[bigshot]

As I said, I don't think mono has anything to do with it. I think it's the sort of distortion created by 1950s era signal processing combined with complex tones of massed violins. If you listen to old popular music recordings with orchestras, you'll find some that have not very natural sounding string sections. They sound good, like in an old movie; but they don't sound like real orchestras. I think the people who created the codecs probably used modern high fidelity recordings to tune the compression, not older recordings like this. The codec wasn't designed to handle that euphonically distorted kind of sound.

In case anyone is interested, this is the album I am talking about. Encode this in 64, 128, 192 and 256 and take note of the sections that artifact the worst in the lowest data rates. Then check those sections in the higher rates and your find the stubborn sections that won't encode as easily. Make sure you have VBR turned off or it will try to fix those sections.

https://www.amazon.com/Decca-Years-Sammy-Davis/dp/B000002OF2

 

[71 dB]

As I said, I don't think mono has anything to do with it.

Okay then.

I think it's the sort of distortion created by 1950s era signal processing

Signal processing? How much digital signal processing did they use in the 1950's?

combined with complex tones of massed violins. If you listen to old popular music recordings with orchestras, you'll find some that have not very natural sounding string sections. They sound good, like in an old movie; but they don't sound like real orchestras. I think the people who created the codecs probably used modern high fidelity recordings to tune the compression, not older recordings like this. The codec wasn't designed to handle that euphonically distorted kind of sound.

I actually like the unnatural, but soft & velvety sound of string instruments in old recordings. That's about the only thing I like about them. Otherwise they often sound like crap. What if the gurgling artefacts are caused by flutter of the magnetic tape machines used to record this music? Maybe flutter is just too fast modulation of frequency for the codecs to follow properly?

 

[bigshot]

They had compressors and equalizers before WW2, and they didn’t have solid state so everything was tube based.

These late 40s recordings weee likely recorded to disc, not tape. And tape flutter would be audible in the uncompressed CD. But the CD sounds fine.

I already explained what causes it. It’s like artifacting during applause.

 

[71 dB]

They had compressors and equalizers before WW2, and they didn’t have solid state so everything was tube based.

That's analog filtering and gain control.

These late 40s recordings weee likely recorded to disc, not tape. And tape flutter would be audible in the uncompressed CD. But the CD sounds fine.

Magnetic tape recorders became available around 1930. The flutter probably contributes with other distortions the the smoot/velvety sound of string instruments in these old recordings, but aren't we talking about "gurgling artefacts" that happen with lossy encoding cause perhaps by flutter not present in lossless audio such as CD?

I already explained what causes it. It’s like artifacting during applause.

Applause is noise that has very rapid dynamic changes. No wonder lossy encoders are on their knees.

 

[bigshot]

I don’t know why you’re arguing because you don’t seem to know much about the subject. From WW2 until about 1952 just about all recordings in the US were recorded to 33 1/3 RPM lacquer masters. The commercial format at the time was 78 RPM shellac discs with a maximum running time of a little over 4 minutes. So the takes would be recorded to 33 1/3 disk (which was hifi) with multiple takes on a disk and dubbed to 78 RPM for release. Magnetic tape was used in Germany during the war years, but it wasn’t used much at all in the US until the LP era. And the first LPs released by Columbia were mastered from the 33 1/3 RPM master disks used to make 78s a few years earlier. They had to join breaks between takes on the fly as they cut the LP master, so there are weird joins on those early LPs. Once the LP format was established, they started using magnetic tape. I’m not sure exactly when that was, but I would guess it wasn’t any earlier than 1954 or so. That is the year of these Sammy Davis Jr recordings. RCA would likely have just been starting to record to tape, but Decca would probably still be recording to 33 1/3 disks because the ultimate destination of these songs would have been 78 RPM shellac disks.

 

[71 dB]

Maybe the US was slow with magnetic tape, and I admit being surprised about that, but in the UK Sir Edward Elgar recorded his works before his death in 1934. The earliest of those electrical recordings being from 1926.

 

[bigshot]

Electric recordings don't mean that they were recorded on tape. It just means they used microphones. In the 78 era, there were two types of recordings. The earliest was acoustic. That was recorded using a horn channeled into the cutting head on a disk lathe. There were no microphones, so it was called acoustic. In 1925, they started recording with microphones wired into an amplifier that fed the signal into the cutting head of the disk lathe. That was called electrical recording. All recording was direct to disk, meaning that there was a cutting lathe in the control booth cutting the recording onto a beeswax or lacquer master. That master was plated and used to create the metal parts... the mother disks and stampers. Those were used to press the shellac pucks into records. This was the process generally used in Europe until around 1954.

Around 1940, Radio Recorders had come up with the technology to make high fidelity long playing records by cutting a lacquer master at 33 1/3 RPM. But the war delayed the introduction of hifi for consumer gear. So as I said, the hifi 33 1/3 recordings were only used as masters to make standard 78s. I don't think Europe ever did it this way. Around 1954 RCA started recording in stereo on tape and that became the standard, replacing recording on disk. It happened sooner at RCA and Columbia and later at the smaller labels, but by the late 50s, everyone was using tape. The changeover in the rest of the world wasn't too far behind the US.

I can only think of a small amount of recordings that were released on 78 and recorded on magnetic tape before the 50s, and those were all in Nazi Germany. I think they were all live radio broadcasts. Furtwangler did a bunch that were recently restored and released. There were some fragments of live operas recorded this way too. Very rough sounding stuff because they were recorded live during public performances.

 

[71 dB]

Yeah, you are right. Electric recording doesn't mean magnetic tape. I am surprised how late magnetic tapes became a thing. No wonder old recordings are so bad.

My knowledge of the history of recorded sound isn't the strongest, because in university those things were clearly considered educationally less important. How GMS phones encode speech was much more important of a topic.

 

[bigshot]

Old recordings aren't necessarily bad. The engineers back then knew how to work within the limitations. The reason they sound bad now is because of the careless way they're mastered for CD release. Even acoustic recordings can make the hair on the back of your neck stand up with realism. It just takes work, and a lot of companies aren't willing to do that for legacy titles.

Like everything else, pre-LP era recordings were designed to be listened to on speakers in a good room. It sounds best with a single full range speaker. With a mono speaker setup, you can get a real sense of depth. Headphones don't do it justice at all.

 

[gregorio]

I think it’s due to euphonic distortion that doesn’t look to the algorithm like music. It’s similar to the artifacting in crowds applauding.

The algorithms aren’t looking for music, just masked freqs in any kind of audio.

Crowd applause is tricky, it’s numerous random closely spaced and overlapping transients (and transient reflections) which almost always suffers from phase inconsistencies and mono-compatibility issues. This can create a sort of interference pattern with lossy codecs’ due (from what I understand) to the overlapping nature of the processing blocks. It’s a sort of phasing effect/artefact that was not uncommon with older and/or lower bit rate codecs and particularly when transcoding. It shouldn’t be an issue with modern encoders at decently high bit rates.

It’s possible a similar effect is occurring with the string section you mention. Again, it’s lots of transients occurring at very slightly different times and often suffers from phase inconsistencies to start with.

G

 

[bigshot]

I can't remember for sure if this was the track that just barely didn't clear below AAC 192, but this is from the same recording sessions. It's a very unique sounding string section. If this was the track it was in the instrumental bridge at 1:30. I encoded it in increasing data rates and took note of where in the track it artifacted at lower data rates, then checked the same sections at higher ones.

 

[Blackwoof]

AAC & Vorbis are transparent at 160 ~ 192kbps VBR while MP3 needs 192 ~ 320kbps to come close.

 

[Blackwoof]

The algorithms aren’t looking for music, just masked freqs in any kind of audio.

Crowd applause is tricky, it’s numerous random closely spaced and overlapping transients (and transient reflections) which almost always suffers from phase inconsistencies and mono-compatibility issues. This can create a sort of interference pattern with lossy codecs’ due (from what I understand) to the overlapping nature of the processing blocks. It’s a sort of phasing effect/artefact that was not uncommon with older and/or lower bit rate codecs and particularly when transcoding. It shouldn’t be an issue with modern encoders at decently high bit rates.

It’s possible a similar effect is occurring with the string section you mention. Again, it’s lots of transients occurring at very slightly different times and often suffers from phase inconsistencies to start with.

Yet Opus & Musepack seem to have less issues at 160kbps VBR than AAC & Vorbis do on samples like that. It actually why I moved to MPC as I failed ABX'ing It at --standard from lossless, Never had to up the bit rate on single song to album like I did on AAC & Vorbis.

 

[gregorio]

Yet Opus & Musepack seem to have less issues at 160kbps VBR than AAC & Vorbis do on samples like that.

TBH, I pretty much stopped following the research/developments of MP3 and AAC about a dozen years ago, because I already found 320kbps and 256VBR audibly indistinguishable from lossless a few years before that. So I don’t really know what the best ones are at the lower bit rates these days because I never use bit rates of 160kbps.

G

 

[bigshot]

The differences in file sizes between 160, 192, 256 and 320 aren't very big. The difference between lossless and lossy is huge, but between different flavors of lossy, it isn't very significant.