Thank you Daniel Weiss of
Weiss Engineering for the
DNA-1, an
audio mastering device designed to remove noise, crackles, clicks, and recover lost ambiance. This is just the perfect thing to help me restore dozens of my old cassettes of
Amiga Music, as well as my mothers large collection of Yugoslavian
45's and 78's from her youth.

Click for a larger pic of the DNA-1
Background noise is worst on old records (which also click and crackle), and pretty bad on most old tapes (tape hiss). It also creeps in when you're recording with mic's (air conditioners, wind against the building). Even vintage analog gear hums a bit and sometimes has a higher noise floor than you'd like. The purpose of this device is to improve the signal to noise ratio while minimizing artifacts. As you mix more sources together, the noise is additive. If you can minimize it at the beginning of a chain, all the better. Even temperature in a room actually causes D/A conversion errors in the converters. Taken to the ultimate level, vacuum energy could be called the background hiss of the universe. You really can't escape noise (at least in our Universe).
The trick of an expensive box like this is that it is good at removing noise without making the original sound wonky. Lots of noise algorithms exist, but this is among the best. This is a challenging problem to solve, as noise is often broad-spectrum and random -- how can a device know what the signal is? How can it differentiate the rush of a cymbal or the tail of a reverb from background noise? Tricky problem. It's not perfect, but the DNA-1 does a good job.
Technically, one of its tricks is a linear-phase convolution engine. I imagine what it does is akin to removing blur from a photograph.
Here are three examples of the noise removal algorithm at work:
Original
Soft noise removal
Medium noise removal
You can hear that the 'original' and 'medium noise removal' samples sound quite different, and see the noise difference in spectral envelopes (guess which is which):
Here are some more examples of the three algorithms working: Denoiser, Decrackler/declicker, K-Stereo Ambiance Recovery:
Aretha Franklin Pre
Aretha Franklin Post
Booker T Declick Example 1
Booker T Declick Example 2
Booker T Declick Example 3
Booker T Full DNA1 Example Pre
Booker T Full DNA1 Example Post
Piano Denoise Pre
Piano Denoise Post
String Denoise Pre
String Denoise Post
Symphonic Denoise Pre
Symphonic Denoise Post
Guitar Solo Declick and Denoise 1
Guitar Solo Declick and Denoise 2
Guitar Solo Declick and Denoise 3
The device is all digital. That is, it takes AES/EBU in and out. You should clock it to a master clock like the Apogee Big Ben. Weiss licensed the noise removal from PureNotes. Also licensed is the K-Stereo ambiance algorithm from the famous and loved Bob Katz. K-Stereo is a bit mysterious, and patent pending. People aren't clear on how it works, only that it does work. Weiss developed the Decrackle/Declick themselves, and are releasing an upgraded algorithm soon. To top it off, the device has POW-R dither, which allows the 24-bit 96khz signal to be truncated to 20, 18, 16 bits (like a CD), with minimal artifacts. It is not noise shaping in the classical sense, and there are various types of algorithms provided. Here's an example of noise distribution from a low level sine wave. If there was no noise, you'd just see the spike. Each -3 db is 1/2 the energy. Anything less than 95 is, for me, pretty great.
The device, in their own words:
Denoiser to reduce all types of wide band and non-stationary noises like tape hiss or air conditioning hum using a novel adaptive approach to noise reduction (no spectral noise footprint required).
Decrackler / Declicker with three independent subsystems to detect and remove a broad range of impulsive noises, for example scratches and clicks from vinyl recordings.
K-Stereo Processor to recover lost or amplify hidden ambiance, space, and imaging; or to generate stereo from mono signals without adding artificial reverberation.
Output Control to regulate stereo balance and stereo width, as well as the overall signal level.
POW-R, the acclaimed technique for transparent and smooth word-length reduction to 24bit, 20bit or 16bit.
All those standard features like 128 snapshots, POW-R dithering, MIDI Control, 24/96 operation, extensive metering etc.
Weiss Engineering will show the DNA1 96kHz/24-bit audio restoration and enhancement processor at the 119th AES Convention. The Gambit Series processor comprises a noise removal section followed by an ambience recovering and gain control section that offers real-time simultaneous denoiser, decrackler and declicker, in addition to the K-Stereo Ambience Processor.
The denoiser, developed in cooperation with the Pure Notes Company (UK), reduces all types of wideband and non-stationary noises, such as tape hiss or air conditioning hum, using a novel adaptive approach to noise reduction that does not require a spectral noise footprint. Three automatic modes, a manual mode and a semi-auto mode may be engaged depending on the level of fine-tuning necessary for the task.
COOL!