Sound Diving #2 - Take On Me
by Ben @ mu:zines | 24th Sep 2021
This "Sound Diving" occasional series will dig deep into some of my favourite songs from my formative years, trying to uncover the sounds and techniques used, what fascinated me about them and why they worked.
When I was young and learning about music, I'd often try to recreate songs I liked on my limited equipment, but I didn't really understand the details of what I was hearing - I'd usually be working out by ear how to play the notes of the latest hits' synth riffs, and that's about all I could do on my little Casio. During the first half of the 80s the UK Charts were full of synth music, and this was what I found new and interesting.
I'll be looking at some of these classic tracks and deconstructing, analysing and recreating some of these, in the hope of learning the secrets that still fascinate me today!
"Take On Me", released in late 1985, was a massive hit at the time - no doubt in part to that video. I always liked the song, and while it would have been the main synth riff that I would be learning to play, I loved the "sound" of the song. It starts with a drum track which is driving and punchy, then some nice synths come in, and of course that bass sound. It's not a "heavy" track, the mix is quite "light", and this made it stand out from other records of the time. The drums sound fabulous - the snare sound in particular - and a recent opportunity to get into the multitracks gave me the chance uncover some answers into why the track sounds the way it does.
Please note: This is a deep dive and a long read - there's a lot to cover!
We are fortunate to know quite a few details about the history of Take On Me from Richard Buskin's excellent 2011 Sound On Sound article, where he gets a lot of detail from producer Alan Tarney, and engineer Gerry Kitchingham (who also did the final mix). I'll be excerpting from this article a fair bit.
The song has a complicated history, and a few attempts were made at recording (and releasing!) it (see the "Evolution of 'Take On Me'" box below to hear all the various versions from initial demos through to the final version), but it's the final Tarney/Kitchingham 1985 version we all know that I'm concerned with here.
The gear used was primarily Tarney's:
So turning to the multitracks we have two mono drum machine tracks, which are both easily identifiable as LinnDrum parts. We then have a "Percussion" track, and then two tracks for a stereo "Snare".
The first drum machine track seems to be carrying the bulk of the drum programming you hear on the track, including the kick, snare, hihats and cymbals, and the various drum fills and so on (the drum programming is quite inventive in places, and Pål's excellent drum programming is remarked on a few times in the SOS article). It's at a stonking 169bpm(!) - although they bump the tempo by +0.5 bpm for the final verse to keep the energy up, and as neither the LinnDrum nor the UMI 2B sequencer support programmed tempo changes, they likely did this in the SMPTE sync box clocking the machines from the Studer tape machine.
As well as progressively building in intensity over the course of the track - the end choruses have all the hihats, tambourines and crash cymbals you could ask for - this drum part has some interesting details. The low-pitched snare on beat 3 gives the kick an extra texture - mixed low it's almost like a bit of a reverb on that kick, and gives some half-time beat cues to the listener in case they were freaked out by the quick tempo! Here's a recreation of a few patterns and fills from that part using standard LinnDrum sounds:
"Helpfully, there are more clues to be found on the multitracks..."
Let's get back to this mystery snare in the main drum machine track. It's super tight sounding with a really defined attack, and it almost sounds too good to be a standard Linn sample, which were often quite dark and grungey due to the quality limitations of sampling at the time - but we know it's on the Linndrum track, so we know that it did come from the Linn. We also know that Tarney had alternate sound chips on that session, otherwise it probably wouldn't have been mentioned in the article.
(On some drum machines in the early 80s, you could open the lid, and replace the individual chips that had the various drum sounds encoded on them, to let you use different sounds. This was an advantage over less flexible drum machines of the time which were permanently stuck with the set of sounds that they came with.)
Helpfully, there are more clues to be found on the multitracks.
On some of the synth tracks, towards the end, after the song has ended, there are some more drum machine parts. It looks to me like these were probably the remains of the original guide drums they put down - nearly 6 minutes' worth, to make sure they didn't run out - and they then recorded over those tracks as they built up the production - but as those synth parts only last for the four minutes or so of the song, they leave the previously recorded guide drum parts exposed at the end for a couple of minutes.
This led me onto a deep dive into the world of replacement LinnDrum chips. Now, I figured that with the popularity and ubiquity of the Linn LM1 and LinnDrum in popular music of the 80s, there would be copious amounts of documentation available online of all the available sound chips, along with audio samples, and probably drum machine sample libraries full of those sounds.
Surprisingly to me, there was none of that available that I could find. So I started to search around, of course trawling my own mu:zines archives, and also reach out to some of my smart synth nerd friends (Hi!) to see what I could turn up. Hunting High and Low, indeed.
Firstly, Linn had a library of replacement sound chips that you could buy, for between about $20-$40 per chip (one chip typically stored one sound, though sometimes longer sounds could be spread over multiple chips). They also had a service where you could send them recordings of your own sounds, and they would encode them on custom chips for your machine.
I was also aware of DigiDrums (who later became Digidesign and developed Sound Tools, the precursor to Pro Tools), another US-based third-party company who started by making a few sets of replacement drum soundchips for the Emu Drumulator (and later, for other machines including the Linn) - some of which were also used on famous records.
It was also possible to make your own custom chips - you could get a "Prommer" or other sampler/chip burner, and record/encode your own samples onto blank chips. I thought it was unlikely Tarney was making his own custom chips, but not impossible - and if the mystery snare was something custom made, then the source was likely to remain a permanent mystery.
Another angle came from the UK - in the mu:zines archives I found a few references to Syco selling replacement Linn and Drumulator sound chips. Syco were the London-based UK distributor and exclusive seller of pretty much all the super high-end stuff - if you were an elite famous producer or artist, and you wanted a Fairlight, Yamaha DX1, Kurzweil K250, Linn 9000 or other fabulously expensive piece of gear - you went to Syco - by appointment only, of course! I wondered whether Syco had their own custom chip library, as Alan Tarney would likely have got any extra LinnDrum chips from a UK source.
Kendall Wrightson, who was part of the team there at the time (you've probably seen that video with him filmed at Syco), confirmed to me that they didn't make any custom chips and they basically imported and sold the official Linn replacement chips, as well as the DigiDrums Drumulator chips - with the "Rock Drums" set being by far the most popular (used very prominently on Tears For Fears' Shout record).
So, the official Linn library was really where I needed to start looking through for alternate sounds, as I wasn't aware of any other companies selling custom soundchips. At least that narrowed the search somewhat!
I reached out to another Linn owner and notable synth historian, Paolo from Synthmania, who was super helpful in looking through his available Linn materials. In particular, he had documentation of the full Linn sound chip library, so I had a definitive list of all the sound chips they did. You've got to love the $20 "Silent chip"! He also had the original Linn demo tape, which had demos of the LinnDrum, and example recordings of all the individual sounds Linn offered.
I went through the demo recordings of the available snares - all 37 of them - multiple times, but couldn't find an obvious match. It wasn't helped by the original recording being quite poor quality. However, after many other searches through sample libraries and other online sources gave no results, I decided to go back through the Linn demos one last time.
What I hadn't realised was just how bad the original recordings on that Linn demo tape were - it was a cassette copy, but the audio was lacking in frequency response, and was rather distorted so the transients were all smeared in that horrible cassette tape over-modulated way. And when I realised just how distorted the sources were, I made more careful comparisons to the dry mystery snare, trying to see past the differences in "shape" of the sound, and the altered frequency response - and I found a match! The tone colour, and the pitch seemed to match.
To check my findings I took the dry mystery snare from the multitracks, and attempted to rough it up with compression, distortion, EQ and cassette emulations, and I could indeed make it sound pretty much like the Linn demo cassette recording.
The source had been found, and I'm super happy to have identified it!
The Mystery Snare is... [er... snare drum roll!] "Snare 7", from the Linn chip library.
(Ok, I guess the discovery doesn't sound all that dramatic, but I'd searched quite hard for this, it's not easily available and it answers a very long-standing question for me about the source of the snare sound. In my quest to definitely answer these type of things, this scores as a big win for me!)
Or so I thought... (and this was to be one of the things that held up publication of this article.)
"...this was to be one of the things that held up publication of this article."
After various approaches trying to source an independent "Snare 7" chip or recording of it (which was a rat hole in itself!) so I could hear a clean sample of it and compare it to the multitrack, I finally obtained one... and instead of the confirmation I hoped for, I began to doubt my findings.
Certainly the pitch was about right, and quite unique because actually few of the Linn snares have a highish tuning - high-pitched snares weren't really that fashionable then! But while tonally it was sort of there, the shape of it - the attack and release, and the high frequencies on the release gave it a quite different feel - much more "squishy" than I'd expect. And comparing this dry "Snare 7" directly to the multitrack showed them to be quite a way off.
Perhaps I was wrong and this wasn't the sample after all? I was dejected - a problem I thought I'd solved seemed to be looking inconclusive after all. I put the project down for some time due to my lack of forward progress and the need to pay attention to other committments.
In the process of coming back and getting this article ready for publication, I needed to finally make a decision on this snare sound, rewrite my conclusions appropriately, and move on - if the snare sound was to remain a mystery, so be it. So I dived back down and had another look.
I went through all the Linn sounds again, looking for matches, and once more, only Snare 7 seemed to be anywhere near close. But it wasn't super close to that dry sound on the multitrack. Maybe that "dry" sound wasn't actually unprocessed at all - maybe, even as a guide track, the machine was going through the SSL desk, with the separate outputs coming up on individual channels, and the engineer had indeed routinely used some SSL EQ and compression on the way to tape to modify the drum sound character?
So again, limiting myself to the same kinds of tools I knew they were using at the time, I took my dry sample, and used only some tape processing and my favourite SSL 4000E channel strip plugin, and tried to see whether I could get that sound from the multitrack - and after an hour of tweaking and trying various approaches, giving up, trying again, rinse and repeat - I got sufficiently close to the source to confirm to myself my earlier result - so if it wasn't a custom chip, I'm pretty sure this was the Linn sample used - or at least, it's the by far the closest chip I can find that gets that sound - nothing else in the Linn library, or from other compatible drum machines gets even close. Key here was some tape saturation, the use of a fast SSL gate to quite aggressively close the longer squishy release down, the SSL compressor to make the snare punchier in the attack, and then I used a little SSL EQ mainly in the lower and upper mids to get close to the recorded snare. I'm not exactly there, but I'm pretty happy with the result.
So it seems even the "dry" guide recording wasn't that dry, after all! It just goes to show how tricky working backwards from a final sound to the original source can be.
When they printed the main drum track with the more detailed programming, they'd tweaked the drum compression to make the snare sound even tighter still. As they were recording a composite drum track rather than a track each for individual drums, they would have wanted to get as much character into the printed drum track as possible, as it wouldn't be possible to separately process the sounds from this track after it was recorded. So the various drum channels would have been processed in the desk and mixed together into this one mono drum track.
And again, my "dry" processed sample gets this extra-compressed sound pretty well to my ear. I'm happy - and a bit relieved, to be honest!
"It just goes to show how tricky working backwards from a final sound to the original source can be."
Armed with the knowledge of the process I'd gone through with the main snare, I took another look at the AMS-triggered stereo reverb snare. I originally thought this was the same sound as the main mystery snare, pitched down a little, and that this would be easy to verify. It became quickly apparent that it wasn't the same snare, and I had to go through a similar process again to identify it.
(With these early drum machines, when you tune a sample up, the sample gets shorter, and with short samples to begin with - memory was expensive back then - it becomes quite obvious when a sound has been tuned up beyond a small amount. And similarly, when a sample was tuned lower, it gets longer, but also it loses the high frequencies as all the sample frequencies are shifted downward. This snare was lower pitched than the main snare, but still had a lot of high end pumped into the reverb, so in that way I knew it wasn't just the main sample pitched down, but was in fact a separate sample.)
After much searching, comparing and recreating, which is difficult because I was trying to compare a dry sample with a compressed and EQ'd (and reverbed!) version, this one seems to be Linn's Snare 2 from their chip library - a more natural-sounding, lower pitched snare for more beef, compressed through the desk and run through a reverb - probably the aforementioned 224 - then sampled into the AMS, and triggered.
This snare is pitched up slightly from it's factory tuning (about a semitone), but in order to match the recorded sound as close as I could, rather than pitching it up and running it through reverb, I found a better match by sampling it at the regular pitch with a slightly longer reverb than required, then pitching up the combined snare + reverb sample - it shortens up the reverb, but moves the tail resonances to be more in line with the recorded version. I wonder if they did this on the original session - slightly pitched the AMS triggered snare + reverb sample up a semitone? Also, there is a slight delay between the left and right side which widens the reverb effect, probably also done in the AMS delay while triggering the sample - this pushes the snare reverb out to the sides of the stereo image, leaving the hard smack in the centre of the mix.
The combination of the higher pitched snare with a fast attack in the main drum track, and this second, stereo, reverbed slightly lower pitched snare filling out the body and tail of the sound, comprises the snare in the mix - although it doesn't sound quite like the finished mix yet!
So we know that the drum track itself was compressed a bit (or at least the snare), but when we mix that with the reverbed stereo snare, it doesn't quite sound like the final mix.
Take On Me is really 90% synths/drums, and 9% vocals, but there are some guitars, so let's take a quick look at this before we get all synthy.
There's one guitar track, consisting of two main parts. The first is an acoustic guitar part playing in the choruses, which is strummed pretty hard - I guess they were really trying to inject some energy for these parts. It's quite boxy sounding, and it's also out of tune on the lower strings (which might be because it's being hit so hard). In the mix, you mostly only hear the top end so what's going on below is less of a problem. By the sound of it, it's open chords playing G, D/F#, Em, C etc with a capo on the second fret.
In the final verse, there's also an electric guitar part on this track too (track sharing was common back in the days where you had a fixed set of tracks available to use) - it's a muted, "thunky", harmonic rhythmic electric part, which "ghosts" the arpeggiated synth part and works quite well to change things up and add something new for extra sonic interest for the last verse.
There's around 8 main synth parts in the song, many of them double-tracked, with some track sharing to make the most use of the 23 tracks available. Let's start with the easy one - the bass, which is that DX7 ROM E.Bass 1 preset that everyone was using in the eighties. Interestingly, comparing this to pristine digital plugin DX7 emulations, the version on the multitrack is a little softer sounding, as if it's been rounded off and the upper frequencies rolled off slightly - probably a result of using the hardware synth plus recording it to tape, rather than a result of deliberate EQ. The bass part is sequenced and hard quantised to 8th notes, and all the notes are at maximum velocity (which on the DX7 means 100, not 127, for reasons of Yamaha misunderstanding the early MIDI spec), and as a result, the effect is a hard, driving bass, with no dynamic variation - the effect is a bit like if it was a DX7 sample sequenced in Page R on a Fairlight.
The first synth part you hear in the intro after the drums have started comes from this main chord track. Like many of the synth tracks, this part consists of multiple synth layers. In fact, considering just how many pads and synth parts there are, it's amazing the mix still manages to feel light and not swamped under huge thick synth layers.
Anyway - those first few notes come from the PPG Wave - I've not been able to find out whether it was a 2.2 or a 2.3 model they had (they're playing one on the video to 'The Sun Always Shines On TV' but I couldn't make out which version it was). The main difference that would concern us here is that the early 2.2 models didn't have MIDI, whereas the later 2.2 models and the 2.3 do - so if the Wave was being sequenced from the computer at all, it was probably a 2.3 or at least a later model 2.2.
So, this is the part that synth nerds traditionally discuss the most. "Mags uses this synth live", "They're playing this synth in the video or Top of The Pops", "I can make this sound on my synth so that's what they must have used" and "I read they used this synth recording the song" (without realising how many different times they recorded this song) - on and on the speculation seems to go.
What we do "officially" know is the gear used from the SOS article, and this recollection from Gerry:-
The next clue we get is in the middle breakdown section - here the riff plays for four bars until the drums kick in and the riff continues, but for that breakdown there's a layer removed from the riff for those four bars, which then returns when the drums kick in. This is a bouncy "pluck" sound - on the left track it's an octave down from the right, and this part sounds very analog & Juno-like to me. It's also quite thick and has a more pronounced stereo effect than the other parts, which I'd attribute to two tracks of the Juno's famous thick internal chorus (the more a similar part differs between the left and right tracks, the "wider" and thicker it will seem).
Remember, these parts, though sequenced, are printed to the multitrack together in one pass each - so there can only be a maximum combination of instrument layers of one (or two) PPG parts, one Juno 60 part, and one DX7 part (neither the Juno or DX7 are multitimbral, so they can only play one sound at a time - although the PPG's can provide multiple sounds at once).
Let's take a closer look at the section in the breakdown and see what we can hear.
In addition to the high PPG part, there is a glitchy attack part which sounds a bit like a fast filter or even pitch envelope "blip" - a common synth programming technique used to get a more pronounced attack on a brass sound or similar voice. There's also a lighter octave down note, so I think indeed there are two PPG sounds making up the high part (I don't think it's a DX7 sound) - either layered in one PPG synth, or two playing at once, in addition to the Juno part.
There are two tracks of hi bells - again, a PPG metallic synth/bell sound, similar to the main chords high PPG sound, with four notes that climb at the end of the choruses. Again, it's two passes of the same sound, and interestingly, the notes on both passes are absolutely tight - there's no way it's played this tight by hand, so this must be sequenced - another piece of evidence for at least one of the PPGs having MIDI. (Perhaps we'll find a part with two PPG's being driven from the sequencer to tell us they both had MIDI...)
Also, the long note tails are again cut off, suggesting a mono mode or layer is at work here, and again it keeps the tails for becoming messy. I didn't find a matching preset for this, but made some minor modifications to the existing PPG sounds I was using for other parts, and got in the ballpark.
The flute and arpeggiated tracks work together in the verses to provide an interlocking rhythmic framework - the flute is the simpler part, almost a repeated ostinato phrase that arpeggiates over the chords, and the "guitar" part is a harsher spikier sound playing arpeggiated versions of the verse chords. Both parts were present in the original demo, albeit with different sounds, again showing how close Tarney's production was trying to replicate the arrangement of that initial demo (although not all the synth parts survived intact).
Both of these parts seemed to me to be sequenced DX7 parts. The flute is Preset ROM-24 Flute 1. While it's only a simple, single part, it's actually recorded in stereo on two tracks, which at first seemed a little strange. Listening to it though, it actually has a sense of stereo motion and it pans left and right slowly, to give some timbral interest - it's subtle, and the levels of each track aren't changing much, but they drift in and out of phase between mono and wide - probably some kind of autopan and thickening setting on the delay unit to give a little extra space and motion in the verses where the mix is a little sparser.
The second arpeggiated "guitar" part underneath was a little harder to lock down, which gives me an opportunity to make a digression on tracking down synth sounds. Synth digression number three...
The DX7 was ubiquitous on many productions through the mid-eighties, and it was also well-known for being difficult (or impossible!) to program (let's not get into why here, this blog is long enough as it is!) but the practical reality is, if you were a producer bringing keyboards into a session, you probably had a DX7 as it was fashionable and in demand, and you probably had a bunch of sounds, likely on cartridges, because the small amount of presets it came with were very overused, fairly generic in nature and not that flexible. Cartridges were about the only way to quickly expand your available sound palette without having to rig up a home computer and faff around with cassettes or MIDI System Exclusive transfers.
So, if you're hunting sounds that seem to come from the DX7, the first thing you do is look through the ROM presets to see if it was one of those sounds - because every DX7 has those. (In this case, the bass sound, and the above flute sound are both in the ROM preset bank that comes in every DX7).
If the sound you're looking for isn't there, then you start looking through the extra official Yamaha DX7 voice library cartridges, because a bunch of those are the first thing a jobbing producer is going to invest in from their Yamaha dealer to expand their DX7's sound palette - because they for sure won't be developing their own sounds, or probably even know anyone that is!
(Extra tip: If you do find a patch on a particular cartridge, then you can more reliably say "Ok, they seemed to have this cartridge available to them" so you can focus your search on cartridges you've deduced they presumably had on the session.)
If you still don't find a match there, then you can expand your search outward to the popular third-party companies offering cartridges of DX7 sounds.
There is a further level of search I might also try if I get this far without finding what I'm looking for. For anyone that's used a DX7 for more than just playing Whitney Houston covers, you'll no doubt know there is a quick way of generating some extra sound possibilities - it's almost like a "cheat" to randomly generate or explore new voices. Let's say you have a plucky sequence going in the studio which you like, but perhaps you want to explore some other tones and timbres - if you go into Edit mode, you can toggle operators on and off (which changes how each operator contributes to make the final sound) or, better still, you can flip through the algorithms, which are the different ways the 6 operators are combined. You can just increment or decrement the current algorithm setting (a number from 1 to 32) and just see how the sound changes as a result of combining the operators with their current settings differently. Hence with very little time & effort, you can do some limited "programming" to get some tonal options or variations of the current patch.
Now, in this modern age of plugin emulations and internet distribution, almost everything is available which is great if you're looking for something specific, but the sheer amount of sounds made for the popular synths out there is rather overwhelming. In the case of the DX7, I had already saved out all the available sysex banks to individual Logic presets for Dexed (a plugin that emulates the DX7) so I could search the whole library by name and by bank or category and so on (that was a major undertaking!) to make it quicker to find particular patches - but even so, my DX7 library has nearly 120,000 patches!
So, if it's a trumpet sound you're after, you can do a bunch of name searches on "trumpet", "trmpt", "brass", "brs" and so on and quickly step through them to track down a match. But when you have a synth sound that you can't easily describe in words - like the arpeggiated sound we're looking for, it's a bit tricker.
The arpeg sound is synthy, but it's plucky, is string-like in the context of a guitar, or harp or even into banjo territory, is percussive like a koto - but I have no idea what to call it. In any case, it's clearly a DX7 part, has a guitar-like attack, and a reverberant sustained sinewave-ish tail which is very typical of many of the DX7 plucked string sounds. It's one track, and therefore initially I thought it must be one single sound.
I spent a bunch of time looking for this patch, but didn't find an exact match. I've come close, but again, it's nice to try and find the actual source if that's possible. And while it's possible Tarney's DX7 was full of edited custom patches, again it's less likely than those voices just being existing presets from somewhere.
Without finding an exact patch in the obvious candidates as well as a deeper search, I went back to the main DX7 presets, and had another look at the ROM 1A "23 Koto" patch, which again has some similar characteristics to the recorded patch, despite being more "mallety" than "synthy". I flipped through the algorithms which didn't bring me much closer, so in the end I modified the stock Koto patch - made it less koto-ey, a bit more synthy and bitey, shortened a few envelopes and changed a few levels - fairly minor edits. I also needed to play the part in octaves, to get both the synth spikiness and the higher tails, and the end result isn't too far off. So yes, maybe they were digging into editing the DX7 in the session...
...until, looking for another sound, I actually stumbled upon a result that makes more sense - and doesn't require any DX7 editing! Going through the factory PPG Wave 2.3 presets, I found PPG Wave 014 A+B which sounded suspiciously like the spitty/bitey synth part in that sound. Dropping the decay down a little and with a few minor tweaks we get the spitty attack sound, and if we then layer the DX7 on the straight, regular Koto preset, we get basically a much closer composite sound (the original has a bit more of an attack, but was close enough for purposes of this blog). So, this seems to be a sequenced combination of the PPG and the DX7 Koto, layered and printed down to one mono track.
"There was no 'vintage drum machine' - it was our old friend the *DX7* all along."
This patch is ROM2B "22 B.DRM-SNAR" (or ROM4B "15 B.DRM-SNAR" - it's the same patch, just also on a different cartridge), and the percussion sound is interesting - it's a bass drum + snare sound, where on the lower part of the keyboard you get a kick drum, and on the upper part you get a snare, but in the middle, you get a gradual transition between the two, and this is exactly where the "guitar" sequence sits, by chance, to give this drum loop effect. You even get the extra snare accent in bar four on the last high note, which is more "snare-sounding" the higher the note gets. As the DX7 is not noted for it's realism in terms of contemporary drum sounds, you get that slight hokey/cheesy naff percussion thing, that on it's own sounds pretty terrible - but in the context of the track here, it's inspired!
I was really happy to answer this mystery by accident (I love these kinds of accidental discoveries!) Now, I'd guess that it's likely that what happened to me here, probably also happened to the band in the exact same way on the recording session - they were flipping through DX7 presets while playing the sequenced "guitar" part and drums, looking for a good sound to use, they flipped onto that percussion patch, and went "Oh, that's interesting!" and promptly printed it to tape. For the whole four minutes - it never stops!
And that leaves us one final synth part, which I always liked - it's a little simple brass solo lead part right at the end in the final fadeout of the song, just before a really nice "skip" drum snare fill - it one of those little interesting parts in the fade out that I liked to hear more of, and to hear it properly rather than it being buried in the fade. And yes, this part was also present at the end of the original demo, though somewhat more prominently.
In the mix, they left it to the fade out and it plays twice, but on the multitracks, it's a sequenced repeated figure that constantly repeats seven times (these tracks continue for about a minute after the final mix has ended). I certainly got my wish to be able to hear it more! At the end, the MIDI loop plays to the end which has a little bit of extra half-hearted playing that wasn't used as the song was over!
So, on listening to this solo'd, it's a really, nasty, horrible cheap brass sound - so I instantly thought it was from the DX7! It does a weird harmonic thing as you play different notes and frequencies phase cancel out. In any case, I went back to the usual DX7 preset search procedure (see above!) and although I found a few close matches, it wasn't conclusive until I found the Yamaha cartridge patch VRC-102(B) 26 - BRASS EN.3 that matched both tonally, and with the correct attack, and also does that weird harmonic thing that sounds like a broken synthesizer when you play multiple notes. Awful sound - but totally works!
But, as we're starting to see in this production, there's almost no synth parts that aren't a layer of multiple sounds. Even this track actually consists of two parts - the DX7 brass sound, and also a simpler analog-style synth brass sound with a bunch of vibrato. I quickly initialised a new sound on the PPG Wave and with some very minor tweaks had something that does the job just fine. It could have just as easily been the Juno 60 on the record - the sound is so simple it can be done by more or less any polysynth, so it would be difficult to know exactly which it was.
Ok, having got this far, the last remaining tracks are for vocals. I'm not going to dwell too much here, there are two tracks for each of the three main vocal parts, and they sound glorious - particularly the washes of angelic "aahs" that feature in the chorus.
"There's a fine line between a demo and a record, but it's a crucial difference."
So what did I learn from all this? Well, as I stated at the beginning, my fascination for this track began with the drums, and I finally have a good handle on them now - the recipe of how that fabulous snare was made and shaped, and what that driving eighth-note pulse was that propels the beat forward, and possibly even how that idea came about.
I've seen how a relatively mundane LinnDrum drum pattern can be morphed into something that sounds great, distinctive and certainly hit-worthy. In fact, they were presumably so pleased with the way the drums sounded that the team decided to open the beginning of a three minute hit pop song with four bars of just those drums as the first thing you hear - and it still manages to be an instantly recognisable intro!
And I still smile at the thought that such a cool, fundamental part of the drum track is actually that cheesy DX7 preset! Much like some other classic instruments, good engineers over the years have worked hard to make these things sound great in the mix - no small task in many cases!
I've been able to look at many of the little keyboard parts I liked, and analyse the source components to the best that my ears can manage. It's fairly clear that the only instruments used on the record were the PPG Wave, the DX7, and the Juno 60, with the LinnDrum doing the drums. There was nothing else on the multitracks I could see that would require some different instrument to achieve, and that corresponds faithfully to the information in the SOS article on the recording of the track.
I've been surprised at how there are no synth parts that aren't doubled, layered, thickened, chorused or stacked in the track. The one or two simple mono parts, like the white noise swoosh and the percussion loop, still seem to have some doubling or chorusing on them to make them just that little bit bigger. Even the bass part is chorused in the mix! I'd guess that this practice of thickening and doubling of the instruments came from Tarney, in order to make the sound more impressive and more like a record - and in fact the band mentions that Tarney did try to make all the sounds bigger and more impressive in the "Making Of" YouTube video for the song.
This track really has the strong signature of the PPG Wave - the chords, bells and high synth motifs were always really attractive to me, and those parts could not easily be replaced by any other synth and maintain the same feel. It's a kind of cold, mysterious vibe that gives some weight and interest to an otherwise lightweight, super-fast pop song.
This character, and it's contributions to many records I love is one of the reasons why the PPG Wave is my second favourite classic synth. It's a kind of quirky, unreliable beast that's not easy to tame, and I'm not so much a fan of it's glitchiness and quirks, but the digital wavetable oscillators and analog filter combination makes it really easy to produce a wide range of colours far beyond the more simple subtractive analog synths of the time. That, and it's blue, of course. Blue synths are cool.
There were a few things that massively helped me in identifying or approximating the PPG sounds used. Firstly, I tend to save out presets from synth plugins as presets in Logic, so I can more easily step through, organise and search for sounds in a consistent manner, regardless of the features any particular plugin supports. Waldorf's excellent PPG Wave 3.V plugin comes with all the PPG Wave individual presets, but only as single layer patches, which limits how good it can sound out of the box. When you use a dual-layer of the A patch and the corresponding B patch, and spread them in stereo, it sounds way better - the preset patches were designed to be used this way. And because I had automated saving out the stereo layered versions of the preset patches I could quickly step through those easily - a great timesaver.
Also, I map the control of my plugins in a consistent way, and in the case of the oscillator controls on the PPG, I have knobs to select the wavetable used, the waveforms for Osc1 and Osc2, the sub-oscillator mode and so on, so when I'm looking for tonal matches, I can do it easily and systematically by sweeping around using multiple hardware controls while playing, just like on the real synth. Doing this onscreen by mouse control is just tedious and painful.
While I would have liked to have got closer to the actual PPG sounds used, this was always going to be a tall order, and from time to time I'll probably continue to search and refine those sounds to home in on where they are. I wonder if Mags' still has his PPG Wave and those original patches saved somewhere, or if those sounds used are lost to him too?
Oh - and I've learned to embrace voice stealing! I've definitely re-assessed some of my opinions about this, and while I do think in most cases it's a quite unattractive-sounding behaviour, I have to admit that there are some positives to it too, and the many synth parts in the track are good examples of this, given the density of the synth layers...
...and despite my feeling that the mix was always a "light"-sounding one, the density of synth layers present was quite surprising to me - the chorus has three complete sets of bass root note + chords from three different instruments (and two of those are dual-mode layered sounds), double-tracked, as well as guitar, and those fabulous backing vocal washes... and yet Gerry Kitchingham's skill at mixing those elements without swamping the mix is clear, albeit no doubt helped by the general good arrangement and sound choices made, which always makes the mix process less challenging.
Speaking of the good arrangement, I'd like to reiterate that this song took two days to go from nothing to a finished mix. They came in on day one with nothing other than perhaps their synths and a few sound programs, and a copy of their Rendezvous demo, built the arrangement out from scratch, programmed the sequencer and all the drum parts, chose and experimented with sounds, recorded all the parts, doubled them, recorded the different guitar parts, backing vocals, and lead vocals, all on day one. This is the benefit of going in to the studio with a well-rehearsed good arrangement and knowing what you are doing, being excited about how it's shaping up, and playing off that energy to achieve good results quickly without getting bogged down into micro-detail.
The engineering and production is all the more evident when you look at how much effects processing was being done to most sounds on the way to tape, and then processed again for the final mix. They had the forethought to think about the context of that sound in the arrangement, and how best to augment it at the time of recording, and creatively use delays, chorus, thickening, layering and compression, commit those things to tape and move on. This is quite different to today's more common working practices of never committing anything you can't undo and making all those decisions at the back end of the process during mixing. There's actually surprisingly little evidence of regular echo-type delay effects in the mix - the ambiences are all from the reverb unit, with the AMS delay mostly being used for it's doubling/thickening capabilities, both to tape and again in the final mix, rather than straight echoes - the only really long echo is recorded to tape on the swirling arpeggiated FX sequences in the middle eight and on the end of the track.
And I guess I regarded - certainly like many people of the time - A-ha as yet another good-looking, lightweight pop band. "For all their teenybopper image, A‑ha were three fantastic musicians," Kitchingham notes, and on reflection I have to agree. The level of detail, maturity, and care taken with the drum programming, synth arrangements and voicing, and of course Morten's vocal chops, says to me that these guys really knew what they were doing even at that young age. I mean of course, it's "just" pop music and is not hugely musically challenging as such, but even so, you've got to recognise that they weren't just phoning it in and getting everything done for them by their producer.
The band must have been super happy with the outcome of their first recording experience with Alan Tarney - a hit single that kick-started their tremendous pop success - as they stuck by him all the way through their first three albums and the most successful stage of their career. Interestingly, during that period, the other artists Tarney was producing were Cliff Richard, David Cassidy and Leo Sayer - how's that for pop cred!
So how did I do? I'm pretty happy with my findings, particularly on the drums as it was a long-time curiosity for me as to how they got that drum sound. I'll very likely use some of the techniques they used here in my own tracks - whenever there's a chance to observe and deconstruct skilled work I regard it as a learning opportunity!
I wonder if that particular LinnDrum and it's box of chips is sitting in a flight-case in some storage unit somewhere, and if the snare chip really is "Snare 7", or whether it was actually something completely different? Or maybe someone out there - or perhaps it's still in Tarney's possession - has the actual 'Take On Me' LinnDrum in their collection? It would be great to find out for sure!
As for the synths, while it's fairly straightforward to find things like DX7 presets used, for analog synths where they weren't using presets, the best we can do is approximate. That main opening PPG sound is a really important part of the track, and while I've been able to refine it and get closer during those process, I haven't been able to nail it just yet - my closest attempt is still a little too smooth and not digital/belly enough - if I make any further progress on that, I will update this blog at a later date.
The process of matching synth sounds on a record is challenging because most people try to emulate what they hear in the mix, on their synth, without quite realising that they are not hearing the sound as made by the synth, but after it's been treated and processed and EQ'd and compressed and mixed - and sometimes these treatments can take the sound quite far from what was originally recorded. Being able to hear the tracks as recorded to tape is a big help, as although in many cases these have also been processed to tape, it's usually easier to extract more detail about the source sounds. But reconstructions in this case are two fold - firstly, I had to match the sounds the first time to the recorded tracks, but having done that, then match them a second time to how they ultimately sound in the final mix (at least the mastered version) - the main riff and the drums in particular, sound quite different in the mix to the recorded tracks. The main riff is quite heavily EQ'ed to bring out the mids a lot more than the raw synth recordings would otherwise suggest, for example.
In some of the more difficult cases of matching the recorded tracks, I'm sure I could have got closer tonally to the recorded parts if I "cheated" and used other tools available to me, like more sophisticated EQ's and so on, but my goal was never to do a perfect sounding recreation, the goal was to understand those sounds and recordings, and to use similar tools that they were using to inform that understanding. And while I could come pretty close to the snares recorded on tape, I couldn't really come close to the final drum sound in the mix, so perhaps there's still some secrets in there that will remain undiscovered for now!
So, the demos and mockups are really to demonstrate my findings, rather than to claim any sort of accuracy to the final record. I always find it very hard to recreate someone else's mix anyway, at least without working on it endlessly in ever more detail, and that really isn't the purpose of this blog series. It would be interesting to find out whether I can even approach that final mix using the original multitracks directly... I suspect it would be nearly as hard as doing it from scratch!
So take the following mockup with that in mind...
Why deconstruct a piece of popular art to it's constituent parts? How useful is it? I can certainly see the argument that some people may find doing this kind of pulling apart of the work can ruin the magic of the final piece - rather like knowing how the magic trick is done can spoil it for people.
For me, it's not like that all. Knowing how the trick is done in no way diminishes the trick for me. If anything, it makes me appreciate it all the more. Even if the "gimmick" is something simple and perhaps even a bit of a cheat, then it just makes me admire the skill of the design, construction and execution of the performance, and the misdirection that makes the gimmick work.
For me, a piece of music truly becomes a piece of art the moment it transcends the sum of it's constituent parts. When it's no longer just some sequences of synth sounds and drums and chord progressions, or a bunch of sounds, but in combination, in purpose, and also often by some ineffable "magic", becomes something greater - a meaningful work that evokes and communicates some emotion in the listener.
Looking closely at those parts doesn't really impinge upon, or reduce the final product for me, but it does give me insight and ideas into some of what went on to make that final product. I learn a lot from doing this, and it makes me understand and appreciate it more deeply. In this way, it has value for me, and I hope, anyone reading along that made it this far!
The next time you hear the song, try listening out for some of the details I've raised here, and hopefully, like me, you'll appreciate it's quality all the more!
Thanks for reading!
...with East End voices | Sep 2021
Another milestone | Jul 2021
Blog entries from 2020...
Part 6 - OCR Part 1a - Contents & Metadata | Apr 2020
Blog entries from 2019...
Synth Patches - The Return | Dec 2019
Part 5 - Outputting the Scans to Use | Nov 2019
*Almost* the first DAW... | Oct 2019
...and Three New Things (Polyphony, Ads, & Stats) | Mar 2019
Part 4 - Processing the Scans | Jan 2019
Follow mu:zines on Twitter: @mu_zines
for updates and other bits and pieces