Voxengo Marquis Compressor VST 1.4 released

[lavoll]

why am i suddenly thinking of battlecruiser millenium?

[Aleksey Vaneev]

Torben, when you are using oversampled peak detection over non-oversampled signal, you are using maximum of all intermediate samples to catch the peak. While signal you are then going to treat is not maximum obviously - as it was not interpolated. Here we have a fundamental deviation. I can't call such approach 'precise' even if it looks like to be precise. Precise approach is to do full oversampling.

[defjamm]

beside voxengo algorithmix use up/downsampling, psp, refined audiometrics and universal audio.

but i think torben's new approach is quite intelligent and needs less cpu.

[defjamm]

The issue is that Aleksey continues to hint towards MPL-1 using bullshit technologies, being a pumper and not for modern music

limiters in general don't get used that much on a lot of modern stuff(not for the main part), it's mostly clipped.

i listen to a lot of 'commercial' hiphop, all the american gold/platinum-lps and they are clipped. there is no other way, if you want it loud but the drums big and upfront and not in the back. in this case people accept the added distortion. some hate that sound, i like it to a certain level, because it fits the music imho.

i think it's the same with rock-stuff, they clip that also to death. people talk more about limiters in online-forums than they are actually used on commercial stuff and wonder, why they can't achieve that sound(maybe that's not bad at all).

on music like the one in bmanics thread, i wouldn't use limiting or clipping at all.

[Aleksey Vaneev]

defjamm, as far as I know iZotope Ozone adopted the same approach long ago.

But it is questionable, unfortunately... To be sure it works you'll have to use oversampled metering tools like Hammerfall DSP tool. It was reported that Voxengo Elephant at 4x oversampling does prevent intersample overshoots.

If MPL performs similarly, then OK, this approach works. Somebody should check this out.

[Aleksey Vaneev]

Here is a proposal for such test using r8brain PRO (demo version will be enough), this is applicable to all brickwall limiters:

1. Take an already loud sound clip.
2. Process it first with MPL (+6 dB), and then with Elephant (+6 dB of gain as well). Elephant should be in 4x oversampling mode. Save both results in different files.
3. Now launch r8brain PRO and select both files for conversion. Specify target sample rate to 4x of the original sample rate (176400 in case of 44.1kHz original sample rate). Then check the 'prevent clipping' checkbox. LinPhase mode should be used.
4. Process the files and open both in a sound editor. You will see that Elephant is dead brickwall while MPL has some intersample peaks appeared (r8brain PRO's clipping prevention dropped overall gain by 0.4 dB in my test).

From this test I conclude that MPL's peak detection - though sophisticated, does not work as expected - it does not suppress intersample overshoots.

BTW, iZotope Ozone passes this test as well. So, maybe the latest Ozone ( 3.08 ) already uses full oversampling - or maybe I just do not know what they know.

[3*s]

Let me just say I own products from both of you and I respect the fact that you can do this out in the open, despite the obvious risks for both of you.

Carry on.

[fgimian]

I also own Voxengo & Kjaerhus products, both compressors, both EQs. I have also tested both mastering limiters. You guys are both brilliant developers !! The GCO and the Marquis are both totally awesome and kick ass in different ways, as do the EQs.

I think both mastering limiters sound awesome, again each have their place, their advantages / disadvantages.

So let's all be happy and get back to making music .. and you guys making great products.



Cheers
Fots

[Aleksey Vaneev]

Thanks for your support, everybody!

[Torben]

Aleksey, we did test it in Matlab and we find that the over sampled peak detection did what it should; it decreased the theoretical overs from 3dB to around 0.3dB. Practical test showed that any D/A overflow were less than 0.1dB. One easy way of proofing why it works is by considering what happens with each sample in the audio signal and in the side chain. sc1 compresses a1, sc2 compresses a2 etc. If you do not produce any frequency components above Nyquist you can down sample just by picking every fourth output sample. Now you will notice that compressing the intermittent samples is waste as you are going to skip them anyway. Furthermore if your up sampler are exact (and FIR) it has not changed the samples that where, only added new ones (and we are going to skip them anyway). So you really end up with not having to up sample the whole audio signal. MPL-1 is not touching the signal unless its peak level is above threshold. We talked to several mastering engineers at the Musikmesse this year and that was how they preferred it.

Torben

[pdxindy]

Besides the bit of negative tone, the informational discussion is great. Way over my head, but interesting nonetheless.

For me there is no doubt that MPL-1 is far superior... for the simple reason that MPL-1 will be available for the Mac!

[VanLichten]

Really no need to fight here ! I guess most people here will agree that you are the best both low-cost professional plugins developers; each of your products is well thought out and useful.Furthermore, similar products are very different in detail and your basic sound is also different IMO, Voxengo has a quality retro touch and Kjaerhus managed to imbed the colour of their plugins into the sound somehow .

BTW, Thorben, any chance of a colouring unit/warmer thing from you?

[4damind]

Thanks for the update. I prefer Voxengo over Kjaerhus, Aleksey is a DSP-God and all of his plugins are outstanding good (and he has released tons of it
I own only the GUP from Kjaerhus because for its easy handling, fantastic analog sounding and easy to use sidechaining. But Marquis is the King, the best sounding compressor I ever bought (I miss only sidechaining this time)

For my ears MPL listens a bit more transparently as Elephant. But it can be with another algo (I use EL3) or other settings for the 2-channels it comes to the same results. Would be interesting to see the results for bmanics little limiter test in another thread.

[Dandruff]

BTW, iZotope Ozone passes this test as well. So, maybe the latest Ozone ( 3.08 ) already uses full oversampling - or maybe I just do not know what they know.

there's a extra switch in ozone for that: "prevent intersample-peaks" or something (works fine in intelligent- und brickwall-mode)

[ObsoleteAcc99]

[DELETED]