Anyone Use Hybrid Crossovers?

[vectorwarrior]

I dont know about the rest of you, but I find the difference between Linear and Hybrid VERY subtle, to the point where I usually can't tell the difference. I can instantly hear the difference between the various analog slopes and the difference between the analog and linear slopes are also very obvious... but that hybrid one compared to standard linear doesn't do anything clear for me.

Personally, I rarely use the Linear slopes because they're so incredibly steep that it makes any differences in processing between the bands quite noticeable due to the harsh slope. I presumed that hybrid would be a smoother, but still linear phase, crossover as this is what the info text implies... but this doesn't seem to be the case to my ears.

Just me?

Also, the latest update is awesome

hybrid.jpg

[MeldaProduction]

Thank you!
And yes, Hybrid is quite similar, but it is far less steep. It could be even less , but well, it is the way it is

[vectorwarrior]

Thanks Vojtech.

Curious to know if other people have thoughts on this. If nobody is using it at all, there may be a case to modify it slightly for an even smoother sound that has a more audible difference. No worries if I'm the odd one out here though

[MeldaProduction]

Hehe sure . We can designed another crossover, but the main question is if it would be useful at all.

[vectorwarrior]

I noticed the crossover video went up today. I haven't watched it yet, but it reminded me that I wanted to look further into my issues with the Hybrid crossover (in that it doesn't sound very different at all).

My tests show that I'm right and it's essentially identical to the linear crossover in terms of frequency response (unless I'm missing some other details about how it works).

Here's the test I did, so you can reproduce for yourself.

1 - Make a 10 second pink noise file (in Adobe Audition)
2 - Do a scan of the 10 seconds in the frequency analysis graph and store it to memory 1 (so you can compare later)
3 - Open MXXX and in an empty preset, drop a single crossover module and move the crossover frequency to 1000hz, deselect output 1 (or mute it in the crossover) so you are only listening to the signal above 1khz as output by the crossover
4 - set the crossover to linear phase
5 - process the 10seconds of noise
6 - do another scan in the frequency analysis and save the scan as memory 2

Repeat steps 4 to 6, changing the crossover and the frequency analysis memory number, so you can compare the output of the crossover between linear phase, hybrid and analog.

Here is the output of the above. In the image below you can see the frequency response of:

1 - The original pink noise (RED)
2 - Linear phase (ORANGE)
3 - Hybrid (YELLOW)
4 - 24db analog (GREEN)
5 - 96db analog (closest slope to the linears) (GREEN)

Freqs 1-5.jpg

You can clearly see the crossovers and that 24dB analog is obviously less steep. The linear phase and hybrid are difficult to tell apart, so below is the same image with ONLY the original, the linear and the hybrid:

Freqs 1-3.jpg

As you can clearly see, they are identical. You can see that I still have memory 2 and 3 selected, they ARE both visible. Vojtech, is this what you intended? I can't see how it could be.

Thanks!

[MeldaProduction]

The hybrid crossover actually changes with frequency. Basically, the linear phase crossover is purely scientific, meaning it has the same response for every frequency, in linear scale! As you know we are actually perceiving things in log scale, which means that the LP's crossover slope is actually steeper for higher frequencies. Analog crossovers on the other hand have the plus minus same slope for all frequencies in log scale. That's what hybrid crossover sort of approximates.

[vectorwarrior]

Ahh, I see. Thanks for the info, so I guess people should avoid linear-phase and stick to hybrid as it's more consistent in a 'natural', if not 'math' sense. I'll always use hybrid from now on!

I tried a quick test with the same technique at 100hz and 6000hz. You can see the test results below. At 100hz it's still identical. At 6000hz it is very slightly different, being slightly gentler.

100hz crossover - original, linear, hybrid.jpg

6000hz crossover - original, linear, hybrid.jpg

I did some more testing and the difference at high frequencies becomes less and less the further you go down, until 1000hz where it's the same. So what I'm gathering from this is that if your crossover is 1000hz or lower then there is no difference in crossover between linear and hybrid. Good to know.

I'd love if there was an overall smoother crossover that was linear phase. Like a 12 or 24db curve. Is that hard/impossible to do with linear phase designs? I've noticed the linear phase EQ's dont have the same filters as the analog style, so I'm guessing it's some sort of technical challenge with linear.

Sorry for all the heavy posting on this, I just never quite understood what hybrid 'did' and I wanted to find out Thanks!

[MeldaProduction]

Actually from your image it is very different! Anyways, it's not that it wouldn't be possible to get the analog LP response, it's sort of performance issue mainly actually. We may do some "exact match" thing, but the question is if it is necessary, from my experience LP filtering in general is sort of specific .

[jemmcauley]

I use the hybrid filters all the time, mainly because you can do multiband processing in parallel, something you can't do with the minimum phase filters. I much prefer the sound of hybrid to the standard linear phase and would welcome another LP option with even more gentle filter slopes.

[pone]

In mxxx/powersynth you can make your own linear phase crossover using convolution, and adjust the steepness with the IR length.

[MeldaProduction]

In mxxx/powersynth you can make your own linear phase crossover using convolution, and adjust the steepness with the IR length.

That's true, though it's not exactly easy to do. Maybe some additional LP filters could be useful...

[jemmcauley]

A big "yes please" to more linear phase crossover options