On long term sustainability for third party plugin devs

(Jonathan Moore1) #61

Well, that pretty much feels like you have just said that to me. Thanks for the warm welcome to your community.

By the way, my studio near-fields are SE Munro Eggs, and my headphones are Sennheiser HD 650’s, used via a Graham Slee headphone amp (and calibrated via Sonarworks), so you’ll have to forgive my placebo influenced opinions. :slight_smile:

I said pretty much nothing but nice things about VCV and it’s community of developers but did highlight some significant differences in the behaviour with regard to your emulations of Braids and Clouds in comparison to both the Softube Modular emulation and the hardware modules. But I did also mention that some of that behaviour difference could possibly be derived from the fact that everything is x4 oversampled in Softube Modular whereas I was listening back to VCV via my DAW’s standard sampling rate of 48k. Plus the Softube emulation is component modeled throughout and the interaction of Doepher EG’s, VCA’s and the really great Doepher VCF (that’s a near-perfect facsimile of Bob Moogs ladder designs across 1, 2, and 4 poles) is going to have a massive bearing of the final timbre produced. Having said all that Vult’s Lateralus running the TH model at 18 dB is a wonder to behold and delivers a fantastic range of liquid sounds to go far further than that certain silver box famous for its (not) 18 dB filter (in reality, it’s a 4 pole design but one pole behaves in an errant manner). The 18dB Vult with a modulated notch on the Braids CSAW is glorious and provides a tone and timbre that would be difficult to achieve in Softube Modular.

The reason I highlighted the differences wasn’t to be negative about the VCV versions of Braids and Clouds per sé but simply to mention that the sonic response is different to both Softube Modular and the hardware but that’s not necessarily a bad thing as VCV is capable of things that Softube can’t; the sonic response isn’t bad, it’s just different but that difference can colour someone’s impressions; in other words, it can create a cognitive bias in a particular direction. As for the UX, this is of course subjective, but the Softube version of Braids includes a set of < > navigation buttons to traverse the waveforms as well as the jog wheel. I personally find the jog wheel on the VCV version of Braids overly sensitive (this may be because I use a Kensington trackball, but nonetheless I find that I often skip two or three programmes instead of a single one).


I’m nearly 51 and have made a living from music production for 30 plus years and regularly work for major artists such as Bryan Ferry (his solo output as well as a special set of remixes of Ladytron and 2HB, which will be released on Record Store Day (13th April this year - to quote " Record Store Day is the one day of the year when over 200 independent record shops all across the UK come together to celebrate their unique culture). I’m hardly in the situation of having “sold like 5 copies yet on your pretentious Bandcamp page in the last 4 years”

I’ve reread my post twice attempting to work out where I was in any way disrespectful of VCV. and certainly said nothing that could be considered “shaming the sound” of what I repeatedly called “a truly innovative product”.

Lesson learned. I’ll obviously stay away from posting any constructive criticism for fear that it will yet again be misconstrued to such a great extent and be met with such a juvenile response.

One last thing whilst I’m here. I noticed a bug in the CVC version of Braids today. I can modulate ‘Color’ with an LFO but ‘Timbre’ won’t respond to the LFO. I’ve attempted with many different LFO’s both directly and via a VCA. I noticed this whilst creating the CSAW patch I mention earlier and didn’t get a chance to see if it’s one of those weird ones that only affects a single programme.

System: Dual XEON E5-2690v2 @ 3.4Ghz (40 core), Windows 10 (latest patches), dual GTX 1080 Ti’s.


(Andrew Belt) #62

That comment was not directed toward you, since you can either 1) make music/sounds other people enjoy or 2) make music/sounds that you enjoy, regardless of what a certain subculture tells you that you should enjoy. I agree with your entire post from On long term sustainability for third party plugin devs, and that is not what I am commenting about. I brought this up because of what an excellent example this is.

What I am commenting about is people who are unable to enjoy themselves while using a musical instrument, or any artistic tool, because a certain subculture (e.g. certain audiophile subcultures) has told them there are certain properties about tools that prevent you from having enjoyment when using them. Happens every day. These people are perpetually unhappy with music production (or any artistic hobby) regardless of what they create because they are constantly trying to create works within some culture’s narrow definition of what is “good”. This causes them to never create creative music at all, thus their empty Bandcamp.

(Jonathan Moore1) #63

Thanks for clearing that up Andrew. I’m so glad as I really am a massive fan of your creation!

Because you quoted @Josep’s comment to me, I took your comment as being directed at me too. You’d of thought that I’d have grown used to the distortion field of forum posts by now. :slight_smile:

(Peter Vos) #64

You need to open the Modulation attuvenator :wink: (attenuverter)

(Leonardo Laguna Ruiz (Vult)) #65

You said many good things about Vult so it would be a shame if you stop commenting in the forum :slight_smile:

Regarding the sound quality, probably both Braids and Clouds are bad examples to compare. These two modules are mostly DSP based. The only analog parts are the DAC and the antialiasing filters. Those parts should be as transparent as possible. In the case of Clouds, the internals run at 32 kHz sampling rate. Most probably Softube used exactly the same code as VCV to create the modules (the original Mutable Instruments code). If you think Softube version sounds better, my guess is that they added some filtering and distortion to make the sound more appealing.

Sound quality is something I think a lot about when I’m modeling some electronics. Often times the more accurate the model is, the “worst” it sounds and the more CPU it consumes. I quote “worst” because it’s completely subjective; analog electronics has lots of quirks. I always try to strike a good balance between something that sounds good and it’s useful by not eating all the CPU.

In Rack you can find plugins that go from the most primitive sound sources and processors (with aliasing or very digital-like) to very complex and carefully crafted. In Softube you can find fewer but more developed modules. If one would want to compare the sound quality of VCV vs Softube it’s necessary to compare only the very best… and still it’s not clear what should be the criteria used to evaluate.

One realm in which Rack is a clear winner is in the number of people it makes happy.

(Josep) #66

From the original Mutable Instruments’ Braids manual, @jonathan.moore1:

F : Modulation attenuverter. This knob controls the amount and polarity of modulation applied to the TIMBRE parameter, from the TIMBRE CV input jack.

So you’re comparing Softube Braids with VCV Rack Braids, and you did not know that? :thinking:

(Josep) #67

Absolutely! :+1:

(Jonathan Moore1) #68

Agreed entirely. Much of what I like about Softube products is their modeling of nonlinearities. It’s the various places where saturation and clipping occur that gives both hardware and software emulations their character - it is of course completely subjective as to whether that character is sonically pleasing.

Something I really enjoy with the Vult filter modeling is the way that underdriving the signal in the filter and overdriving at the VCA (the Bogaudio VCAMP is great for this) or post VCA, in the DAW produces very different results (my preferred route is post-VCV, using the Sly-Fi AXIS (DSP modeled by UBK) as it’s saturation behaviour also changes in a very nonlinear manner depending on whether it’s slightly or massively overdriven). The TH model of Lateralus has a wonderfully liquid quality with the drive around 10’o’clock and the AXIS both compresses and saturates it’s output in a very complimentary fashion. The reason I love working in Live more than any other DAW is that I can make plugins behave like CV controlled devices (via Max4Live) that act as natural extensions of hardware or software emulations of hardware.

With regard to aliasing, I love it’s sonic qualities with certain digital kit/emulations, especially when paired with a decently saturated signal path. I’m not a purist who oversamples everything x16 at render-time. Sometimes its the gnarly nature of aliasing that provides the core sonic signature to a production. :slight_smile:

I’m very open with regard to sonic sources and even use iOS devices as sound sources. The performance of the A12X surpasses most common i7’s and i9’s on the market and it seems to be especially performant with DSP. The Korg Odyssey emulation brings my workstation to it’s knees whereas I can run numerous instances via Gadget on my iOS devices (it’s so annoying that Korg do everything in a walled garden manner rather than releasing AU3 versions). One of the major differences of the performance is that Odyssey is multi-core when polyphony is used on iOS but remains single core in both Windows and OS X. The code is exactly the same on iOS and desktop versions, the only difference is that the iOS version is maxed at 8 notes of polyphony/unison whereas the desktop versions default to 16 voices.

I’m really hoping that the introduction of polyphony to VCV is matched with multi-core capabilities in those modules that form the building blocks for polyphonic patches.

Outside of virtual modulars, I’m a long term user of U-He Diva and Bazille and have used Repro 1/5 since they were in beta (IMO Repro 5 is the current benchmark for polyphonic virtual analog emulation). These are only able to provide polyphonic performance as each note is effectively handled by a separate core (the programmatic reasoning is that each note is a separate signal path according to the U-He developer). Attempting anything of merit polyphonically in Softube Modular is always a challenge as you’re always held back by the single core performance of your workstation. In my case I have 40 core’s on my main Windows workstation (16 faster cores on my Mac) but they have a maximum single core performance of 3.4 Ghz. I find with VCV I manage comfortably in most occasions on my PC as my 1080 Ti GPU is more than capable of driving all my system needs and I’ve found that even though my Mac has faster single core performance, it croaks more easily due to Apples weird and wonderful ways with graphics drivers/hardware.

Hopefully, it’s clear now that what I was highlighting with regard to the differences with the sonic response of the Mutable Instruments emulations was more the possible false perceptions it could lead to with VCV triallists who are familiar with either or both the hardware modules and the Softube emulations. Reading it back I could have been clearer about that intent.

As I mentioned before, I think that VCV in general and the Vult modules provide a decent compromise between CPU and sonic performance (I’m very intrigued by your bespoke VULT development platform as this on the surface would seem less performant than native C++ but your plugins are impressively lean).

If you’re totally uncompromising with the modeling of the signal path that limits the usefulness of a virtual modular with today’s hardware. With Softube Modular you hit far more brick walls than you do with VCV, especially when using modules such as the Buchla 259e - but considering the Buchla hardware retails at $1600, I’m happy to put up with it’s software brethren occasionally crippling my poor overworked workstation… The joy of VCV is that you’re hard pushed to enter the realm of coughs and splutters and the vast range of options keeps your creativity flowing - as long as your not crippled by the paradox of choice - but that’s a whole different discussion!

(Nik Jewell) #69

That’s some pretty hot kit you have in the hyperdrive of your spaceship Jonathan :wink:

(Leonardo Laguna Ruiz (Vult)) #70

Yes, as you mention, is hard to talk about quality when even the defects like distortion or aliasing can be sonically pleasing.

The Vult language is exactly as fast as C++. This is because The Vult compiler generates C++ that is very easy to optimize for the C++ compiler. The Vult compiler does a many extra task to simplify writing the code and also impose restrictions that avoid me writing unnecessary complex code.

(Lars Bjerregaard) #71

Very well said Jonathan, I largely agree.

(Lars Bjerregaard) #72

Hear hear!


At least for Cloud I noticed a behavior that does not happen on softube’ version:
VCV version: Any input audio signal comes out with a significant loss of volume even with the blend knob all the way dry. The “in gain” knob will only create a limiting effect.

It does not happen in the softube version, and it changes a lot when you send a guitar into cloud, the sound comes out as loud and warm at it comes in.

@jonathan.moore1 It would be awesome if you could check the behavior of the hardware.

(Jonathan Moore1) #74

The 40 core is showings its age (purchased in 2013). But until Threadripper sort out their multicore performace for applications outside of 3d rendering, it will soldier on. ;j

I’m hoping never to have to pay Intel over inflated prices for a XEON again. But having said that, you really do get what you pay for. The XEONS in my workstation were launched in 2012 and still provide reasonble firepower by modern standards. I’m also hoping the old girl’s got a few more years in her too!

(Patman / NYSTHI Manual) #75

If you hold Ctrl, it (or any other control) chooches along at a leisurely pace… :wink:

(Nik Jewell) #76

@pyer I have that problem with VCV Clouds too. The output levels are too quiet and I have never been able to fix that to my satisfaction.

The Softube one doesn’t, as you say, but it doesn’t have any of the alternative modes or firmwares. “Too confusing for the user” was Softube’s response when I asked them about this :wink:

I too would be interested to know what the hardware does.

(Leonardo Laguna Ruiz (Vult)) #77

I ran a quick test in my HW Clouds (DIY clone) . I’m inputing a 5 Vpp signal and the output clips about 3.5 Vpp. The output is quite low.

If some out has a real one please confirm this behavior.

(Jonathan Moore1) #78

If anything the Clouds hardware module can easily be overdriven if you’re not careful with your CV’s. It certainly doesn’t suffer from being too quiet.

One of the things to remember with the Softube version of Clouds (and Softube Modular in general) is that nonlinearities are built into all parts of the signal path right up to the overall gain knob. This means that overdriven signals tend to distort in a ‘pleasing’ manner.

Plus if you’re using a DAW like Ableton it has shed loads of headroom above 0dB (and that’s most modern DAW’s AFAIK) due to the internal 32/64 bit floating point resolution. By rights you should gain stage in just the same manner as you would with hardware, but DAW’s are very forgiving these days. However if you mix everything with overshoots your master will sound like mush, even though it’s not technically been clipping.

This is one of the reasons it’s hard to compare hardware modular to software. The signal path is very different. Softube Modular is a great facsimile of a Eurorack system because it ‘component modelled’ throughout all the way to the final master gain. It’s very probably the reason that I like the Vult modules so much in VCV (as I understand things, they’re component modeled too). However, I suspect that the Softube engineers have purposely built in ‘pleasing’ distortion where real hardware would let you know you’ve over egged the pudding! I’d guess there’s an assumption that folk using a software modular have a very different learning path than those learning with hardware, so the software has to be a little more forgiving.

(Jonathan Moore1) #79

I was using the attenuverter but borrowed an output from an LFO from another part of my patch and forgot that I had the Scale set very low on that LFO (a BOGAUDIO LFO). Total user error on my part, the LFO was already attenuated so the attenuverter on Braids was applying minimal modulation.I should have checked things out properly before tacking a ‘bug report’ on the end of a post. Major doh! moment I’m sure I’ll commit many others too. :wink: