Upgrading the SSL through MIDI: an epic calibration

This forum is for discussing Reason. Questions, answers, ideas, and opinions... all apply.
User avatar
Raveshaper
Posts: 1089
Joined: 16 Jan 2015

30 Jan 2015

I'm doing something strange again. What else is new.

I had the brilliant idea of messing around with the SSL in ways that you can't normally do with CV, and I have developed a workaround using MIDI loopback that effectively unlocks the filters and the EQ so that they respond to audio in real time (with latency of course).

Fine tuning the patch that makes this possible is requiring an epic amount of calibration and is somewhat project-specific to each track it's used on.

Some of the things you can potentially do with this arrangement are along the lines of dynamically high-passing things out of the way of a kick, creating complex sidechain networks that dip EQ, and other advanced mixing abilities that all respond to each track passing through an instance of the patch.

Due to limitations, it is best to use this on buses and not individual tracks, but the power it provides is there regardless of the application.

Is anyone interested in this topic? It will take me some time to figure out the best settings overall so a proper tutorial will be an eventual arrival at best, but so far it looks very promising.
:reason: :ignition: :re: :refillpacker: Enhanced by DataBridge v5

User avatar
ScuzzyEye
Moderator
Posts: 1402
Joined: 15 Jan 2015
Contact:

30 Jan 2015

Sounds neat, I'd like to see where it goes.

Here's a thought: Since the manipulation of the controls is coming via MIDI instead of CV, it is possible to record the automation of the controls. Then you can bump that automation track forward to compensate for the latency.

User avatar
Raveshaper
Posts: 1089
Joined: 16 Jan 2015

30 Jan 2015

I found a bug in the program working on this.
If you control multiple control points in the analyzer window via MIDI, it scitters around trying to render every manipulation at the same time by blinking between them really fast.
It causes a huge amount of system lag when this happens. Apparently these points were programmed with the expectation that they would only be manipulated manually one at a time, even though they were built with midi control capabilities in mind. The malfunction stems from trying to render the manipulations all at once even though the display can only render one at any given time. Strange thing to program a continuous curve in E Mode and not allow for simultaneous rendering of all control points.

I'm generating CV to then send over MIDI. It is very hungry and likes to devour DSP so it may prove to be a "you can do this if you want, but..." sort of thing.
I'm excited for it anyway. It's like the whole SSL has become a giant RE.
:reason: :ignition: :re: :refillpacker: Enhanced by DataBridge v5

User avatar
orthodox
RE Developer
Posts: 2286
Joined: 22 Jan 2015
Location: 55°09'24.5"N 37°27'41.4"E

30 Jan 2015

QwaizanG wrote:I had the brilliant idea of messing around with the SSL in ways that you can't normally do with CV, and I have developed a workaround using MIDI loopback that effectively unlocks the filters and the EQ so that they respond to audio in real time (with latency of course).
Do you really need MIDI loopback to control the EQ knobs? Are you controlling only EQ gains? If so, the same thing could be done by controlling the gain of parallel channels with filters in them.

User avatar
Raveshaper
Posts: 1089
Joined: 16 Jan 2015

30 Jan 2015

I am using 12 signals to control all knobs between LPF and LF dB (12 knobs). I know I can do it with stock devices, this is an experiment.
:reason: :ignition: :re: :refillpacker: Enhanced by DataBridge v5

KEVMOVE02
Posts: 267
Joined: 26 Jan 2015

30 Jan 2015

Okay, for those of us that are clueless (namely me), what is the practical application of this process? Is there an audio example that illustrates the potential of this technique? 

User avatar
Raveshaper
Posts: 1089
Joined: 16 Jan 2015

30 Jan 2015

I will setup a demo using this method and more traditional static techniques. That way you can hear the difference between the two. The main reasons I built this are I learn better visually and like to see what's happening as well as hear it, and the audio processing in the SSL sounds different than stock devices. I'll get this submitted along with a proper write up as soon as I can.

Edit: another reason I build it is I haven't seen this anywhere so I fed my curiosity. Who knows, maybe this will be a big thing. Maybe not.
:reason: :ignition: :re: :refillpacker: Enhanced by DataBridge v5

User avatar
orthodox
RE Developer
Posts: 2286
Joined: 22 Jan 2015
Location: 55°09'24.5"N 37°27'41.4"E

30 Jan 2015

QwaizanG wrote:... the audio processing in the SSL sounds different than stock devices.
I think that's an illusion. SSL EQ bell filters seem to be the same as those in MClass EQ and PEQ-2. You can even get the same result by adding/subtracting BP12-filtered signal to/from the original.

User avatar
Gaja
Posts: 1001
Joined: 16 Jan 2015
Location: Germany
Contact:

31 Jan 2015

orthodox wrote:
QwaizanG wrote:... the audio processing in the SSL sounds different than stock devices.
I think that's an illusion. SSL EQ bell filters seem to be the same as those in MClass EQ and PEQ-2. You can even get the same result by adding/subtracting BP12-filtered signal to/from the original.
are you saying props recycled thenmclass eq algorithms for thenssl and then lied about modelling the ssl eq after the ssl eq, or rather that thenmclass eq was modelled after the ssl eq in the first place?
Cheers!
Fredhoven

User avatar
esselfortium
Posts: 1456
Joined: 15 Jan 2015
Contact:

31 Jan 2015

This is a really interesting experiment, though probably a bit mad for practical use in actual productions.  :)

It's a shame (and somewhat surprising) that there's no way to flip the SSL around to afflict it with CV modulation from the rack. I'd be far more interested in using it, then!
Sarah Mancuso
My music: Future Human

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

31 Jan 2015

QwaizanG wrote:... the audio processing in the SSL sounds different than stock devices.
orthodox wrote: I think that's an illusion. SSL EQ bell filters seem to be the same as those in MClass EQ and PEQ-2. You can even get the same result by adding/subtracting BP12-filtered signal to/from the original.
+1

There may be subtle differences in the shapes, but it's been pretty much agreed in the audio engineering world that if you match the shapes, you match the sound. The problems arise when folks think that by matching the VALUES, you automatically match the shape, which couldn't be further from the truth. 

The exception is when there is saturation involved like with the Trident EQ as one example. 
:)
Selig Audio, LLC

User avatar
orthodox
RE Developer
Posts: 2286
Joined: 22 Jan 2015
Location: 55°09'24.5"N 37°27'41.4"E

31 Jan 2015

Gaja wrote:are you saying props recycled thenmclass eq algorithms for thenssl and then lied about modelling the ssl eq after the ssl eq, or rather that thenmclass eq was modelled after the ssl eq in the first place?
The look and feel was modeled in the first place. The filters are the same simple linear.

User avatar
Raveshaper
Posts: 1089
Joined: 16 Jan 2015

31 Jan 2015

Yes, I agree this is a bit mental to even think about trying, but it seemed straight forward: Control the EQ section and the analyzer window becomes a powerful real time display like iZotope and similar suites. Didn't think I would discover a super massive black hole of system lag from the rendering routine losing its mind. Yes, you can do all of this in a much simpler way. It's not that this is the only way to do dynamic EQ or filter adjustments, I've already done that in combinators. But this is the only way to see everything you're doing to a signal in one place (that I know of).

I stand corrected that the SSL sounds different, letting your eyes trick your ears is a good argument against using this workaround on principle alone, but if I train myself to get a balance between spotting frequencies and hearing when it hits, I feel that this method will speed up my process and results.

I'll get that audio example done soon. But even so, you could use the methods employed without midi looping the SSL.
:reason: :ignition: :re: :refillpacker: Enhanced by DataBridge v5

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

31 Jan 2015

QwaizanG wrote:... the audio processing in the SSL sounds different than stock devices.
orthodox wrote: I think that's an illusion. SSL EQ bell filters seem to be the same as those in MClass EQ and PEQ-2. You can even get the same result by adding/subtracting BP12-filtered signal to/from the original.
Gaja wrote:are you saying props recycled thenmclass eq algorithms for thenssl and then lied about modelling the ssl eq after the ssl eq, or rather that thenmclass eq was modelled after the ssl eq in the first place?
There are many other possibilities than those two, no?!?

The filters used to construct EQs are often similar - it's how they are applied where the differences crop up. The number of bands and the ranges of each control are all a part of modeling an EQ like the SSL, which contribute nothing to the actual "sound" of the device. In other words, they could easily have used some old code if it happened to work for the SSL. There are only so many filter designs that work well for audio…

:)
Selig Audio, LLC

User avatar
Gaja
Posts: 1001
Joined: 16 Jan 2015
Location: Germany
Contact:

31 Jan 2015

QwaizanG wrote:... the audio processing in the SSL sounds different than stock devices.
orthodox wrote: I think that's an illusion. SSL EQ bell filters seem to be the same as those in MClass EQ and PEQ-2. You can even get the same result by adding/subtracting BP12-filtered signal to/from the original.
Gaja wrote:are you saying props recycled thenmclass eq algorithms for thenssl and then lied about modelling the ssl eq after the ssl eq, or rather that thenmclass eq was modelled after the ssl eq in the first place?
selig wrote:
There are many other possibilities than those two, no?!?

The filters used to construct EQs are often similar - it's how they are applied where the differences crop up. The number of bands and the ranges of each control are all a part of modeling an EQ like the SSL, which contribute nothing to the actual "sound" of the device. In other words, they could easily have used some old code if it happened to work for the SSL. There are only so many filter designs that work well for audio…

:)
Thanks for explaining selig.
I always found the SSL EQ to sound better in my setups (well most of the time anyway), so I assumed there was an actual measurable difference in the Q curves or "sound" in some way. There's a good chance that my assumption was wrong of course, so I'll try and compare the Mclass to the SSL when I find the time (My company just bought a post production company and there's lots to do to equalize workflows and other stuff). I'm interested to see how different I perceive the EQs in an A/B comparison. Actually I think I'm going to set up an experiment with the apprentices, so we can all learn...
Do you have any suggestions for an analyzing tool to compare visually as well?
Cheers!
Fredhoven

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

31 Jan 2015

Gaja wrote:Thanks for explaining selig. I always found the SSL EQ to sound better in my setups (well most of the time anyway), so I assumed there was an actual measurable difference in the Q curves or "sound" in some way. There's a good chance that my assumption was wrong of course, so I'll try and compare the Mclass to the SSL when I find the time (My company just bought a post production company and there's lots to do to equalize workflows and other stuff). I'm interested to see how different I perceive the EQs in an A/B comparison. Actually I think I'm going to set up an experiment with the apprentices, so we can all learn... Do you have any suggestions for an analyzing tool to compare visually as well?
What I've done in the past is to use FuzzMeasure (Mac only) to compare curves and tweak parameters until the curves matched - see the results below. What I'll do here is to give you some settings that will closely match (though the actual values will not). 

First I'll start by saying the G EQ in the SSL will never go as narrow as the widest setting on the MClass, so don't even try to compare those - they'll never match!

Try these settings and see if you can tell the difference. For all graphics, the MClass is shown in yellow and the SSL in blue. 

Here's the first set, the Parametrics (you can't even tell there are two graphs here they are SO freaking close!):
Image 

MClass EQ "Param 1"
+6.0 dB
627.2 Hz
Q=1.0

SSL LMF
>>>"E" MODE!!!<<<
+6.03 dB
628.1 Hz
Q=1.0


Here's one for the high shelf (notice things aren't lining up exactly with values OR curves, you'll possibly hear a slight difference):
Image 

MClass
+6 dB
6.0 kHz

SSL
+10.48 dB
5.08 kHz


Image  And finally, the Low Shelf (notice things aren't lining up exactly with values OR curves, you'll possibly hear a slight difference):


MClass
+6 dB
154.9 Hz

SSL
+11.11 dB
40 Hz



Attachments
Screen_Shot_2015-01-31_at_5.39.09_PM.png
Screen_Shot_2015-01-31_at_5.39.09_PM.png (56.65 KiB) Viewed 2157 times
Selig Audio, LLC

User avatar
orthodox
RE Developer
Posts: 2286
Joined: 22 Jan 2015
Location: 55°09'24.5"N 37°27'41.4"E

31 Jan 2015

selig wrote:Here's the first set, the Parametrics (you can't even tell there are two graphs here they are SO freaking close!):
Image 

MClass EQ "Param 1"
+6.0 dB
627.2 Hz
Q=1.0

SSL LMF
>>>"E" MODE!!!<<<
+6.03 dB
628.1 Hz
Q=1.0
Adding to that, here is the phase response for those parameters:
Image


They also match.

User avatar
Raveshaper
Posts: 1089
Joined: 16 Jan 2015

31 Jan 2015

I like the fact that the SSL shelving doesn't introduce dipping. Is there a benefit to dipping like the MClass does?

I'm going to investigate FuzzMeasure.
:reason: :ignition: :re: :refillpacker: Enhanced by DataBridge v5

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

31 Jan 2015

QwaizanG wrote:I like the fact that the SSL shelving doesn't introduce dipping. Is there a benefit to dipping like the MClass does? I'm going to investigate FuzzMeasure.
Indeed, it's what makes the famous Pultec boost/cut trick work so well! McDSP EQs even include a dip and peak control so you can dial it in to taste. That being said, there's no inherent reason to have the dip and it's easy enough to design shelf EQs with no dip.
:)
Selig Audio, LLC

User avatar
Raveshaper
Posts: 1089
Joined: 16 Jan 2015

31 Jan 2015

Can you start a pultec thread explaining what it is and why/when someone would use it? I suffer from confusion.
:reason: :ignition: :re: :refillpacker: Enhanced by DataBridge v5

User avatar
orthodox
RE Developer
Posts: 2286
Joined: 22 Jan 2015
Location: 55°09'24.5"N 37°27'41.4"E

31 Jan 2015

QwaizanG wrote:I like the fact that the SSL shelving doesn't introduce dipping. Is there a benefit to dipping like the MClass does?
If I remember well, dipping naturally occurs when the filters Q is too big. In most cases, when it's more than 0.7.

User avatar
orthodox
RE Developer
Posts: 2286
Joined: 22 Jan 2015
Location: 55°09'24.5"N 37°27'41.4"E

01 Feb 2015

I was wrong about the Q. Here is what happens when you construct a shelf just by summing LP12 gained by 12dB and the original:
Image
The LPF Q is varying from 0.5 to 2.

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

01 Feb 2015

QwaizanG wrote:Can you start a pultec thread explaining what it is and why/when someone would use it? I suffer from confusion.
So sorry, I keep forgetting that our past conversations were recently vaporized. I'll try to dig up what I came up with - I still have the graphics but not the text. It was in a thread comparing the Softube Pultec model to the Kussa MP5. Then we went on to try to replicate those curves with other EQs, always a good time! ;)

As for the effect…
Pultec EQs have a boost AND a cut knob, so you can use them both at the same time. Since they are not 100% symmetrical, you get interesting effects when you use both. Specifically on the low shelf section, using boost and cut in equal amounts gives a curve like in the following graphic. Each color represents one of the four frequency choices, Red = 20 Hz, Green = 30 Hz, Blue = 60 Hz, and Yellow = 100 Hz. All examples have max "Boost" and max "Atten".
Image 

Notice the dip frequency in each example. Turns out the 30 Hz setting is a popular choice for the Boost/Atten trick on kick drums because many folks dip around 200 Hz on kick anyways (this one dips @ 245 Hz). So whatever frequency you choose, you have essentially two bands of EQ in one: a low end boost, and a "mud" range dip. Add some top end from the high band parametric and adjust the width to taste - you've got a classic kick drum EQ setup! :)
Attachments
Screen_Shot_2015-01-31_at_11.18.50_PM.png
Screen_Shot_2015-01-31_at_11.18.50_PM.png (64.35 KiB) Viewed 2130 times
Selig Audio, LLC

User avatar
Raveshaper
Posts: 1089
Joined: 16 Jan 2015

02 Feb 2015

After looking into the matter, I found something rather ugly that has set me back from investing heavily in this as a tool for my work and it may be bad enough to force me to abandon this setup.

I could be mistaken, but this bug appears to be specific to the SSL mixer. It may exist elsewhere.

I was experiencing strange behavior when mapping remote overrides during playback with my CV signals running. Each control I tried to map detected all MIDI CC at the same time instead of treating each message coming through the dedicated channel of the MIDI Bus as a discreet control change. I would guess that this means the entire channel strip of every channel in the SSL doesn't distinguish one MIDI CC from the other no matter what the mappings are. Considering most knobs and buttons can be mapped with ROs, that's horrible. But wait, there's more!

When I investigated using the EMI and scrolled up from MIDI CC 0 through 127, the CV gained context with every button and knob and changed their values all the way up the channel strip with each higher CC value controlling something further up on the interface. The change was one directional, whereby the state of the button or knob was altered and retained after the CV was broken from that control. If a knob was turned on, it stayed on, etc.

This either means that ROs are specifically mapped within the software itself, such that using the wrong CC for the wrong knob or button will cause crosstalk and confusion, or the whole thing is broken.
I will need to try the laborious process of enumerating incoming CC values so that I can find which ones correspond to the proper controls, then see if they remain discreet once "properly" assigned.

Is there a MIDI input list for the SSL in the manual somewhere?
:reason: :ignition: :re: :refillpacker: Enhanced by DataBridge v5

User avatar
ScuzzyEye
Moderator
Posts: 1402
Joined: 15 Jan 2015
Contact:

02 Feb 2015

QwaizanG wrote:The change was one directional, whereby the state of the button or knob was altered and retained after the CV was broken from that control. If a knob was turned on, it stayed on, etc.
That's how MIDI works. It isn't like CV where a constant voltage equals constant value, and removing the voltage returns the value to 0. If you send a MIDI command to change something, it changes. If you want to change it back, you have to send a command to change it back. Even notes need a note-off for them to stop. So yeah, if you're flipping through all the controls, while feeding a CV signal, it'll adjust the control, and then you move on to the next control without sending the original value, it's going to stay where you left it.

An aside: Sometimes, especially with older, flaky hardware, you'd end up with notes continuing to hold even after the sequencer was stopped. That's why there's a MIDI panic button, that sends 128 note-off events, one for each MIDI key. The Stop button in Reason does that, and is the cause for the banging bug in Radical Piano. It plays the key release sound when it receives a note-off. So when you hit stop, and it send 128 note-offs, you get the sound of all the dampers banging down at once.

Post Reply
  • Information
  • Who is online

    Users browsing this forum: Ahrefs [Bot] and 27 guests