fix echo effect reducing volume of input

[Be-ing]

The input samples were being divided by 2.0 along with the echoed signal. This changes it so only the echoed signal is divided by 2.0. Now, with the send parameter all the way down, toggling the effect enable switch does not alter the sound at all. If the echoed signal is not attenuated, it can clip somewhat easily.

[timrae]

Now it gets louder when you press the enable switch :-/

[Be-ing]

With the send knob all the way down? If the send knob is not all the way down, it should get louder when enabled.

[timrae]

If it wasn't all the way down it was close to it.... Enough to be annoying

…

On 7 Feb. 2017 8:58 am, "Be" ***@***.***> wrote: With the send knob all the way down? If the send knob is not all the way down, it should get louder when enabled. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1166 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACsA4gfXB8pXt1ds1RF_jI7CPepVLXc5ks5rZ7OTgaJpZM4L14t5> .

[timrae]

Btw reverb doesn't have this issue... Not sure if it will be of reference to you, but might be worth checking

…

On 7 Feb. 2017 9:07 am, "Tim Rae" ***@***.***> wrote: If it wasn't all the way down it was close to it.... Enough to be annoying On 7 Feb. 2017 8:58 am, "Be" ***@***.***> wrote: > With the send knob all the way down? If the send knob is not all the way > down, it should get louder when enabled. > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#1166 (comment)>, or mute > the thread > <https://github.com/notifications/unsubscribe-auth/ACsA4gfXB8pXt1ds1RF_jI7CPepVLXc5ks5rZ7OTgaJpZM4L14t5> > . >

[Be-ing]

I tested this again and it sounds fine to me. What you are expecting to happen @timrae? If the send knob is up a little bit, it gets a little louder when enabled, as I expect. When the send knob is all the way down, there is no change at all when enabled. Anyone else care to test?

[daschuer]

Yes, the result is better. However, this changes the clipping behavior.

Before
(In + clipped(delay)) / 2 -> not clipping if input is not clipping

Now
In + clipped(delay) / 2 -> clipping (going to red) can still happen

This is basically the same as mixing two tracks. So it should be not an issue isn't it?
Or should we add an other clipper to the output, like we do in the phaser?

I think we can improve the sound by using a tanh clipper that models the sound of a clipped transistor circuit. A bit faster tanh approx. is used in the moog filter:

mixxx/src/engine/enginefiltermoogladder4.h

Line 238 in 75e85d3

inline float tanh_approx(float input) {

[Be-ing]

This is basically the same as mixing two tracks. So it should be not an issue isn't it?

Kinda. It's not quite the same. It's like mixing in another track that is ~6 dB quieter. So yes, clipping is still possible and it is up to the user to be mindful of that. I tried not dividing the echoed signal by 2, but IMO it was too easy to clip that way. I think this PR should be merged as is. Afterwards, if you have ideas for improving the effect further, please open a new PR.

[daschuer]

LGTM Thank you.