Understanding filters negative effect on transients

[stonerbox]

Well, all worthwhile.

I'm surprised there's such a difference.

From my standpoint I'm surprised it's solved. Learned a few things from you. Thank you. From now on I'll choose and design EQs and filters carefully.

[Digital Larry]

I have been following this thread - so what is the bottom line/conclusion?  I seem to have missed it.

DL

[Fancy Lime]

I have been following this thread - so what is the bottom line/conclusion?  I seem to have missed it.

DL

That constant voltage crossovers solve the phase problem introduced by frequency-splitting and recombining a signal, at least to a degree. I think...

[stonerbox]

I have been following this thread - so what is the bottom line/conclusion?  I seem to have missed it.

DL

The Sallen-Key crossover altered the signal heavily pushing the original signal "into the background" by adding triangle shaped top and bottom waves. Terrible masking. The constant voltage crossover leaves the original signal fully intact and retains all the fine details and dynamics. Difference is absolutely massive. Comparison video.


[Click for full sizes]

Sallen-Key Crossover (Transients Altered)


Constant Voltage Crossover (Transients Unaltered)

[ElectricDruid]

Honestly, I'm not hearing any significant difference, and there's no sign of it on the frequency spectrum graph either. I know my ears aren't great (too much Hendrix listened to at Hendrix volumes earlier in life), but seriously? Do you really think that's "absolutely massive"? I was hoping for something at least *audible*....sigh...maybe getting old is a sh!t as they say....

[stonerbox]

Honestly, I'm not hearing any significant difference, and there's no sign of it on the frequency spectrum graph either. I know my ears aren't great (too much Hendrix listened to at Hendrix volumes earlier in life), but seriously? Do you really think that's "absolutely massive"? I was hoping for something at least *audible*....sigh...maybe getting old is a sh!t as they say....

To me it's a big, big difference in dynamics and clarity. I do listen on a pair of ATC SCM25A but it is clear on regular headphones too. However if you listen to it on a phone speakers it gets lost in translation.
The CVC has outstanding dynamics and much better presence in the low midrange and bass. But to be fair the high midrange and treble comes out unaltered too. It just sounds the same as it did pre the filter. Magic. The Sallen-Key on the other hand just sucks the life right out of the signal, at least when it is configured as a crossover.

there's no sign of it on the frequency spectrum graph either.

Be sure to check the frequency response again. The peaks might be similar but both the volume increases (despite the purple clips having a smaller "print") and dynamics of the peaks reaches both higher and lower than the triangle-masked green clips.

[Digital Larry]

OK I took a closer look at both the waveforms and the circuit.

It's hard to judge pure phase shift effects, but I boil waveforms down to the following:
- a quickly rising edge is indicative of extended high frequency response (this is what I call a transient)
- the ability to hold a DC level (flat tops on square waves, e.g.) is indicative of good low frequency response.  We see the rise time after each transient.  My guess is that the signal changes sign before it has time to fully rise.  I'm guessing that with a lower frequency test signal we would see the flat tops being held as there are no caps in the series path on that side.
- ringing on transitions is indicative of resonant (second order or higher) behavior

The circuitry of the better one looks like a common DSP implementation, namely:

highpass = input - lowpass

so that adding highpass to lowpass = input.  It's so simple!  It doesn't even matter what "lowpass" is!

Whereas the one which sums the outputs of two separate resonant sections has a fairly big notch in the middle where the transitions overlap because of phase cancellations.  And it will clearly depend on what the lowpass function is.

I've also used a 2-pole state variable filter structure in DSP and there, adding the HP to the LP is how you get a notch.  And you can do something else to get BP.  Input - notch.  But you can also do the input - highpass trick to get lowpass, even on an SVF, though it's not clear why you'd want to.  It doesn't cost components, only more instructions.

Is that pretty much it?   

DL

[FiveseveN]

I was hoping for something at least *audible*....sigh...maybe getting old is a sh!t as they say....

Maybe it's your speakers? The difference is exclusively in the low end.

What I'm hearing is a signal that's high-passed/low shelved vs one that isn't. It's consistent with the waveforms we see and the FFT I ran to double check.

[antonis]

I was hoping for something at least *audible*....sigh...maybe getting old is a sh!t as they say....

My signature second part feels for you, Tom..

[ElectricDruid]

I was hoping for something at least *audible*....sigh...maybe getting old is a sh!t as they say....

Maybe it's your speakers? The difference is exclusively in the low end.

Probably. I was listening on my laptop. I'll try headphones and see if I can hear it.

[Digital Larry]

I listened on my AKG240s (sorry) and yeah I perceive a bit less low end.  Clarity I associate with the high end and is not changed.  Dynamics I associate with compression, and since it's a distorted sound yes there's compression.  I generally can only hear compression on clean sounds where it becomes more obvious, or in the actual playing.

Lacking low end, there is less oomph.  This is one of the reasons I like trying to be objective since we all have words which mean different things to different people.  I suppose if you were motivated enough you could pass the exact same sample through both circuits and then try to subtract one from the other. 

DL

[ElectricDruid]

yeah I perceive a bit less low end.  Clarity I associate with the high end and is not changed.

This is making less sense to me. I thought the problem was supposed to be *transients* getting affected. Low end is the *opposite* of that, pretty much!

[Rob Strand]

1:19:
Old man, how is it you hear these things.
Young man, how it is you do not.

[Digital Larry]

yeah I perceive a bit less low end.  Clarity I associate with the high end and is not changed.

This is making less sense to me. I thought the problem was supposed to be *transients* getting affected. Low end is the *opposite* of that, pretty much!

I was not the one with the original comment about transients.  I agree with you.

DL

[ElectricDruid]

yeah I perceive a bit less low end.  Clarity I associate with the high end and is not changed.

This is making less sense to me. I thought the problem was supposed to be *transients* getting affected. Low end is the *opposite* of that, pretty much!

I was not the one with the original comment about transients.  I agree with you.

No criticism of you intended, DL.

[Digital Larry]

No criticism of you intended, DL.

None taken.