RoomFit™ is Dumb as Rocks

[EddNog]

I ask for Harman, it gives me this (I used red to mark the resulting curve based on its automatic correction algorithm) when setting correction range up to 250Hz:


Hilariously bad. This worked better months ago the last time I tried it. Whatever, “enhancements,” have been made seem to have broken it. These results are absolutely terrible and would completely wipe the bass. Better to just use no correction at all.

The algorithm is clearly failing to look at the average level of the curve and then properly aligning the Harman profile to the natural response above the correction cutoff. Look at the distance between the yellow and the blue lines to the right of 250Hz.

Am I supposed to turn down the speakers myself to manually align the blue and yellow? The system is not smart enough to realign these by itself?

-Ed

 

[EddNog]

Okay I figured it out.

It has to do with how LinkPlay implemented cut-only mode. Well this stinks because it makes cut-only mode unusable. Basically have to fall back to limiting boosts to like 0.5 or 1dB and disabling cut-only mode.

-Ed

Edit: meant cut-only mode

 

[slartibartfast]

Okay I figured it out.

It has to do with how LinkPlay implemented boost-only mode. Well this stinks because it makes boost only mode unusable. Basically have to fall back to limiting boosts to like 0.5 or 1dB and disabling boost-only mode.

-Ed

Cut-only mode? . It seems to choose a better target level if you use a wider frequency range.

 

[EddNog]

Cut-only mode? . It seems to choose a better target level if you use a wider frequency range.

I ended up just forgetting RoomFit™ all together and applying just one cut at 45Hz/-4.5dB/Q=3.0 to fix this:



And I’m all set.

-Ed

 

[dominikz]

I ask for Harman, it gives me this (I used red to mark the resulting curve based on its automatic correction algorithm) when setting correction range up to 250Hz:
View attachment 26744

Hilariously bad. This worked better months ago the last time I tried it. Whatever, “enhancements,” have been made seem to have broken it. These results are absolutely terrible and would completely wipe the bass. Better to just use no correction at all.

The algorithm is clearly failing to look at the average level of the curve and then properly aligning the Harman profile to the natural response above the correction cutoff. Look at the distance between the yellow and the blue lines to the right of 250Hz.

Am I supposed to turn down the speakers myself to manually align the blue and yellow? The system is not smart enough to realign these by itself?

-Ed

This is IMHO incorrect implementation of automatic target level match function in RoomFit, I fully agree.
This kind of behaviour was already discussed a bit in this thread with @jiaxin.li (WiiM), and at least part of the issue was acknowledged, but unfortunately not all.

 

[EddNog]

This is IMHO incorrect implementation of automatic target level match function in RoomFit, I fully agree.
This kind of behaviour was already discussed a bit in this thread with @jiaxin.li (WiiM), and at least part of the issue was acknowledged, but unfortunately not all.

Yeah it appears to be a half-assed attempt at forcing the algorithm to be cut-only, but this is only in any way, shape, or form functional for people applying full-band correction, so for people like me looking to correct only in the lower frequencies where room/boundary interaction occurs, it’s not going to work at all.

-Ed

 

[dominikz]

Yeah it appears to be a half-assed attempt at forcing the algorithm to be cut-only, but this is only in any way, shape, or form functional for people applying full-band correction, so for people like me looking to correct only in the lower frequencies where room/boundary interaction occurs, it’s not going to work at all.

-Ed

IMHO the issue is not with the cut-only mode per-se (since filter calculation part works fine), but with the unjustified assumption that only the response within the RoomFit correction range should be used to set the automatic target curve level.

It should be pretty obvious that this assumption only works well in cases where the fullrange measured response shape already roughly matches the target curve shape.
In other cases you can get some kind of unexpected behavior. Your example in post #1 is IMHO a very graphic illustration of why this was a bad design choice by WiiM.

As mentioned in the other thread, I'd personally suggest to WiiM to use a wider and possibly even fixed frequency range (e.g. 80Hz to 4kHz) to find the target curve level, unrelated to the RoomFit correction range. Also, I'd allow manual adjustment of target curve level by the user, useful if the algorithm misses the mark.

Let's hope that @jiaxin.li (WiiM) and @WiiM Team consider this.