Mercuriall released Spark

[fortunjj]

Barmaleus,

One more interesting piece to this puzzle... Ableton assigns one core per track. Period. So since one core is used per track even a multi-threaded plugin such as Spark will still be using a single core - entirely defeating the benefits of the architecture parallelism.

The only reasonable answer I have is to live with the performance of the 32 bit plugin (which is still much worse than it is in Reaper but usable in Ableton) or figure out how to map key presses to plugin parameters in another host. This still doesn't explain why the 64-bit performs much worse, but that depends on how the 64-bit version was written. If it is more or less a recompile, there shouldn't be much difference. But if, for some reason, the 64-bit version relies on threading differently (again, I don't know why it would), then the constraint to run on one core could easily be the answer.

[Barmaleus]

fortunjj,

This is interesting. In this case there is no chance Spark would run flawlessly in Ableton, as it relies on multi-core processors. Unless there is a way to force Ableton to use all the cores per track.
The 32 and 64 bit versions are different indeed. As far as I remember, the 32 bit version may perform better. I don't know the technical reasons behind it (I am not a programmer), but that's how it is.

[Barmaleus]

Hi everyone,

Please fill in this short anonymous survey we are carrying out. It takes around 10 seconds to fill it in. Thanks a lot!
https://goo.gl/forms/S3vPs69y9a06Hfew2

[Barmaleus]

Hi everyone!

We've recently updated our web shop and here is what's new:
- PayPal buyer protection is now active
- PayPal orders are now accepted from more countries
- We can now send manual invoices through Yandex Checkout - another popular payment system. You can pay through it with your credit card, for example. To get a manual Yandex Checkout invoice, please contact our support through the website form at https://mercuriall.com/cms/contact.

Check out and subscribe to our SoundCloud channel: https://soundcloud.com/mercuriall
We have lots of new content and will be adding more!

[gbtommasi]

Tried the demo and I have to say it is a fantastic plugin, awesome for marshall sounds, even wonderful cleans. Very though to tell the difference in tone and response with a real JCM800 or a Superlead (never played a real AFD, still Sparks one is wonderful).

That said, @Barmaleus, could you clarify the cab section? I am very curious about it. Is it based on IRs?
I switched the whole cab section off and dropped an IRs loader after Spark, loading some picks from Ownhammer Marsh UK collection. As I always do with any plugin. But for the first time ever, it sounded worst, just not as good as Spark's cabinet.
This really impressesed me.

Thank you

PS: no performance problems here, Macbook Pro i7 3.7->4.5 Ghz, 16Gb RAM, Apogee Duet, Mainstage and Logic Pro X. But I can confirm other users issue with the noise gate: no matter how I set it, it does nothing...

[Barmaleus]

gbtommasi,

Glad you like Spark!
The cab section is indeed IR based. It's fed with multiple IRs captured in a specific way and then a special algorithm "fills in the empty spaces". This way there is no need to endlessly scroll through the list of single IRs - just move the mic around and find the sweet spot.

Re. the noise gate - do you know what's the strength of the signal that comes into Spark? Maybe the the noise floor is very low?

[gbtommasi]

Thank you for the explanation, it works very well indeed

For the noise gate: no, single coils and quite a lot of hum and buzz with high gain and overdrives. This evenings I will do more tests, possibly I made something wrong

[Barmaleus]

Pssst!
ReAxis annonuced =)
https://www.kvraudio.com/news/mercurial ... -aax-38935

[tappert]

No cab based on Mesa? For a Triaxis amp simp, it's a bit surprising

[EvilDragon]

Yeah, a couple of Mesa cabinets are an absolute must to go along with Triaxis...


Well, gonna have to use 3rd party IRs, but it's never gonna sound the same as cabinets modelled the Mercuriall way... :/

[alteregoxxx]

@Burillo
Thanks a lot for the detailed feedback, appreciated!
1. Which dials were crackling for you when moved?
There is no easy immediate fix. Each increment of the dials' position is studied by neural networks. For each increment there is a separate model of differential equations. When the "surfaces" of two neighboring models' solutions are not identical, the crackling may appear.
Basically, neural networks need to be trained again (and this physically takes a lot of time), but we need to know the exact plugin state to reproduce it. If you can make a GUI-less screenshot, that'd be amazing.

2. Have not heard about lagging so far. Need to check.

3. Yes, CPU usage will be above average, that's a trade-off for modeling the entire thing without any wave-shapers or EQ-matching.


@Baxterbrown
1.Could you tell me the exact models of the AMD and Intel processors you have? Unless it's a secret =)

2. To clarify - the correct way to check the CPU usage is by looking at the Task Manager. You should not be worrying about the DAW meters jumping around, especially in Studio One. Reaper's meters look much more robust to me.

3. To clarify - are you experiencing bad plugin performance (cracks and pops and similar) or just worried about high CPU usage? This part is not clear to me yet, sorry.

4. See #3 above in my reply to Burillo =)

I'm following this topic with a smile on my lips; it's years that I spread the word in saying that Mercuriall modeling approach is leaps and bounds over the concurrency: you can felt an hyper realistic responsiveness in tone and dynamic under the fingers while playing, a feeling that is simply missing (or not so "alive") in (almost) all the amp sims.
I'm a Mercuriall addicted since their first releases (Tomato, Harlequin etc etc), and their free JCM 800 is still my choice for that Marshall sound. Finally I've again some time to playing guitar and I'm evaluating from some days their Spark suite: state of the art.

@Barmaleus
Well, I'm particularly "involved" in the technical aspect of amp sims modeling, hardware in my case: I'm working in my spare time, from a couple of years, on a low voltage, 18V, hardware "clone" of a 12AX7 realized using solid state parts and already "cloned" some famous preamps using this approach.

So, I've a tech question, and I would really appreciate your feedback on it: since the coupling between the guitar and amp sims in the DAW world is, in the best case, a simple DI with a resistive input impedance of more or less 1Mega Ohm while in the real world tube amps have all but a resistive input impedance (Tube amp input impedance is not flat and not constant at 1meg in the audio band) how can you model this aspect?

Moreover, the voltage transfer at the tube amp input is also dependent on the guitar output impedance (strongly inductive-capacitive), while in an amp sim you have the DI output (again, usually a resistive, and low, output impedance) impedance feeding the input of the amp sim.
And this aspect has a massssssive effect on the tone. Honestly I'm not able to see a way of modeling this phenomena in the software world; the only way I can think of is making assumptions, for example modeling a "generic" type of pickup, feed it with the signal of the DI output, and feeding the amp sim with the signal of this pickup model. Still in this case, we all know a LesPaul is immensely different ffrom a Tele...so I really can't see a way...

[Barmaleus]

Hey guys!

I've answered ReAxis questions here (let's discuss ReAxis stuff there): viewtopic.php?f=6&t=493529&p=6906524#p6906524

@alteregoxxx
Thank you for your support and kind words! It reassures us we are on the right path! =)
As for your technical question - I have no idea myself. Need to check with Vladimir, the mastermind behind the code. He should know.

[cknapheide]

An amp has a 1 Megaohm resistor at the input, that is all the guitar is seeing. The inductive part in the equation is only the coil(s) of the pickup.

So just be sure your interface has this 1 Megaohm at the input (hi-Z) and the modeler will behave right

[alteregoxxx]

An amp has a 1 Megaohm resistor at the input, that is all the guitar is seeing. The inductive part in the equation is only the coil(s) of the pickup.

So just be sure your interface has this 1 Megaohm at the input (hi-Z) and the modeler will behave right

Nope sir.

The 1 Megohm resistor is in parallel with the input impedance of the tube itself.
And if you believe that is infinite, you are forgetting to consider the parasitic capacitance effect between Grid and Cathode (Cgk) of a Vacuum Tube first and, much mure important, the Grid to Plate capacitance (Cgp): the latter, because of the strong Miller effect present in the common cathode configuration (read first stadium of a lot of preamps, JCM800, Soldano SLO100 etc etc etc), plays a significant role in forming a low pass filter when a source has a finite impedance, that is always.

P.S.: Permit me a wise suggestion: be wise and avoid to be peremptory when you talk about things that you are not sure you mastered perfectly.

[cknapheide]

I agree in all you describe, but this happens inside the amp itself and has to be modeled.

The guitar is not affected because its is just loaded by the 1 meg resistance.