Now, I don’t know about you, but the onset of this song’s prechoruses feels like a bit of a let-down. Yes, the arrangement tries to fabricate a sense of musical arrival by adding piano and bass parts on the first occasion (at 0:27), and by adding piano and synth-pad parts second time around (at 1:45), but fundamentally I think the prechorus’s impact is being undermined by its choice of opening chord.

You see, moving from one chord to a different chord carries its own inherent musical emphasis, so it usually feels most natural to us if the rhythmic placement of these chord changes (or ‘harmonic rhythm’) lines up with other stress patterns in the music. Indeed, one of the less well-known ‘rules’ of traditional harmonic theory is that you should try to change chord when moving from a weak beat to a strong beat – thereby syncing up the stress patterns of the harmonic rhythm with the musical metre’s stress pattern. This is why you’ll find lots of situations in pop songs where a chord is played on a strong beat (eg. beats one and three in a four/four time signature) and then sustained over onto the following weak beat (eg. beats two and four in four/four), whereas it’s much rarer to hear a new chord arriving on a weak beat and then sustaining onto the following strong beat.

Moreover, this principle operates on a structural level too, where chord changes tend to feel most natural when moving from a weak bar to a strong bar within the prevailing phrase structure, and that’s where I think the prechorus of ‘Die With A Smile’ comes unstuck. While the song starts off by swapping regularly between A^maj7 and D^maj7 chords on the (more strongly stressed) first and third bars of each four-bar musical section, as you’d naturally expect, when we get to the (strong) first bar of the prechorus, the chord doesn’t change, maintaining the same D^maj7 harmony we heard in the (weak) fourth bar of the verse, and thereby robbing some musical emphasis from the beginning of that section.

On the plus side, though, this song does provide a great illustration of how reverb automation can be used for section differentiation. If you listen to the first chorus (0:45-1:17), there’s a long-tail reverb on the snare drum, although it’s dull-sounding enough that the rest of the arrangement masks it a lot of the time – soloing the mix’s Sides channel reveals it much more clearly. However, when we hit the second verse, this reverb is suddenly removed, which subtly (but profoundly) reduces the sense of subjective ‘size’, making the presentation feel more intimate at that point. Then the reverb begins to creep back in for the second prechorus, before returning full-force for the rest of the song. What’s particularly interesting to me in this scenario, though, is that the reverb character is fundamentally fairly unnatural-sounding, and in any case starkly in contrast to the snare sound itself, which I suspect is why I don’t get a sense that the chorus snare is merging very much with its reverb, which means it doesn’t appreciably recede in the mix’s front-back depth perspective as a result.

I’ll confess right upfront that I always disliked this tooth-rottingly syrupy song, with its relentless D-Bm-Em-A harmonic loop, plodding groove, and plasticky ’80s synth textures. Given its extraordinary longevity, though, I recently felt duty-bound to have a closer analytical look at it, and I have to say that I now hold it in a lot higher esteem. Many musicians worry that analysis will ‘kill the magic’ of their favourite music, but personally I think it’s the other way round – not only does analysing songs I already love deepen my appreciation of them, but I also find that analysis of music that doesn’t immediately appeal to me often uncovers qualities that I initially overlooked.

So what was it that I found in this case? Well, one of the first things I noticed was the unusual distribution of the programmed drum fills, where the same pattern of fills almost never appears in any similar song section. On further investigation, it seems as if George Michael (who not only wrote this song, but also apparently created and performed the entire arrangement himself) built this production around a 20-bar LinnDrum sequence constructed out of eight different two-bar programmed patterns chained in the following order: 1-2-3-4-5-6-5-7-1-8. If I now superimpose the song structure over this repeating chain of drum patterns, you’ll see that the song’s eight-bar musical sections don’t fit neatly into the 20-bar loop length…

…and if I now rearrange that picture to line up all the similar song sections vertically, it’s clear to see that this almost completely avoids any similar song section from having an identical pattern of fills.

The sharper-eyed amongst you may, however, have spotted that my song structure here is actually missing a couple of sections, namely the Solo after Chorus 1b and the repetition of Chorus 2. Well, it turns out that both of these sections just repeat the drum patterns of the section before them, like this…

…which leads me to suspect that they were probably added midway through the production process after the LinnDrum had been committed to tape, by editing the multitrack. And I can see the reasoning too, because those copied additions give the main body of the song a more regular structure, comprising two iterations of Solo-Verse-Prechorus-Chorus-Chorus.

But it’s not just the drums that vary between song sections, because the bass part is similarly nuanced. Despite the repeating chord progression, every single eight-bar section of the song delivers a slightly different bass line, with just one notable exception: Chorus 2b, where the bass appears to have been copied from Chorus 2a alongside the drums, suggesting to me that this tape edit was done later in the production process than the edit that created Solo 1. The fact that George Michael put such effort into adding musical variety to this arrangement, a pale imitation of which could easily have been thrown together in a fraction of the time via lazy copy/paste repetition, speaks volumes to me about his work ethic. Plus, assuming my hunches about the tape editing are correct, I think it’s also fascinating to get a glimpse into the creative process from which this song’s final structure developed.

The arrangement overall is very smartly managed too, building up progressively from each of the two verse sections by virtue of progressively richer/brighter synth layers and vocal doubletracks/harmonies. But my favourite aspect of the arrangement is (to formulate a sentence that I never, ever thought I’d find myself writing) the sleighbells! You see, they don’t just jingle all the way, whatever you think you may remember about this song – they actually only appear during the Solo sections. This is genius, because if there had been sleighbells constantly battling against the lead vocal in the upper spectrum, there’s no way the mix engineer would have been able to bring out the characteristic breathiness of George Michael’s voice as effectively. In fact, I might even hazard a guess that one reason for editing in the extra Solo section near the start of the song might have been to provide the sleighbells an earlier opportunity to signpost the song’s seasonal nature without conflicting with the voice.

All these fun things notwithstanding, there is one aspect of this song that does leave me scratching my head. When I lined up the single version against the longer ‘Pudding Mix’ version on Wham’s contemporaneous Music From The Edge Of Heaven album, it immediately became apparent that they were drifting out of sync with each other. Tempo-mapping them both revealed that the album version remains at a pretty regular tempo throughout the song, whereas the single version speeds up by around 1bpm during Solo 1 (0:55-1:11), and then very gradually slows down (by roughly 0.5bpm) over the next 32 bars before stabilising from Solo 2 onwards. Given that the LinnDrum was one of the earliest commercially available drum machines, without the benefit of today’s sophisticated tempo-mapping facilities, the simplest method of implementing such tempo changes would have been to tweak the final tape’s varispeed control in real time during the mastering process – a speculation that’s corroborated by the overall pitch rise (of roughly 15 cents) you can hear if you compare the first and second choruses side by side. Chorus 1: play_arrow | get_app Chorus 2: play_arrow | get_app

Regardless of how the single’s tempo profile was created, though, it does seem rather odd to me. If the idea was to subliminally give the verse a bit more pep, then why not do a similar move during Solo 2 as well? Or if you’re not going to do that, then why slowly reduce the tempo after Solo 1 at all? After all, if the extra pace helped the first verse, shouldn’t it help the rest of the song too? I can’t imagine that any concerns about the single’s overall duration are pertinent here either, because comparing the single with the album shows that the tempo changes make less than two seconds’ difference to the song’s running time. In fact, even if the higher tempo at the end of Solo 1 had been maintained for the rest of the song, that still wouldn’t have shortened the overall duration by more than about three seconds.

It’s a Christmas mystery…