Login

Timed mode how much ms?

Tips & tricks, working results, technical support

Re: Timed mode how much ms?

Postby Cupwise » Sat Aug 10, 2013 1:52 pm

i guess i was wrong about my rayphlex comp being perfectly flat even with a 2.8ms kern length. i zoomed in on the graph for it and there is a ripple/bump in the midrange, where it goes up a bit in level for the freqs above that. you can barely see it on the graph unless you zoom in and i think it's less than half a db which is hardly noticeable. some programs, if you lower the h1 length to 2.8 you get a cut of several db. so there is some degree of dependence on the hardware here.

like i said, you can have NAT generate a single sine tone sweep and then deconvolve that into an impulse without running it through hardware, and the impulse you get fades to total silence in well under 1ms. you can't even see any bass cycles. if you load the program using that single impulse, you can lower the h1 length to 1ms and still have a perfectly flat response, with no 'bump' or ripple, or dip in the bass anywhere. perfectly flat. so somehow the impulse less than 1ms is still accurately recreating a flat response.

i'm not sure how that makes any sense, but i think it proves that there is a degree of hardware dependency here, and this is why the rayphlex comp has such a hardly noticeable dip. something about the aphex compressor being as clean as it is. because the impulses that are in those programs look a lot like the ones you get from deconcolving an unprocessed sine sweep.

so to me, this issue is less important for some hardware, and more important for other hardware, in terms of getting an accurate bass response.
Cupwise
Expert
Expert
 
Posts: 773
Joined: Tue Nov 23, 2010 2:03 am

Re: Timed mode how much ms?

Postby Tim Petherick » Sat Aug 10, 2013 3:02 pm

I can see why G picked the standard ms for eq's

Also on harmonics; 10ms is generally used. So, if we take a 50hz sine wave ,first harmonic will be 100hz.10ms kernal length covers that.

So if the standard lengths have 50ms k1 and 10ms on harmonic players.That's 20hz on fundamental and 100hz on first harmonic so that covers the above.I guess it makes sense for higher order harmonics to not need long kernal lengths.

On a side note,Ahd feature in nat sampler limits the length of harmonic orders. But these are or can be used for higher orders. Im guessing Ahd was developed to capture those higher orders.

Also someone asked about truncation on samples, if the kernal length is not long enough to cover the capture. I believe that nebula reverb stops this from happening but nebula regular does not. That could be a reason why nebula regular sounds slightly different!? I think some mentioned this on the transient loss thread.

I'd like to just say that my opinion on transient loss page about nebula regular has changed. I think it maybe on page 1 or 2. My opinion changed when I actually started developing presets.


By the way just to make it clear , the plot graphs I posted had the same results for both timed and freq d player at the same lengths.

If you want to Work out bass response covered ,take 1000 and divide by the chosen ms.

Hope that reads ok. I have to check things over and over.
My posts are normally very short! :lol:
User avatar
Tim Petherick
Expert
Expert
 
Posts: 1352
Joined: Sat Apr 17, 2010 4:07 pm
Location: Bath , Uk

Re: Timed mode how much ms?

Postby yr » Sat Aug 10, 2013 5:21 pm

Short clean kernels can cause more artifacts, I keep it at 50ms for preamps to minimize aliasing, bass clicks etc. Here is the VCult T:
clean 2ms.png
clean 2ms.png (198.95 KiB) Viewed 1068 times

clean 10ms.png
clean 10ms.png (187.82 KiB) Viewed 1068 times

clean 50ms.png
clean 50ms.png (146.39 KiB) Viewed 1068 times
Reuven | post-production & sound-design | scenography |
website | nebula presets
User avatar
yr
Beta Tester
Beta Tester
 
Posts: 438
Joined: Mon Mar 29, 2010 9:04 am
Location: Amsterdam

Re: Timed mode how much ms?

Postby Tim Petherick » Sat Aug 10, 2013 5:34 pm

Wow that's not subtle eh? :lol:

Nebula reverb?
User avatar
Tim Petherick
Expert
Expert
 
Posts: 1352
Joined: Sat Apr 17, 2010 4:07 pm
Location: Bath , Uk

Re: Timed mode how much ms?

Postby yr » Sat Aug 10, 2013 6:44 pm

yes- reverb.
Reuven | post-production & sound-design | scenography |
website | nebula presets
User avatar
yr
Beta Tester
Beta Tester
 
Posts: 438
Joined: Mon Mar 29, 2010 9:04 am
Location: Amsterdam

Re: Timed mode how much ms?

Postby richie43 » Sat Aug 10, 2013 7:54 pm

When you guys are differentiating between Nebula regula and Reverb, that is confusing to me. If you are tweaking the xml's to suit your needs, it doesn't matter which dll you start with since the only difference between the regular and reverb dll's is the settings in the xml. So why does it matter which you use for these tests that you are doing the graphs for? I am not questioning your knowledge, by the way, I am always trying to expand my own understanding of this technology.
The Sounds of the Hear and Now
http://soundyaudio.com/
richie43
Expert
Expert
 
Posts: 4868
Joined: Fri Feb 04, 2011 8:47 pm
Location: Minnesota, USA

Re: Timed mode how much ms?

Postby Cupwise » Mon Aug 12, 2013 9:54 am

as far as i know, the dlls for nebula 'standard' and nebula reverb are exactly the same. the only difference is a few settings in the xml. i think the main one is that reverb is set to be able to load and convert longer kernel lengths. maybe they have different buffer settings too.

but back to the bass cycles thing- if there is a significant phase shift in the bass, then that could possibly add several ms to the required length you'd need to capture them. i've seen h2+ impulses where the impulse took up to or over 100ms to fade below the noise floor. i just don't think the measuring bass cycles thing is the last word on the issue.
Cupwise
Expert
Expert
 
Posts: 773
Joined: Tue Nov 23, 2010 2:03 am

Re: Timed mode how much ms?

Postby enriquesilveti » Tue Aug 13, 2013 9:44 am

Some clarifications:

1. Volterra Kernels are FIR filters not audio.
2. Preamps and Comps and EQs are complete different NAT sampling models. Nebula be with all, Giancarlo is a Genius.
3. You can not sample something that not exist. If your DA filter from 10/20 Hz to DC, you will not able to sample. Check for example Mammoth HPF or try to sample a software perfect HPF, you will get -30/50 dB but not to -oo. NAT need something to compare.
4. The best way to get good quality samples is using repetitions and you get around -6 dB on each if the device under sampling is stable. This help to sample more kernels.

if you have NAT generate a tone sweep offline, and don't run that through anything and just make a program from it without touching it, the impulse you get out of that is very very quick. it fades to digital silence pretty much instantly, and it still has a good bass response. but in some cases, some hardware requires more time for an accurate bass response. i think it's probably something to do with the hardware introducing phase shifts on the low bass, so that extends the data for the bass out beyond the typical impulse length. so you have bass cycles going out up to 50ms to 100ms or even beyond that.


Correct, and in addition Nebula make a peak synchronization to handle this issue.

Maybe enrique Can chime in But I thought that nebula will play a flat frequency response on the low end if the kernals are not long enough to represent the actual capture. So if you take a compressor it will be accurate down to 100hz at 10ms kernal length either on timed or freqd. But will flatten out along the low end within reason.


Switching between SPLIT-H and non SPLIT-H will send Nebula to a unstable state, SAVE and RELOAD before compare. I don't know if is a bug, issue or a feature, ask Giancarlo.

There is some information about this topic in compressor tip of the month plus we are preparing some new information...

8.2 Vectorial Volterra Kernels Technology

Vectorial Volterra Kernel Technology is a combination of multiple advanced technical processes that Acustica Audio team develops including: Vectorial Kernels Technology, Dynamic Volterra Kernels Technology, Multi Dimension Kernels Technology, Time Varying Kernels Technology and Split Hybrid Kernels Technology.
The Volterra series is a model for non-linear behavior. It has been applied in the fields of biomedical engineering and biology, especially neuroscience. It is also used in electrical engineering to model intermodulation distortion.
In mathematics, a Volterra series denotes a functional expansion of a dynamic, nonlinear, time-invariant functional. The Volterra series, which is used to prove the Volterra theorem, is a series of infinite sum of multidimensional convolutional integrals.
Estimating the Volterra coefficients individually is complicated since the basis functionals of the Volterra series are correlated. This leads to the problem of simultaneously solving a set of integral-equations for the coefficients.
Enrique Silveti.
Acustica Audio customer and technical support.

MBP 11.5 (i7-4870 | 16 GB | 512 SDD)
SP4 (i5-6300 | 8 GB | 256 SDD)
UFX | Lyra2 | USBPre2
VM U15 | VM Win10 | VM OSX 10.12
N4/NAT4 | SPX3 | RX5 | LN2C | Smaart8 | R5 | PT12 | PX10 | NIK5
User avatar
enriquesilveti
Expert
Expert
 
Posts: 2663
Joined: Sun Mar 28, 2010 9:00 pm
Location: Lodi | Madrid | Buenos Aires

Re: Timed mode how much ms?

Postby enriquesilveti » Tue Aug 13, 2013 9:47 am

8.2 Vectorial Volterra Kernels Technology

Vectorial Volterra Kernel Technology is a combination of multiple advanced technical processes that Acustica Audio team develops including: Vectorial Kernels Technology, Dynamic Volterra Kernels Technology, Multi Dimension Kernels Technology, Time Varying Kernels Technology and Split Hybrid Kernels Technology.
The Volterra series is a model for non-linear behavior. It has been applied in the fields of biomedical engineering and biology, especially neuroscience. It is also used in electrical engineering to model intermodulation distortion.
In mathematics, a Volterra series denotes a functional expansion of a dynamic, nonlinear, time-invariant functional. The Volterra series, which is used to prove the Volterra theorem, is a series of infinite sum of multidimensional convolutional integrals.
Estimating the Volterra coefficients individually is complicated since the basis functionals of the Volterra series are correlated. This leads to the problem of simultaneously solving a set of integral-equations for the coefficients.


Again G. is a genius!
Enrique Silveti.
Acustica Audio customer and technical support.

MBP 11.5 (i7-4870 | 16 GB | 512 SDD)
SP4 (i5-6300 | 8 GB | 256 SDD)
UFX | Lyra2 | USBPre2
VM U15 | VM Win10 | VM OSX 10.12
N4/NAT4 | SPX3 | RX5 | LN2C | Smaart8 | R5 | PT12 | PX10 | NIK5
User avatar
enriquesilveti
Expert
Expert
 
Posts: 2663
Joined: Sun Mar 28, 2010 9:00 pm
Location: Lodi | Madrid | Buenos Aires

Brain Blowing!!

Postby richie43 » Tue Aug 13, 2013 6:15 pm

That's it Enrique...... Now I have to retreat from the rest of my life and re-read your last few posts and the documentation you referred to.... Giancarlo is a genius, but he also has some very awesome people in his support "staff".......
Attachments
BrainOverload.jpg
BrainOverload.jpg (73.1 KiB) Viewed 1019 times
The Sounds of the Hear and Now
http://soundyaudio.com/
richie43
Expert
Expert
 
Posts: 4868
Joined: Fri Feb 04, 2011 8:47 pm
Location: Minnesota, USA

PreviousNext

Return to Working with Nebula

Who is online

Users browsing this forum: No registered users and 3 guests