The NAT tip of the month (#16 - December 2011)

N.A.T. tips and tricks
User avatar
Posts: 2861
Joined: Sun Mar 28, 2010 9:00 pm
Location: Lodi | Madrid | Buenos Aires

The NAT tip of the month (#16 - December 2011)

Post by enriquesilveti » Fri Dec 16, 2011 9:28 pm

Creating R2R and Tape Booster Plus

NAT Sampling and the creation of Reel to Real
Many Nebula Pro users find it interesting to work with NAT and learn how to create programs. Each category of program within Nebula has its own unique process. Creating a reverb program is a different process than creating a flanger or an equalizer, etc. Each type of program comes with its unique challenges. With equalizer, the challenge can involve consistency, the patience required to sample settings over and over again, until hundreds of settings are created, and then editing each one to align with its analog equivalent. This process can take weeks and weeks to perfect for a single equalizer.

Sampling tape is perhaps the most challenging process of them all. Where the concept may seem simple enough, the challenges are numerous. For example, when sampling a microphone preamp, a console channel, or a typical high end tube or solid state device, there is a consistent signal generated that remains constant enough that the job of replaying the results through the Nebula engine is straight forward. That is not to say that it is easy, but one can expect consistent results with enough careful testing and sampling. With tape machines, there are similar qualities to a preamp program. But, by design, the moving parts of the tape machine and the nature of recording media are such that alignment and accuracy are challenging to the point that for a long time it was assumed that any high quality sampling was just not possible. This was true in early impulse response formats that used very fast, short, bursts of full spectrum, as is the case with more complex sampling techniques. When sampling a very short burst or sweep, it is easier to capture the frequency spectrum of a single setting with tape. Because the inconsistent non-linearities are reduced by sampling very quickly, one can make an equalizer sound like tape, but only in a single instance, and only with limited results.

Longer sweeps can generate a more accurate example of the harmonic content of tape, but as tape records and aligns to its self-correcting bias frequency, the tape also moves along the various mechanisms, flange, heads, etc., bending, stretching, and losing magnetism. As tape is played, it changes and eventually can lose its original spectral response. Timing characteristics can become elongated and stretched out, which not only can make sampling a challenge, but used to make multi-machine synchronization a well-acknowledged nightmare. Songs recorded between more than one machine were aligned using a synch tone. When this would become stretched or lose fidelity, the synch would become lost and a new tone had to be 'striped' in hope of synchronizing tracks once again. Ah, the miracles of modern digital alignment!

So, with Nebula, we have a unique opportunity to measure not just one spectral response, but multiple layers at multiple volumes, so that the entire dynamic range can be recreated. We can replay the different ways that equipment plays back quieter signals and louder signals, and the changes to distortion and frequency character at all of these levels. The absolute dynamic range of tape is one of its major obstacles as well. Where a high quality microphone preamp may have a noise floor as low as -80 dB and be able to boost louder than +3 dB with a relatively clean signal before distorting, tape has a great deal of non-linearity and hiss anywhere from -70 dB to -40 dB depending on the machine. The window for getting a perfect dynamic capture series requires a great deal of testing, as noise and hiss do not translate accurately and can ruin correct readings.

With tape, alignment becomes an issue at every volume, and between every take. So, we have to establish an accurate reading to begin with. By digitally recording source material first, we have a sonic match that the programs must recreate. We have to make sure recordings are not just accurate in small examples or only at certain layers, but at all volumes. When tape changes, fluctuates, these changes do not overlap correctly, and can cause frequency shifts that do not occur in the original machine. As an example, something consistent yet non-linear, like an A*I preamp, can perform within an expected range for hours and hours and hours. The small changes in frequency will only occur if the level sent to the device is greatly louder or different than other audio. So, a very long sweeping signal can play over and over again, one volume at a time, for literally hours, and as long as the recording continues to save and digitally transfer correctly, there is no trouble. The loud frequencies remain similar, the quiet passages similar, and the harmonics reflect the change from being pushed lightly to hitting overloaded signal to a certain degree.

With tape, this same process cannot occur at faster, more accurate tape speeds, because first the tape reel ends before sampling is finished. Secondly, there is slightly different behavior from even the best tape media at the center of the reel than that of the exterior. Even a very high end synchronization bias machine does not lock perfectly through this procedure. The pressure and resistance shifts with the amount of tape remaining. For analog recording, it remains stable enough to reproduce frequencies consistently enough to sound great over time. But imagine if you recorded the entire reel with the exact same signal, constant, and then needed that signal to align perfectly over itself without going any faster, slower, higher, lower, at any level. As we know from calibration practices or lessons in recording, if two tracks of identical material are set apart even by a micro-second, imperfections like phase dissonance occur. The same is true for dynamically recreating tape in Nebula. So, we have to test and retest, reel by reel, every possible combination until finding the perfect result between the number of samples, the number of repetitions, dynamic range, and length of sweep, until the ideal scenario is reached, for each unique machine.

A good analogy for how Nebula needs to see a dynamic program is to compare a tape program to traditional graphic design. The professional four color process required that an image be created with Cyan, Magenta, Yellow, and Black, each on its own layer of film. A mark is placed to align the layers so that each color fits directly on the next layer. If they are off even slightly, the image becomes distorted and has artifacts. The same is true if dynamic redundant layers of volume return inconsistent responses that have a noticeable shift from each portion of each sample.

So, the answer to accuracy is creating many master samples for comparison, testing and retesting, checking alignment, testing different tape stock and learning as much from what does not work as what does. I have found that every machine has its own unique range that responds best, from different volume ranges that respond well, to different tape media, to different portions of the reel at different speeds. By running many instances many times, and doing alignment and editing by ear and by visualization, hundreds of routines and samples are reduced down to the best collections.

The final results are documented and edited to match the response of the original machine at the same settings. If there is an inaccuracy the samples are discarded. In some rare instances, a finished routine can be created by using some ideal samples from one series matched to another. In the case of the Reel to Real collection, the challenges were very unique with every machine. The Wollensak was probably the biggest challenge for accuracy, consistency, and also because it showed such a unique response to every volume and tape media, I tried every possible combination to make sure it was represented with as much of the desired personality as possible.

Creating the stand alone Reel To Real and Tape Booster Plus Audio plug-ins
Creating an audio plug-in based upon the Nebula technology (V.V.K.T.) comes with its own challenges. It is a rewarding process that comes with its own unique challenges. Our philosophy is that every stand alone version of a Nebula program or library benefit Nebula Pro users in addition to those who are new to the technology. With Nebula, we offer as wide and complete a collection of program options as possible, so that the user can experiment with virtually any setting of a device that they want to. When using the stand alone version, the collection may focus on a "best of" from the collection, encouraging users to look into purchasing Nebula as a host at a later time once they experience the unparalleled quality from the technology. With Reel To Real, the stand-alone plug-in contains all of the same machines in the Nebula library except for the TEAC cassette player. Every tape stock is included in the stand-alone as well, along with all tape speeds, but some additional program settings, like multiple volume ranges, are only available for Nebula. Tape Booster Plus offers 44.1kHz and 96kHz versions in Nebula and stand-alone, with additional volume program presets in the Nebula Library designed to make it easy for the Nebula user to simply 'load-and-go' without concern for setting levels. With the stand-alone the plug-ins are edited to load at their maximum saturation to make it easy to reduce the levels if needed.

Where stand-alone users enjoy stunning GUI visualization, we always offer a substantial discount for these plug-ins for the Nebula owners, making it an easy decision to add the GUI stand-alone version if they wish to enjoy the visuals. Other benefits are that the stand-alones are edited specifically to run as optimized for the specific type of program run, and based upon the average current CPU resources. Where Nebula may be tweaked by the user for use on the widest range of programs, they have the ability to tweak their stand-alone versions until they get the maximum use of their resources.

One thing that users may find interesting and important, is that sometimes newer systems and operating systems don't offer the huge CPU performance boost they hope for. One of the most common reasons for this is the assumption that a faster computer with more memory will automatically reduce the amount of latency and memory buffers they should use in their audio host. The fact remains that a little bit of latency is required from the host in order to offer bus speed resources for signal flow to and from the processor and RAM memory. With zero latency and a super fast computer, a user is likely to get less performance than an average machine with only a little memory with well-tuned low latency. By setting one's host to match the resources of their sound card or audio interface and the sample rate used, they can make use of things like faster RAM and 64 bit processing and reduced RAM limitations. So, if you have made the plunge for Windows 7, 64 bit, more RAM, or a faster processor, make sure that you do a lot of bench testing with the chosen sample rate and audio host latency and buffer settings. You may find you can get as much as 200 to 300% performance boost by getting these settings right.

CD Sound Master also seeks to benefit stand-alone users within and outside the Nebula community by offering easy flexibility with the computer system of choice. If a customer purchases a 32 bit collection and moves to 64 bit, they can simply ask for an update at no charge. Changing to a new computer is equally quick and easy. Our current aim is to bring this same flexibility and power to MAC OS in the immediate future.

Michael Angel from