Voxengo Elephant 3.0 as brickwall

[Aleksey Vaneev]

Do not forget please that you can change the knob response in the global settings window. If they are feeling too swift, you may decrease the sensitivity.

[macc]

Indeed, I need to fiddle about with that.

One thing I don't like about the new version is the way the CPU hit varies so dynamically according to whether it is limiting or not. In quieter sections the CPU backs right off to 'idle' levels, but then shoots up when it gets back to business. Perhaps it is just my machine (dual-core Athlon, a little old ish I suppose), but no other plug does this even with oversampling.

Any particular reason for this behaviour?

[Tarekith]

I like it a lot, I was waiting for the new Sonalksis limiter to drop so I could compare all the native AU limiters, and Elephant came out on top for me. I had forgotten about it as in the past Voxengo plugs were all PC VST, but I'm very glad now this was ported to OSX. Wish there was a way to set the final output level to something to like -0.2dBFS though.

[kjg]

Originally Posted by Tarekith

Wish there was a way to set the final output level to something to like -0.2dBFS though.

Set "out gain" to -0.2dB?

[electro]

Originally Posted by macc

Indeed, I need to fiddle about with that.

One thing I don't like about the new version is the way the CPU hit varies so dynamically according to whether it is limiting or not. In quieter sections the CPU backs right off to 'idle' levels, but then shoots up when it gets back to business. Perhaps it is just my machine (dual-core Athlon, a little old ish I suppose), but no other plug does this even with oversampling.

Any particular reason for this behaviour?

Could it be disengaging oversampling during passages where audio is below the threshold?

[Tarekith]

Originally Posted by kjg

Set "out gain" to -0.2dB?

I was under the impression this knob had a different purpose than being an output ceiling, I still need to go over the manual again though.

Also, would agree that I wish the metering was a little more responsive, seems kinda sluggish. Not a huge deal though.

[macc]

Originally Posted by electro

Could it be disengaging oversampling during passages where audio is below the threshold?

That's what I thought, but I was interested to see what Aleksey had to say...

[Aleksey Vaneev]

Setting Out gain to -0.2dB is the same thing as setting ceiling to -0.2dBFS since limiting occurs at 0dBFS. Of course, the ceiling will still be staying at 0 dBFS, but you are affecting the output signal level and it turns out to be the same as ceiling at -0.2dBFS.

CPU load may vary due to knee processing: it takes less CPU time when level stays below some threshold. It should not be a large change, though - maybe a dozen percent relative to overall plug-in's CPU load.

Sluggishness of UI is not really a 100% plug-in related thing: audio hosts may reduce UI update frequency when audio part is taking a lot of CPU time. I've noticed this especially on Mac OS X.

[Tarekith]

Thanks for the clarification Aleksey

[macc]

Originally Posted by Aleksey Vaneev

CPU load may vary due to knee processing: it takes less CPU time when level stays below some threshold. It should not be a large change, though - maybe a dozen percent relative to overall plug-in's CPU load.

:(

Seems to be a lot more than that here. Will do some further testing though thumbsup

[kjg]

Originally Posted by Tarekith

I was under the impression this knob had a different purpose than being an output ceiling, I still need to go over the manual again though.

True. Technically it is not an output ceiling, but an output gain.
What I understood from when I checked out Elephant2 (haven't looked at 3 yet), is that it is a fixed threshold design where you set the amount of limiting by adjusting the input gain, and adjust the output level (and effectively the ceiling, like Aleksey confirmed) by setting the output gain.

Regards,
Klaas-Jan Govaart

[E-Irizarry]

My belated respect to you, A.V. Thank you for your plugins.

Voxengo Soniformer 2.x, Elephant 3.0/3.01, and Voxformer 1.9 are the truth y'all!!!

[macc]

Originally Posted by macc

:(

Seems to be a lot more than that here. Will do some further testing though thumbsup

Seems to be related to channel linking in some way... Need to test more in depth - but got work to do!!

[Torg]

I have just demoed the plug.

but when in "clip mode" if i select any oversampling setting different than 1X the signal is not brickwalled at all.

There are massive overloads above the output ceiling

the other modes work ok with these settings, no overloads.

i read the manual already but couldnt figure out...

since the thread title is "Voxengo Elephant 3.0 as brickwall" it seems the right place to post this question since i am getting no brickwall at all when using these settings...

[Bob Yordan]

Hiya

As I remember trying out a demo of Elephant v1, it was not so easy to set (tweak with the knobs) a defined out level like 0dB or -0.05dB or -1dB etc. That the limiter did not exceed however hard it was pushed on the input.

Has that changed now??

On the UAD PL GUI, I just type in the output level.

[Tarekith]

Originally Posted by Bob Yordan

Hiya

As I remember trying out a demo of Elephant v1, it was not so easy to set (tweak with the knobs) a defined out level like 0dB or -0.05dB or -1dB etc. That the limiter did not exceed however hard it was pushed on the input.

Has that changed now??

On the UAD PL GUI, I just type in the output level.

You can define the knob sensitivity now (in multiple ways in fact), so it's very to have precise control with the mouse, the mouse wheel, via the shift command, etc. You can also enter the exact values via typing them, and the GUI even has a little pop up keyboard that appears to help with this if your host prevents certain keys from being used due to key command assignments.

[bmanic]

Originally Posted by Torg

I have just demoed the plug.

but when in "clip mode" if i select any oversampling setting different than 1X the signal is not brickwalled at all.

There are massive overloads above the output ceiling

the other modes work ok with these settings, no overloads.

i read the manual already but couldnt figure out...

since the thread title is "Voxengo Elephant 3.0 as brickwall" it seems the right place to post this question since i am getting no brickwall at all when using these settings...

This true for ANY clipping plugin. As soon as you oversample, you can't set an exact output ceiling. You can always add another instance of elephant after the clipper to keep it a a dedicated level.

What I'd like to see in future Elephant updates is the ability to have clipping AND limiting at the same time and to chose the order, but for now one can use two instances.

Cheers!
bManic

[Torg]

Originally Posted by bmanic

This true for ANY clipping plugin. As soon as you oversample, you can't set an exact output ceiling. You can always add another instance of elephant after the clipper to keep it a a dedicated level.

Not really. Gclip for instance.

You can oversample(2X) and you dont get overloads.

"This true for ANY clipping plugin." --> Can you further explain why ?

thanks

[Bob Yordan]

Originally Posted by Tarekith

You can define the knob sensitivity now (in multiple ways in fact), so it's very to have precise control with the mouse, the mouse wheel, via the shift command, etc. You can also enter the exact values via typing them, and the GUI even has a little pop up keyboard that appears to help with this if your host prevents certain keys from being used due to key command assignments.

Thanx, I will check out the new Elephant demo.

[Aleksey Vaneev]

Originally Posted by macc

Seems to be related to channel linking in some way... Need to test more in depth - but got work to do!!

This is to be expected. When you set Ch.Linking to below 100%, limiter has to process limiter algorithm for all channels instead of just one grouped channel.

[Aleksey Vaneev]

Originally Posted by Torg

Not really. Gclip for instance.

You can oversample(2X) and you dont get overloads.

"This true for ANY clipping plugin." --> Can you further explain why ?

thanks

I'm sure Gclip hardclips output after performing oversampled clipping. Overshoots after clipping in oversampled mode are simply unavoidable otherwise. It's mathematics, not some bug.

bmanic, I hear your desire to have an additional clipping stage. But as you've noted yourself, inserting two Elephant instances is a better choice especially considering you still need a full control over both algorithms.

[macc]

Originally Posted by Aleksey Vaneev

This is to be expected. When you set Ch.Linking to below 100%, limiter has to process limiter algorithm for all channels instead of just one grouped channel.

Thought so thumbsup

Any chance of explaining a little more about what the 'dyn' control actually does? I like mode 5 mostly and mode 2 on some things. I have no idea why though 5 seems nicely smooth, 2 nicely aggressive. I really really like Elephant 3!

[Torg]

Originally Posted by Aleksey Vaneev

Overshoots after clipping in oversampled mode are simply unavoidable otherwise. It's mathematics, not some bug.

Show the math please!

[kjg]

Originally Posted by Torg

Show the math please!

Aleksey is right.
Go do your own research or at least ask politely. tutt

[Torg]

Sorry.

didnt mean that...

will do

[Nordenstam]

PST: you'll find the answer in the intersample peaks.

[kjg]

Originally Posted by Torg

Sorry.

didnt mean that...

will do

demanding an explanation from someone (even using an exclamation mark) is just a bit too eager

back on topic:

clipping a waveform and then resampling (downsampling back to the original samplerate) will result in overshoots.
when the waveform is re-constructed the value between two clipped samples will be estimated and this will result in value higher then the clipped value.
it is pretty much the same as with inter-sample peaks on DA conversion.

hope that helps.

EDIT: I see Lupo gave you a hint while I was typing. He is right. Just google and read a bit on this subject and you will see why Aleksey is right.
EDIT2: I've attached an interesting pdf.

[macc]

Originally Posted by macc

Thought so thumbsup

Any chance of explaining a little more about what the 'dyn' control actually does? I like mode 5 mostly and mode 2 on some things. I have no idea why though 5 seems nicely smooth, 2 nicely aggressive. I really really like Elephant 3!

Bump - This Dyn control, again... Please explain what it does!

It literally just made the difference between passable master and a really good, smooth, deep master. I can tell it seems 'kind of' like a rlease control, lower numbers meaning it emphasises the snare in a way, like a short release time might... but there's some thing more than that.


I still really like E3 thumbsup

[Ermz]

From the manual:

The “EL Dyn” control selects additional dynamic response type used when the “EL-1”
or “EL UNI” algorithm is engaged. Depending on the selected type Elephant will
reveal or hide subtle features of the sound when limiting happens: this control mainly
affects the aspect of temporal masking. The “Off” option can be used to disable
additional dynamic response. Note that sonic difference between all the offered
response types is subtle – you may choose the type in accordance with your subjective
feelings alone.

Sure, I suppose it says a lot but doesn't explain much for the layman. I'm quite happy just considering it the 'x-factor' as the math is undoubtably way beyond me.

I've been trialing version 3 on a few different things lately and as always what I love about it is the wide array of options that you can use to tweak the limiter to work with the source material.

So far it's doing better for me than version 2, although I have to say that I really don't like the GUI - especially the GR meter.

I'll continue to keep using this one and maybe shoot it out against the Sonnox in the near future.

[macc]

Yeah man, I have read the manual, but thanks

I was after more nitty gritty details though, regarding the 'this control mainly
affects the aspect of temporal masking' bit.