View Single Post
Old 12-07-2018, 06:27 AM   #110
nitsuj
Human being with feelings
 
nitsuj's Avatar
 
Join Date: Nov 2017
Posts: 292
Default

Quote:
Originally Posted by JamesPeters View Post
I notice ReEQ adds some latency. I'm guessing there is a filter in it that requires this? I try to avoid plugins that add any latency, except where necessary. (I'm picky that way. )

I see what mawi means. The "fall" speed of ReSpectrum is a bit slow for monitoring things like sine sweeps (and how a plugin affects its shape, a useful test for plugins). The shape of the wave on the trailing edge "drags". Besides that though I think it's perfectly usable the way it is.
It only adds latency in oversample x2 mode. To be precise, it's 16 samples of latency. Oversampling is used to allow more accurate filter curves near Nyquist. Without that, you'll find that the curves compress as they get towards the top end of the spectrum. Try it in oversample x1 mode and you'll see what I mean. It's the way Reaper's ReaEQ works and also the way that Ableton's EQ8 works.

Oversampling gives more frequency space so that the curves aren't affected. However, the final downsampling can introduce severe aliasing artifacts so it's necessary to apply a steep low pass filter near the end of the audible spectrum to remove that. ReEQ uses a FIR filter to do so which is a common tool used in oversampling situations. But these do introduce latency.

I want to stress that when either in x1 or x2 mode the SVF filters are extremely good - much better than the RBJ cookbook ones that a lot of EQs used to use. Clean, stable, low noise and transparent. So, at least now you can make a call. If you're ok with a little latency, go for x2. If you require no latency then stick with x1. It won't be a problem if your working with low-mid eq'ing and if you're doing any eq'ing in the very top end you'll probably be notching or shelving anyhow. Good to have options.

Clearly there are ways of having more accurate filter curves near Nyquist without oversampling, for example Fabfilter Q2/3, DMGAudio EQs and others. I do read that adjusting the curve shapes near Nyquist by other means can introduce phase issues near that part of the spectrum although I didn't get around to trying. Trade offs I suppose.

EDIT: I meant to say, if you're working above 48kHz sample rate then the oversampling doesn't happen regardless of whether x1 or x2 is selected. There shouldn't be any latency either.

And sure, the spectrum analyser lossy decay integration is geared towards tracking frequency peaks which I found most useful for working with music. It wouldn't be hard to add an optional fast integrator. I implemented the deep spectrum floor range because I too like to see what plugins are up to.

Last edited by nitsuj; 12-07-2018 at 06:32 AM.
nitsuj is offline   Reply With Quote