OverTone FC-70 Review - Fairchild 670, Really?

[bmanic]

and thinking about it some more, i have to question exactly how much room there is for sensing differences in shape when you are dealing with attacks going maybe below a millisecond. i'm not saying you couldn't tell a blocky attack from a smooth one, at those speeds, but if one is fairly well curved and another one has a perfect curve/shape, i think it would be very hard if not impossible to tell them apart.

You can test this yourself very easily. Just create a few very short impulses.. only a few samples long. Now move around the sample positions and listen to the changes. You might be surprised.

Cheers!
bManic

[Cupwise]

to me it looks like the glue one without the oversampling has its attack cycle completed sooner than the one with, or in other words that the attack is faster. if you look at the lone snare towards the front, you can see that some of the body after the peak is louder in the oversampled one than in the one without.

edit- actually in some of the other snares it's the opposite. it's not a 100% consistent change you can point at between the two to single out what the cause of the oversampled one sounding better may be. a lot changes between them, so i don't know that you could just chalk the change of sound up to being a smoother attack shape necessarily, is all i'm suggesting.

[meloco_go]

and thinking about it some more, i have to question exactly how much room there is for sensing differences in shape when you are dealing with attacks going maybe below a millisecond. i'm not saying you couldn't tell a blocky attack from a smooth one, at those speeds, but if one is fairly well curved and another one has a perfect curve/shape, i think it would be very hard if not impossible to tell them apart.

Actually, the faster the attack the easier to spot differences between samplerate of the sidechain in compressor. The reason being is that on shorter attacks non-linear effects play much larger role and compressors start to generate more harmonic and intermodulation distortion products. It's not the timing that ear spots, it's the tone they generate.
If you consider sidechain that runs at normal sampling rate and that it must act on only a handful samples (especially at attack times below 1 ms) and also that some instruments might have very fast transients with frequency components going right up to fs/2 it is pretty clear that sidechain works in very hard conditions and may generate a lot of aliasing. This aliasing is basically a voltage error, that is being added to the transient. It is interesting test to subtract non-oversampled render of the Glue with the render at very high oversampling, the difference contains clicks at the positions of transients. Ans this is not only the character of the Glue, many compression plugins exhibit the same behavior when you compare their action at 44.1 vs 96.

[meloco_go]

edit- actually in some of the other snares it's the opposite. it's not a 100% consistent change you can point at between the two to single out what the cause of the oversampled one sounding better may be. a lot changes between them, so i don't know that you could just chalk the change of sound up to being a smoother attack shape necessarily, is all i'm suggesting.

Just subtract one from the other and you'd see and hear

[Cupwise]

edit- actually in some of the other snares it's the opposite. it's not a 100% consistent change you can point at between the two to single out what the cause of the oversampled one sounding better may be. a lot changes between them, so i don't know that you could just chalk the change of sound up to being a smoother attack shape necessarily, is all i'm suggesting.

Just subtract one from the other and you'd see and hear

ok, i just did that with the wavs bmanic linked, and i don't see that at all. a lot more is left over besides just some clicks in attacks, and you can tell just by looking at them that the clicks aren't super drastically different between the two. some attacks have bigger louder clicks in one version, some in the other. and like i said there is clearly more differences between the two versions than just the attack shape, which you can see by looking at them.

[meloco_go]

ok, i just did that with the wavs bmanic linked, and i don't see that at all. a lot more is left over besides just some clicks in attacks, and you can tell just by looking at them that the clicks aren't super drastically different between the two. some attacks have bigger louder clicks in one version, some in the other. and like i said there is clearly more differences between the two versions than just the attack shape, which you can see by looking at them.

You should subtract oversampled vs non-overampled versions of the same plugin to hear the effect on attacks.
Different compressor plugins surely may have differences beyound that, i.e. in compression curve shape, timing constant, non-linearities modelled and maybe frequency response to some extent.
However, the perception of compressor action on attacks is clearly the function of how envelope follower works there.

[Cupwise]

the clips were both of the glue, with and without the oversampling. check them out yourself. if you null them a lot more is left than just the attacks.

[meloco_go]

the clips were both of the glue, with and without the oversampling. check them out yourself. if you null them a lot more is left than just the attacks.

I did this before, and the most of the difference was on attacks (also a bit of residual top end that oversampling killed, but that can be compensated).
The differences might arise if other plugins were used which exhibit random behavior, and if the files were lossy coded.
Later today I can create another set of files, or if you have the Glue, you can check for yourself.

[Kingston]

And the plugin is based on this? Interesting!

I didn't say that. I meant the image on the review website is wrong.

[Liero]

Here are the IK vs Native Instruments SSL bus compressor examples. All settings identical on both. Because of how different they are I had to set the makeup gain by ear so please consider this while listening. The issue is with the way the attack sounds.

Forgot to do the 670 render which I promised and I didn't save the project so drumloop lost. Oh well, it was a shitty 30 second job anyway..

IK SSL, fastest attack, fastest release, 10:1 ratio

NI SSL, same settings

IK SSL, attack opened up to 1ms

Native-Instruments SSL, attack opened up to 1ms

Note how different they sound! Personally I think the NI/softube version sounds much closer to how I remember it.. the attack that is! I actually prefer the release action of the IK SSL but the attack bothers me a lot. It's not what I'd call "phat" and controlled. It's clicky.

EDIT: the more I play with these the more I think "the truth" might lie somewhere in between! We had a go with Kingston changing some of the internal components in his SSL clone and one of the clear changes was in how the transients were either "precise" or "chunky" in addition to one sounding much more hi-fi and loosing less top end. I don't remember exactly what we were changing.. the VCA's? But there was a definite change in character between the "expensive" components and the original ones. I think we both ended up preferring, or rather "understanding", the cheaper ones due to the chunkyness. It's very possible I remember it all wrong though. It's a couple years ago.

Cheers!
bManic

Links are not working.

[ObsoleteAcc99]

[DELETED]

[meloco_go]

the clips were both of the glue, with and without the oversampling. check them out yourself. if you null them a lot more is left than just the attacks.

So here's two files, the Glue is set at 0.1ms attack, 0.4s release, 10:1 and threshold pretty low, so there's something like 10-15 dB GR happening.
One file is the GLue set at x64 oversampling on render, the other is no oversampling, followed by another instance of the Glue that does not do any GR, but is also set at x64 render to compensate for filters.

https://dl.dropboxusercontent.com/u/279 ... ue_64x.wav
https://dl.dropboxusercontent.com/u/279 ... _no_OS.wav

What do you think?

[Cupwise]

What do you think?

maybe you adding the extra OS version after the non OS one to compensate for filters explains why these two clips cancel each other out better than the ones bmanic posted.

but
you can't really know for sure that the extra loud peaks left over (and some of the body of the drums is left too) is all from the non-OS version. because you are nulling the two together. it could come from either one, or it could come from a combination of both. and some of it could even be phase distortion/shifting difference between the two which would leave stuff behind but isn't 'as bad' as a newly introduced element like harmonic/inharmonic distortion etc.

also, if you just load the two wavs up, and switch from one to the other, the oversampled one actually looks to have the louder peaks than the non-oversampled one. that makes me think that the bulk of that peak material left over from the null actually belongs to the oversampled one, not the other. the fact that lots of the 'body' of the drums is left over also means that, again, there is more difference between the two than just the attacks themselves and that could contribute to one sounding better than the other.

i DO think the oversampled one sounds better. but i'm not seeing any evidence from this null that it's because the oversampled one has a smoother attack. that may be the case but i don't see it proven here.

[Aiynzahev]

Still loving this comp

[meloco_go]

maybe you adding the extra OS version after the non OS one to compensate for filters explains why these two clips cancel each other out better than the ones bmanic posted.

Actually the difference because of that is very very little.

but
you can't really know for sure that the extra loud peaks left over (and some of the body of the drums is left too) is all from the non-OS version. because you are nulling the two together. it could come from either one, or it could come from a combination of both.

Yeah, but actually it doesn't matter. The idea is that a) there's audible difference between how two sounds b) the actual difference is mostly on attacks/high frequency sources.
I made another test -- I set a sine wave generator feeding two instances of the Glue (one at no OS, other at x16). There is a bit of difference on the steady sine, so maybe somewhere the gain is not 100% correct, but that difference is VERY small (like -75 dBFS, when the signal is about -3dBFS). And the most importantly it does not depend on frequency up to about 20kHz, before the AA filters kick in.
However, when I set up a saw wave modulation on the sine, there appeared a signal bursts on the transition regions.

the fact that lots of the 'body' of the drums is left over also means that, again, there is more difference between the two than just the attacks themselves and that could contribute to one sounding better than the other.

Well, I wouldn't call that "lots" of body, I'd say "some"=)
The differences on the attacks is about 10-20dB higher than on the body.

i DO think the oversampled one sounds better. but i'm not seeing any evidence from this null that it's because the oversampled one has a smoother attack. that may be the case but i don't see it proven here.

You don't like the term smooth?=)
I'd say the OS version sounds "deeper", less "flat".