of amplifiers do not apply the IEC amendment [2] , while formally the 1976 amendment remains valid [83] .

In the 1970s, critics of the IEC amendment drew attention, first of all, to the undesirable nonlinearity of the “corrected” frequency response of the through channel. At a frequency of 20 Hz, the AFC block off relative to the linear one was –3.0 dB, at 40 Hz –1.0 dB, at 60 Hz –0.5 dB [85] [16] . High-quality reproduction of such low frequencies was the destiny of professionals and a few wealthy lovers, and they did not want to part with the acquired [83] . The infrasonic rumble in systems of this level was minimal, and for the reproduction of warped disks, if necessary, long-known switchable filters were used [83] .

The IEC amendment also had objective defects. A first-order filter at 20.02 Hz more or less effectively suppressed only the fundamental tone of distortion from warping (−14.2 dB at 4 Hz) [85] [16] . At the frequency of the main resonance of the tonearms (approximately 13 Hz), the noise suppression was reduced to −5 dB [85] [16] . To protect the bass reflex speakers, which are extremely sensitive to the passage of infrasound, this was not enough; it is no coincidence that this type of speaker became widespread only after CDs replaced vinyl [16] . Another specific problem for the 1970s and 1980s was the need to use electrolytic capacitors in the feedback circuit. In those years, capacitors of the required nominal size had an unacceptably high scatter in the initial capacity (−20% ... + 50%), and introduced audible distortions into the audio signal [13] .

The Neumann Pole

In 1995, among lovers and developers of equipment, it was claimed that, after filing the manufacturer of recorders , an additional pole with a time constant of 3.18 ms (cut-off frequency 50.0 kHz) was introduced into the standard anti-RIAA function . According to Keith Howard’s investigation from Stereophile magazine, Allen Wright, Distinguished Australian Electronics Engineer, was the first to report the “news”; after him the news was repeated by the no less authoritative Jim Hegerman [83] . Soon, preamplifier manufacturers supplemented their devices with a circuit that “compensated” for the allegedly used “Neumann pole” recording. Its influence on the frequency response was small (+0.64 dB at 20 kHz), but it could introduce a significant, audible phase error in the upper octave of the sound range [83] . Worse, the ultrasonic click components amplified by this circuit could overload subsequent amplification stages and acoustic systems [40] .

In fact, the “Neumann pole” never existed [40] [13] . The real Butterworth filter used by this company only protected the cutter drives from high-frequency interference. The cutter itself, in principle, was not able to record frequencies lying above the frequency of its own resonance (22 kHz) [40] [86] .

Implementation Examples

Typical RIAA OA corrector circuit
with switchable “IEC amendment”

 
Schematic diagram [87] . The input circuit (Rin = 47 kOhm according to DIN 45547) is shown schematically and in the general case requires individual adjustment for the used pickup [5]
 
Frequency response. Green dotted line - frequency response at the output of the op-amp (without low-pass filter R4C4)

RIAA

Frequency correction when playing records can be implemented traditionally, by analog filters, or in the digital domain. For example, in 2005, the Audacity program included 12 historical correction schemes, including the standard RIAA scheme [88] . For high-quality sound reproduction, according to 2008, digital signal processing was unsuitable; the prospect of switching to digital correction appeared only with the introduction of 24-bit ADCs [89] . In serial preamps-corrector, traditional analog filters are still used - both passive and active filters with frequency-dependent feedback circuits. Passive circuits require larger signal amplitudes, greater overload margin, higher supply voltages, they are extremely sensitive to the input resistance of the load of frequency-dependent circuits [90] [91] . These requirements are easily met in vacuum tube amplifiers, and active filters prevail in transistor devices [90] [91] .

Of the many configurations of active filters, most authors recommend a circuit on a single low-noise operational amplifier (op amp) in a non-inverting switch [92] [86] [91] ; when executed as a separate device, it is usually supplemented with an output voltage follower , and when a low-sensitivity pickup with a moving coil is connected, with an input amplification stage or a step-up transformer [93] . The alternative op amp circuit in inverting switching, popular in the 1970s, has an unrecoverable disadvantage - the noise level is about 14 dB worse - and therefore is practically not used [94] . In the past, similar circuits were widely used on specialized low-noise ULF specialized sound microcircuits (for example, LM381 and its clone K548UN1), but as the sound equipment sales fell, these ICs were discontinued, and the designers were forced to return to universal op amps [95] .

There are four basic, equivalent configurations of the frequency-dependent feedback loop (R1C1R2C2) covering the opamp. In the above version (“Lipschitz chain A”) R1C1 = T 1 = 3180 μs, R2C2 = T 2 = 75 μs, (R1 || R2) (C1 + C2) ≈T 3 = 318 μs [96] . The capacitance C0 together with R0 forms an HPF not provided for by the standard with a cutoff frequency of 3.3 Hz, which prevents the amplification of the bias voltage of the op-amp; switchable high-pass filter “IEC amendments” R3C3 is passive. Since the gain of the op amp in a non-inverting switch never drops below unity, in order to suppress the passage of ultrasonic frequencies to the output, a passive low-pass filter R4C4 with a cutoff frequency of 63 kHz is additionally introduced into the circuit [87] . To compensate for the attenuation introduced by this filter in the audio range, the time constant (R1 || R2) (C1 + C2) was chosen somewhat different from the standard 318 μs.

In a high-quality preamp-corrector, the overload margin should be at least 28 dB at sound frequencies and at least 34 dB at ultrasonic [97] . To fulfill this condition, the gain of the given circuit is set to the minimum possible, only 30 dB at 1 kHz [87] . To reduce the Johnson noise of the resistances, their values ​​are chosen as low as the output stage of the op amp allows [32] . In the worst case, when ultrasonic frequencies are amplified, the op-amp load resistance drops to R0, which should not fall below the acceptable for an op-amp. In the above example, the value of R0 (220 Ohms) is selected according to the standard series E3 ; its derivatives R1, C1, R2, and C2 inevitably have non-standard values [32] . When choosing the closest values ​​from the standard series E12, the frequency response deviation from the standard, excluding technological variation, is 0.7 dB; for the E24 series, it decreases to 0.12 dB and only when using the components of the E96 series does it reach acceptable 0.06 dB [98] . The best (but also the most expensive in serial production) solution is an individual selection of R1, C1, R2, and C2 from standard resistance and capacitances connected in parallel [32] .

Anti-RIAA Filters

Passive anti-RIAA filter

 
The classical scheme of Williamson (1971), with the specified capacities and resistances specified by Lipschitz and Jung (1980). The attenuation of the generator signal at a frequency of 1 kHz is 44.1 dB (160.3: 1) [99]

For debugging and checking the frequency response of the preamp corrector, oscillating frequency generators (GCF) with frequency response identical to the standard frequency response of the RIAA recording channel are used. In the XXI century, specialized digital generators with the possibility of external frequency response programming [100] are best suited for this task. In amateur practice, analog “anti-RIAA filters” are still used, which are connected between the output of a normal GKCh and the input of a preamp-corrector. These filters, like the correctors themselves, can be active or passive, with a frequency-dependent circuit concentrated in one cascade or with cascading filtering. From the point of view of the convenience of fine-tuning the frequency response, passive circuits with cascade filtering are preferred, in which each frequency-dependent first-order link is isolated from the next link by a voltage follower with a high input resistance [101] . From a cost point of view, concentrated passive filters similar to the R0R1C1R2C2 circuit from the preamp-corrector circuit shown are preferable [99] . When using high-quality, thermostable components with an acceptable deviation from the nominal no worse than ± 1%, the maximum deviation of the frequency response of the circuit from the standard is about ± 0.2 dB [99] . The best accuracy is achievable only by tuning the filter using professional measuring instruments [99] , while the cost of precision capacities and resistances can reach prohibitively high values [100] .

Comments

  1. ↑ In the literature, decimal fractions are usually not given. In practice, they are not significant (the rounding error is imperceptible by ear), but it is fractional frequencies that are standardized - derivatives of integer time constants.
  2. ↑ The thickness of the lacquer layer is 0.15 mm, the thickness of the aluminum base is 0.5-1.0 mm [19] .
  3. ↑ The combination of nitrocellulose, which was introduced into the practice of studios in 1934, and a forcedly heated cutter was and remains fire hazardous, but there was no replacement for nitrocellulose in studio recordings. Safe, but noisy substitute compositions were used only in household recorders [20] .
  4. ↑ The three main faces of the incisor are the front working surface and two symmetrical rear surfaces. In addition, two chamfers are removed between the working and rear surfaces, which forms two narrow polishing faces [22] .
  5. ↑ Neumann, Ortofon and other manufacturers chose precisely helium (gas, not liquid helium) for its high specific heat , which minimized the mass of the heat carrier compared to conventional air cooling [24] [25] and increased the efficiency of heat removal. For example, in Ortofon DSS732 recorders, replacing air with helium allows increasing the current of the recording coil from 0.8 to 1.0 A [26] .
  6. ↑ The standard width of the recording zone is 86 mm [35] . With a step between the grooves of 200 μm, 430 grooves fit on it, with a step of 65 μm - 1320 grooves.
  7. ↑ A mixture of related concepts of tracing and tracking is common in English literature. The first of them relates to the rounding of the microscopic displacements of the groove by the needle (non-flexing errors), the second to the accuracy of the orientation of the pickup needle (angular errors) [50] .
  8. ↑ For attempts to standardize such a standard and its connection with real recordings, see Elyutin, A. Diet for speakers. The spectrum of the music signal. // Auto sound. - 2001. - No. 11 . - S. 34–42 .
  9. ↑ Hoff expresses the same relationship asω3/2 {\ displaystyle \ omega ^ {3/2}}   , that is, 9 dB per octave [53] .
  10. ↑ Western Electric designers were the first to curb the fatal resonance of the cutter, usually lying in the region of 2 ... 10 kHz, using rubber shock absorbers . However, natural rubber quickly aging, losing damping properties, which gave rise to inevitable shifts in the frequency response of the recorder [59] .
  11. ↑ Alan Blumlane applied this scheme, but was not its author. It is not known whether he used precisely the frequency of 250 Hz, and not any other. The main merit of Blumlein was the development of an electromagnetic system for the depreciation of the cutter, which became the de facto European standard [60] .
  12. ↑ This decline was characteristic of “cold” incisors. Forced heating of the incisors, which eliminated this drawback, was introduced only in the 1950s [8] .
  13. ↑ In 1958, it was Blumlein’s patent that became the basis of the stereo recording standard. None of the solutions competing with him reached the serial production [65] .
  14. ↑ Copland gives an example of an original varnish disc marked immediately by three mutually exclusive systems: AES, CCIR and Orthophonic. In fact, it was recorded according to the RIAA standard [57] .
  15. ↑ Moreover, the plates themselves, turntable drives and electromagnetic pickups of that time have already reached a fairly high level [79] .

Notes

  1. ↑ 1 2 3 4 5 6 7 Moyer, HC Standard Disc Recording Characteristic // RCA Engineer. - 1957. - Vol. 3, No. 2 . - P. 11-13.
  2. ↑ 1 2 Jones, 2012 , p. 586.
  3. ↑ Vogel, 2008 , p. eleven.
  4. ↑ Vogel, 2008 , p. 12: "this is nothing else but sequence of ..." (for inverse playback function).
  5. ↑ 1 2 3 4 5 Vogel, 2008 , pp. 11-12.
  6. ↑ 1 2 3 Galo, 1996 , p. 46.
  7. ↑ 1 2 3 4 Galo, 1996 , p. 48.
  8. ↑ 1 2 3 Galo, 1996 , p. 49.
  9. ↑ 1 2 3 Apollonova and Shumova, 1978 , p. 50.
  10. ↑ 1 2 3 Vogel, 2008 , p. 12.
  11. ↑ Self, 2010 , p. 169.
  12. ↑ 1 2 3 4 5 Galo, 1996 , p. 50.
  13. ↑ 1 2 3 4 5 6 Self, 2010 , p. 167.
  14. ↑ Self, 2010 , p. 167: "the most popular cutting amplifier."
  15. ↑ Vogel, 2008 , pp. 12-13.
  16. ↑ 1 2 3 4 5 6 7 8 Howard, 2009 , p. one.
  17. ↑ Vogel, 2008 , p. 13.
  18. ↑ Eargle, 2012 , Fig. 10.15.
  19. ↑ Apollonova and Shumova, 1978 , p. 112.
  20. ↑ Copeland, 2008 , p. 51.
  21. ↑ 1 2 3 4 Capel, 2013 , p. 52.
  22. ↑ Apollonova and Shumova, 1978 , p. 102-103.
  23. ↑ Apollonova and Shumova, 1978 , p. 104.
  24. ↑ Apollonova and Shumova, 1978 , p. 97.
  25. ↑ Jan Szabo. Cutting it Close (unopened) . Ensemble HD (2013). .
  26. ↑ Apollonova and Shumova, 1978 , p. 95.
  27. ↑ Copeland, 2008 , pp. 66, 67, 111, 119.
  28. ↑ Eargle, 2012 , Ch. 10.4.2.
  29. ↑ Apollonova and Shumova, 1978 , p. 72, 88.
  30. ↑ 1 2 Sapozhkov, 1989 , p. 226.
  31. ↑ Sapozhkov, 1989 , p. 223.
  32. ↑ 1 2 3 4 5 6 7 8 9 Self, 2010 , p. 165.
  33. ↑ Apollonova and Shumova, 1978 , p. 77.
  34. ↑ Copeland, 2008 , p. 214.
  35. ↑ Sapozhkov, 1989 , p. 227.
  36. ↑ 1 2 3 4 5 Apollonova and Shumova, 1978 , p. 45.
  37. ↑ 1 2 3 4 5 6 Arshinov, V. Records. State Standards // Radio. - 1977. - No. 9 . - S. 42-44 .
  38. ↑ Eargle, 2012 , Ch.10.9.2.
  39. ↑ Apollonova and Shumova, 1978 , p. 216.
  40. ↑ 1 2 3 4 Howard, 2009 , p. 3.
  41. ↑ 1 2 Apollonova and Shumova, 1978 , p. 216-217.
  42. ↑ Eargle, 2012 , Fig. 10.1.
  43. ↑ 1 2 Self, 2010 , p. 212.
  44. ↑ 1 2 Apollonova and Shumova, 1978 , p. 42.
  45. ↑ Apollonova and Shumova, 1978 , p. 43-44.
  46. ↑ 1 2 3 4 5 Sapozhkov, 1989 , p. 225.
  47. ↑ Vogel, 2008 , p. five.
  48. ↑ Self, 2010 , p. 211.
  49. ↑ Sapozhkov, 1989 , p. 224.
  50. ↑ Copeland, 2008 , p. 43.
  51. ↑ 1 2 3 4 Copeland, 2008 , p. 99.
  52. ↑ Apollonova and Shumova, 1978 , p. 46.
  53. ↑ Hoff, 1998 , p. 128.
  54. ↑ Sapozhkov, 1989 , p. 225-226.
  55. ↑ 1 2 Hoff, 1998 , p. 129-130.
  56. ↑ Copeland, 2008 , p. 153.
  57. ↑ 1 2 3 Copeland, 2008 , p. 101.
  58. ↑ 1 2 3 4 5 6 Eargle, 2012 , Ch. 10.1.
  59. ↑ Copeland, 2008 , pp. 113-114.
  60. ↑ Copeland, 2008 , pp. 104-105, 127.
  61. ↑ Copeland, 2008 , pp. 104-105.
  62. ↑ 1 2 3 Copeland, 2008 , pp. 101-102.
  63. ↑ 1 2 3 Galo, 1996 , p. 47.
  64. ↑ Copeland, 2008 , p. 157.
  65. ↑ Copeland, 2008 , p. 57.
  66. ↑ 1 2 3 4 5 Copeland, 2008 , p. 155.
  67. ↑ Copeland, 2008 , p. 152, 155.
  68. ↑ 1 2 Copeland, 2008 , pp. 155-156.
  69. ↑ Copeland, 2008 , pp. 153-154.
  70. ↑ Copeland, 2008 , p. 154.
  71. ↑ 1 2 Copeland, 2008 , p. 100.
  72. ↑ 1 2 Copeland, 2008 , p. 156.
  73. ↑ Copeland, 2008 , p. 158.
  74. ↑ Copeland, 2008 , pp. 158-159.
  75. ↑ Copeland, 2008 , pp. 150, 151.
  76. ↑ 1 2 3 Copeland, 2008 , pp. 148.
  77. ↑ Copeland, 2008 , pp. 148, 150.
  78. ↑ Copeland, 2008 , pp. 150, 158-159.
  79. ↑ 1 2 3 4 Jones, 2012 , pp. 591-592.
  80. ↑ Self, 2010 , p. 184.
  81. ↑ Self, 2010 , p. 187.
  82. ↑ Self, 2010 , p. 186.
  83. ↑ 1 2 3 4 5 6 Howard, 2009 , p. 2.
  84. ↑ Self, 2010 , p. 175.
  85. ↑ 1 2 3 4 5 6 Self, 2010 , p. 166.
  86. ↑ 1 2 Self, 2010 , p. 168.
  87. ↑ 1 2 3 Self, 2010 , p. 170.
  88. ↑ Fries, B. Digital Audio Essentials. - O'Reilly, 2005 .-- P. 269-271. - ISBN 9780596008567 .
  89. ↑ Copeland, 2008 , pp. 39-40.
  90. ↑ 1 2 Vogel, 2008 , pp. 228-230.
  91. ↑ 1 2 3 Jones, 2012 , pp. 599.
  92. ↑ Jung, 2005 , p. 6.17.
  93. ↑ Vogel, 2008 , pp. 6-7.
  94. ↑ Self, 2010 , p. 171.
  95. ↑ Hood, JL Audio Electronics. - Newnes, 2013 .-- P. 127. - ISBN 9781483140803 .
  96. ↑ Lipschitz, 1979 , Fig. 1.
  97. ↑ Jones, 2012 , p. 594.
  98. ↑ Self, 2010 , pp. 164-165.
  99. ↑ 1 2 3 4 Lipschitz, S. and Jung, W. A High Accuracy Inverse RIAA Network // The Audio Amateur. - 1980. - No. 1 . - P. 23.
  100. ↑ 1 2 Self, 2010 , p. 179.
  101. ↑ Self, 2010 , p. 178.

Sources

  • Apollonova, L.P. and Shumova, N.D. Mechanical recording. - 2nd ed. - M .: Energy, 1978.
  • Sapozhkov, M. A. Acoustics: A Handbook. - M .: Nauka, 1989 .-- ISBN 52560001876.
  • Capel, V. Newnes Audio and Hi-Fi Engineer's Pocket Book. - Newnes / Elsevier, 2013 .-- ISBN 9781483102436 .
  • Copeland P. Manual of Analogue Sound Restoration Techniques - British Library , 2008.
    <a href=" https://wikidata.org/wiki/Track:Q29471641 "> </a> <a href=" https://wikidata.org/wiki/Track:Q23308 "> </a> <a href = " https://wikidata.org/wiki/Track:Q7173383 "> </a>
  • Eargle, J. Handbook of Recording Engineering. - Springer, 2012 .-- ISBN 9789401093668 .
  • Galo, G. Disc Recording Equalization Demystified // ARSC Journal. - 1996 .-- P. 44-54.
  • Hoff, PH Consumer Electronics for Engineers. - 1998. - (Wiley Series in Practical Strategy). - ISBN 9780521588171 .
  • Howard, K. Cut and Thrust: RIAA LP Equalization // Stereophile. - 2009. - No. Mar 2, 2009. - P. 1-4.
  • Jones, M. Valve Amplifiers. - Newnes / Elsevier, 2012 .-- ISBN 0750656948 .
  • Jung, W. Op Amp Applications Handbook . - Analog Devices / Elsevier , 2005. - ISBN 0916550265 . - ISBN 0750678445 .
  • Lipschitz, SP On RIAA equalization networks // J. Audio Engineering Society. - 1979. - Vol. 27. - P. 458–491.
  • Self, D. Small Signal Audio Design. - Focal Press / Elsevier, 2010 .-- ISBN 9780240521770 .
  • Vogel, B. The Sound of Silence: Lowest-Noise RIAA Phono-Amps: Designer's Guide. - Springer, 2008 .-- ISBN 9783540768838 .
Source - https://ru.wikipedia.org/w/index.php?title=RIAA curve_&oldid = 95283149


More articles:

  • Yelnya (Noginsk district)
  • Molokova (Municipality Alapaevskoye)
  • Water polo at the 2016 Summer Olympics
  • Hold On Pain Ends
  • Borshosh-Kumyatsky, Julius Vasilievich
  • Suslonger City Settlement
  • Stripes (Leningrad Oblast)
  • Girac, Kenji
  • Australia at Junior Eurovision
  • So Hyang Song

All articles

Clever Geek | 2019