It's the original 16/44.1 file that has this incread high frequency energy.
So, I'm wondering if the 16/44.1 file has been created with a bad sample-rate converter, which has a lot of ringing around 20 kHz. Or could it be something else?
It's very unlikely that the original used a bad sample rate converter. There were some dodgy sample rate converters around the turn of the millennium but that situation only lasted for a few years and you probably wouldn't expect to see that amount of ringing unless you used a test signal specifically designed to highlight it. Far more likely is that it's the result of noise-shaped dither, where quantisation error is effectively converted to white noise and "shaped", IE. The dither noise is reduced in the frequency band/area where our hearing is most sensitive and moved to areas where we're least sensitive, typically from around 16kHz - 22kHz.
[1] I have foobar2000 (no special converting plugin), is this sufficient? (and audacity to check).
[2] Is there a link to a page where most of this stuff is easily explained?
1. Foobar should be fine. Using a dedicated converting plugin with noise-shaped dither would be more representative of how commercial music releases are handled and is theoretically better. Although as castleofargh stated, standard (triangular) dither or no dither at all (truncation) is very unlikely to audible except in rare cases and at very high (uncomfortable) listening levels. If you're going to do it manually without a dedicated plugin, reduce the sample rate first and then reduce the bit depth, not the other way around.
2. No idea, I never really use foobar but quite a few here do and can point you in the direction of a simplified page, if there is one.
Intuitively, we would expect that. But in practice it's a lot messier. The best argument in favor of getting the hires file is often the worst reason of all. That a sound engineer was asked to make a different version for the different formats. While that was completely justified for vinyls vs digital because of the physical limitations of vinyl, it makes no sense for various digital resolutions. And yet it's not uncommon. In fact I'd argue that pretending to have audibly improved resolution was such a fail that they have no other choice but to manufacture the audible difference.
I've tried to get some understanding of this and watched/read as many interviews of sound engineers on this topic as I could. A few argue that the hires version will be purchased by an elite with better more neutral gear and a taste for more dynamic more blablablah elite audiophile flattery. While the rest of the pleb will listen to the album in their car or some cheap stereo system without extended frequency response. So the argument was to make CD versions that will still work fine with those poor uneducated people's circumstances. And this is ironically about the most reasonable explanation I've seen beside the guys who simply admit that they were asked to make different versions so they did their job.
In practice, it's even messier than you've explained. The membership here on head-fi are generally very serious about sound quality, which is largely why they're members in the first place. Engineers are very serious about SQ too, though typically not in the same way that the membership here is. To us engineers, SQ is dependant on the consumers' receiving media/equipment but most here on head-fi have a rather simplistic, narrow/exclusive notion of SQ.
There is, potentially, a justification for quite a number of different versions/masters. For example, a version for analogue radio playback, which accounts for the time restrictions and the severe multi-band limiters typically employed on music broadcast stations. So going back to SQ, the goal with a radio edit/master is to achieve the best SQ for radio listeners. However, if you were to play this audio file from disk (rather than receiving it as a radio transmission) on a good sound system (rather than a typical radio system), most/all the membership here would consider this version to be significantly "worse", while us engineers consider it to be "better" (for the intended purpose). In fact, when radio play was the single greatest driver of record sales, those engineers with the specialist knowledge, experience and skill to create a master that wasn't degraded by the radio broadcast chain (and compared favourably to other broadcast tracks), were highly sought after (and rewarded). From the narrow audiophile point of view, these skilled engineers were rewarded for effectively making "worse" masters! Likewise, a Youtube version/master is quite commonly desired, with the requirement that the track/s at least work to some degree when played on the internal speakers of laptops and other mobile devices, which is difficult to achieve if you want that master to still compare decently when played back on more reasonable equipment. There may also be another version/master required for TV broadcast and another for film, etc.
Another consideration, unfortunately, is that the worth of most major record labels is largely dependant on their back catalogue. All the corporations that own the labels have back catalogues valued in the billions of dollars but of course, the actual worth of a back catalogue is dependant on the revenue it can generate and one of the ways that can be achieved is by issuing new versions, re-mixes and re-masters. This can be difficult from the engineers' point of view, especially if a particularly good (or particularly loved) mix/master has already been released previously.
So, we've got a messy set of scenarios, circumstances and considerations, only one of which is primarily aimed at audiophiles. The set of digital "resolutions" is therefore (potentially) just an extension of an already messy set that started in the 1950's, which record labels will exploit if they feel it's worth the cost/effort, baring in mind that it is specifically aimed at the small (and misinformed) audiophile market.
G