Convolution 'strength' adjustment

The suggestion is to have a feature similar to the ‘Mix adjust’ in Foobar2000 added to Roon’s convolver:
image

or the ‘AutoEq strength’ in Wavelet:

I find it very useful in those apps, because while the headphone correction filters often have a generally positive impact on my SQ perception, some aspects are too pronounced (e.g. the PortaPro filter at 100% results in an unpleasant bass punch, while at 70% it’s spot on imho).

I believe this would really improve the usability of Roon’s DSP feature.

If I’m not mistaken, once you determine the optimal strength of the convolution, you can obtain another convolution that achieves the same result, right?

No doubt it’s possible, but AFAIK it requires additional software. Moreover, I adjust it quite often, depending on how the source material was mixed. I would go insane if I had to generate a new filter each time:) It should be a simple slider/dial, like the ones implemented in freeware apps like Foobar long ago.

I think Roon guys don’t want to get into the convolution generation game. This is arguably a relatively cheap feature, but I remember that Roon has to re-initialize the signal path whenever the filters change, so that may be an impediment to providing a slider working in real time.

U sure that what I’m suggesting (mix adjustment) involves new filter generation? How can a light Android app such as Wavelet do it in real time with no playback interruptions? At this price I’d seriously expect Roon devs to try and catch up with freeware, especially considering that this feature has direct and profound impact on sound quality perception, which apparently Roon is all about,:wink:

They’d have to change the filter or support adjustable mixing of the original signal in the convolution output. Either way, it would have to be done in real time.

Bumping this one up. I believe this would be a useful feature:)

What happens when your convolution filter is your crossover?? hmm

No idea as I have no use case for a crossover in my setup. I gues you’d be inclined to leave the mix at 100% maybe?

What I was trying to suggest is the request technically makes little sense, convolution DSP is used for more tasks than just target curve correction.

In this scenario, users would typically create multiple convolution filter setups and save them on to a PC for easy loading as needed. I access about 4 convolutions in my setup.

In terms of software, convolution filters are easily created in freeware packages including:
REW, which has a direct Roon FIR export (Its free) and/or together with rephase (free).

Anyway, the feature request is your to make, just presenting an option as a work around for you.

Well, I’m glad that the devs of Foobar and Wavelet didn’t share your approach :wink: Apparently, they saw sense in putting this in their free apps so I see it reasonable to expect a similar functionality in an audiophile player priced at hundreds of $.

Other uses of convolution would not be impacted in any way so I don’t get that bit. It’s a request for an addition that would benefit the headphone listeners. I think we can assume they are a considerable subset of Roon users. XOs, room correction etc. need to be done differently and no one argues with that.

In my case, the Harman curves work subjectively fine for some music material, but not so much for other - we all know there are huge differences between albums in terms of mix and mastering. For instance, some bass-heavy music will produce nasty distortion in my PortaPros with full-scale AutoEQ applied. With Wavelet, it takes me 10 secs to operate the slider and get satisfying sound again. Following your suggestion, I should get out of bed, start the PC, launch REW and fiddle with the filters just to enjoy a single album, then do it again for another release with unconventional production.

Don’t get me wrong, I appreciate your suggestion. I am also aware of tools like REW and their capabilities. They just aren’t fit for my use case. And again, I feel that at this price point and with software that claims focus on sound quality, we really should get features that freeware delivered years ago.

1 Like

Thanks for you reply.

FYI, Check out tools curve.
This is what I use. Sits in between flat and Harman. Harman is way too bass heavy for me.

I have no issue with your request.

But, I would still argue convolution is best executed at chosen sampling rate and with different target filters/curves. Not by %.

1 Like

Thanks! Do you mean Floyd Toole? Can you point me to a source for parametric data? All I can find are plots taken from his book, I guess that’s not enough as a starting point. I have relied on automated solutions so far, but I’m eager to study the subject and learn stuff.

No doubt about that! I’m just trying to balance the work I put into this hobby and the enjoyment I take out:) Not much free time for tinkering these days. Wouldn’t the ideal solution (for HPs at least) be a convolution with 3 parameters: (1) headphone response (2) target curve and (3) the frequency spectrum of the input signal? Essentially a solution that would pre-compute a target response modified by the characteristics of the record I’m about to play?

Sounds like what you after is a base convolution, with a variable loudness curve? No convolution filter will analyse content and adjust EQ.

Dacs like the RME adi2 dac Fs will provide variable loudness, customisable by user.

Otherwise simply select different convolution EQ filters to suit.

You are essentially asking for AI configurable/learning EQ.

Yeah, I was fantasizing there :slightly_smiling_face: but with stuff like ANC and surroundings awareness in modern headphones, I guess it might not be that far away? And I’m not thinking of a real-time EQ, just something that would scan a sound file and compute the specific correction in advance.

1 Like

But to which target? Or preference?

Whatever I choose: Harman, Toole, or some other response curve (there’s an UAPP plugin called MorphIt that will emulate another headphone for you based on your device and the target), but the input signal would also be taken into consideration so that no filter ‘over-/undershoots’ the target.