Filter automation is too low rate

Using Cubase 11 Pro on Windows. I want to automate filter sweeps at a rate of 1/16ths or faster at 150 bpm.
i’ve tried feeding the Monark filter in Reaktor, as well as the EQ4 filter (low pass) of the Channel Strip, but it responds way too slowly to be useful.
(I draw automation on the timeline, not recording over MIDI, so the settings in the Automation Settings panel don’t seem to help)

Is there some kind of setting that lets me send automation faster to one of these filters? Or if not, is there another filter that’s part of either the Cubase Pro setup, or the Native Instruments suite, that will let me do this?

When I try to play this, I hear about three “beats” and the ripple of four 16th hits at the end is just a single filter-open passage.


(Also: “image may only be 700 pixels wide” – what is this, 1998?)

Anyone? Bueller? Bueller?

You need to set the buffer size as small as your computer can manage.

Audio Mixdown, Freezing and Render in Place can be accelerated by increasing latency. This might reduce the resolution of automation though. > VST 2 plug-ins calculate one automation value per parameter for each ASIO block > (which equals buffer size in samples), whereas VST 3 plug-ins decide independently if automation data is being rendered blockwise, sample-accurately or at a self-defined resolution.

Wait, what? Is automation tied to buffer size?
My buffer size is 256 and I really can’t make it smaller than that currently.

Pretty sure some important new feature of VST 10 years ago was “automation can now be sample accurate” so how come this would still be the case?
Also, each of those hits are much longer than a single buffer, so I should actually be hearing each of the hits if automation was applied at buffer rate. It sounds like it reacts even slower than that.

Yep Cubases automation is unfortunately not sample accurate. Its buffer chunk dependant. Internal filters like the channel strip or their internal vsts dont cut it for such precise stuff youve got in your screenshot. Also the steeper transitions with Frequency and StudioEQ are producing terrible glitches on transitions- the most usable (but not perfect) VST ive found is VST3 version of Meldas fre MBandpass, try it maybe youll get better results.

I do alot of such automation stuff with electronic music and moved it to BItwig because of such issues.

FL Studio does it too, and I do have a license for that. But I feel more at home in Cubase :frowning:

Not doing sample accurate automation of the built-in standard tools seems like a very significant oversight!

At a minimum, it could process in mini-chunks (16 or 32 frames) and update the automation along each of them. That’s not even hard, and doesn’t have particularly high cost for most algorithms (other than FFT based ones.)

It’s a trade-off between accuracy and cpu consumption.

You only need to burn the CPU when the automation actually is accurate, though. Compared the wild flora of virtual function calls each and every plugin has to go through for each and every buffer processed, a single if statement per block of samples to check whether it’s automated or not is not going to be a measurable difference.

So, the solution for me was to download the latest VST 3 SDK, and whack at it for a bit until I had a sample-accurate gain plugin.
Solves my immediate problem, good enough for me!
Also, it sounds like some hosts just send stairstep automation instead of properly interpolated line segments as recommended in the VST Spec, which generates definitely crunchy zipper noise. Oh, well.

If someone else finds it useful:
https://watte.net/Accurate_Gain.vst3.zip

Hey jwate thanks for this. It will come in handy. I personally dont have any programming skills so I couldnt do anything like this

Regards J

And because what I really wanted was a filter… (This one isn’t particularly clean, which is fine for my use. And, you know, might be interesting in itself :slight_smile:

https://watte.net/Accurate_VCF.vst3.zip

Thanks. Yore really fast with building this things.
Ill give it a go.

Edit: Ive tested it on a part that whas giving me trouble with internal filters and is accurate as it should be. Well done mate.


Regards J

I found the same, but why not perfect what’s about?

Hey. Look at my screenshot one post before- Melda works better than other filters but you still have to make little ramps before the bar start to avoid glitches- you never know depending on the buffer size either it will cut in to soon or too late. a great example is it eats the first transient of a piano for instance.
on the other hand the filter that jwatte posted i can draw a ramp directly on bar start and it will be accurate.

Glad someone likes it :slight_smile:
There’s a bug where the filter may re-use automation from the last buffer if the DAW doesn’t send new automation for the next buffer – I’ll post an update soon.

Version 1.0.1 available at https://watte.net/Accurate_VCF.vst3.zip

Version 1.0.1 2020-11-22

  • Fixed bug where automation would be repeated for next buffer if host stopped sending
    automation
  • Normalized “Bite” so at “0” it’s transparent and at “1” it doesn’t clip.
  • Fixed display of cutoff frequency (It’s not 2e+4, it’s 20000!)

Alright, I scratched this itch, I can automate weird sounds into trippy percussion, I’m likely to let this dog sleep from now :slight_smile:

In 2020, it would really be nice if Steinberg actually automated their EQ and such at (close to) sample rate. Image-Line does it. (Hint, hint, :smiley:)

Thanks for improving this jwatte. The frequency now shows real value.

Hopefully thew internal stuff will get sample accurate automation someday.

Regards J

Also fixed a glitch that made automating resonance and bite at the same time cause bad output.
Which let me crank up the resonance more without losing stability, so, yay!
No, really. I’m done. I’m going to keep my hands away from this now. There’s work tomorrow!