Your favorite "versatile analog-modelled/emulation" synth plugin?

[_leras]

Softube have deals pretty fequently.

Might grab the Model 72 next time as well... just to compare against Monark a bit.

Or you could do that with the demo? Just sayin'.

[LoveEnigma18]

Yeah, I was tempted by Model 72 recently, but I don't use mono synths much so I skipped it, especially considering I have Monark too - which is excellent. Softube emulations are definitely on another level, especially the 84 for me when the times come to get one. Might sound counterintuitive, but the sound actually feels like analog/hardware even though I have never touched the actual hardware

I will compare Model 84 with Repro and Diva too to decide which one I like better. Model 84 doesn't have as many features as Arturia synths do, arpeggiator for example, but I realized I could drive it with Pigments' arperggiator/sequencer if needed.

I agree, Model 84 is really fantastic. It was on special discount until yesterday, $79 with a code. But it will be on sale again... considering that Model 72 was sold for $39 recently I'm guessing we will see a similar deal at some point.

Yeah, missed the sale by few minutes. Hopefully it will be back soon because I am really enjoying the demo. As you said, I hope it gets the similar, < $50 sale price soon.

In fact, I maybe an odd guy but I am liking it more than Repro and Diva for its simplicity and sound both. Not to say Repro/Diva are less, it's just a personal taste.

Not odd at all. Diva was a bit too complicated for me as well, that's why I sold it. Not necessarily architecturally too complicated, but, all those tweaks and modifiers simply went unused most of the time, and, I don't like it when I don't use 90% of the stuff. Think I'm also more into a traditional layout. Diva was too modular for me. Not really a big fan of combining a Roland filter with Moog oscillators etc. I'm sure it's a nice playground when you ARE into that though. I don't know. It didn't really gel with me.

Yes, exactly the same for me. Massive X and even Dune 3 would be great if we want more filters with typical VA/wavetable synthesis. When things start to get modular/complex, even to the slightest extent, I start getting uncomfortable. I am mostly a preset tweaker that has improved a little bit with my understanding of sound design (no way near to you guys here), so as long as it is simple enough to change something, I can do it to fit in the song. I also had Diva and Repro during the first big sale they had with NI. For some reasons, I did not get along with them. Not to mention some of their patches yesterday spiked up my i7-7700 like nothing has before. As I said, it is more of a personal thing. Both sound great, Repro might be just be ahead for me relatively, but Model 84 has really impressed me. Simple and limiting, but has everything to create great bread and butter sounds of all types (relatively) fast. I kinda ignored it until now, but demoing it removed that hesitation. It would awesome to see it on sale for ~ $50 next time and I would pick it up the second I see it.

[chk071]

Softube have deals pretty fequently.

Might grab the Model 72 next time as well... just to compare against Monark a bit.

Or you could do that with the demo? Just sayin'.

I already demo'd it, so I can't anymore.

[Markus Krause]

This is what really defines analog sound (you can switch most of this on/off in the Config menu of Saurus3:
- Filter with self-oscillation
- slight oscillator-drift
- slight instabilities
- Soft saturation
- Damping of frequencies over 8Khz in the OSCs
- a little noise
- no aliasing
- filters are slightly 'off-tune'

Markus, what do you mean by "soft saturation"? A modest amount of harmonic distortion due to imprecise waveform rendering, maybe? If so, does that imprecision vary over time?

A soft clipper is similar to a hard clipper, but with the corners smoothed. A common choice of soft-clipper is the cubic nonlinearity
https://ccrma.stanford.edu/~jos/pasp/Soft_Clipping.html

[Markus Krause]

The other techniques were familiar to me, but this took me by surprise. The more I think about it, the less sense it makes. There are multiple points in the signal path where synths have to model saturation. Pre-faking it with a wavetable technique could reduce the CPU hit, but would surely sound wrong on patches with multiple oscillators. I think I must have misunderstood something. Can you explain further?

Many other analog-emulations use a common wavetable technique called 'lookup table'. It is used to prevent aliasing and to keep the CPU low while still providing a analog-like sound.

[Ploki]

intriguing, thanks for your insight.

I knew hive uses wavetables for the VA part, but wasn't aware DIVA (and i assume repro?) does as well.

Edit: IT DOESNT

I do remember trying to make a decent saw-wave in MaxMSP years ago, and anything synthesised not WT was a pain in the ass to do to sound good (because you either need to filter it severely or it aliases)

[YnJ]

Granted it was created when Saurus was at 1, but this is my favourite Saurus bank. Very Vintage Analogue(y).

http://www.electric-himalaya.com/saurus-old-volt.html

Excellent sound set, thanks for the link!

[gentleclockdivider]

- F0 partial of a pre-rendered sawwave that is stored in a lookup-table is over-attenuated to simulate saturation on the OSC and make it sound more 'warm' (DIVA and many analog emulations do this)

The other techniques were familiar to me, but this took me by surprise. The more I think about it, the less sense it makes. There are multiple points in the signal path where synths like Diva have to model saturation. Pre-faking it with a wavetable technique could reduce the CPU hit, but would surely sound wrong on patches with multiple oscillators. I think I must have misunderstood something. Can you explain further?

DIVA as well as most other analog-emulations use a common wavetable technique called 'lookup table'. It is used to prevent aliasing and to keep the CPU low while still providing a analog-like sound. In DIVA the 'single-cycle waveatable' stores not a linear, but a soft-saturated sawtooth. It is a sawtoth with 'a little more bass'. The F0 harmonic is slightly boosted, which results in a sawtooth with a 'belly'.
You can validate this for yourself when you take a look at it with an oscilloscope. There you can see ringing-artefacts (https://en.wikipedia.org/wiki/Ringing_artifacts) which are the result of the wavetable-technique (http://mtg.upf.edu/node/485) and you can see the bass-boosted sawtooth.

Are you sure about that , diva uses waves stored in look uptables , and not mathematically calculated ?

[Markus Krause]

I do not think Diva uses BLIT for the oscillators. I used It for Saurus. This approach makes most sense when you want to model synced oscillators. However it is more difficult to implement as you need to take care of DC offsets.
I suggest you ask the developer himself what approach he used

[mystran]

You can validate this for yourself when you take a look at it with an oscilloscope. There you can see ringing-artefacts (https://en.wikipedia.org/wiki/Ringing_artifacts) which are the result of the wavetable-technique (http://mtg.upf.edu/node/485) and you can see the bass-boosted sawtooth.

I have no clue what Diva does (nor do I care to make a guess), but ringing artifacts in general only mean that the waveform is bandlimited in one way or another. You get those with wavetables, you get those with BLEPs, you get them even with additive synthesis and if you're oversampling, then the ringing you actually see would typically be from the oversampling filters, whether or not the actual method used produces some ringing of it's own at some higher frequency.

[mystran]

I do not think Diva uses BLIT for the oscillators. I used It for Saurus. this approach is necessary when you want to to synced oscillators. however it is more difficult to implement as you need to take care of DC offsets.

Nobody in their right mind uses BLIT anymore, 'cos the BLEP (preintegrated version of the same) theory has been well-developed for like 10+ years and doesn't have any DC offset problems whatsoever.

[Markus Krause]

I am refering to pre-ringing in the oscialltor shapes and not ringing in general. You don't get any significant pre-ringing when you use a BLIT with a minimum-phase windowed-sinc pulse. See figure 6 here:
https://www.cs.cmu.edu/~eli/papers/icmc01-hardsync.pdf

nobody in their right mind uses BLIT anymore, 'cos the BLEP (preintegrated version of the same) theory has been well-developed for like 10+ years and doesn't have any DC offset problems whatsoever.

No. That's not true. As soon as you use BLEP + synced oscillators you'll get a slight DC offset.
Apart from this the BLIT+integrator is a more accurate emulation of the analog OSC circuit than BLEP or pre-rendered waverforms.

As I mentioned before you should ask the developers themselves what technology they used.
I did use these technologies:
Saurus: BLIT
Electra's 'analog' oscillators: waveforms stored in a lookup table

[mystran]

nobody in their right mind uses BLIT anymore, 'cos the BLEP (preintegrated version of the same) theory has been well-developed for like 10+ years and doesn't have any DC offset problems whatsoever.

No. That's not true. As soon as you use BLEP + synced oscillators you'll get a slight DC offset.
Apart from this the BLIT+integrator is a more accurate emulation of the analog OSC circuit than BLEP or pre-rendered waverforms.

I couldn't remember for sure if U-he had stated it publicly (although it is rather easy to see from some subtle spectral details if you've ever worked with BLEPs), but whyterabbyt kindly posted (in the other thread on U-he forum questioning your claims) a link to a thread where Urs explicitly describes which methods which synth of theirs is actually using (ie. BLEP in case of Diva).

In the case of synced oscillators though, the only "DC offset" you get (in a correct implementation) is the expected behaviour you also get in analog synced oscillators, which is actually highly important if you want to correctly model the interactions between the DC blockers and the filter input non-linearity (especially for something like a ladder or cascade with a strong input non-linearity).

The claim about BLITs being more accurate is also objectively wrong. Tabulated BLIT would theoretically produce the same results as a BLEP if it wasn't for the horrible numerical error accumulation from integration. While analytical BLIT would theoretically avoid the (rather theoretical) compact support limitation of a tabulated kernel in the steady state case (while still suffering from the horrible numerical error accumulation), the result is only correct if the frequency is held constant and in practical use it can perform even worse than a tabulated BLIT.

That said, I'd imagine that in reality you already knew all this.

[Markus Krause]

Alright. You convinced me that also BLEP (without minimum phase) could cause the pre-ringing. I updated my posts.

BLIT has an advantage compared to BLEP in the aspect of aliasing. You get a 6db better rolloff with it. BLIT (with minimum phase) also does not introduce a delay and pre-ringing. Something like pre-ringing or linear-phase filters (FIRs) simply does not exist in the analog world. Every filter that you can find in an analog synth is is a minimum-phase filter.

'horrible error accumulation from integration' is also happening in a very similar way with the analog capacitors used for sawtooth circuits. It results in instabilities which can be clearly heard when you play a very low sawtooth on an analog synth. Saurus emulates this effect by adding slight noise to the capacitor/integrator.

https://www.electronics-tutorial.net/an ... generator/

https://www.falstad.com/circuit/e-sawtooth.html

[vurt]

intriguing, thanks for your insight.

I knew hive uses wavetables for the VA part, but wasn't aware DIVA (and i assume repro?) does as well.

diva doesnt.