Advice on creating a graph for iMac 17,1 (Prelim with UMIK-1 attached)

[taprobane99]

Setup:

late 2015 iMac 17,1 on a glass desk (with 4 downfiring-speakers). UMIK-1 on desk, horizontal orientation, at 50cm from iMac screen

Motive:

Personal - to flatten speaker response on my iMac in Linux
More broadly - this should help the Asahi project when you need to tune the M3 iMac which also has large downfiring speakers.

Attached:

1. Screenshot of all 4 speaker responses measured with UMIK - purple/red = woofers, yellow/blue = tweeters
2. Screenshot of combined response using my pipewire graph
3. Pipewire graph
4. Audacity waveform/spectrogram of left tweeter FIR (x axis seconds)

Questions

1. What are the best settings in the REW EQ window? Should I use a match range of 100Hz-20kHz or something shorter?
2. I had to use a gain of 3 on the tweeter FIRs, is there a way to avoid this?
3. What do you think of my filter chain parameters (attached as a text file)?

j413-imacmod-public.txt

[taprobane99]

I managed to flatten the iMac speakers fairly well but did some things very different to the Asahi approach:

• Create FIRs for each speaker
• Peaking filters at 850 Hz and 3.5 kHz to boost and attenuate those freqs
• Crossover (24 dB Linkwitz-Riley) filter at 3.8 kHz
• Peaking filters on tweeters at 5.2 kHz and 8.4 kHz
• Woofer FIRs weighted to gain 0.4

Any comments? Graph attached

j413-imacmod.txt

[chadmed]

Sorry, this is not good enough. Those peaks and troughs will be extremely noticeable at +/- 7/8 dB in the worst cases. Keep in mind that 3 dB is a doubling in perceptual volume. You ideally do not want more than 3 dB variation across the spectrum, at least between 200 and ~17,000 Hz. It's okay (and pretty much unavoidable) to have some rolloff outside this band, and that is what Bankstown is for. The rolloff needs to be gradual and smooth though, there cannot be dead zones or peaks in the rolloff.

Can you explain why you made the design choices which differ from our existing approach?

[taprobane99]

If I just use the FIR filters alone the response is far less flat.

I am using the following settings in the REW Wizard EQ window:
Target: Full range speaker
Match range: 1 kHz-20kHz (Tweeters), 80 Hz-4kHz (Woofers)
Max Boost: 6 dB
Flatness target: 2 dB

Perhaps I need to be tweaking the EQ filters manually rather than relying on the auto "Match response to target"?

I also found this info on macrumors from someone who built a physical crossover for the iMac speakers:

I did a frequency sweep of the tweeter, and there is zero output just under 2KHz-ish, but a +3-6dB uplift from 2.5-3.8KHz, so I thought to put the crossover at ~3.8KHz, so the louder peak below that would be taken down in the 1st order crossover point roll-off.

I'm not understanding how Asahi Audio sets a crossover frequency without using an actual crossover filter as I have done?

[chadmed]

I set the crossover frequency to be the point where the woofer's response starts to fall off, and try to get the tweeter and woofer to meet at ~ -3dB off the target.

And yes, using the automatic function is probably the big mistake. Unfortunately it doesn't work very well, so you'll have to manually go through and set up a combination of crossovers, peaking filters and shelf filters. I also usually turn off the match range/target stuff and just give myself a flat target dB somewhere around the woofer's flattest passband, then try to match everything to that.

[taprobane99]

I'm now following exactly the Asahi audio method. Manually creating combinations of crossover/peaking/shelf and using your graph structure. I'm not sure my result is that much better, although it does sound great!

It would definitely be helpful if you could share a REW Wizard .mdat workspace file that shows the EQ window filter settings used for each of the drivers in a Macbook?