DSP Robotics Support

I'm sorry Martin,
I just had to move the knobs for the three bass strings,
and close and open the schematic.
It's not noticeably any heavier for this component now.
Thanks a whole bundle- very appreciative.
It doesn't flutter on/off even without smoothing.
Cheers!

--edit: actually my device fluxes around a lot in CPU, nothing to do with Martin's unreal device

Hi,

I still hadn't digested all of this, but let me thank you Martin for the great answers and help.
Its like someone switched on the light in a dark, cold and lonely room...

Cheers !

Thank you for those words, Rocko. I am glad that you find my considerations useful. I was a bit worried that I kinda hijacked your thread - didn't mean to.

On the contrary - I've uploaded this example to the forum for knowledgeable people to high jack it and take it to the next level.

So, now I'm digging into it and learning the advantages of FIR for EF.
Still, when I use some 'high end pro audio' compressors in my DAW, some of them have zero latency (no compensation).
Do these probably use a simple IIR first-or-second order filters for EF?

It's up to you, Rocko, to tell the host what compensation you think it should have,
just in case you didn't know that.

Another thing you can do instead of filtering is use peak?
This takes 1 sample delay. Just registers that it is falling now.

Hi,

Thanks.
My question is more a theoretical one, less about the DAW setting.

I mean, now I better understand why to use an FIR filter for E.F. which causes a latency of (say) 10ms.
So, what do the 'pro audio' grade compressors (like Abelton Live Compressor, or FabFilter ProC, as example) run at their zero-latency setting.
My guess is that they run a simple IIR filter of first or second order with all its shortcoming.

What do you think?

Hey all,

Can I ask how to implement the Envelope Follower with 4th order Butterworth?

Would it be this:

x[n] --> ABS ( x[n] ) --> 4th order Butterwoth Low Pass --> attack and release mechanism

In attack and release mechanism I mean, something of the sort:
y[n] = a * x[n] + (1-a) * y[n-1]

Or is the idea to implement the 'attack and release' into the Butterworth LPF itself ?

Thanks,
Rocko

Hi,

I'm summing up some of the stuff discussed and explained here, mainly by Martin (a thousand thanks)...

The topic is focused on 'pro audio level' envelope follower (myself very interested in low frequency following ability).
Based on Martin's work, I've gathered three envelope followers (FIR moving average, 4th order Butterworth IIR, 1st order IIR).

The attached FSM file is for playing with it 'live' and viewing the ripples and overshoots/undershhots at real time.

Notice that the 1st order IIR uses the attack/release as the LPF mechanism, while the 4th order version, uses an LPF on signal, followed by a attack/release mechanism.

Any comments are welcomed !

One thing I noticed is that your 1pole IIR envelope is partly negative. That may cause problems further down when you take logarithms or powers for gain calculation.

Hi,

For both IIR's I need to add 'ABS' module after the filter itself and before the log part.
I didn't add this to this specific schematic.

Trying to understand your work on the moving average filter... That's my next step