Cytomic 'The Glue' Compressor

[Mercado_Negro]

As you can see, Im very surprised how the native ssl duende compress the 808 low end. that's what I wanted to get.
But with the same settings, I am a little disappointed about The glue. The transcient is above everything.

The Glue is my go-to compressor for everything, not just for buses but individual tracks as well. I've been using it for over 2 years now and one thing I've always missed is the controlled and tight low end I'm used to with SSL processors. The Glue thins things out too much sometimes. I've tried to control this by adjusting the sidechain filter but it doesn't get me there either. The way it treats transients is spectacular and the range knob is genius but that 'lack' of low end control/tightness really saddens me sometimes.

Andrew, if you're reading this, maybe you could take a look at this issue since it's, in my opinion, the only 'issue' that makes The Glue '95% there'.

Cheers

[ObsoleteAcc99]

[DELETED]

[Mercado_Negro]

nothings perfect. Not even the $1000+ lexicon pcm bundle which is my fave soft reverb of all time. I'd still wish and change things about it. I doubt Andy will be making any major sound changes to the glue at this point as it will affect the hundreds of users who have current settings in current projects.

I think price doesn't have anything to do with 'perfection'.

There's always room for improvement while considering current behavior/sound. New options via menu could be added, just like it happened with all oversampling modes.

[seanturco]

If I understand, ssl has added a soft clipping /limiter , before or after the compression . therefore it over compress transcient ?

Digital means softclipping for you ?
At this Time, The glue sound nice and the GUI is understandable !
if you add a "digital option", The Gui wouldn't be understandable as now . ( I don't know if you understand what I mean )
( it Would be a nighmare for a newbee to understand what is digital or soft clip )

However it could be very interesting to add a digital option, In case I would like to "push" The low end without losing the ssl character. but also switch off "digital option" when I want a "clean" and "slappy" sound.

Edit : Im JeromeD , I loose my old account this morning

[kvaca]

At first it looks like you dont need ovesampling at all
I have compared to waves ssl comp and it exhibits almost same aliasing as Duende too.
But another situation is when you increase attack to 10 or 30ms with otherwise same setting,than it exhibits much higher amount of aliasing-could you please show us the spectrum plots in that case to be objective and tell us why,becouse all other compressors I know behave rather opposite way.

Hi kvaca - it sounds like there is something wrong with your settings. Do you have the Peak Clip enabled or are you clipping the output of you DAW somehow? This is the only way I can think of that slowing down the attack could produce more aliasing, since the level of compression will decrease, so the output level will increase and possible clipping could occur.

Here are my plots of 0.1 mS and 10 mS attack, as you can see the only different is less harmonics on the 10 mS which is as expected:

http://cytomic.com/files/forums/Sin993_ ... 6_4_-8.jpg
http://cytomic.com/files/forums/Sin993_ ... 6_4_-8.jpg

hmmm ,I have used Wavelab for test tone generation and also Wavelabs internal FFT meter instead of Span which I dont have installed here /and I tend to think that should be at least as good as the Span/,Ive used Blacman-Harris 9th windowing and it shows aliasing at -150 db for 0.01ms attack and -125 db for 10 ms.Off course you are still better than your competitors in that particular behaviour.
edit-off course no peakclip used and 32 bits so no clip anyway

[penguinfromdeep]

Do you think you can hear aliasing that happens at -125db?

[kvaca]

Do you think you can hear aliasing that happens at -125db?

no, but that was not my question-I have asked why aliasing of The Glue increases with attack length if in other compressors decreases...
I have no problems with aliasing in Glue at all-just be curious...

[penguinfromdeep]

If I remember right Andy told that when you use the shortest attack, the signal path will be oversampled (even in the first versions where there was no 'oversampling' as an option included)

[kvaca]

If I remember right Andy told that when you use the shortest attack, the signal path will be oversampled (even in the first versions where there was no 'oversampling' as an option included)

yes this can explain that behaviour...

edit-but he told that no oversampling was used?

[michu]

If I remember right Andy told that when you use the shortest attack, the signal path will be oversampled (even in the first versions where there was no 'oversampling' as an option included)

actually i remember it as sidechain being oversampled on attack stage, while signal path is running at host rate all the time.

[Artgiver]

Hey thx for the helpful posts. So then, the Glue is what i'll be 'stuck' with (NPI) if Elysois Alpha and Slate's bus comp don't change my mind. Spending one or two hundred more is worth it to me if the other two are better than the Glue FOR VOCALS mind you.

[andy-cytomic]

As you can see, Im very surprised how the native ssl duende compress the 808 low end. that's what I wanted to get.
But with the same settings, I am a little disappointed about The glue. The transcient is above everything.

The Glue is my go-to compressor for everything, not just for buses but individual tracks as well. I've been using it for over 2 years now and one thing I've always missed is the controlled and tight low end I'm used to with SSL processors. The Glue thins things out too much sometimes. I've tried to control this by adjusting the sidechain filter but it doesn't get me there either. The way it treats transients is spectacular and the range knob is genius but that 'lack' of low end control/tightness really saddens me sometimes.

Andrew, if you're reading this, maybe you could take a look at this issue since it's, in my opinion, the only 'issue' that makes The Glue '95% there'.

Cheers

Hey Mercado_Negro. Thanks for your post. I agree that the analog hardware will have more detail and interest due to the multitude of non-linearities and self noise in the signal path. I have only modelled the main sidechain non-linearities in The Glue. I am interested in doing more detailed models in the future, but this will not be added to The Glue as it is intended as an efficient and ultra-clean compressor.

Modelling things like slew rate limiting, the internals of the "That" chip, self noise, and other non-linearities in the main signal path would also require considerably more cpu and research and development time, which is all beyond the scope of The Glue, but for future products I am interested in looking into all these things, and also the possibility of switching in more advanced models on render if they take too much cpu for realtime use.

[andy-cytomic]

hmmm ,I have used Wavelab for test tone generation and also Wavelabs internal FFT meter instead of Span which I dont have installed here /and I tend to think that should be at least as good as the Span/,Ive used Blacman-Harris 9th windowing and it shows aliasing at -150 db for 0.01ms attack and -125 db for 10 ms.Off course you are still better than your competitors in that particular behaviour.
edit-off course no peakclip used and 32 bits so no clip anyway

As already pointed out -125 dB is completely inaudible, and it is around the noise-floor of the most expensive analog kit. I can only hear contant to around -85 dB below a peak of 0 dBFS myself, and even with high quality dither you can only squeeze a perceived -110 dB out of a CD. However, thanks for pointing this out, I will look into this further and let you know what I find out. Aliasing, along with harmonics, should always be higher faster for a faster attack, so although it may not be audible, there may be something interesting to investigate in the 0.01 mS case that I can transfer over to the other times and make them even lower aliasing

[Mercado_Negro]

Hey Mercado_Negro. Thanks for your post. I agree that the analog hardware will have more detail and interest due to the multitude of non-linearities and self noise in the signal path. I have only modelled the main sidechain non-linearities in The Glue. I am interested in doing more detailed models in the future, but this will not be added to The Glue as it is intended as an efficient and ultra-clean compressor.

Modelling things like slew rate limiting, the internals of the "That" chip, self noise, and other non-linearities in the main signal path would also require considerably more cpu and research and development time, which is all beyond the scope of The Glue, but for future products I am interested in looking into all these things, and also the possibility of switching in more advanced models on render if they take too much cpu for realtime use.

Hi Andy, hope you're doing fine.

Thanks for you reply, now I understand Like I said, The Glue is still, after over 2 years now, my go-to compressor for almost everything. I was very curious and always wanted to talk about this subject with you, that's why I appreciate your response and of course, it makes sense.

Cheers

P.S. Hey, please don't forget what we talked about your new plug-in

[andy-cytomic]

Hi kvaca, I checked into it and this seems to just be an oddity of where the energy of the aliasing is being pushed with random constructive and destructive interference. It looks like more energy is being pushed under the "skirt" of the main 993 hz peak, and so you can't see it, but it is still there. Try setting the release to 0.4 and look at the difference between 0.01 mS and 0.1 mS and you will see that the 0.1 mS has less aliasing. Remember also that the windowing used in FFT's will produce notches and other slight differences in the spectrum since it is a low pass fir filter. I suggest for practical use you set the range on your FFT display to -120 dB since anything below his is just not practical