Curve Shapes

[Rich Walsh]

I've been looking at these again, having done a lot of work on making custom ones for v2 several years ago. Compare the two attached captures: v3 seems to snatch toward the end – I preferred the shape (and sound) of v2. These are with the same handle settings in the Inspector – everyone's favourite, handles to the corners.

v2:

[micpool]

What do you mean by "handles in the corners". I can get many different  curve variations with handles in the corners

Mic

[Rich Walsh]

Dragging the handles to the corners, as far as they will go – something I've seen many designers do, and something that used to give a good general purpose fade shape for everything except slow fade-ins.

I don't really know what the "linear" curve is supposed to be now: it used to be linear in dB, but now there seems to be some non-linearity about how % on the curve equates to level in dB.

Rich

--
--
Change your preferences or unsubscribe here:
http://groups.google.com/group/qlab
 
Follow Figure 53 on Twitter: http://twitter.com/Figure53
---
You received this message because you are subscribed to the Google Groups "QLab" group.
To unsubscribe from this group and stop receiving emails from it, send an email to qlab+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/qlab/afc4f929-dd9f-47cf-a705-512154ad8520%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[micpool]

Ahh, I see

Do you want to send me a workspace with the  audio, and fade that gave you that v3 envelope history.  I can't replicate your findings, Here's my plot of the handles in the corner shape.

Mic

[micpool]

The curve shape and smoothness is identical in v2 and v3, but the effective fade length is 15% shorter in v3 (with both custom curves set to handles in corner and min level at -120dB

Here's  V3 at 20s in pink overlaid on the v2 at 17s

[Rich Walsh]

Mic and I are looking at this off list, but I am not seeing the same results. I am recording the output of QLab and looking at the resulting audio files, and consistently seeing kinks in v3 that I didn't/don't in v2: here is a 30s fade from v3 recorded via Soundflower into TwistedWave – notice the kink at c20.5s.

You might argue that you wouldn't hear this subtlety, but I think it makes it even harder to know what the fade handles are going to do. I spent a long time in v2 building a fairly good approximation to an equal power fade – which is great for panning between speakers; it'd be a pain to have to do all that again to workaround this non-linearity of the fader shape.

I've always thought it strange that QLab presents fades in percentage of travel of a linear fader, ie: linear in dB. Everything else presents fades in percentage of level, or directly in dB. Consider fading a 0dBFS sound: halfway through a traditional linear fade the level is down 6dB; halfway through a QLab linear fade (straight line on the display) it should be at -30dBFS if your minimum audio level is -60dB. I think.

It would be more intuitive to me if the curve you draw represented the resulting shape of the waveform – as with a DAW. It is currently abstracted from the shape of the waveform by the logarithmic relationship with dB on the fader.

I am only about 90% sure I've got all this right though…

Rich

[micpool]

I can now repeatedly reproduce your recorded bird beak envelopes and my perfectly smooth ones.  There would appear as you say to be a fundamental problem with the maths!

So it would appear that any fade whose range is greater than 120dB goes linear after 120dB, or something along those lines.   

Mic

[Rich Walsh]

Aha! It's because I was using a -12dBFS test tone and sticking the master up to +12dB to compensate. I can't remember if I did that in v2 or not; have you checked whether this bug was in v2? I'll test again with the gain on a fader other than the one I'm targeting with the fade.

Ironically, the shape of the bird beak is exactly the shape I'm looking for in a "linear" fade!

Thanks for spending the time to figure that out, Mic.

Rich

[micpool]

The problem is not present in QLab 2 which exhibits smooth curves throughout the 132dB range of the fade.

Mic

[Chris Ashworth]

The percentages are because the fade shape is applied to arbitrary things, not just decibel levels.

After feedback about how QLab 2 fades worked we changed it for version 3.  Instead of fading through the decibel domain it fades through the same domain that is used for the volume sliders. 

Last but not least, there will be new parametric fade shapes in the future.

-C

[Rich Walsh]

What's the fader law then, ie: how do I compensate for the curve?

Usefully, the default linear shape is now very similar to my "50% @ 1/5" curve in v2 that was perfect for slow fade-ins, while the QLab S-Curve is very similar to my "handles to the corners" curve – which are the two curves I use for almost everything except pan fades.

Incidentally, I took this screenshot with the minimum level set to -120dB; changing it to -60dB doesn't scale the curve, it just makes all the faders below -60 jump to out. I'm not sure that's consistent with the point of the minimum volume is it? I can't quite get my head round it though… Mic?

Thanks.

Rich

On 29 Jan 2016, at 18:12, Chris Ashworth <ch...@figure53.com> wrote:

Instead of fading through the decibel domain it fades through the same domain that is used for the volume sliders.

[Christopher Ashworth]

Well, it depends on what you're trying to do, but you likely can't; it's not intended to support achieving a specific functional match to a given equation. It's intended to sound nice. Which, based on the feedback from v2 and v3, more folks think it does now compared to v2. The beak there at the end of large db changes does look like unintended behavior, which I'll put on the list to investigate.

There's room for more specific / powerful / custom curve functionally and it is indeed coming.

C

(mobile)

[Rich Walsh]

Sorry, I'm not being critical – just poking around under the bonnet 3 years too late. I did this with v2 about 6 years ago, when I first joined this forum.

So, in the short term there isn't an easy way for me to calculate how to draw an equal power fade in v3, as I had in v2. Fair enough; I don't use it that often – and analysis shows that the 2 default fades now built in to v3 are essentially what I'd arrived at in v2 as things that "sound nice", so I agree that you've made a step forward.

However, I don't think the logic of the minimum level is working – at least in an intuitive way. If I set the level to -120dB I get a nice smooth fade that lasts almost the whole duration; if I change the level to -40dB (the maximum) I'd expect to get the same, with the fade ending at a higher level. What I see, however, is exactly the same duration of fade but it's truncated early.

Compare these two screenshots: a very similar shape fade from v2 & v3. The top channel is a 10s fade with -120dB minimum, the bottom with -40dB. v2 is doing the right thing; what v3 is doing is essentially pointless I think...

Am I right in thinking that in v3 your fader law is effectively applied by inserting a VCA, so you'll see the same shape whether fading out from 0 or -60 – ie: the fader law doesn't start at -60 in the latter case? Testing implies this is true.

Thanks.

Rich

[micpool]

That looks familiar

Is this the gating behaviour that was discussed and fixed relating to in fades:

https://groups.google.com/forum/#!msg/qlab/iE5gIUd7_VE/ZZswJU3LBAAJ

but this time gating fade outs?

Mic

[Rich Walsh]

You know you're getting old when you forget you've already come across a bug and been part of a discussion about it!

I need more sleep.

Rich

[Chris Ashworth]

On January 30, 2016 at 7:12:14 AM, Rich Walsh (rich...@mac.com) wrote:

Sorry, I'm not being critical 

Oh, no worries; as often than not, the detailed investigations into these things shine light on errors on our end.  Was just trying to (probably too) quickly describe the intentions & context of the current behavior.

– just poking around under the bonnet 3 years too late. I did this with v2 about 6 years ago, when I first joined this forum. 

So, in the short term there isn't an easy way for me to calculate how to draw an equal power fade in v3, as I had in v2. Fair enough; I don't use it that often – and analysis shows that the 2 default fades now built in to v3 are essentially what I'd arrived at in v2 as things that "sound nice", so I agree that you've made a step forward. 

However, I don't think the logic of the minimum level is working – at least in an intuitive way. If I set the level to -120dB I get a nice smooth fade that lasts almost the whole duration; if I change the level to -40dB (the maximum) I'd expect to get the same, with the fade ending at a higher level. What I see, however, is exactly the same duration of fade but it's truncated early. 

I’ll dig into this and will report what I find.  Have a couple of fairly urgent support tickets to help on, but will see what I can find after they’re tended to