Analogue modelling

Forty-nine analogue plugins do not add up to a night and day difference#

The myth goes like this: one instance is not enough, but put 20 or 30 across a session and all the little changes add up until the mix is completely transformed. So I put it to the test with one of my own songs.

The analogue mix had a tape machine emulation on every audio track, a Neve-style channel strip on every track and two buses, and an 1176 emulation for every compressor. That is 49 instances of paid analogue-modelling plugins. For the digital mix I took all of that off and matched it using only free plugins, mainly the stock Reaper EQ, the one a lot of people hate on, and my free compressor. No saturation or distortion anywhere in that mix. Both mixes shared the same free plate reverb and slapback delay.

On good monitoring I could hear very subtle differences on the snare and the lead vocal. But a 1 dB boost on the digital EQ would do more than those 49 plugins did. I expected a bigger difference and I was surprised at how small it was.

514 listeners mostly picked the wrong mix#

Then I let other people decide. I played the two mixes blind in a video, put uncompressed 24-bit files in the pinned comment, and asked which mix was the analogue one. 514 people voted.

There was a statistically significant tendency to misidentify the analogue mix. What people expected to hear from the tape emulation and the Neve-style EQ was actually a stock EQ. What they expected from the 1176 was my free compressor. So people could hear a difference. Their idea of what "analogue" should sound like just doesn't seem to match what the analogue-modelling plugins actually do.

Hearing a difference is not preferring it#

Which mix you can identify and which mix you prefer are different questions. You might spot the analogue mix and still prefer the digital one. See blind testing.

On preference there was no statistically significant result either way. So whatever sound "open, rich, vintage, authentic" is meant to describe, there is no evidence, at least none I have seen, that people actually prefer it. The full method and statistics are in the PDF. It has not been through academic peer review, but the study is free to download, so try to replicate it with your own audience and make a response video.

Stacked emulations do not compound#

In my EQ tests I matched analogue-modelling EQs against a plain digital EQ and listened to the delta. Then, to simulate the cumulative argument, I boosted that delta by 20 dB, which is like stacking ten identical instances with the same setting on the same sound. Nobody mixes like that. Even so, on real music most of the differences stayed under the just noticeable difference. Pink noise showed more variation than music. In a real session every track has different settings, so the errors do not stack that way anyway. See EQ.

Free plugins and a plan reach further#

Unlike EQs, I do think compressors have different flavours. But unless a compressor does something radical with feedback, you can simply put a distortion unit after the compressor and get the same effect. I pushed an LA-2A emulation to its most extreme, most distorted setting and set it against a free chain: stock EQ, my free compressor and a free distortion plugin. It should have been night and day. When I revealed which was which, the clip some people called warmer and fuller in the low end was my free chain. Going the other way, I built a drum sound from a clean compressor, pre-emphasis EQ and distortion, and the most extreme setting on an analogue-modelling compressor could not get anywhere near it.

I don't want unspecified mystery distortion bundled into my compressor. I want explicit processing I can control.

The scam is the adjectives, not the modelling#

Is it snake oil? Not really. The developers are not lying. They genuinely try to model the hardware and maybe they do a good job, and they have to make money somehow. My problem is the copy: "warm", "silky top end", "honey oozing out of your speakers". That hyperbolic language is there to make you think it will give your mix a different colour, and it sets you up to hear colour that may not be there. The pretty vintage interface does the rest, because what you see changes what you hear. See perceptual bias.

In my personal opinion most of this stuff is just not really worth the money. It will not make or break a mix, and turning tiny knobs on a skeuomorphic EQ is a lot slower than using a normal one.