I have made excellent progress using REW and a UMIK-1 in improving my small office system, but I am struggling to replicate the same success on my main system. My assumption is that it is easier to make significant improvements in a near field set up in comparison to the complexities of dealing with the more complicated reflections and room modes with the larger room due to the larger distances. Is my intuition correct?
I struggled with similar questions a few months back but found a fairly simple formula that IMHO works pretty well in mid/farfield:
- Use MMM with periodic pink noise to measure L and R individually, and then L+R together
- In REW set the EQ target curve to your preferred target (I use either the -0,8dB/octave slope or B&W 1974 target in my main system) and generate cut-only filters for individual L and R channels (set maximum individual and combined boost to 0). Typically I limit the correction to the area below ~300Hz or so (so below the room transition / Schroeder frequency) and I allow high-Q filters here
- After you apply the filters use MMM again to remeasure L and R individually, and then L+R together to validate your results. With some luck there will be no new notches in the combined L+R response after EQ
- Listen
- If the combined L+R response after EQ sounds ands looks more ragged, you can instead use the original L+R measurement to generate the EQ that applies to both channels and repeat steps 3. and 4.
Second question. I have currently been applying the same eq to left and right channels (mainly due to my time constraints). My rooms are both reasonably symmetrical. Is there likely to be a significant benefit in eq'ing left and right separately? Is it something that most people here do?
In my experience both combined and separate L/R EQ can work pretty well, thought strictly speaking each is a bit of a compromise IMHO:
- EQ-ing L and R individually with IIR filters (minimum phase) can cause new peaks and dips in the L+R response due to the introduced phase shift. Though I haven't experienced any issues, I assume it might be audible with recordings that have bass frequencies mixed to mono.
- EQ-ing L+R together will give a smoother mono bass response, but now individual channels might still have some peaks and dips. This might be audible with recordings that have stereo bass content.
If you're interested I've documented
my experiences with REW generated EQ and MMM here (the same thread also cover various other DRC tools I tried).
Another note - if your speakers are not neutral in tonality but have good directivity, you can also EQ above the transition frequency based on anechoic listening window (LW) response from measurements such as provided by
@amirm or
@hardisj. Do note however that there is typically some variability within speaker production runs so depending on production tolerance precise high-Q EQ corrections might or might not apply to your specific unit (even if anechoic measurements are available) so low-Q tonality corrections will likely be more universal. You could also try to measure LW response of your specific loudspeaker units using
this guide by
@napilopez to be able to generate custom speaker-correction EQ filters in HF - but again note that quasi-anechoic method loses precision in the mid frequencies so you would again be partially limited there.
There are also some arguments that doing full-range EQ corrections (i.e. above transition frequency) directly based on in-room steady state response (such as measured with MMM or single-point measurements with sweeps and gated by FDW) but although it can sound pretty good, so far I can't really see how that approach can avoid introducing new resonances to the loudspeaker direct-sound - it is something I plan to investigate more going forward.
