The Development of Virtual Circuitry Modeling Audio Effects

– From the birth of physical modeling to Portico emulations –

– From the birth of physical modeling to Portico emulations –

Physical modeling technology, a general term for software emulation of natural acoustic phenomena and electronic circuitry for use in musical instruments and audio effect processors, began at Yamaha in 1993 with the development of the VL1 Virtual Acoustic Synthesizer. At the time there were essentially no other electronic music products based on physical modeling, so the VL1 became the world’s first commercially available physical modeling synthesizer. Since the technology was being used to model acoustic phenomena, we referred to it as a “VA synthesis,” with “VA” standing for “Virtual Acoustic.”

K’s Lab, the team assembled to develop the VA tone generator, was established at the Yamaha research and development center under the leadership of Mr. Toshifumi Kunimoto (affectionately known as “Dr. K”) in 1987. The name “K’s Lab” actually came later, but Mr. Kunimoto has been and continues to be the driving force behind the development of physical modeling for Yamaha musical instruments and audio processing.

In addition to the virtual acoustic modeling used in the tone generator itself, physical circuitry modeling was applied in some elements of the VL1 effects as well. The VL1 was followed by the VP1 with physical modeling of string instruments, the AN1x with modeling of analog synthesis, and the EX5 and EX7 which combined algorithms from both the VL1 and AN1x with a range of new physical modeling concepts. The technology was constantly evolving and being refined at K’s Lab. The term “VCM” was first used to describe the modeling technology used in the “Add-on Effects” series of audio effects developed for a digital mixing console that was released in 2004. From that point onward K’s Lab has been devoted to developing physical modeling technology that effectively emulates a variety of audio effects, including the types of analog outboard processors used in recording studios and guitar processors, and implementing that technology in a wide range of products. In addition to independent processors, VCM technology is currently used in Yamaha MOTIF XS and XF synthesizers, CP series electronic pianos, and numerous other products.

In the next section we’ll take a look at the VCM based Vintage Open Deck effect that was original developed for Yamaha digital consoles.

When the Open Deck project began, the Yamaha K’s Lab team had almost wrapped up development of some great sounding compressor and EQ processors that used VCM technology to model physical electronic circuitry in software, and they were looking for a new challenge. The subject of tape reproduction came up, and the involved discussions that followed began to focus on the significant sonic impact that the reproduction machine has on the sound of an analog production chain that might also includes microphones, a mixing console, and outboard processors. Considering the situation and technology available at the time, magnetic tape was in many ways a “necessary evil” that was unavoidable for the process of recording multiple audio sources, some of them at different times, and bringing them together again in the final mix. But at the same time the distortion and compression that were characteristics of tape reproduction enhanced the sound of the instruments and vocals in a pleasing way, becoming an integral part of the music and production style of the day. Even today some engineers and producers will “print” rhythm tracks to analog tape and then bring them back into the DAW (digital audio workstation) in order to capture that analog vibe.

Figure 3 shows an overview of the tape reproduction process. Of course the physical phenomena involved were well understood at the time, and were taken into account in the design and manufacture of tape recording and playback equipment. To accurately emulate the process digitally it is necessary to use VCM technology to model the characteristics of the circuitry from the recording amp to the record head, the properties of the magnetic tape itself, and the circuitry from the playback head to the playback amplifier with its NAB or similar playback equalization. Numerous elements are involved, but none are particularly difficult to model individually. The analog recording and playback process shown in Figure 3 is also the block diagram of our VCM Open Deck model.

When the model is ready the measurements and testing begin. One of the problems we faced at the time (around 2002) was that DAW recording had already become mainstream and it was difficult to locate professional analog recorders that were in good working condition. It was also difficult to find engineers who were adept at carrying out the important azimuth and EQ adjustments that were necessary to elicit optimum performance from analog recorders. We were lucky, and managed to capture the data we needed using great sounding analog recorders that had been kept in top-notch condition.

Since the record and playback amps in the VCM model are separate components, different record and playback devices can easily be combined for more variety. We often describe VCM technology as being a “component level” emulation, but in reality it is a mix of emulation at the macro and micro component level. If the characteristics of an amplifier, for example, can be modeled as a whole, then macro level component emulation is sufficient. But if it is necessary to model the individual diodes and transistors that make up the amplifier’s circuitry to achieve the desired effect, then micro level component emulation is called for. Most commonly available modeled effects attempt to model the entire effect as a whole on the macro level, but at K’s Lab we often go deeper, modeling individual electrical components at the micro level as necessitated by the quality and overall efficiency we are trying to achieve.

Development was complicated by the fact that tape magnetization and playback from that magnetic field is a rather ambiguous process that is difficult to measure and define. Even with the basic model finished, the task of embedding individual tape recorder characteristics in the model proved to be a challenge. Which characteristics were essential to the effect we were after, and how should the parameters required for those characteristics be implemented in the model? Even more importantly, how could those parameters be made to realistically approach the desirable, music-enhancing sonic qualities of an actual tape recorder? We had a lot to think about. Eventually these problems were solved by a combination of science and craftsmanship, bringing the technical expertise of our signal processing engineers together with the talent and taste of eminent sound engineering consultants. An enormous amount of time was spent adjusting and tweaking until we ended up with four outstanding data sets based on four professional recorders: Swiss ‘70, Swiss ‘78, Swiss ‘85, and American ‘70. The primary goal was to recreate the musical functionality of these decks rather than to recreate their technical characteristics exactly, so they were tuned to bring out the most desirable sonic qualities of each deck.

Let’s consider frequency response as an example. Figure 4 shows the actual reproduction frequency vs. amplitude response of the four analog tape recorders. Figure 5 shows the response of the corresponding VCM Open Deck models. You can see that the response is accurately recreated, right down to the finest details. Please note, however, that the Open Deck effects model a number of the nonlinearities introduced by the analog tape reproduction process, and the data shown here represents only one small part of the overall Open Deck effect.

Our efforts were rewarded with generally high acclaim, and Jun Toyama, a sound producer who took part in the evaluation phase of the project, went as far as to say that we had created a “valuable reference library for future generations.” The end result was versatile Vintage Open Deck effects that not only provide realistic tape compression that can be invaluable in consolidating rhythm tracks as well as other elements of a mix, but record bias that can be shifted from the standard setting to control harmonic content as well. The ability to combine different record and playback recorders is another popular feature. In studio production the recorder used for final mastering is often different from the recorder used for the mix. The fact that the Vintage Open Deck effects make it easy to replicate that environment was greatly appreciated and quickly embraced by production professionals.

The VCM technology based “Add-on Effects” series was initially released in 2004 as audio effects for Yamaha digital mixing consoles. Other effects in the series include the Comp276 and EQ601 effects created under the direction of recording engineer Shinichi Akagawa. All are highly regarded for their outstanding musicality. Figure 6 shows the Open Deck and Comp276 interfaces in the Studio Manager application that allows overall management of Yamaha digital mixing console data.

The emulation accuracy and musicality of VCM technology received generally high praise in evaluations, from DM/O series console users, PM5D console users, M7CL console users, and a broad spectrum of other end users. But although the response from collaborators and users was positive, the K’s Lab team was determined to set the highest possible goal for validating the technological and musical value of their creations. The question was … what would that goal be?

One possibility that was immediately obvious was to emulate a highly regarded and recognizable piece of analog gear to everyone’s satisfaction. It would not, however, be sufficient to attempt an emulation that others had already done. It had to be a “Holy Grail” emulation at which no other vendor had yet succeeded. The subject was discussed at length, and the Portico series analog outboard processors produced by RND (Rupert Neve Designs) was nominated. We were under the impression that Rupert Neve was a devoted champion of analog audio who would be unlikely to collaborate with a digital audio vendor on the basis of “average” digital signal processing technology. But as luck would have it we received word that Mr. Neve had begun to show an interest in developing digital technology right around that time. It wasn’t long before we had arranged a meeting with Rupert Neve at RND.

We arrived at Mr. Neve’s private residence on a summer day in 2007. In addition to conveying our strong desire to prove our technology by modeling an RND equalizer and compressor, we had done our homework and were well prepared. At the time we had already researched past and present equalizers, both stand-alone and console types, and had reached the point at which we could accurately model just about any analog equalizer ever made. What we actually had was more “know-how” than “technology.” There is simply no technology in the world, including sampling, that will accurately model all equalizers at the flip of a switch. It is not possible to capture and reproduce the characteristics of all vintage equalizers using one piece of software. Our experience had taught us that our know-how must be applied to individually optimize each model, and that considerable time and effort is required.

At that meeting at Mr. Neve’s residence we carefully described VCM technology and its background. As an example of the type of results that can only be achieved with VCM technology, we explained how the K’s Lab team had already succeeded in modeling the characteristics of a Neve equalizer that was popular in the ‘60s and ‘70s. We also explained that we were able to accurately digitally model the feedback type compression of the 33609 and other vintage compressors, which is a feat that few others had achieved. We brought that up because, judging by the specifications of the Portico 5043, we believed that Mr. Neve was partial to the feedback compression sound. It turned out that we were right, and Mr. Neve actually expressed his dissatisfaction with most current plug-in emulations of feedback type compression at the meeting. We also expressed our belief that modeling was only half the job, and that at K’s Lab we spared no effort in fitting and tuning our models to achieve the most musically appropriate sound.

The meeting was a success and our preparations paid off. That evening we received word that Mr. Neve would like to collaborate with Yamaha and K’s Lab. We subsequently learned that Mr. Neve had previously been approached by numerous developers and had rejected them all. The opportunity to collaborate with RND came to us because Mr. Neve understood that we not only had the technology to accurately emulate the measurable characteristics of analog gear, but that we were devoted to achieving the best possible sound as well. Without that understanding the collaboration between RND and Yamaha K’s Lab would not have begun.

The VCM modeling process can be broadly divided into two stages: reproducing the functionality of the equalizer or compressor being modeled, and adding the sonic character of the amplifiers, transformers, and other physical elements involved. In most cases function comes first.

A notable characteristic of the Portico 5033 equalizer filter (transfer function) is the distinctive response of the Neve shelving filter. Figures 8, 9, and 10 show the frequency response of the low shelving filters used in the well-known Neve 1073 developed by Mr. Neve in the late ‘60s and the current Portico 5033 by RND. One of Mr. Neve’s significant achievements at the time was the effective use of overshoot at the “shoulder” of the shelving filter curve to create an optimized overall response. Rupert Neve was a pioneer in engineering this type of filter response at the time, and the same approach was used from the early 1073 through to the Focusrite ISA equalizers. The famous SSL 4000 series and other consoles were also influenced by the original Neve concept, but the current RND Portico units take the merits of that design to new heights with refined, up-to-date circuitry. In broad terms, Mr. Neve’s ideal is a trapezoidal shelf shape that can be approximated by carefully engineering the filter’s overshoot. Moreover, the overshoot changes subtly according to the filter’s frequency setting, so its affect on the sound is multi-dimensional and complex. This meticulous control of frequency response is just one aspect of Mr. Neve’s genius.

Although the transfer function of the 4th order filter in the Portico 5033 equalizer has been carefully engineered to extremely demanding standards, approximating the filter’s characteristics using VCM technology was not difficult. Compare the response of the original analog filter and the VCM emulation in Figures 9 and 11, respectively.

Asked why the filter was originally designed that way, Mr. Neve explained that back in the ‘60s sound engineers needed to be able to adjust the mix balance to a degree after the tracks had been mixed to 2-track tape. The equalizer’s shelving filter was designed to allow the balance between guitar, bass, and drums, for example, to be adjusted in the stereo mix, and that same capability has proven to be a valuable tool for adjusting mix balance as well as refining individual tracks in today’s production environment as well. The basic design is timeless, and the talent that created it so long ago deserves our greatest respect.

Now, looking at the Portico 5043 compressor we note that it has a unique FB switch that changes its gain control method. Compressors consist primarily of a section that detects level and determines the amount of gain reduction to be applied, and a section that actually reduces the gain of (compresses) the input audio signal. Communication between these two sections is accomplished either via “feedback” or “feed-forward” circuitry. In the Portico 5043, when the FB switch is engaged the compressor functions in feedback mode, and when the switch is off the compressor functions in feed-forward mode.

Most modern compressors are feed-forward types, while most vintage compressors use feedback. The turning point was around the mid ‘70s when the development of VCA chips made it easier to implement feed-forward operation. The dbx 160 and similar contemporary compressors are feed-forward types, while older units like the Urei 1176 and Teletronix LA-2A rely on feedback operation. Among the numerous factors that contribute to the sound of a compressor, the choice of feed-forward or feedback control has a significant effect. Feed-forward compressors tend to have a more natural, open sound, while feedback compressors can create a more powerful, assertive sound.

The Portico 5043 FB switch is a product of Rupert Neve’s deep understanding and insight. Compressors he developed in the late ‘60s, starting with the 2254, were feedback types. Mr. Neve is thoroughly acquainted with the benefits of both vintage feedback and contemporary feed-forward compression, and we believe that is why the option to use either is provided on the 5043.

The ability to select between feedback and feed-forward compression has been accurately modeled in the VCM model. Figures 14 and 15 compare the attack waveform of the original with the VCM model. Even the very slight dip in the envelope due to over-compression when the FB switch is engaged has been precisely emulated (although it’s difficult to see in the graph). The functional models of the Portico equalizer and compressor were soon completed and presented to Mr. Neve with a direct comparison between the VCM models and the original hardware. Mr. Neve gave us his approval, and it was time to move on to the next phase: adding the distinctive sonic character of the devices.

The amplifiers and transformers used in the input and output stages of the Portico 5033 and 5043 are an essential part of their sound, and therefore had to be modeled with precision. The amps and transformers remain in the signal path even when the EQ or compression is bypassed, resulting in a more musical tonality when the Portico units are simply being used as line amplifiers. The sound of the amplifiers and transformers is the foundation of the “Neve sound,” and that foundation combines with musically appropriate equalization and compression in the Portico 5033 and 5043.

Mr. Neve provided documents in addition to the circuit diagrams that were invaluable in modeling the amplifiers and transformers. Those documents included text that gave us some insight into his design philosophy. It was a rare and valuable view of his engineering approach and intent, as well as his unswerving dedication to the pursuit of “good sound.”

Rupert Neve’s engineering brilliance and his intimate understanding of sound are evident in the amplifier circuitry and transformers as well. The key is in the creation of “musical-sounding distortion.” The designs of the amplifiers and transformers are exquisitely balanced to enhance and support the music. The amp circuits are discrete designs with push-pull output stages that are carefully configured so that crossover distortion does not occur. The “pure” sound achieved is a testament to Mr. Neve’s extensive experience and knowledge.

Audio designers and manufacturers who also design their own transformers are a rarity, and the fact that Rupert Neve does just that – he began his career as an engineer at a transformer manufacturer – is another contributing factor to the “magic” of the Neve sound. He has always designed the transformers used in his gear, from the earliest Neve company products right up to the present day. The input and output transformers used in the Portico units are no exception, but they are not simply old designs rehashed. They are totally new designs that have been engineered to optimally support today’s sound. The same goes for all related circuitry as well, ensuring that the entire package is musically in tune with the times.

The amplifier circuitry, transformers, and peripheral circuitry, and the balance between them, are the cornerstone of the Portico sound. Mr. Neve describes that sound as having “depth and perspective.” Others refer to it as the “sweetness” that is so desirable in top-class analog equipment. All of the music, including the most delicate reverberations, is reproduced with care, as is the environment in which the sound was recorded. Mr. Neve pays meticulous attention to such details when designing any piece of analog gear, including the Portico series, and those details make up the core of the renowned Neve sound.

All of this meant that our most important mission at K’s Lab was not only to emulate frequency characteristics, but to achieve that “sweetness” in the digital format as well. We threw all of our resources and effort into realizing that all-important Portico depth and perspective in our modeling of the amplifiers and transformers. We started by using basic amplifier and transformer models that we had created for other effects, but were unable to accurately recreate the sound of the Portico circuitry and transformers. The problem was that the models were biased by the characteristics of components and circuits that were originally created with a different sonic approach to realize the tonality of different gears. So we started from scratch, creating a large number of new models and components using VCM technology, and trying out totally new methods wherever necessary. Evaluating each and every detail from both the engineering and sonic perspective, we began building a model and refining the parameters that could emulate the Portico sound. We would create a prototype, adjust, evaluate, and then go back to the prototype stage, and that cycle was repeated a great many times. More than thirty parts were created and discarded during the development process.

We eventually succeeded in creating a sound to which Mr. Neve gave his full approval. Directly comparing the original hardware and the model, Rupert Neve provided the following written evaluation:

“The behavior and response are identical. The performance and the sonic signature are accurately captured.”