How much did ADCs in the 80s affect the sound quality of CDs?

[Hayabusa]

reel to reel, not real to real.

please!! its like kindergarten allover again

 

[AnalogSteph]

I remember many CD's which sucked compared to their cassette releases!

I guess sometimes two wrongs do make a right. The late '80s in particular were the time of exaggerated treble in pop/rock recordings, and it kind of was en vogue in speakers as well. Tape saturation taking the edge off just enough under these circumstances is entirely believable.

Once you know what you're starting out with, it becomes obvious that old analog formats aren't exactly super faithful to the original (if occasionally flattering):

 

[MaxwellsEq]

The late '80s in particular were the time of exaggerated treble in pop/rock recordings

SSL consoles with compressor and/or gate.

 

[JIW]

The thing is, as long as your DAC can comfortably drive your power amp into clipping (which they will usually do), you might just as well be applying extra gain on the digital side if need be. This is exactly what ReplayGain is for.

The real tough nuts are those recordings with super high crest factor. I think my record holder here is a recording of Mahler's 4th from 1998 (Birmingham Symphony, Simon Rattle)... album gain +7.83 dB but peak at pretty much exactly 1.00, peak / average nearly 26 dB, DR18. Reproducing that at 85 dB SPL average without clipping would take quite the system... even with 88 dB / W / m speakers you would be looking at a 250 W amp.

Anyway, I have a bit of a soft spot for early digital recordings and will pick them up whenever I happen to come across them. They can be really good - there is absolutely nothing to suggest that Beethoven's Fidelio with the Chicago Symphony under Solti dates from 1979, for example (recorded with DECCA system). Even in those days, there were machines recording at 16 bits @ 50 kHz (Soundstream, 3M) or 18 bits @ 48 kHz (DECCA). (Hence why SRCs were becoming necessary by the time the CD took off. Did I mention Denon with 47.25 kHz?) The one big classical label with rather mixed results in those days supposedly was DG... they would pick up the slack in later years and by the early '90s had highly advanced recording systems of their own. EMI seems to have had good results even with the comparatively ancient Sony PCM 1600 in the early days... a model which seems to have been quickly superseded by the 1610 or the multitrack DASH PCM 3324 (which could also do 48 kHz).

Speaking of Sony, the later PCM 1630 (1986) was the first model to explicitly generate wideband noise to dither the ADC, so if you find studio recordings using the 1610 that sound crunchy at low levels that's probably why. (The fade-in at the start of Peter Gabriel 4 ("Security") comes to mind.) Acoustic recordings tended to have plenty of analog noise by themselves, but in the studio it was a different story.

The highest crest factor I have come across is Respighi's Roman Trilogy performed by the Pittsburgh Symphony Orchestra conducted by Lorin Maazel released on Sony Classical (SK 66843). Peak level is -0.17 dB while RMS level is -30.44 dBFS giving a crest factor of 33.28 dB. See here for individual track data. From Discogs comment (quoted without correction):

Orchestra:
recorded with a pair of Bruel & Kajer (B&K) 4009 omni microphones (8' behind Maazel, 15' up, 5' apart).
Sony PCM 3348 digital tape recorder.
Off stage brass (Pines IV, Feste I) - (3) Neumann TLM-170's
Heinz Hall - April 26-27, 1994

Organ:
Mander, pair of B&K 4009 omni microphones
Church of St. Ignatius, Loyola, NYC - July 20, 1996

Also, note the use of 20 bit recording technology.

 

[j_j]

I'm not keen on analogies, but this can be easily seen when stretching the dynamic range of an 8-bit photograph (e.g. moving black and white points) the histogram then shows missing colours as gaps.

It's the exact same mathematical result, indeed.

 

[AnalogSteph]

The highest crest factor I have come across is Respighi's Roman Trilogy performed by the Pittsburgh Symphony Orchestra conducted by Lorin Maazel released on Sony Classical (SK 66843). Peak level is -0.17 dB while RMS level is -30.44 dBFS giving a crest factor of 33.28 dB.

I checked it out an Amazon Music (a search for "respighi maazel" should turn it up). That recording is no joke indeed, I really have to crank it up. I'm not surprised it's from 1996... the mid-late '90s to early 2000s seem to have produced some of the most demanding classical recordings out there. Let's see what ReplayGain has to say once the CD arrives.

 

[JIW]

I checked it out an Amazon Music (a search for "respighi maazel" should turn it up). That recording is no joke indeed, I really have to crank it up. I'm not surprised it's from 1996... the mid-late '90s to early 2000s seem to have produced some of the most demanding classical recordings out there. Let's see what ReplayGain has to say once the CD arrives.

I set my volume control such that 0 dBFS is 7 dBV for my HD600. The highest Peaks should be around 115 dB SPL that way. For speakers in a room that have a sensitivity of 90 dB at 1 W at the main listening position that would require around 300 W peak power.

The DR at the site I linked seems to be the average of the track DRs not album DR.

For a more recent classical recording, Ravel: Daphnis et Chloé - Pavane by Yannick Nézet-Séguin conducting the Rotterdam Philharmonic Orchestra released on BIS in 2015 has a peak of -0.16 dB and RMS of -28.16 dBFS for a crest factor of 31.01 dB (See here).

For an even earlier recording, Tricycle by Flim and the BB's from 1983 has a peak of -0.06 dB and a RMS of -25.18 dBFS for a crest factor of 28.13 dB (See here). The 2020 re-release is even more dynamic with peak of -0.57 dB and RMS of -27.51 dB for a crest factor of 29.95 dB (See here). Note that this is not classical but a small jazz ensemble. Recorded using Mitsubishi X-80.

Pictures of full booklet available on Discogs.

 

[AnalogSteph]

Let's see what ReplayGain has to say once the CD arrives.

Album gain is "just" +3.78 dB, with a "true" peak of 0.988001. It set a new local record for track gain though, at a maximum of +25.22 dB for La fontana di Valle Giulia all’alba. Its "true" track peak is a measly 0.07303. La fontana di villa Medici al tramonto is no slouch either, at +23.41 dB / 0.082841, and I pini del Gianicolo is still near the top of what I had previously (+21.27 dB / 0.091644).

On a per-track basis, it's not actually an exceptionally demanding disc (the aforementioned Mahler no. 4 needs something like -6 dB worth of pre-gain on no less than two movements, vs. around -3 dB here, e.g. La Befana @ +3.36 dB / 0.988001), you just need to keep riding that volume control or go by track gain. The quiet tracks reveal a substantial (microphone) noise floor when doing the latter though, so clearly the performance really was that dynamic.

Oddly enough, I cannot reproduce the RMS level mentioned:

Peak level is -0.17 dB while RMS level is -30.44 dBFS giving a crest factor of 33.28 dB.

I am seeing an average of -25.08 dBFS and a peak of -0.10 dB TP according to True Peak Scanner for Foobar2000 (AES +3 dB mode on), which agrees with the DR Meter plugin (-25.0/-25.2 dB). And yes, it's the same exact disc as pictured in the Discogs entry, same catalog number, barcode and "Made in Austria".

For an even earlier recording, Tricycle by Flim and the BB's from 1983 has a peak of -0.06 dB and a RMS of -25.18 dBFS for a crest factor of 28.13 dB (See here). The 2020 re-release is even more dynamic with peak of -0.57 dB and RMS of -27.51 dB for a crest factor of 29.95 dB (See here).

Quite of its time I shall say, but definitely nice and punchy sounding, very "clean studio". I wouldn't be surprised if the remaster had gained its higher dynamic range from a bit of a bass boost, either way it's nice to see one that isn't squashed.

Do keep in mind that a fairly dry studio recording tends to result in high measured dynamics, so one cannot compare that to a classical orchestra recording 1:1.

 

[JIW]

Album gain is "just" +3.78 dB, with a "true" peak of 0.988001. It set a new local record for track gain though, at a maximum of +25.22 dB for La fontana di Valle Giulia all’alba. Its "true" track peak is a measly 0.07303. La fontana di villa Medici al tramonto is no slouch either, at +23.41 dB / 0.082841, and I pini del Gianicolo is still near the top of what I had previously (+21.27 dB / 0.091644).

On a per-track basis, it's not actually an exceptionally demanding disc (the aforementioned Mahler no. 4 needs something like -6 dB worth of pre-gain on no less than two movements, vs. around -3 dB here, e.g. La Befana @ +3.36 dB / 0.988001), you just need to keep riding that volume control or go by track gain. The quiet tracks reveal a substantial (microphone) noise floor when doing the latter though, so clearly the performance really was that dynamic.

Oddly enough, I cannot reproduce the RMS level mentioned:

I am seeing an average of -25.08 dBFS and a peak of -0.10 dB TP according to True Peak Scanner for Foobar2000 (AES +3 dB mode on), which agrees with the DR Meter plugin (-25.0/-25.2 dB). And yes, it's the same exact disc as pictured in the Discogs entry, same catalog number, barcode and "Made in Austria".


Quite of its time I shall say, but definitely nice and punchy sounding, very "clean studio". I wouldn't be surprised if the remaster had gained its higher dynamic range from a bit of a bass boost, either way it's nice to see one that isn't squashed.

Do keep in mind that a fairly dry studio recording tends to result in high measured dynamics, so one cannot compare that to a classical orchestra recording 1:1.

Curious. I may try to see what the files on Qobuz are like. Did you check the RMS levels of the individual tracks?

 

[AnalogSteph]

Curious. I may try to see what the files on Qobuz are like. Did you check the RMS levels of the individual tracks?

 

[bmc0]

Here's a pretty dynamic recording: Stravinsky - Three Greek Ballets (Naxos). Stats for all tracks:

Channel                       L            R
DC offset           -0.00012341  -0.00018903
Minimum             -0.99499512  -0.81936646
Maximum              0.99948120   0.88305664
Peak level (dBFS)       -0.0045      -1.0802
RMS level (dBFS)       -29.0429     -29.2255
Crest factor (dB)       29.0384      28.1453
Peak count                    1            1
Peak sample           196465406    196980829
Samples               205660056    205660056
Length (s)              4663.49      4663.49

Of the three works, Orpheus has the highest crest factor:

Channel                       L            R
DC offset           -0.00021987  -0.00038726
Minimum             -0.99499512  -0.79965210
Maximum              0.99948120   0.88305664
Peak level (dBFS)       -0.0045      -1.0802
RMS level (dBFS)       -30.7505     -31.7478
Crest factor (dB)       30.7460      30.6675
Peak count                    1            1
Peak sample            66023486     66538909
Samples                75218136     75218136
Length (s)              1705.63      1705.63

Oversampling reveals two very small intersample overs in track 38 (+0.002dB).

Note: The RMS values shown are not +3dB (e.g. full scale sine wave is -3.01dBFS).

 

[JIW]

View attachment 421638

Those are near enough identical to the ones I posted.

DR15 -6.38 dB -26.96 dB 2:42 01-I. I pini di Villa Borghese
DR13 -6.78 dB -28.09 dB 7:10 02-II. Pini presso una catacomba
DR13 -20.76 dB -39.98 dB 7:25 03-III. I pini del Gianicolo
DR14 -0.32 dB -20.26 dB 4:55 04-IV. I pini della Via Appia
DR13 -22.74 dB -42.98 dB 5:33 05-I. La fontana di Valle Guilia all'alba
DR15 -9.38 dB -30.55 dB 2:20 06-II. La fontana di Tritone al mattino
DR14 -4.02 dB -21.56 dB 3:04 07-III. La fontana di Trevi al meriggio
DR13 -21.63 dB -41.93 dB 5:03 08-IV. La fontana di Villa Medici al tramonto
DR14 -2.61 dB -21.30 dB 4:50 09-I. Circenses
DR15 -5.86 dB -27.35 dB 7:28 10-II. Il Guibileo
DR16 -9.82 dB -33.01 dB 7:29 11-III. L'Ottobrata
DR17 -0.17 dB -22.05 dB 5:36 12-IV. La Befana

Do you know how the album RMS is calculated?

I took the track duration weighted average of the track RMS values in dBFS originally and got -30.44 dBFS. I calculated it again now using absolute sample values -
10^((RMS dBFS - 10*log10(2))/20) - and now get 0.0301 which is -27.42 dBFS. Simply taking the average with equal weights gives -26.78 dBFS.

 

[danadam]

I set my volume control such that 0 dBFS is 7 dBV for my HD600. The highest Peaks should be around 115 dB SPL that way.

You mean Sennheiser HD600? Aren't they only 97 dBSPL/1V?

 

[Sokel]

View attachment 421638

A nice one from 1978-1988 which also feels dynamic is this one:

Loading…

www.discogs.com

Edit:link

 

[JIW]

You mean Sennheiser HD600? Aren't they only 97 dBSPL/1V?

No. 97 dB is Sennheiser's spec at 1 mW. They measure around 105 dB at 1 V.

 

[danadam]

No. 97 dB is Sennheiser's spec at 1 mW. They measure around 105 dB at 1 V.

Maybe a mistake on their page then, they show "97 dB (1V)":

 

[JIW]

Maybe a mistake on their page then, they show "97 dB (1V)":

View attachment 421652

Yes. For the HD650, it is 103 dB at 1 V which is closer to measurements.

 

[pablolie]

So the parting premise here is that throughout times, recordings were always engineered to be perfect yet were regularly held back by the limitations of new tech fads?

 

[MaxwellsEq]

So the parting premise here is that throughout times, recordings were always engineered to be perfect yet were regularly held back by the limitations of new tech fads?

In my opinion, "mastering" is what everyone is complaining about here. I read an interview with Thomas Dolby where he'd finished recording multiple tracks over quite a long period and sent them off to be mastered. In his opinion, each of the tracks sounded a bit different and some had more peaks than others and some sounded louder than others. He was expecting the mastering engineer to level everything out and make the tracks a coherent album (and presumably fixing mistakes such as DC offset).

Meanwhile Steven Wilson considers that he is now sufficiently competent in eliminating mistakes that he no longer needs the support of a mastering stage.

Finally, for a while Kate Bush sold a high res version of '50 Words For Snow' on her site, but there was an added bonus (that she was perhaps not aware of and certainly didn't mention), it had not been through mastering!

 

[Blumlein 88]

Yes, mastering differences are what people are hearing. It still irks me that so many people believe the old myth about LP masters being used for CD. Do you realize roughly using 1 khz as 0 db. Do a straight EQ down to 20 hz which will be at -20 db, and a straight EQ up to 20 khz which will be +20db. Do that EQ to some of your modern CDs and listen to it. That is the RIAA curve such a CD would have when you play it back flat. Try that to a few and you'll see just how ridiculous this idea is.