Historical Recording Characteristics

History

The need to alter the frequency response of the audio programme when cutting a record has been recognised since the dawn of electric gramophone (or phonograph) discs in the mid-nineteen-twenties. The choice of frequencies at which to cut and at which to boost the audio as it's recorded onto the disc is known as the recording characteristic. And there were many!

The seemingly deeply confusing situation concerning different recording characteristics and when and where they were employed is simplified by having a historical perspective. Like most complicated situations, it didn't start out complicated: it evolved into something complicated, and ultimately pretty chaotic at the point that LPs and shellac records co-existed, before it was brought under control by the, now ubiquitous, RIAA recording characteristic.

The 1930s

In the case of the early, electric 78 RPM records developed first by Western Electric and later by British Columbia Gramophone (the latter system was developed by Blumlein to circumvent paying royalties to Western Electric after the Wall Street crash), a simple RC bass-cut circuit was employed before the cutter amplifier so that the recorded groove was recorded with constant-velocity above 300Hz and constant-amplitude below. This reduced the size of the grooves thereby avoiding over-cutting, and it increased the playing time of the disc. This type of recording characteristic could be defined by a simple time-constant; although the fashion was to use frequency as the parameter. This, in record-collecting circles is usually referred to as the turnover and is expressed in Hertz (or cycles per second).

For the designer of the electrical reproducer, the task was simple too: to introduce a complementary characteristic by means of a bass-boost filter; entirely defined by the turnover frequency of the recording filter. With one difference, whilst a filter may easily be arranged to "cut" frequencies with no limit: it isn't possible to have an infinite "boost". So the compensating bass-restoration filter on the reproducer side had to be a shelving type with the shelf determined by the maximum boost at a certain frequency. We'll see later that this had ramifications later on.

The 1940s

The situation above lasted about ten years, until the late years of the nineteen-thirties when high-frequency pre-emphasis started to be included in the recording chain. The idea was that, by boosting the treble signals a better signal-to-surface-noise was secured when a complimentary de-emphasis was applied in the record player. Thus the typical recording characteristic took on the form in the picture below: with a mid/low frequency turnover and high-frequency pre-emphasis.

There is some difficulty in determining the date at which recording pre-emphasis started to be employed in the USA. But Russell's article (Stylus in Wonderland, by O.J. Russell, Wireless World October 1954) gives us an important data points. Firstly, he tells us that the idea of pre-emphasis for phonograph recordings came from FM broadcasting (where it is always employed). He also says,

Accordingly, the recording characteristic of Fig. 1 [with a simple turnover and no pre-emphasis] was all the pre-war " hi-fi " addict bad to worry over.

(Russell is, of course speaking of the 1939 - 45 European war).

Now we know that Armstrong first demonstrated FM to the FCC in June 1936 and that the very first commercial stations were opened in January 1941. So it seems we can reasonably infer that pre-emphasis started to be employed at the tail-end of the thirties, around the time of the onset of the European war.

In Europe, only Decca embraced pre-emphasis of 78 RPM discs (as part of its FFRR recording system). EMI continued to use the original Blumlein characteristic until it discontinued production of 78 RPM records in Britain in the 1960s.


Illustration from Stylus in Wonderland, by O.J. Russell, Wireless World October 1954

Pre-emphasis too was not defined as it is in the modern way with a time-constant. Instead, it was defined as the amount of boost (or cut on replay) at a specific frequency; usually 10kHz. Thus, a recording characteristic was defined, for example, as: turnover at 250 Hz with +5 dB boost at 10kHz, or, for short, 250-5.

All recording characteristics, up until the nineteen-fifties are simple variations of these two parameters: turnover and boost at 10kHz. The Devil is in the detail that the record companies all chose different turnover frequencies. And all chose different pre-emphasis curves from 0dB at 10kHz, to +16dB at 10kHz; a staggering range! Tone controls on preamplifiers dates from this period; they were necessary not to offer tonal choice, but to make the majority of records listenable at all!

Technology and sociology

Now, the idea of cutting the bass, to improve playing time, and to pre-emphasise the treble, to cut surface-noise, are both really good ideas (given certain assumptions) and more of both were gradually applied to records as the nineteen-forties progressed; especially in America. In fact, it is not a caricature to say that the main difference between pre-RIAA records from America and those from Europe in this period is that the European labels were much more cautious concerning the degree of equalisation applied to 78 RPM records than were their American counterparts.

This wasn't simply fusty, European conservatism. America during this period wasn't a colonial power as were many of the European nations (especially Britain and France). In wealthy, domestic America, electricity had reached most homes by the 1930s and electrical reproduction of sound was the norm. Not so for European manufacturers for whom a substantial market for records lay in the far-flung outposts of the empire where mechanical gramophones remained the predominant method of reproduction well into the fifties. The European record companies were only too aware that the more equalisation applied to records, the worse they sounded on a mechanical gramophone.

Nevertheless, in the beating-heart of their massive entertainment industry, American record labels were especially enthusiastic in pursuing more bass-cut and more treble-boost until the two ideas practically "met in the middle" resulting in an almost entirely sloping characteristic with a gradual rise from bass to treble frequency extremes, with little or no level middle part; as in the following diagram.


Illustration from Stylus in Wonderland, by O.J. Russell, Wireless World October 1954

The 1950s

By the fifties, the whole greater-and-greater equalisation fetish was generally felt to have got to be too much of a good thing. The zenith was the NAB characteristics (Curve-B on the Quad 22 controller and Stereo Lab) which has a turnover of 500Hz and +16dB boost at 10kHz; a treble level so elevated that it often causes stylus tracking problems. Clearly it was time for a standard, and RCA had introduced a recording characteristic which they called the RCA New Orthophonic curve, defined at the time as turnover 500 Hz with +13.7dB boost at 10kHz. This was generally felt to be a reasonable choice of parameters and in 1952 the Recording Industry Association of America (RIAA) adopted this curve and proposed it as a standard. Conventional wisdom has it that, by the mid-fifties in a belle epoque of international cooperation, most American labels and most major European labels had adopted the new RIAA standard. However, many labels - especially in Europe and Asia - were much slower to adopt the RIAA characteristic and disc recording characteristics were not effectively standardised until the late 1960s, or possibly even later.

The bass-shelf

Remember the issue that, whilst it's possible to cut low-frequencies all the way to 0Hz when recording the disc, it isn't practical to boost all the way to DC on replay? In the early days, the choice of how much (and thus to what frequency) the boost would be correctly applied was simply the designer's choice. But as the post-war craze for "high-fidelity" got going, the need to define this became apparent. Moreover, it was appreciated that, if a shelving filter, instead of a simple low-pass filter was used on the record-side, it would be possible to match the equalisation exactly on the replay side. So a third parameter was introduced which defined the ultimate boost level of the bass-shelf.

Thus to the two numbers that defined a recording characteristic (turnover frequency and emphasis at 10kHz) was added a letter to mean:

  • X = +12dB
  • C = +14dB
  • A = +16dB
  • B = +18dB
  • R = +20dB
  • N = No bass-shelf; equalisation goes to DC!

Which is why RIAA is sometimes referred to as: 500R-13.7, meaning a 500 Hz turnover with a +20dB limit to the bass-shelf, and a 13.7 dB rolloff at 10 KHz. This nomenclature helps decode tables such as the one from High Fidelity Magazine in October 1955.

Time constants

Nowadays, the preference is not to use this code system, but to describe the recording characteristic with the time-constants of the electrical networks needed to encode or decode the characteristic. Thus, if the time-constant governing the HF pre-emphasis is t3, that governing the turnover is t2, and the that governing the corner of the bass-shelf is t1, then the overall response of the replay equaliser is given by the expression,

Which curve to use...?

Of course, this is the important question and hopefully most questions are answered simply here. However a few of the finer points for discussion are covered here, especially in relation to the characteristics of early LPs. The recording characteristics of 78rpm (shellac) records is covered elsewhere.

Research

Our research on the recording characteristic curves comes from two major historic references:

  • The Radiotron Designer's Handbook, F. Langford-Smith, Wireless Press, Sydney, Fourth edition, 1952. and
  • High Quality Sound Reproduction by James Moir, Chapman & Hall Ltd., London, 1958

as well as the relevant standards documents and a wealth of other articles and on-line resources. Including Peter Copeland's wonderful (if flawed)

  • Manual of Analogue Sound Restoration Techniques, The British Library London, 2008 and available on line from the British Library site.

Finally, our grasp of the historical perspective was greatly enhanced by the article,

  • Stylus in Wonderland, by O.J. Russell, published in Wireless World magazine, October 1954. (Our illustrations of the development of recording characteristics are taken from this article.)

In general we have always tried to source the original reference to verify any modern information.

Langford-Smith essentially presents the various recording characteristics in two sets of curves. Both of which are reproduced here.


Original (and somewhat overambitious) caption: Recording characteristics used by all English, Australian and the majority of European record manufacturers for 78 r.p.m., together with B.B.C. transcriptions.


Original caption: Recording characteristics used by most American record manufacturers.

Moir, on the other hand, tabulated his data (and included useful correction circuits which we could SPICE!) The key table (Table 7.4) from this reference is reproduced here with four columns added to cover the equalisation characteristics included in Stereo Lab and inspired from the QUAD 22 controller.

The problem with historic recordings and how to equalise them is threefold:

  • The recording characteristics weren't really standards: they were "house standards" - employed by each company (and sometimes closely guarded secrets at the time.)
  • They weren't defined, or at least communicated to the public, in any consistent way. Sometimes we have turnover and pre-emphasis figures, but sometimes all we have is a graph on log-log paper, or a recommended de-emphasis circuit. And there are lots of mistakes in the literature; especially on-line. This is the main reason we went back to the old references.
  • The standards weren't, and probably for the above reasons, very accurately specified. Normally ±2dB was considered acceptable.

There is thus the necessity to be archaeologist, detective, reader of runes and pragmatist when deciding upon entering double-precision floating-point coefficients demanded by software equalisation!

To be sure, applying the wrong equalisation can have some fairly disastrous results. The differences between some recording characteristics is not subtle. For example, a Decca 78 equalised with RIAA equalisation will give +8dB too much bass at 100Hz and 10dB treble cut at 10kHz! Clearly we need to do a lot better than that!

However, not all mismatches are quite so disastrous, and here is where the QUAD 22 circuit and table were so clever because they apply intelligent and informed compromise equalisations to cover a wide range of different characteristics. Careful analysis of the QUAD table and Table 7.4 of Moir is informative and worthwhile.

However, with only four switchable equalisations, the Quad 22 still represents something of a compromise because there were certainly more than four recording characteristics! So, we have added some of our own. Our observations concerning the original and different characteristics employed at the end of the 78 and early LP era are given below.

Click this link for specifics and details of equalisation for shellac-era records.

EMI (HMV) to July 1953

Comparing Langford-Smith's EMI curve with Moir's tabulated data for HMV 78 we get a reasonable match. This is the old Blumlein 300-0 characteristic which is the appropriate equalisation for the reproduction so many European shellac-era records. Setting EQ-G in Stereo Lab matches this characteristic exactly. We believed 300-0 was a "78-only" characteristic, but Copeland (2008) informs us that this characteristic was used for LPs and 45s until mid July 1953.

BSI 33 & BSI 78 (500R-13.5 and 350-10.5)

Moir doesn't attribute these characteristics to a particular standard. But, from the dates, it must relate to BS. 1928:1955, that for Microgroove and Coarse-Groove records. This is the standards document which introduced 3180uS/318uS/75uS or RIAA equalisation for LPs, and 3180uS/450uS/50uS Coarse-Groove equalisation for 78s. The slight differences in the table compared with the values in the columns RIAA and Microgroove are thought to be calculation errors. This was, after all the age of log-tables and slide-rules (the time-constants referred to in the standard are identical).

NAB (500B-16, often simply annotated 500-16)

Here we have excellent agreement between references. We can therefore be confident of this characteristic and subsequent equalisation. Quad's Curve-B is exactly the NAB characteristic; as is the equalisation in Stereo Lab. Record manufacturers who used this standard included: Artist, MGM, Tempo and Westminster (sometimes, see dust jacket). The National Association of Broadcasters (NAB) changed their name to The National Association of Radio and Television Broadcasters (NARTB) for a short time after the introduction of television broadcasting. But they reverted to NAB again in 1958. This is why you will sometimes see the NAB characteristic called the NARTB characteristic. Note that this curve has its origins in equalisation of transcription discs used at radio stations. The NAB curve thus only ever applied to lateral-cut 33⅓ RPM discs: it was not applied to 78 RPM records.

Note: The NAB characteristic employs time constants of 3180µS, 318µS and 100µ exactly as the QUAD 22's B-curve. This truth is obscured by NAB's decision to put the 0dB reference at 700Hz rather than 1kHz which confuses everyone and is what appears to skew the table values slightly from Quad Curve-B.

RCA 33, 45 and 78

Here we have excellent agreement between Moir and Langford-Smith, but the frustration as to when and to what this characteristic was applied! It appears to be a curve sometimes referred to as Old Orthophonic (500N-12.7) which was possibly applied to 78s from 1947 - Aug 1952. It is possible it was applied to 33⅓ and 45 RPM recordings. In any case, it is very close to RCAs "New Orthophonic" standard of 1952*, so use RIAA (EQ-A) for all 33⅓ and 45 RPM RCA records.

*We think New Orthophonic was a "tidy-up" of Old Orthophonic. The Old curve (500N-12.7) fails to specify a bass-shelf and this is supicious because the replay equaliser (as explained above has to implement a limit to the bass boost.

AES (400-12)

AES was not a recording characteristic at all: it's a replay standard. The American Audio Society's laudable idea was to try to get playback equipment manufacturers to standardise on a playback characteristic, thereby forcing the record manufacturers to "sort out their act" on the recording side. Their plan worked well, at least in the USA, and American manufacturers embraced this idea with vigour and many American labels cut their LP records to this standard prior to adopting RIAA as the Quad table and the Pspatial Audio Equalisation Guide indicate. Here, Moir lets us down as his tabulated values for AES are wrong, and don't even agree with his own published curve. Happily, other, older references survive.

The characteristic is, in fact, very close to RIAA which was doubt one of the factors helping to encourage the adoption of the RIAA standard when it was eventually proposed (see below). By way of proof, the a graph of the error in the volume/frequency characteristic of an AES disc when replayed with the RIAA curve is shown below. The results are inside the AES standard limits of ±2dB. Interestingly, the results are much worse if an AES encoded disc is replayed with the Coarse Groove de-emphasis, so it remains a mystery why Quad recommend EQ-C for discs cut to the AES standard.

Nevertheless, the errors when an AES disc is replayed with RIAA are not inconsequential. And this is a big deal when you consider that AES was the most widespread characteristic used by American record manufacturers after the NAB standard. The treble attenuation caused by RIAA replay is certainly perceptible. For that reason, Pspatial Audio have developed a specific AES de-emphasis algorithm (EQ-A3) which will guarantee the best sound from these discs.

Here's a final thought...... As the graph of the error in the volume/frequency characteristic of an AES disc when replayed with the RIAA curve illustrates, the results are inside the typical standard limits of ±2dB. This would have meant it would have been possible for cutting amplifiers equalised for AES (400-12) to have been left entirely un-modified and the claim to have been made that any discs were cut to the "new standard". Certainly, experiments made with discs cut well after the claimed RIAA implementation date often sound better with AES de-emphasis.


The frequency response of an AES equalised record replayed with RIAA curve. (Idealised response; this was obtained by subtracting one transfer function from the other).

American Columbia LP microgroove (500C-16)

The Columbia LP characteristic is the NAB curve with a revised bass-shelf time-constant; from 3180µS (NAB) to 1950µS (Columbia). This change to the bass-shelf is the origin of the term 500 modified or 500M you will see in older equalisation guides. HMV 33s produced in England were cut using this standard, as were LP records by Vanguard, Bach Guild, Cetra and Vox. If Columbia characteristic is replayed with the standard NAB curve (as with the Quad 22), the bass is too heavy. Pspatial Audio have therefore introduced a variation of NAB setting (EQ-B2) which is optimised for records cut using the Columbia characteristic.

CCIR or Coarse Groove (350-10.5)

Drafted in Geneva in June 1951 and agreed in London in 1953 for the international exchange of radio programmes on disc, this characteristic is usually simply called CCIR. It applied to 78s and microgroove recordings of all formats. However, this characteristic is important because it was used on all the great pre-1955 Deutsche Grammophon LPs as well as on (British) Philips and Caedmon LPs, and thereby an archive of important recordings. Telefunken continued to use this characteristic until 1962. With Time constants of 3180µs*, 450µs (350Hz), and 50µs (3180Hz), this is identical to the British standard Coarse Groove characteristic (EQ-C).

*In truth, the original CCIR standard do not have a bass-shelf, this was added by DGG and the other record companies.

London (usually meaning 500C-12)

Here we have a problem with classification, because the London name was adopted for different standards at different times. No London characteristic appears in the Langford-Smith's European curves or in Moir's table. But Langford-Smith describes a characteristic which he calls London in the text of which we have no other trace. Nevertheless, general useage for this characteristic is to describe the Decca 33⅓ characteristic tabulated by Moir. Quad suggest EQ-A (RIAA) for British LPs recorded with this characteristic. It's true this gives a pretty good match at high-frequency, but the bass is over-boosted. Pspatial Audio's EQ-A2 is a modified version of Quad setting A which is optimised for London records. A German language reference (kindly provided to Pspatial Audio by Wolfgang Leister) FUNKSCHAU, vol. 15 and vol. 16 of 1958 confirms that this 500C-12 characteristic is the London curve.

RIAA, RCA New Orthophonic, IEC No.98 (500R-13.7)

The RIAA standard equalisation is applied upon selecting EQ-A. This characteristic covers the vast collection of vinyl records from the middle sixties forward.

RCA New Orthophonic, IEC No.98 are alternative names for exactly the same recording equalisation, as is 500R-13.7. It's true that the later IEC standard added an additional LF breakpoint, but this was only for reproduction. See Pspatial Audio's position on the, so called, IEC Amendment.

Enhanced RIAA

The origins of the, so called, Enhanced RIAA characteristic are explained here. The term is entirely a misnomer and Stereo Lab does not include this characteristic for these reasons.

Decca Recording characteristics

We first believed that Decca obsessively fiddled with the equalisation of their records as many references quote at least three characteristics before the 1955 adoption of RIAA. However, these mysterious recording-characteristics, for which little no historical evidence seems to exist, are based on misinterpretations of the evidence as we'll show here.

Decca (often called 250-5, 3dB slope)

The curve published by Langford-Smith (which he refers to as the 1949 standard) is badly drawn: with sharp points of inflection from the flat, central section. Nevertheless, the curve and Moir's tables match well confirming that this is good data. Moir too has a version of this graph, and showing the same, suspicious idiosyncrasies, along with a recommended replay filter. It is probably these two drawings (and, perhaps a common predecessor) which have caused all the problems.

You will find everywhere modern references to a 1949 Decca characteristic of 640µS/53µS with a 3dB/octave pre-emphasis. But there are several problems with the 3dB/octave theory. A 3dB/octave filter is not an easy thing to construct and, whilst this wouldn't have been impossible at all on the recording side, it's hard to imagine how Decca would have envisaged the listener at home profiting from adopting a 3dB/octave characteristic. And, if it had existed, wouldn't it have been interesting enough that contemporary books such as Moir's or Langford Smith's would have covered it?

The key to the mystery is this......

Someone, whose identity is lost in the sands of time, interpreted these inflection points in the Decca 78 curve as the time-constants. And this is the origin of the entirely fictitious Decca equalisation dubbed 640µS/53µS. This faux-pas is highlighted in our version of Moir's curve. The indicated frequencies are not related to the time constants: they relate to the point at which the curve departs from the flat, central section. The coincidence is simply too marked for this not to be the logical explanation! Unable to explain why pre-emphasis which onsets at 3kHz only reached +5dB at 10kHz (when you would expect 11.5dB) gave rise to an erroneous belief in the mysterious 3dB/octave Decca curve!

In fact, there is no mystery. All Decca's 78s are equalised with a pre-emphasis which onsets at 6.1kHz (time-constant of 26µS), a boost which is entirely consistent with a 5.6dB rise at 10kHz; exactly as illustrated. Moir's de-emphasis circuit further confirms this theory illustrating as is does that the HF pole is defined by the 16k resistor in series with a source impedance of 10k in combination with a 1nF capacitor. (The reactance of the 100n is capacitor insignificant at these frequencies). Thus,

26,000 × 1 × 10-9 = 26µS.

Finally, to guild the lily, here is the SPICE frequency characteristic of Moir's de-emphasis circuit which confirms all the above (the 3dB points are indicated).

On the basis of Langford-Smith's curve, Moir's tabulated data and curve and this SPICE analysis we reject the existence of a 3dB/octave characteristic and recommend EQ-H for all Decca 78s.

Decca LPs (500C-11.5 in UK)

Some very early Decca LPs (only those before June 1950) used the 78 characteristic described above. After this date Decca used the Decca 33⅓ characteristic which is tabulated by Moir and which we could label 500C-11.5. Both the turnover (a perfect 3dB at 500c/s) and the pre-emphasis are as clear as day from the table. We have determined the time-constants of this curve to be 1590/318/60µS. The Quad 22 table refers to Decca UK LPs and suggests RIAA. It's true this gives a pretty good match at high-frequency, but the bass is over-boosted because of the 100Hz bass-shelf in the 500C-11.5 characteristic . Pspatial Audio's EQ-A2 is a modified version of Quad setting A which is optimised for Decca records. There is some evidence that Decca in the US used various characteristics: including 500C-10.5, 400N-12 and 500B-16 so EQ-A, A1, B or C may be applicable in certain cases.

TELDEC (500R-10.5)

Decca also founded a record company with Telefunken called Telefunken-Decca Schallplatten GmbH with TELDEC as the label principally to release classical recordings with the Bamberg Symphony Orchestra, the Berlin Symphony Orchestra and the Berlin Phil'. Important recordings therefore! In 1958 Funkschau magazine reported that TELDEC had opted to manufacture their discs to a German DIN-Standard recording characteristic. If this is true, this was particularly outré given that this was several years after the lauded, so-called "standardisation" of the industry on RIAA. Certainly the German Standards Authority DIN did try to introduce this standard (called DIN 45533) in 1957 as a kind of "European RIAA" but it was never formally adopted as a standard. The time constants of this proposed DIN standard were 50, 318 and 3180 microseconds: close to RIAA, but different enough for recordings recorded with the former and played back with the latter to sound a little dull.

Certainly, the TELDEC technical team writing in Audio Magazine in November 1958 present results (in this case for vertical tracing distortion) which indicate quite clearly that they were using 50uS pre-emphasis during the development of the stereo cutter-head system in conjunction with Neumann.

The RIAA Wikipedia article says this of this characteristic, "The extent of usage of this curve is unclear" which is probably about right. But it may have been quite extensive. Deutsche Grammophon made at least one stereo test disc using this standard and we personally believe that all DGG recordings - well into the 1970s - used this equalisation. It's quite possible that this choice accounts for the poor reputation that DGG discs have for sound quality compared with discs from DECCA and HMV. If replayed with RIAA equalisation these discs lack "lustre" and "life".


Deutsche Grammopon records are usually regarded by record collectors to lack the sound quality of recordings by DECCA or HMV. But this is possibly due to the DIN recording characteristic being used instead on RIAA

In order correctly to equalise these recordings, Pspatial have made the appropriate modification of the RIAA characteristic in Stereo Lab called EQ-A3. This is the de-emphasis we recommend for TELDEC discs, post 1955 DGG discs and possibly other German-cut LPs of the same vintage.

I'm not sure what recording characteristic was applied to a certain record. How can I conveniently audition the various equalisations to compare them and see which sounds best?

It is possible to audition any converted file directly from the application by right clicking on the file as it appears in the main window. This is invaluable when comparing different conversions; for example of different phonograph equalisations. Simply make a version of each of the candidate equalisations and right click to hear each conversion.

If you would like to see other, characteristics included in the Stereo Lab software, please contact support@pspatialaudio.com.


Appendix - Relationship between Time Constants and attenuation at 10kHz


A picture paints a thousand words...


Links

Home page

For all support issues, go here.

For Pspatial Audio sales, email: sales@pspatialaudio.com



© Pspatial Audio 2015 - 2017. All rights reserved. Apple Certified Developer. Stereo Lab, Aria 51, Aria 20, Head Space, Groove Sleuth, iLOOP and FRANCINSTIEN T-Sym are trademarks of Pspatial Audio. FRANCINSTIEN and Bride of FRANCINSTIEN (BoF) are trademarks of Phaedrus Audio. Macintosh and the Mac logo are trademarks of Apple Computer, Inc.