Thanks to collaboration between Sound Devices, LLC and Dan Dugan Sound Design, Inc., the Sound Devices 688 Field Production Mixer, which already featured 12-channels of automixing with Sound Devices' MixAssist, introduced the renown Dugan automixing as another powerful option for audio professionals to choose.

This article explains in more detail the history and development behind these industry-leading automix features.


For as long as people have been electronically mixing microphones using mixers, people have struggled with achieving excellent results. Good, clean, intelligible speech in a mix does not happen by accident, and in fact takes quite a bit of practice and experience. As Film and TV productions continue to use more and more microphones simultaneously, this problem of producing a clean mix becomes more and more acute for the production mixer. One tool of the audio trade—the automatic mixer—has long had success in other areas of the audio market, and can now be applied to production audio for Film and TV. There has been a bit of misunderstanding regarding the theory and principles of automatic mixers, and this paper seeks to clarify this by examining the two most widely-accepted and high-performing automatic mixer algorithms on the market.

The term “automatic mixer” is a bit of a misnomer, as an automatic mixer does not automate everything pertaining to sound mixing. What an automatic mixer does do -- and does very well -- is to turn down (attenuate) microphones which are not actively being used. There are two benefits to this seemingly simple action: first, this reduces the background noise in a mix, making the desired speech cleaner and more intelligible; second, if used in conjunction with a PA system, more gain can be achieved before the system starts feeding back. Note that automatic mixers do no “mixing” per se -- the user still has full control of the gain of microphones via the faders on their mixer. These algorithms operate post-fader such that if a fader is turned all the way down by the user, the microphone is effectively removed from the operation of the automatic mixer.

The earliest attempts at automatic mixers were just glorified audio gates. If the incoming sound was higher than a fixed threshold, the microphone turned on and conversely, if the audio dropped below the threshold, the microphone turned off. There have been several variations of this method: an adjustable threshold shared between all mics; a user-adjustable threshold per microphone; a separate microphone dedicated to generating the threshold. All of these methods of course work poorly as the threshold is never quite right and the resultant audio is either choppy or always-on. No good automatic mixer today uses such primitive principles for operation. Enter the work of Dan Dugan.

Dan Dugan invented the first real automatic mixer in the early 1970s. Dugan worked as a sound designer in theater starting in the 1960s and experimented with several unsuccessful methods before inventing his groundbreaking method for automixing for which he was awarded a patent in 1976. The core concept of this mixer has been unchanged since then, and it is totally automatic - no thresholds or other adjustments to set.

Principle of the Dugan Speech System™

Dugan’s system operates on a very elegant principle: each individual input channel is attenuated by an amount (in dB) equal to the difference (in dB) between that channel’s level and the sum of all channel levels. The gain of all channels is adjusted immediately and continuously based on what each channel is receiving. The neat aspect of this mathematical construct is that the total gain through the system never changes (always stays at 0dB)!

Here are some examples for a system with four microphones:

All talkers at the same level (or no talking):
Microphone Level (dB) Attenuation (dB)
1 0 -6.02
2 0 -6.02
3 0 -6.02
4 0 -6.02
Two talkers at 20 dB above the other mics:
Microphone Level (dB) Attenuation (dB)
1 0 -3.05
2 0 -3.05
3 -20 -23.05
4 -20 -23.05
A mix of incoming levels:
Microphone Level (dB) Attenuation (dB)
1 0 -1.61
2 -5 -6.61
3 -10 -11.61
4 -15 -16.61

In all of these cases, the total gain through the system adds up to 0dB, as each signal adds as an uncorrelated source (root of the sum of squares).

Principle of the MixAssist™ System

Another very well-regarded method for automatic mixing was invented by Steve Julstrom while working for Shure, and has been used extensively by Shure Inc for many years in their popular FP and SCM automatic mixers. This has been the basis of Sound Devices’ MixAssist algorithm as well. Compared to the Dugan system, MixAssist conceptually operates more to a very sophisticated gate - a decision is made to turn a given microphone on or off, albeit with very smooth transitions and finite off-attenuation. MixAssist achieves excellent results, and operates on very different principles compared to Dugan.

Noise-adaptive threshold (NAT):

For each microphone, an ever-changing and automatic threshold is continuously calculated. This per-channel threshold has a slow-attack and very fast decay. This characteristic makes essentially makes the NAT a good ‘hole detector’ of the audio - the logic being that the room noise is always going to be found at the trough of the audio envelope. When an incoming microphone’s signal is instantaneously above this threshold it can be turned on. Steady-state sounds (air conditioning etc) will not turn on a microphone, only varying speech-like signals.

Maxbus:

The envelope of all microphone signals is logically or’d together to get the instantaneous peak of the loudest microphone. Each mic’s envelope is continuously compared to this Maxbus; if the envelope is greater (and it meets the NAT criteria), then it is gated on. The very unique result of the maxbus is that if a talker speaks into two or more microphones, it will only gate on one of the microphones, eliminating any comb filtering.

Last mic lock-on:

this simple yet very effective concept just holds the microphone last used, on. This creates a very seamless mix.

Off-attenuation:

When a microphone is turned off, it does not turn all of the way off, instead it is attenuated by a certain amount, typically 15dB. This results in a more transparent sound, as there is little benefit (noise build up-wise) to attenuating a microphone totally off. Some implementations allow the user to adjust this parameter.

Number of open mic attenuator (NOMA):

For each doubling of the number of open microphones, the gain is attenuated by 3dB. Similar to the Dugan concept, this maintains the total gain through the system at 0dB.

Comparison and Conclusion

Both of these systems have been used by thousands of people with excellent results. In the end, both systems accomplish very similar goals: they turn down the microphones not actively in use, and they maintain the overall system gain between microphones to 0dB. Each offers near-instantaneous (~2mS) turning on of microphones in use.

  • The Dugan is described by some to sound a bit smoother in operation, as all of the microphone gains are continuously and smoothly changing. The mics are neither all the way on or all the way off. Each is continuously mixed in various proportions.
  • MixAssist is described by some to offer a lower noise floor as microphones which are picking up constant noise will be attenuated at all times. MixAssist also due to its Maxbus will only turn on one microphone even if several are hearing one source.

Sound Devices now offers both of these excellent choices in its 688 field mixer/recorder, selectable by the user in the 688’s menu. Each system can be used in many different situations, and Sound Devices encourages users to experiment with both, as users will hear subtle differences between the two algorithms. Sound Devices is proud to offer both of these industry-leading algorithms to its users to help them produce better and better audio.