Login

Transient loss

Tips & tricks, working results, technical support

Re: Transient loss

Postby alpsessays » Sat Mar 16, 2013 4:54 pm

Ok, so are we def saying it is purely the Timed switch that is what makes the most difference to the sound - quite a few people seem to be releasing libraries with HQ Timed versions - so there appears to be general consensus on this. What I mean is, what about things like evf 17? Is that neither here nor there now in terms of best quality (in other words, the Timed thing makes the most difference).

Chris
alpsessays
User Level I
User Level I
 
Posts: 17
Joined: Wed Sep 05, 2012 7:53 pm

Re: Transient loss

Postby Cupwise » Sat Mar 16, 2013 10:20 pm

i've recently come to a decision about that which goes against what may be general consensus and what i previously thought was the best way myself. this is just me expressing my thoughts and why i think them, and people can consider them however they want, or not. but here it is- instead of evf 17, i think that for 'preamp style' programs, the best thing to use for more realism may be peak mode. i've been releasing any of those types of programs i've put out recently with that mode selected (such as pass through programs made from the comps i've sampled). and i've released updates for a few of my older sets where one of the things i updated is switching from evf17 to peak.

here's my reasoning behind this- in an actual hardware compressor, there is a detector circuit. it detects the level coming in and applies compression if it exceeds the threshold. but in something like a preamp, there is no detector circuit. nebula has to always have the level detector though, because it needs to know which dynamic sample to play, regardless of whether you are using a compressor program or a preamp/console/etc. so to get results closer to hardware like a preamp which has no detector circuit, the obvious thing would be to try to minimize any influence by Nebula's detector. if you send a loud signal into a preamp or a tube amp, it distorts instantly. there isn't a detector circuit there, measuring the signal to decide whether it should 'saturate' or not. and there definitely isn't an attack control that can delay the onset even more (which nebula also always has for dynamic stuff). it happens instantly. the closest detector mode to that type of behavior is peak. evf 17 is like a modified RMS detection. RMS measures the signal over a length of time and does something like averaging the level over that time and uses that result to decide what to do. this is slow and means that fast transients don't register. a tube amp does nothing like that at all. peak detection mode is closer to instant, so to me it makes sense to use that.

and i feel that once i thought about it and realized this, i could actually hear that when sending a drum through some things like a tube amp/preamp program, there is a bit of an un natural quality there with RMS or EVF 17 mode. because what's happening is that the transients are too fast to be detected by those modes, so they will be using the lower dynamic samples. a drum transient could peak at -8db, but because evf17 mode reacts slowly, it might be using the samples that were taken from the hardware at -40db, or whatever the lowest samples were taken at. i felt i could hear results that seem like that's exactly what's happening (but i know factually that this is exactly what's happening). in a real tube amp, a drum transient at a high level would instantly produce higher levels of distortion, but a nebula program using evf17 made from that tube amp, that won't happen. the transient will use the cleaner, lower dynamic samples. and for something with tubes, depending what it is, at higher levels you might have a bit of compression with the distortion, but you won't get that for those fast peaks either. to me that's the part i think i can hear that doesn't seem right. even if the compression/saturation/distortion is still kind of subtle, if you run drums through some hardware peaking at a fairly high level, you will hear the peaks getting a bit saturated. with evf17 in nebula, you wont. i'm not just guessing here either, i know that this is how nebula works.

so what's happening is that the fast transient peaks which are louder than the body of the drum, and should be saturating more, are actually saturating LESS than the body of the drum. because by the time the body of the drum is happening, the evf 17 detector mode is finally telling nebula to use higher level dynamic samples, after it missed the loud transient. even if you switch to peak mode the default attack setting for these programs is 15ms, which means that nebula won't use the higher input level samples until after that attack stage is complete. again, a tube amp or preamp doesn't have that delay. i think to me it usually sounds better around 7ms. the fastest speed would seem logically to be the most realistic setting but i think for me it usually sounds best around 7ms, and i usually set release to around 2ms. that's just my preference but i do know what sometimes at the fastest positions a program may produce artifacts.

for equalizers, i'm pretty much 100% certain that it won't matter at all which mode you use for detection. evf 17, peak, rms, whatever, it won't matter- because they aren't going to be used. these are detection modes and that page of nebula settings is all about detecting the input level so nebula can decide which dynamic sample to use. with equalizers there are no dynamic samples. at least with MOST equalizers. the only exceptions to this i know about are henry's titan eq which he sampled like the 'preamp style' programs. there are no adjustable controls, he just sample dynamic captures of the controls at different settings. then i released a couple of what could be called dynamic eqs in my frequency tweakers sets, using some weird old tube equipment. but your typical nebula eq with adjustable controls has no dynamic samples, so the detection mode does absolutely nothing. this is a 100% true fact and you can look in the FUN page right in nebula to see this. usually if not always, if a program has dynamic samples and uses the evf1 controller, it will show up in the FUN page. load a preamp program and go to the FUN page and look for evf1. now look for it with an eq program. not there.

so huge long rant summarized- in my opinion peak mode with faster than default attack setting, is better for getting transients more authentic to hardware, with amp/preamp/console/etc type programs. anyone can feel free to disagree but i feel like i've come to understand pretty well, a lot about how nebula works under the hood, and i've explained why i feel the way i do. evf17 is slow and doesn't catch transients to have them use the higher level dynamic samples that they should use.
Cupwise
Expert
Expert
 
Posts: 773
Joined: Tue Nov 23, 2010 2:03 am

Re: Transient loss

Postby Melon_Jack » Sat Mar 16, 2013 11:08 pm

Cupwise wrote:i've recently come to a decision about that which goes against what may be general consensus and what i previously thought was the best way myself. this is just me expressing my thoughts and why i think them, and people can consider them however they want, or not. but here it is- instead of evf 17, i think that for 'preamp style' programs, the best thing to use for more realism may be peak mode. i've been releasing any of those types of programs i've put out recently with that mode selected (such as pass through programs made from the comps i've sampled). and i've released updates for a few of my older sets where one of the things i updated is switching from evf17 to peak.

here's my reasoning behind this- in an actual hardware compressor, there is a detector circuit. it detects the level coming in and applies compression if it exceeds the threshold. but in something like a preamp, there is no detector circuit. nebula has to always have the level detector though, because it needs to know which dynamic sample to play, regardless of whether you are using a compressor program or a preamp/console/etc. so to get results closer to hardware like a preamp which has no detector circuit, the obvious thing would be to try to minimize any influence by Nebula's detector. if you send a loud signal into a preamp or a tube amp, it distorts instantly. there isn't a detector circuit there, measuring the signal to decide whether it should 'saturate' or not. and there definitely isn't an attack control that can delay the onset even more (which nebula also always has for dynamic stuff). it happens instantly. the closest detector mode to that type of behavior is peak. evf 17 is like a modified RMS detection. RMS measures the signal over a length of time and does something like averaging the level over that time and uses that result to decide what to do. this is slow and means that fast transients don't register. a tube amp does nothing like that at all. peak detection mode is closer to instant, so to me it makes sense to use that.

and i feel that once i thought about it and realized this, i could actually hear that when sending a drum through some things like a tube amp/preamp program, there is a bit of an un natural quality there with RMS or EVF 17 mode. because what's happening is that the transients are too fast to be detected by those modes, so they will be using the lower dynamic samples. a drum transient could peak at -8db, but because evf17 mode reacts slowly, it might be using the samples that were taken from the hardware at -40db, or whatever the lowest samples were taken at. i felt i could hear results that seem like that's exactly what's happening (but i know factually that this is exactly what's happening). in a real tube amp, a drum transient at a high level would instantly produce higher levels of distortion, but a nebula program using evf17 made from that tube amp, that won't happen. the transient will use the cleaner, lower dynamic samples. and for something with tubes, depending what it is, at higher levels you might have a bit of compression with the distortion, but you won't get that for those fast peaks either. to me that's the part i think i can hear that doesn't seem right. even if the compression/saturation/distortion is still kind of subtle, if you run drums through some hardware peaking at a fairly high level, you will hear the peaks getting a bit saturated. with evf17 in nebula, you wont. i'm not just guessing here either, i know that this is how nebula works.

so what's happening is that the fast transient peaks which are louder than the body of the drum, and should be saturating more, are actually saturating LESS than the body of the drum. because by the time the body of the drum is happening, the evf 17 detector mode is finally telling nebula to use higher level dynamic samples, after it missed the loud transient. even if you switch to peak mode the default attack setting for these programs is 15ms, which means that nebula won't use the higher input level samples until after that attack stage is complete. again, a tube amp or preamp doesn't have that delay. i think to me it usually sounds better around 7ms. the fastest speed would seem logically to be the most realistic setting but i think for me it usually sounds best around 7ms, and i usually set release to around 2ms. that's just my preference but i do know what sometimes at the fastest positions a program may produce artifacts.

for equalizers, i'm pretty much 100% certain that it won't matter at all which mode you use for detection. evf 17, peak, rms, whatever, it won't matter- because they aren't going to be used. these are detection modes and that page of nebula settings is all about detecting the input level so nebula can decide which dynamic sample to use. with equalizers there are no dynamic samples. at least with MOST equalizers. the only exceptions to this i know about are henry's titan eq which he sampled like the 'preamp style' programs. there are no adjustable controls, he just sample dynamic captures of the controls at different settings. then i released a couple of what could be called dynamic eqs in my frequency tweakers sets, using some weird old tube equipment. but your typical nebula eq with adjustable controls has no dynamic samples, so the detection mode does absolutely nothing. this is a 100% true fact and you can look in the FUN page right in nebula to see this. usually if not always, if a program has dynamic samples and uses the evf1 controller, it will show up in the FUN page. load a preamp program and go to the FUN page and look for evf1. now look for it with an eq program. not there.

so huge long rant summarized- in my opinion peak mode with faster than default attack setting, is better for getting transients more authentic to hardware, with amp/preamp/console/etc type programs. anyone can feel free to disagree but i feel like i've come to understand pretty well, a lot about how nebula works under the hood, and i've explained why i feel the way i do. evf17 is slow and doesn't catch transients to have them use the higher level dynamic samples that they should use.


In theory, close to truth.. That's, imo, can explain low frequency loss. It's all about knowing original hardware and testing settings, until u find closest point. Or listening and choose what fit your style, your sound most.
Sorry for my bad english =))
User avatar
Melon_Jack
Member
Member
 
Posts: 280
Joined: Thu Apr 01, 2010 12:55 am

Re: Transient loss

Postby Cupwise » Sat Mar 16, 2013 11:37 pm

somehow this post got duplicated when i edited it...
Last edited by Cupwise on Sun Mar 17, 2013 12:22 am, edited 1 time in total.
Cupwise
Expert
Expert
 
Posts: 773
Joined: Tue Nov 23, 2010 2:03 am

Re: Transient loss

Postby Cupwise » Sat Mar 16, 2013 11:38 pm

Melon_Jack wrote:In theory, close to truth.. That's, imo, can explain low frequency loss. It's all about knowing original hardware and testing settings, until u find closest point. Or listening and choose what fit your style, your sound most.

well, i agree that some people might prefer to have things set up in any different particular way, regardless of how authentic to hardware that setup is. you might just like the way it sounds.

but, i can state that it is 100% fact and not opinion at all, that evf 17 mode is 'slow', that it doesn't catch transients, and that this means that a loud transient will not be using a sample that was taken from the hardware at a loud level, like it should. instead it will use a quieter one. this is fact. i've analyzed test tones with oscilloscopes and have seen it in action. it's common knowledge that this is how RMS detection works, that it's slower than peak, which means it will miss transients. and evf 17 is like a modified RMS detection. these things are not opinions they are basically like mathematical facts.

now on the hardware side, a preamp has absolutely no detector circtuit, and no attack/release controls. a loud peak going into a tube preamp will trigger saturation. you can sample that preamp with nebula, and your program will have those higher dynamic levels that saturate in it, but if you set that program up to use evf 17 or rms modes, a fast loud peak will not be caught by the detector and it will not trigger nebula to use those saturated samples quickly enough.

these are things that can be measured and looked at, and it's possible that somehow something i'm saying may be a bit off, but someone would have to explain specifically why or how it is. i'm being very specific here. a tube amp does not have a detector to decide when to trigger saturation. nebula does. a tube amp saturates instantly if the signal is loud enough to saturate. nebula with evf 17 definitely will not. so anyone can like or prefer whatever they want, but for accuracy to hardware, any kind of RMS mode won't get you there with amp type stuff.

the only times RMS or evf 17 are going to be accurate to hardware is if it is for hardware that used an rms or similar detector circuit, such as an RMS compressor.
Cupwise
Expert
Expert
 
Posts: 773
Joined: Tue Nov 23, 2010 2:03 am

Re: Transient loss

Postby gurth » Sun Mar 17, 2013 12:23 am

Really sounds logical to me, thanks for sharing these things, Tim.
User avatar
gurth
User Level X
User Level X
 
Posts: 104
Joined: Sun Jul 04, 2010 10:48 am
Location: Holland

Re: Transient loss

Postby djomardiaz » Sun Mar 17, 2013 12:34 am

Thanks for the info Tim
Paz hermanos.....
User avatar
djomardiaz
User Level IV
User Level IV
 
Posts: 44
Joined: Sat Dec 18, 2010 6:27 am
Location: Valencia (Spain)

Re: Transient loss

Postby RJHollins » Sun Mar 17, 2013 10:21 am

Cupwise wrote:i've recently come to a decision about that which goes against what may be general consensus and what i previously thought was the best way myself. this is just me expressing my thoughts and why i think them, and people can consider them however they want, or not. but here it is- instead of evf 17, i think that for 'preamp style' programs, the best thing to use for more realism may be peak mode. i've been releasing any of those types of programs i've put out recently with that mode selected (such as pass through programs made from the comps i've sampled). and i've released updates for a few of my older sets where one of the things i updated is switching from evf17 to peak.


Hi Tim,

For the sake of space, I only quoted a you're initial statement, but I too wanted to express my thanks for providing your working experience from working at the beginning of the sampling/evauluating/construction process. It is from the actual 'Developers' that 'their' observations can educated the whole community. [I do wish that other Devs would be even more proactive on this]. 8-)

Getting straight to my question would be this ... 'if' PEAK mode provides the most acurate translation of voltage level into the tiers of Nebulas' convolution layers, then WHEN or WHY would we use any other Mode? Is there an issue if PEAK Mode is used on a library that does NOT have dynamic sampling??

Another 'factor' that plays to this area are the 'TIMED' settings. We have several different 'playback' modes to choose from. Any additional insights/comments/observations would be most welcomed!

As you say, the concept of better, or best setting choice may be subject to many personal opinions ... which is fine ... the point I'd look at is a basic one. As 'Users', we only have the library to listen and compare ... the Devs have the prime position of direct comparison with the hardware, and that being focus on 'tuning' the library to said piece.

Thanks Tim for sharing your insights, and some wonderful library creations.

All Respect.
Sincerely.
i7-5820k, MSI X99A Plus, 16 GIG Ram, Noctua NH-D14, Win-7 Pro [64-bit], Reaper-64

NVC [Nebula Virtual Controllers]
RJHollins
Expert
Expert
 
Posts: 2631
Joined: Sun Mar 28, 2010 5:53 pm

Re: Transient loss

Postby Cupwise » Sun Mar 17, 2013 12:24 pm

RJHollins wrote: 'if' PEAK mode provides the most acurate translation of voltage level into the tiers of Nebulas' convolution layers, then WHEN or WHY would we use any other Mode? Is there an issue if PEAK Mode is used on a library that does NOT have dynamic sampling??
i already kind of answered these questions in my posts but they were really huge and rambling so i can't blame you for missing it. ;) i think that the other modes like RMS and EVF17 would only be authentic to hardware if the hardware had an RMS detector in it, such as a compressor that uses RMS. a library that doesn't have dynamic sampling, such as an EQ, will not be affected by which mode you use, at all. i did a paragraph explaining why this is towards the bottom of my 1st long post, but basically its because they don't have dynamic samples. the peak/evf17/rms modes are for detectors which detect incoming input levels and tell nebula which dynamic sample to use. if there aren't any dynamic samples, the detector does nothing.
RJHollins wrote:Another 'factor' that plays to this area are the 'TIMED' settings. We have several different 'playback' modes to choose from. Any additional insights/comments/observations would be most welcomed!
well between timed and freqd is something that i don't know quite as much about on the technical side, but picking which to use is a more simple and straightforward thing. they are just two different modes for the convolution, and have slightly different quality levels. giancarlo and people who have compared earlier in this thread say timed is a little better, but it takes more cpu.
RJHollins wrote:As you say, the concept of better, or best setting choice may be subject to many personal opinions ... which is fine ... the point I'd look at is a basic one. As 'Users', we only have the library to listen and compare ... the Devs have the prime position of direct comparison with the hardware, and that being focus on 'tuning' the library to said piece.
there are two different things here.

1st one is that a person may like RMS, EVF17, or PEAK mode best, with any given dynamic program, in any particular situation. they will sound different, and sometimes maybe one may sound better to you than another. this exists in the realm of personal preference/opinion. there is not 'wrong' answer here. if you like something, you like it.

2nd is how accurate is a program to the hardware. factually, this isn't opinion at all. what i'm saying is, accuracy to hardware is a measurable thing, it's more scientific, and it's not really an opinion. you can have an 'opinion', but it can be wrong. so at the end of the day, it still kind of falls back to being an opinion with the masses, because people have to choose between several different plugin versions of a 1176. many of the consumers of those plugin recreations will be people who used 1176 comps, and can say 'plugin X sounds more like the real deal to me than plugin Y', but there are mountains of extraneous variables that can and do influence their opinion. were the 1176 comps they used for years in the studio maintenanced at all? did they have their capacitors replaced? which revision did they have? these and other things can cause two actual hardware 1176s to sound different. how long has it been since the person heard their real 1176, before they auditioned the plugins? etc etc etc. so in reality, one 1176 plugin may be scientifically more accurate to the actual hardware 1176 piece that it was modeled from, compared to another plugin that isn't quite as close to the piece IT was modeled from, but even the most experienced studio expert may prefer the one that's least accurate to the 1176 it modeled. that plug version may sound closer to the 1176 that THEY used for years, or that person may just like the way that one sounds better. but the fact would still remain that the other plugin is actually more scientifically accurate to the hardware it modeled.

so i guess you could say that because of that, the accuracy to the hardware isn't THE most important thing. whether or not you like something is. but the theory with nebula and with other analog sim plugs has always been, that if you get your model closer to the hardware, more people will like it. that won't always be true but it seems like it is true more often.

so people can prefer evf17 mode which is like RMS, in a console or preamp emulation. but it's a matter of 100% fact that those modes are slower than peak because they operate on blocks of time, whereas peak mode tries to measure the level of input at any given moment and use that. it's a simple question-
does a tube amp have anything remotely like an RMS detector in it that decides when to saturate?

no. it saturates instantly. peak mode is faster and closer to instant. with peak mode nebula tries to measure the signal for exactly what level it's at at any given moment, and plays the dynamic sample to match that level. with rms nebula is taking blocks of time and taking one value from each block, and using that value to pick the sample to use. a tube amp simply does not behave like that.

i keep saying tube amp because the difference will be a lot less noticeable with a super clean solid state console, which is designed to not saturate, and which is very clean up until you pass the point where you are overdriving it. most or all of the dynamic samples in those nebula programs are going to be very clean and unsaturated/uncompressed. so the fact that the fast transients loud transients are 'incorrectly' using clean low level dynamic samples won't be as noticeable because the higher level samples are similarly clean also. try loading some tube programs with a fair amount of color to them, set attack to around 5ms, release around 2ms, loop some drums through it, and then switch back and forth between peak and evf17 modes. one thing that may sound contradictory to what i've been saying is that with peak mode you will get the nasty distortion that nebula gives you if you hit it with a signal that's too loud, BEFORE you get that with EVF17 mode. it's because in peak mode nebula is actually detecting those louder peaks and trying to play the loudest samples, but evf17 mode is missing those peaks so they don't distort. if that happens just back off on the input volume a bit.
Cupwise
Expert
Expert
 
Posts: 773
Joined: Tue Nov 23, 2010 2:03 am

Re: Transient loss

Postby alpsessays » Sun Mar 17, 2013 1:10 pm

Cupwise wrote:try loading some tube programs with a fair amount of color to them, set attack to around 5ms, release around 2ms, loop some drums through it, and then switch back and forth between peak and evf17 modes.
Ok, so you're saying then that to get the closest to the hardware - peak mode should be used on everything (but esp on stuff that is transient heavy).

As somebody who uses nebula - as most people do really - to get closest to the hardware sound as possible, this a given then for me. This is what is great about this forum!

Thanks!
alpsessays
User Level I
User Level I
 
Posts: 17
Joined: Wed Sep 05, 2012 7:53 pm

PreviousNext

Return to Working with Nebula

Who is online

Users browsing this forum: daniddd and 6 guests