KVR Audio

mustgroove · Post by **mustgroove** » Wed May 17, 2006 6:38 am

I'm in the process of mixing down some tracks at the moment, and I've just spent ages trying to figure out why the start of the tracks are all clipping... I thought initially I'd missed some automation errors, but everything was fine automation-wise...

So after much testing, it turns out it's Marquis, and I'm not sure if this is a bug, or part of it's normal behaviour, but it's slightly annoying either way...

Basically, when offline-bouncing, Marquis takes a while to "get started" at the beginning of a track - it needs to be fed a signal for a while before it settles into consistent compression, and the length of time it needs seems to depend on the Program-Dependent Release Contour setting...

Check out these screenshots from Wavelab... These pics are from the same loop, with different settings:

This one is where the release contour knobs are all set to 0.28... The output is hotter at the start, but settles in after a while... Remember that the loop being fed into Marquis is a proper loop - same thing over & over with no changes in level...

This one is where the release contour knobs all set to 1.80... It definitely starts out hotter than where it ends up, but it takes a shorter period of time to "settle in"...

This effect is very difficult to manage when mixing down... When mixing tracks with Marquis as last plugin in the Master chain (i.e. no limiter), if you set your levels to be as hot as possible without clipping, then this effect invariably means that the start of the track clips anyway... Using a limiter is not an option when you're providing final mixes which will then go onto proper mastering...

I should add these clips were bounced in Ableton Live 5.2, I have not tested other hosts as yet...

Just wanted to post this for people's information, in case people haven't noticed this effect yet... Is this behaviour expected/normal Alexey? Is there any setting that can be used to eliminate the "settling-in" period, without having to use high release-contour settings (which may not suit the programme material)?

/edit Both clips are 31 seconds long... they are both the same 32 bar loop bounced through Marquis, with the only difference between the two being the release contour knob settings

egbert · Post by **egbert** » Wed May 17, 2006 8:06 am

A clear indication of the timescale for those samples would help.

mustgroove · Post by **mustgroove** » Wed May 17, 2006 8:07 am

Both clips are 31 seconds long

ST54 · Post by **ST54** » Wed May 17, 2006 8:35 am

looks & sounds like pretty normal behaviour to me.
Maybe check out the voxengo forum , the release countours have been discussed there.
It's all about your "attack" times, not ?
Sometimes I do this on purpose, to get me more dynamics in a loop or track.

egbert · Post by **egbert** » Wed May 17, 2006 8:53 am

mustgroove wrote:Both clips are 31 seconds long

Which mode is chosen - T3?

And what are the attack and release times and the feedback/feedforward setting?

mustgroove · Post by **mustgroove** » Wed May 17, 2006 8:54 am

Sorry should have filled in the details...

T3
Forward mode
Sharp (but Force at 0%)
Round detector
Phasy

/edit corrected - was actually in Forward mode

mustgroove · Post by **mustgroove** » Wed May 17, 2006 8:58 am

Should also say the clips are stereo, I've just posted only the screenshot of the left channel to save space... But the way the peaks "taper" away is the same for the right channel as well

egbert · Post by **egbert** » Wed May 17, 2006 9:02 am

mustgroove wrote:Sorry should have filled in the details...

T3
Feedback mode
Sharp (but Force at 0%)
Round detector
Phasy

OK - but are your attack and release settings long?

mustgroove · Post by **mustgroove** » Wed May 17, 2006 9:04 am

Attack is 10ms, release is 1000ms or 3000ms, can't remember which...

But between the two clips, the only parameters that were changed were the program-dependant release knobs... In the first one they're all 0.28, in the second one they're all 1.80... Attack/release/T-mode/etc. were identical for both

fgimian · Post by **fgimian** » Wed May 17, 2006 9:16 am

I've come across this before too, especially on bass drums, the first one snaps a lot more than the following ones. This must be algorithm related, and as you said, it just takes a little while to kick in and reacts too drastically to the initial transient. I find that the Kjaerhus GCO is a lot more predictable which is why I use it in those particular situations.

Cheers
Fots

Rune L-H · Post by **Rune L-H** » Wed May 17, 2006 9:17 am

This is of course only for Aleksey to say for sure, but I would guess the Feedback mode (especially used with offline rendering) could result in something like that. Try doing your test with Feedforward mode.

Edit: I don't have Marquis myself so I'm really just guessing here.

Cheers,
Rune

mustgroove · Post by **mustgroove** » Wed May 17, 2006 9:19 am

The thing for me is, Marquis is utterly unbelieveable on the master bus as a "make shit sound hot" machine...

A possible solution is to play 16 or 32 bars of a dud sound until the compressor "settles in" and then just edit the crap out in Wavelab before burning/delivery...

But that's a bit more arsing-around than I'd like!

I'd also like to point this out to people and especially to Alexsey, just so we can get a soundbyte from him about whether this is normal, and if so, what purpose it serves...

fgimian · Post by **fgimian** » Wed May 17, 2006 9:26 am

Another idea is to render the loop as a new WAV file with the compression applied so that it always sounds the same. What I like / dislike about Marquis is that it's a little more human, it doesn't always react to the sound the same, so with Bass Drums and drum loops which are to be consistent, a render might be a good solution. I used it for the rendering of a bass drum recently.

Let's just say that Marquis gives it THE sound, well worth it.

Cheers
Fots

P.S.: Theres usually a Marquis on my master bus too

Melts the sound together, soooo good

mustgroove · Post by **mustgroove** » Wed May 17, 2006 9:33 am

soundpalace - the screenshots I've posted are of the beginning of an entire song with Marquis on the master bus, rather than a loop track with Marquis on it... That's the way I discovered the problem - when I was bouncing my "final" mix, the start was clipping because I'd set the output to be as hot as possible without clipping, and then when I bounced it the start was clipping because it was exhibiting this effect...

For the screenshots, however, I turned down the master output so it wouldn't be clipping, so the extent of the gain "gradient" would be more obvious...

So yeah, it's tough to counter this effect when Marquis is on the master bus in any way other than just looping the intro for a while so it gets settled, and then cut out the crap...

Unless Alexsey can fill us in on something we're/I'm missing!

For mine, using Kjaerhus compressors isn't the answer... Marquis T3 has a "glue" thing going on like I've never heard in a plugin before...

egbert · Post by **egbert** » Wed May 17, 2006 9:35 am

I would reduce the release time and see if you get a more stable level.

Marquis issue - takes a while to "get started"