Analog synthesizer against digital emulation
The discussion is as old as digital technology in the music field. Analog versus digital. Which sounds better? But that’s not the point here, because both technologies have their advantages and disadvantages. It becomes interesting, however, if one follows the approach of emulating a coveted analog synthesizer on a purely digital level. Of course, this is also nothing new. Starting with the Clavia NordLead via the Roland JP8000 and Yamaha AN1X, the first plug-in emulations of the Minimoog (Steinberg Model-E) and Prophet-5 (NI Pro5) soon followed. In the meantime, the standard has been set so high by Arturia products and U-He Diva that you have already come very close to the perfect emulation of the analog circuits. Even in 1: 1 In broad comparison, digital emulations can hardly be distinguished from analog originals. I would like to attribute the best emulation currently to U-He’s Diva. Hear and read for yourself whether TAL is also expanding into these regions!
Roland’s Juno-60 is, along with its siblings Juno-6 and Juno-106, the most popular synthesizer in the Juno series to date. This is largely due to the very good punchy and excellently tuned sound and the extremely simple operation. Since the Juno-60 was delivered without MIDI as standard and the DCB interface only provided rudimentary MIDI functions, an upgrade is now offered with the Minerva kit, which replaces the entire CPU of the Juno-60 with a new one. This gives it a comprehensive MIDI implementation and even new functions such as monophonic velocity and a second LFO. On eBay, depending on the condition, it now reaches prices of around EUR 1,000 with an upward trend. Since I own a Juno-60 and have restored it technically, I appreciate the device and can also understand it well.
I would like to briefly go into the synthesis of the Juno-60: The oscillator is digitally controlled (DCO) but generates the waveforms on an analog level without band limitation. It can generate a modulatable pulse oscillation, a pseudo sawtooth (folded sine wave), and a mixable sub-oscillator. White noise can also be added. All waveforms are added and are always phase locked apart from the noise. The individual voices vibrate freely in the phase and are not retriggered when the key is pressed. The filter is an analog resonant 24 dB low pass filter (Roland IR3109 chip). An LFO and an ADSR envelope must satisfy the Juno-60. The analog bucket chain chorus rounds off the characteristic sound. A four-stage switchable static high-pass filter is also available as a candy.
In addition to his products under the company label TAL (Togu Audio Line), Patrick Kunz has been emulating various analog devices such as the Roland SH-101, the Juno-60 Chorus, and the Roland SVC-350 vocoder for some time and with very good results. With the freeware TAL-U-No-62, the first and already quite good Juno-60 emulation has now been replaced by the chargeable TAL-U-NO-LX. While the U-No-62 tended to achieve the character of the Juno-60, it was nowhere near able to build up the “analog pressure” of the original. This has now changed with the LX version, and I would like to have the original compete against the emulation for a sound duel.
The Juno-60 is controlled via the Minerva Kit via MIDI and converted via the RME Fireface 400 and 88.2 kHz stereo output. The TAL-U-NO-LX is hosted in Nuendo 5 at 88.2 kHz and is exported directly as an audio file without conversion. The audio files attached to the test are converted from 88.2 kHz to MP3 at 320 kbps. This format is sufficient for a relative sound impression. It was tested with version 1.79. Another important point is that the TAL-U-NO-LX was modeled after Patrick Kunz’s unmodified Juno-60. My Juno-60 model differs somewhat, of course, by analog component tolerances. Likewise, the envelope times and LFO frequencies could differ slightly with the Minerva Kit. I would, therefore, like to focus primarily on the oscillators, the filter, and the chorus.
The sawtooth and the sub-oscillator are very easy to compare and convince of the level of detail compared to the original, although the sub of my Junos is a little more brilliant. The comparability of the modulated pulse oscillation is somewhat more difficult. The pulse oscillation of my Juno-60 thins out a little with lower duty cycles in the low frequencies, while the TAL-U-NO-LX remains steadfast there. The LX also performs very well in the high pitches. Where other emulations already cause violent aliasing, however, it remains very calm.
The filter can also convince with low and medium resonance and is emulated very well according to the character of the Juno-60. The closer the resonance comes to the self-oscillation range, the greater the differences between the original and the simulation. The simulation sounds good, but the real analog Juno-60 filter is somewhat nobler, coherent, and crisp in sound.
The chorus is also strongly reminiscent of the original bucket chain chorus of the original. In the 1: 1 comparison, differences can be easily identified. At least my Juno-60 sounds a bit wider and “fatter”.
In the tested version 1.79, the algorithm for voice theft is not yet fully developed or does not correspond to the behavior of the Juno-60. The TAL-U-NO-LX can produce up to 12 voices, but it would be nice if it behaved like the original with six voices.
However, my criticism is “complain at a high level”, if you hear the sound samples, you know what I mean.
It would be a shame, of course, if the TAL-U-NO-LX couldn’t do a little more than the Juno-60. As a plug-in, it makes sense to increase the number of votes to a maximum of 12 and also to offer polyphonic velocity on the envelope and volume. The LFO has also been given various forms of vibration and portamento has also been added. I wouldn’t spend much more on the TAL-U-NO-LX so as not to water down the “natural” character of the Juno-60.
Interview with developer Patrick Kunz
How did you go about emulating? Have you analyzed the waveforms and also measured and compared the distortion values?
Patrick: Analyzing the shape of the vibration was certainly an important point. I recorded the distortion with frequency analysis. As always with my products, the focus is on the sound. Hours of listening tests and experimenting with the algorithms finally led to this result.
Thorsten: Is there any nonlinear distortion somewhere in the TAL-U-NO-LX?
Patrick: At the core of the filter are nonlinear components. This was the only way to get the typical two distortion peaks that can be seen in the Juno frequency analysis with the oscillators switched off and full resonance.
Thorsten: What is a zero delay filter and how does it differ from conventional digital filters? What could further development of the resonance behavior look like?
Patrick: The good thing about these filter types for the listener is above all the fast response times of the resonance frequency, for example with fast envelopes, which is otherwise rather sluggish, especially at low sampling rates. The exact result is calculated for each sample without having a sluggish feedback loop. With the increase in computing power, it will certainly be possible to include more and more properties of the individual electronic components in the model in the future.
Thorsten: The oscillators are band-limited to minimize aliasing?
Patrick: That’s right. The waveforms have practically no aliasing. In contrast to my other synths, instead of a minBlep, asymmetrical blep (band-limited step) was used to create the most perfect waveforms, as they are also available with Juno-60.
Thorsten: The filter sounds much more open, lively, and transparent at a sampling rate of 96 kHz. Couldn’t the TAL-U-NO-LX be oversampled internally at 44.1 kHz? Maybe as an option?
Patrick: The filter is already 4x oversampled. This was necessary because of the non-linear components in the filter. It might also be possible to do more oversampling here. It was also important to me that the synth did not consume too much CPU. I think I will be able to make improvements in this direction in the future.
TAL-U-NO-LX is without question the most realistic emulation of the Juno-60. The character of the Juno gets to the point and even the direct comparison between the original and the “fake” shows how advanced the emulation is. So if you are looking for a Juno sound, in practice you will get along perfectly with the TAL-U-NO-LX. With high resonance values of the filter, however, there is still room for improvement in terms of authenticity. Certainly, the emulation does not match the original 100%, but I think that 99.8% offered here is also completely sufficient. Everyone should be able to answer the question of whether they are selling their hardware and swapping it for the plug-in with ease.
➕ Excellent emulation of the Roland Juno-60
➕ Other new features
➖ Resonance behavior at high values is not yet optimal