Flashback Friday: Mark Barton on Analog and Modern Digital Synthesis

Our "Flashback Friday" series looks back at guest articles and interviews from the past five years at Cherry Audio. First up, from 2019: Mark Barton makes the case for digital synthesis in software being "more analog than analog" in hardware.

Mark Barton is a brilliant audio engineer, with a list of accomplishments a mile long. In the 1970s, Mark designed the Pollard Industries Syndrum, which defined the sound of electronic drums for the disco generation. He's also done groundbreaking work in speech synthesis, including writing the MacInTalk text-to-speech system for the first Apple Macintosh computer, and designed the innovative Zeroscillator hardware synthesizer module for Cyndustries. For the past five years, Mark has been brining his talents to Cherry Audio with modules and bundles for Voltage Modular under the MRB Labs and Cherry Audio/MRB brands, and has contributed to the wildly popular instruments Miniverse, Novachord + Solovox, GX-80, Lowdown, Eight Voice, Rackmode Signal Processors, and, most recently, Synthesizer Expander Module and Pro Soloist.

Why Modern Digital Synthesis Is More Analog Than Analog - by Mark Barton

I have been designing analog synthesizer electronics for much longer than I care to admit. I built my first modular system from scratch in 1973, and I’m still engaged in the practice. I like digital stuff too, and both certainly have their place. The complement of synthesizer modules in most peoples’ systems is generally a mix of the two. The purpose of this article is to dispel some myths and bring a little insight into what it means to be digital. I will not get into the religious argument about which is “better” or the joy to be had by turning real knobs or plugging in patchcords.  These are the synthesists’ physical experiences and have little to do with the resulting sound, or the ultimate goal: the listener’s experience.

When I started authoring modules for Voltage Modular, I came upon the realization that digital synthesis in the modern age was actually more analog than analog electronics. I’m going to make that case here, and perhaps after you read this, you will agree. I promise no transistors were harmed in the writing of this article.

First, let’s discuss the difference in audio quality and get that out of the way in a hurry. Have you ever listened to a recording of an analog synthesizer on a CD or your computer? Did it sound all right? Was there a discernible difference between the live monitoring and the digital recording?  I thought not. OK. Done.

So now what do we mean when we say “analog”? The first online dictionary that popped up defines analog as: “relating to or using signals or information represented by a continuously variable physical quantity such as spatial position, voltage, etc.” That is the modern definition of the word, all set to be used next to its antonym buddy, “digital”. The Merriam-Webster dictionary has a better definition which is: ”something that is similar or comparable to something else either in general or in some specific detail : something that is analogous to something else”. So you see, both methods of modelling sound are analogs. One uses a variable quantity of electrons (voltage), the other uses variable numbers (values) to represent sound. So which is a better analog?

I’m not going to engage in hairsplitting or have someone accuse me of playing with semantics, so I will stick with the first definition. I’m doing so because that definition has the magic words in it: “continuously variable”. This is where folks believe the two methods split. One is continuously variable, the other is not. We’ll see about that.

On a good, no, a very good day, analog synth electronics can manage 100db of dynamic range. That means the signals inside can vary by about 100,000:1 from the smallest meaningful signal (not just noise), to the loudest. On a Eurorack system, that’s a voltage range of 10 volts down to 100 microvolts (millionths of a volt). Wow, sounds pretty good, and it is.

Early 8-bit digital sounds (like on the Fairlight CMI and other early samplers) had a range of 256:1, or 48db. Horrible next to analog. 16-bit digital, and even 24-bit floating point digital has an accuracy range of 65,536:1, which is 96db. Getting better, but still not quite as good. I’ll cut to the chase. Modern digital arithmetic, such as is used in Voltage Modular, uses the IEEE 64-bit floating point format commonly known as “double precision floating point”. Whatever does that mean? That format represents a value using 53-bits for the number itself and 11-bits for the position of the decimal (actually binary) point. That translates to about 15 decimal digits with the decimal point able to swim around a bunch. So numbers like 577,534,038,953,904, 0.0000000000849372894872037, and 943,634,928,264,836,000,000 are all representable. You can see that in each number, there are 15 meaningful (significant) digits at all different orders of magnitude. That’s what floating point means.

Let’s forget about the decimal point for now, because that only scales a number up or down and has nothing to do with its accuracy. It’s the 53-bits that is important, so let’s get a feel for that. 53-bits represents a difference of 9,000,000,000,000,000:1. Yeah, that’s 9 quadrillion to one, from the smallest to the largest quantity, or a whopping, no, staggering, 319db, and we haven’t even moved the decimal point yet! If we squish that range into the 20 volt range of a Eurorack system, that means that the smallest quantity that can be represented or “modelled” is 20v / 9,000,000,000,000,000 or 2 thousandths of a trillionth of a volt. A microscopic number well beyond human comprehension.

So from the signal-to-noise perspective, modern digital is 219db better than analog at doing its job of modeling a quantity faithfully. Oh, that’s just 89 billion times better, (no charge for the sarcasm). If an analog signal bumps and wiggles around noisily 89 billion times more than a digital signal on its journey from one value to another, which is more accurate, smooth, and continuous? Which is a better analog?

I’m actually playing a little game with the concepts here because that incredible 319db of dynamic range can never be output into the real world. We are still at the mercy of digital-to-analog converters, amplifiers and speakers to deliver the results to us. However, that giant Voltage Modular synth you just patched up does all of its work in that exquisitely detailed modelling space - more analog than analog.