TOPIC:

I will definitely keep trying. Just did another listening session. Keep preferring dirac off. Imaging is more holographic and “real” is the only way I can (subjectively I know) describe it. Leonard Cohen is RIGHT THERE in three dimensions with dirac off. Turn it on and he flattens out and now he’s on a screen. 🤷‍♂️ 

I did 9 measurements for focused imaging, roughly 30cm apart. I will do a remeasure with points exactly 50cm apart and be more precise when I get the time. Shall report back. 

tallbeardedone wrote:

I did 9 measurements for focused imaging, roughly 30cm apart. I will do a remeasure with points exactly 50cm apart and be more precise when I getthe time. Shall report back. 

I realise I was a bit imprecise with my description above. You have 8 points surrounding your central one and spacing between these 8 of 30 cm is OK but it's recommended for them to fall on a sphere of minimum radius 50 cm centered on the MLP.

I wish there was a more official source for this but checking back I was remembering this info. being given by Flavio Fellah (of Dirac) as reported on another forum here:

www.hifiwigwam.com/forum/threads/minidsp.../page-6#post-1948878

Very helpful post thank you. I will do a more precise re measure and try again. 

Team - I'm experimenting based on @Ultrasonic recommending limiting Dirac to the lower frequencies but I still want to try full range EQ.

So with Dirac I've created two configs with the same target curve but limited one config to ~350Hz.
I've used the SHD as speaker and mic method to measure the Dirac filter of both configs.

 

Showing one channel only, the dark red at the top is the dirac filter when limited to 350Hz while the blue (tucked in behind it) is the full range filter.
Down at the bottom we have the trace arithmetic A/B result.

What I was thinking of trying is converting the A/B to a biquad for the mains. (Yes in this example I am boosting the 1kHz-4kHz range so i can clearly see some difference and know the approach works). But I can't figure out in REW though how to convert this trace directly to biquad. What I get is the inverse of what I want! REW matches the trace to zero rather than zero to the trace.



Any ideas?
Thanks,
Alex



EDIT: Ignore me. In the cold light of day the answer became obvious. Rather than A/B, I did B/A. That gave me the difference as a negative. Then in EQ I set the target as 0dB and that gave me the filters I need. Will try this out later!

 

asx77 wrote:

Team - I'm experimenting based on @Ultrasonic recommending limiting Dirac to the lower frequencies but I still want to try full range EQ.

What is your thinking behind trying to compare what you are? Not sure if you may have misinterpreted something I wrote... 

From my perspective I can't see a way that the biquad result is likely to be better than using Dirac Live.

What you've plotted is the amplitude response of the filters, which shows how the filter changes the relative amplitude of different frequencies. Filters also change the relative timing of the different frequencies, and this is what is shown by the filter's phase response (I plotted one example for a DL filter earlier in this thread). DL deliberately makes phase changes targeted at improving the impulse response, whereas biquad filters will also introduce phase changes but not ones that are specifically designed to be desirable. 



 

@Ultrasonic, Looking at the dirac measurements I noticed my speakers have a little dip around the 2kHz-3kHz region. I likened this in severity of problem, to when you said that if a speaker is too bright you could choose to do a simple eq to fix and avoid using dirac up to too high a frequency. My experiment was just going to compare (subjectively) the two approaches but I hadn't appreciated that I might be messing up the speaker's phase by doing this. 

asx77 wrote:

@Ultrasonic, Looking at the dirac measurements I noticed my speakers have a little dip around the 2kHz-3kHz region. I likened this in severity of problem, to when you said that if a speaker is too bright you could choose to do a simple eq to fix and avoid using dirac up to too high a frequency. My experiment was just going to compare (subjectively) the two approaches but I hadn't appreciated that I might be messing up the speaker's phase by doing this. 

OK, two key differences to what I had in mind above but could certainly have explained more fully:

1. I was referring to a system sounding bright rather than what any graph suggested. A key point here is what we hear is not the same as what we see on a graph (our brains handle direct and reflected sound differently in the way that a standard REW frequency sweep measurement does not).
2. Where a difference is to be made I was not therefore thinking of one directly trying to fit to a measured response, but rather something more like a traditional tone-control, adjusted by ear. Or perhaps introducing a 'BBC dip' in the midrange, but again as a smooth change judged by ear.

As a side-note I have previously experimented with making time-gated measurements outside to try to measure the true speaker output independent of the room. In principle good data obtained like this would give a curve where correcting to a desired shape would make sense, I'll be honest I won't say my personal experiments in this regard were hugely successful though.