Hardware v ITB

Can plug‑ins ever match the sound of outboard and an analogue console? Our unique comparison recreates a 48‑channel hardware mix using software alone.

Hardware versus in‑the‑box mixing has been a hot topic for almost as long as any of us can remember, but I don’t think there has ever been a true A/B comparison. Has anyone completed a full 48‑channel mix using hardware, then meticulously replaced it all with plug‑in equivalents, matching the levels and carefully replicating EQ curves — not just using equivalent plug‑ins, but also using test tones and spectrum analysers in an effort to make them sound, as far as possible, identical? I don’t think so. In fact, I don’t believe there has been an A/B test of any sort on a full 48‑channel mix. Until now: I’ve done just that, and in this article, I will take you through the process I used and report what at least some of you should find to be pretty interesting findings. But first, it’s probably helpful if I explain just how I got to the point where I would be insane enough to attempt such a painstaking comparison...

Hardware Crazy!

I’ve used hardware by preference for a while. Earlier in my career, though, I mixed exclusively ‘in the box’, and never quite managed to get my mixes to that point where they sounded ‘like a record’. I was happy that I knew what I was doing, and was certainly getting good results. But I still felt that the final product could sound better, so my thoughts turned to hardware — and when I saw a used SSL XLogic SuperAnalogue channel strip in a music store at a bargain price, I bought it. Back in the studio, I used the hardware insert feature in Pro Tools and, lo and behold, my lead vocals now sounded like a record. Whoa!

The author’s Neve 8424 mixer, with some of his SSL X‑Racks above, and (below) just some of the outboard he has acquired over recent years in his hunt for sonic perfection!

Of course, that bargain turned out to be my most expensive buy ever! I called the store to ask if they had any other analogue gear in stock and, sure enough, they had a Neve 8801 EQ as well. When I first used that, I almost cried, and my gear obsession then very quickly ran out of control. I wanted more and more channels of analogue processing, and tracked down the last remaining stock of what I consider an extremely underrated product line: SSL’s X‑Rack, a modular system they created before the 500 series became ubiquitous. The UK prices were fair and the exchange rate with the British pound favourable, so I ended up with five X‑Rack chassis, each filled with modules, and during the next 18 months, I tracked down even more units from as far afield as Mexico.

Of course, I needed a lot of A‑D/D‑A converters to send audio between my DAW and all these real‑world channels and, although I’d never really bought into the analogue summing hype (it might change the sound subjectively, but I don’t perceive any ‘shortcomings’ in digital summing), I figured that with all this audio already in the analogue domain, it would make sense to have an analogue summing option available so I could mix in the analogue domain, or pass through fewer stages of conversion. To that end, I ordered a Neve 8424 console (reviewed in SOS October 2020), a versatile, high‑quality mixer with no onboard mic preamps or processing. That was perfect for me, since those bases were already covered by my outboard.

This brings us very close to today, but there was one more step in my productions that compromised two and half albums for me! In recent years, I’ve been working with a rock project called Symphony Of Sweden, which has had some minor success on Spotify. Even though I was mixing their tracks in the analogue domain, I felt that my mixes lacked the openness and the airy high end of the best albums. Partway through the third album, I tracked this down to my mastering processor, Softube’s Weiss MM‑1 Mastering Maximizer plug‑in (which uses exactly the same algorithms as the legendary Weiss DS1 Mk3 hardware).

If you’re reading this, Softube, don’t worry, because I have much better things to say about this plug‑in later! But, at this stage, MM‑1 was causing real problems for me. Until the point at which I captured the final mixes, everything was sounding great. I used the Neve 8801 EQ and SSL G‑Bus compressor on my analogue master bus and then, as the final processing stages, added the Weiss MM‑1 and a dither plug‑in before bouncing the final 16‑bit/44.1kHz master. Part way through the third album, I listened very carefully and could hear that the Weiss plug‑in was doing something I didn’t like: the energy it added was great, but it also ruined the high end for me. To solve this, I put my trust in analogue gear again: I was one of the first in the world to buy the expensive‑but‑wonderful LAAL, an analogue look‑ahead limiter made by Polish company HUM Audio Devices (reviewed in SOS February 2024). Half of the songs on our third album, Haunted, have the LAAL in place of Softube’s MM‑1.

Oversampling Awakening

At this point, I was a happy camper: having chosen my analogue path, I could finally make mixes and masters that I was genuinely proud of — that sounded ‘like a record’ — and I could easily have continued with that approach. But, about a year ago, I was making a sound library for a plug‑in/virtual instrument company, and in the process, I discovered something that would make me revisit this analogue/digital debate.

The core idea for this library was to use my analogue synths, but I also wanted to include one of my Firechild trademark sounds from a PPG Wave 2.2. As I didn’t have that synth any more, I tried to create it using Waldorf’s plug‑in version instead. Initially, I thought it seemed nothing like I remembered the hardware sounding, but for reasons I can’t really explain — it was just pure luck — I gave it a shot with Pro Tools running at 88.2kHz instead of 44.1kHz. The PPG V3 plug‑in came alive at this sample rate. “What the heck?!” I thought. “Is this the solution to all my problems?” I tried the same thing with Arturia’s range of synths, which I didn’t think had sounded great at 44.1kHz, and yes, it worked for these too! Of course, many soft synths already feature internal oversampling and, for example, with Xfer Records’ Serum the difference to my ears is huge. But could I apply this ‘oversampling’ to standard audio plug‑ins as well? And might it give me the same improvement?

Back in the day, I had recorded some songs at 88.2kHz, but had found it a hassle. I needed to use double ADAT cables for eight channels, projects took up double the disk space and used double the CPU power, and I was always having to sample‑rate‑convert third‑party audio loops. At that time, the outcome just didn’t seem to justify that hassle, and I’d remained on 44.1kHz ever since. But I now had good reason to revisit the issue, and was pleased to find that working at higher sample rates has become much more convenient in the intervening years.

Obviously, lots of plug‑ins now feature internal oversampling options, but it turns out that they don’t all do a transparent job of it. If you send pink noise through a plug‑in and then use a spectrum analyser, you can see in some that the high‑cut filter used to eliminate aliasing also brings down the audible high end. Different plug‑ins sound better on different oversampling multiples (some offer more multiples than others), but the perfect ‘middle way’ seems to be 2x. I was a little late to the party, but my experiments led me to DDMF’s Metaplugin, which has been a lifesaver. Metaplugin is itself a plug‑in, but it can host and chain other plug‑ins in a really flexible way. Importantly, it can also oversample the whole chain while your DAW still runs at 44.1kHz, which makes oversampling a lot easier to manage.

The sound of many plug‑ins benefited from finding the optimum oversampling setting, often with help from DDMF’s Metaplugin. Oversampling uses more computing resources, of course, so more use was made of this on critical signals, such as lead vocals and, in particular, the mastering chain.

So, how about that ‘evil mastering plug‑in’ — did oversampling improve the performance of Softube’s Weiss MM‑1? It sure did. More specifically, when hosting MM‑1 in Metaplugin and running at 16x oversampling, it sounded like a dream (I don’t know why, but this particular plug‑in definitely works best at 16x!). Having spent what I did on the LAAL, I should probably have been crying at this point — but I was definitely smiling! Don’t get me wrong, the LAAL is a wonderful analogue processor and it sounds great, but on hearing the MM‑1 plug‑in do for my mixes what I’d always hoped it could, my world turned upside down in an instant!

The final piece of this revelatory jigsaw fell into place just a couple of weeks before writing this article. I was having dinner at a friend’s house, where he has a little music studio in the back yard. I thought of him more as a virtuoso keyboard player (he’s the first call for that here in Sweden) than as a producer, but he played some mixes he’d done recently and I fell off my chair. These were 100 percent in‑the‑box mixes and while he employs some interesting tricks (such as using Studio One’s binaural panner to make the mix really wide and spacey, which impressed me), overall his mixes just sounded great.

Between that experience, and the results of my own initial tests with oversampling, I had to know if I could mix as well entirely in the box as I could with my collection of analogue gear. So I conceived an insane, hugely time‑consuming experiment: I would create an all‑digital carbon copy of a complete 48‑channel analogue hardware mix, replacing every single bit of hardware with plug‑ins, and summing in the box.

The Hardware Mix

The mix in question is a reworking of the Symphony Of Sweden song ‘She’s In My Head’, first released back in 2020. (For this new version we added the word ‘Reloaded’ to the title.) We used the original vocals but chose to re‑record most of the instruments, including the drums, for which, of course, we used high‑end hardware, including API 512C mic preamps, and classic ‘usual suspect’ drum mics such as AKG C414s, SM57s, Neumann KM184s, and so on.

When it came to mixing, I started by performing a full mix with outboard gear only — a ‘big console’ kind of setup — with the sole aim of trying to create a good, engaging mix. Forty‑eight separate channels came out of Pro Tools through my Avid HD I/O converters and they were each routed through an analogue EQ, which, in turn, fed an analogue compressor. Five of these channels had analogue de‑essers at the end of the chain (the routing in the studio is hardwired this way) and I used some other gear where the mix required it, including some SPL Transient Designers on the drums. If you’d like to see more detail about precisely what processing was used, check out the chart towards the end of this article. There, I’ve listed all the hardware I used for the first mix, along with the plug‑ins that replaced it in the second.

After the analogue outboard channels, the audio signals were fed to the Neve 8424 console. Though ostensibly a 24‑channel console, this has are multiple inputs, and in this case I routed 24 channels to the console’s 24 main input channels, and the rest to the 24 ‘input C’ channels. The former fed the stereo bus directly, whereas the latter were summed by the stereo cue bus, which in turn was routed to the master bus. However, before hitting the C inputs, those 24 signals passed through a pair of Wes Audio ngLevelers (which SOS reviewed in March 2023). These analogue, VCA‑based, automatable level controllers allowed me to perform detailed level automation using my DAW, complete with recall (though for the C inputs only). I completed the mix with this analogue setup as I normally would, though digital processing was involved in the form of my send effects, which included Lexicon, AMS and Bricasti hardware reverbs and two Universal Audio DSP‑powered reverbs: a Lexicon 480 and an AMS gated reverb.

The Software Challenge

Once I was happy with my outboard mix, my challenge was to achieve the same result working entirely in the box. I did not want to improve the mix, and neither did I want to prove that analogue was better: all I wanted was to see just how similar I could make the software mix sound to the hardware one. I hope you think I’ve been meticulous in my approach to this, because it was not easy! It took me a full four days to get this mix right, just like an old‑school mad scientist. I stopped only to eat and to sleep (pretty badly!) and at one point I told my wife I was slowly going insane!

When it came to EQ, rather than rely on listening alone, I first ran pink noise through my analogue EQs, and used the Crave DSP EQ2 plug‑in, which has a great spectrum analyser, to get a visual ‘stamp’ of the hardware EQ settings. I set the analyser to its longest averaging time, which made it easy to see just how I should set the controls on the plug‑in to match the curve of the hardware EQs. Many channels hadn’t required particularly extreme settings, but the drums were a big challenge, in terms of both EQ and dynamics. For the hardware mix’s drums I’d mainly used Neve 8803 EQs, but I didn’t have an 8803 emulation plug‑in, and I chose to use the Brainworx bx_console SSL 9000 J to emulate the Neve settings. This worked very well, although in a few cases (mainly kick, snare and tom close mics) I was surprised by just how extreme the plug‑in settings needed to be to match the hardware EQ’s curves.

For compression, I found that I could sometimes just dial in identical values on equivalent analogue‑modelling plug‑ins, and I must say that Universal Audio are really nailing it with many of their plug‑ins (the Distressor and API 2500+ in particular really are spot‑on emulations of the hardware!). For other compressors, I put the analogue hardware track on the left and an emulation plug‑in on a duplicate track on the right, so that I could then see on the meters if I needed to tweak the ratio, attack or release. When I had both level meters on the master dancing identically, I knew I’d matched the compressor settings pretty well (though, obviously, I listened to the effect to confirm that).

The most time‑consuming task — but also the most interesting to me, so it’s what I chose to work on first — was definitely to pin down the drum sound. Not only did that involve the most radical EQ settings, as I’ve explained, but I’d also used an SPL Transient Designer on four tracks. There are plug‑in versions that do a similar job, but I found that they were just not giving me quite the same ‘punch’ as the hardware. To compensate for that, I found that I had to be a bit more assertive with those channels’ compressor settings. And, of course, as well as the channel dynamics I had to work on the drum bus compression, where I’d used two compressors in series.

Though the original mix was done using only outboard, there were several digital effects involved. Among the highlights when it came to the plug‑in replacements were a Relab reverb, which did an excellent job of replicating the effect of not only the Lexicon but also the Bricasti hardware, while Wave Alchemy Glow was an admirable stand‑in for an AMX gated reverb.

The next most time‑consuming task was the reverbs. Reverbs are really important for my mixes, so I wanted them to be totally identical, and I happened to use three hardware reverbs on this song. I had to clone two Bricasti M7s and one vintage Lexicon 300 reverb. I also chose to stay away from ‘DSP‑powered’ plug‑ins for this test and stick with native processing, so I needed something to fill in for the UA reverbs, too. Incidentally, I’m not a big fan of convolution reverbs for music production, as they don’t quite offer the same complexity of texture. Happily, I found that the Relab LX480 v4 plug‑in could match the sound of both the Bricastis and the Lexicon 300. I found that extremely interesting — not so much with the Lexicon 300, since that’s ‘half a 480’, which is what the Relab recreates, but that the Bricastis were so close in sound to the 480 was a surprise. (Brian Zolner and Casey Dowdell must have listened to the 480 a lot when designing their M7 algorithms!). As a substitute for the UA AMS gated reverb, I bought Wave Alchemy Glow, which sounded great, and for UA’s 480L I just dialled in similar settings in Relab’s version.

Mastering Chain

Of course, I still had to replicate the analogue mastering chain, which I usually apply during the final mixdown to avoid unnecessary A‑D/D‑A conversion. This comprised a Neve 8803 EQ, an SSL X‑Rack G‑Bus compressor and, finally, the HUM Audio LAAL.

I tried a couple of EQ plug‑ins but, again, my focus was not on finding accurate models of the hardware units, but rather on matching the sound. My eventual choice was Plugin Alliance’s SPL PQ plug‑in, and to match the curve I again ran pink noise through the hardware EQ, first matching extreme settings one band at a time, so as to better understand each band’s frequency and Q values, and then backing it off again before replicating the EQ settings I’d used for this track. I’d used very small amounts of EQ really, with a 1dB dip at around 280Hz, another 1dB cut around 3.5kHz, and just 0.7dB of high‑end boost.

For the G‑Bus compressor, I opted for the UAD version and dialled in the same settings as on the hardware: the slowest attack with a 2:1 ratio and auto release (yes, the classic master‑bus setting that always works!). Finally, to emulate the LAAL I turned back to the Softube Weiss MM‑1 Maximizer. To my ears it sounds a bit like the LAAL, actually, and matching the LUFS level was not difficult at all, but when listening by ear there was a touch of low bass missing, and perhaps some subtle analogue distortion going on in the LAAL too. I looped one bar and used the Crave EQ2 to measure the EQ curve, and found that in order to match them, I had to boost 0.5dB with a 96Hz peak filter. I then dialled in a little saturation courtesy of the SPL PQ’s THD feature until I was happy.

To get the right result, I ran the SPL PQ and the UAD G‑Bus compressor in Metaplugin set to 2x oversampling, while another Metaplugin instance hosted the Weiss MM‑1 at 16x. But while the mastering is kind of close, you can see that as a bonus — my focus for this test was very much on seeing whether I could recreate the mix.

The above table shows the different sources in this song, and the processing used for both the hardware and software versions of the mix.

Takeaway Thoughts

The expression on the author’s face when it dawns on him just how good software can sound...To my ears the two mixes sound very similar, and if you’d like to hear them yourself, you can find the files in the box below. I was — and still am — surprised, upset, sad and happy at the same time. After all, I invested a lot in hardware after my earlier experiences with plug‑ins!

As I’ve hopefully proven through these tests, when you have a target to aim for you absolutely can reach it when working 100 percent in the box, as long as you take the time to learn your plug‑ins and try them with different oversampling settings. Metaplugin really has been a lifesaver for me, particularly for critical instrument and synth parts, but many processor and effects plug‑ins also sound better in Metaplugin running at 2x, 4x and sometimes even higher multiples. I also like that when using plug‑in chains Metaplugin reduces the number of stages of oversampling (and associated filtering) compared with using plug‑ins’ built‑in oversampling options.

But while I could match the sound to my satisfaction, working in the box was a very different experience from working with hardware. In particular, I found it interesting just how extreme some of the plug‑in settings had to be to match what the hardware was doing — this test has given me more confidence to turn those virtual knobs all the way if it genuinely sounds better. (This even extends to panning, by the way: a ‘three o’clock’ setting on the Neve console’s panner sounded almost like maximum right on the Pro Tools pan knob — this probably explains why panning in Pro Tools has previously led me to mix rather narrower than when I mix on the Neve!)

While I could match the sound to my satisfaction, working in the box was a very different experience from working with hardware.

I don’t think I would ever have been able to make a mix like this in the box without the prior experience of working with hardware, either. Learning the hardware boxes, and the nature of the physical interaction — being able to dial in the right settings with such ease — has been a great school for me. It encourages you to judge things by ear, without the distraction of what’s on the screen. I’d also dare to say that analogue hardware generally has a wider sweet spot, and that means that almost everything you dial in will sound good in a strange sort of way, making it easier to achieve a good mix more quickly.

I imagine that the big question many of you will be asking right now is this: am I now thinking of selling my analogue hardware? No, not just yet. But if my studio were on fire, I would now almost certainly rescue my computer before my X‑Racks, and that was definitely not the case six months ago.