Hi Drizt, good to see you're onto this! You'll be overwhelmed once you realise the optimisation potential acoustic measurements offer.
- Is berun12 right in suggesting you measured both speakers at the same time? As he said - never do this. Treat each side fully independently, but work towards achieving identical transfer function at the LP from both sides. This normally means different treatment for both sides.
- Don't use 1/3 oct smoothing. You cover all important details. 1/20 oct is ok to remove noise from the FFT and retains all the detail. I tend to look at a FR at different smoothing widths to better understand details and global trends.
- Interesting is the comparison LP vs 30cm front/top offset. To interpret fully you should look at the impulse responses (especially the first 20ms), the energy-time curves and the waterfall plots.
- You showed the big dip is a room interaction - but it's definitely not good, you will perceive it as a coloration. The causes are manifold, but I'm confident that optimisation of speaker and LP position plus selective treatment of problems arising from reflections will allow improvement.
Although I'm not endorsing dead room designs such as yours, investigations are way easier. The following dot points are about how I would tackle a room such as yours. In a nutshell, I recommend that you start optimisation with the room mode dependencies and work out the placement of your main speakers and the LP:
- Take an inverse measurement approach - place the main speaker at the LP (midspan between tweeter and midrange driver at ear height) and measure the transfer function at the possible speaker positions with microphone at speaker midspan height. Take measurements at a grid of say 30cm over an area as wide as 1m and as deep as 2m for each speaker position (mark mic positions with tape on the floor).
- Then shift the LP forth and back in say 30cm increments and repeat the full set of measurements.
- Import all in Excel, discard all frequency bins above say 400Hz (that's all reflection dependent), calculate the magnitude, and normalise so that the mean is zero dB. Calculate the variance from zero mean for each response, and rank the responses based on the variance. Plot the 10..15 best responses.
You have now identified a range of positions resulting in the flattest frequency response in the modal frequency range. Inspection of the transfer functions and their waterfall plots should will reveal the Schroeder frequency. Optimise the LF response at the LP using narrow parametric EQs. Remember you are in the modal region - assuming that the speakers have constant power output with near monopole source directivity, any dips in the FR at the LP are due to fluctuations in the radiation impedance of the driver. The room-speaker system is thus minimum phase and EQ can be applied to correct!
Integrate your subs. I suggest inverse measurements again to position the subs. Asymmetric placement has higher chances of achieving near constant FR at the LP. Use a procedure such as Welti & Devantier propose to optimise the seat-to-seat variances.
Undertake meaningful nearfield measurements to achieve matching acoustic transfer functions for the subs and the main speakers. Otherwise you'll be struggling to achieve coherent summation. Pay attention to phase alignment - you'll need to delay the main speakers to achieve coherence with the subs. A word of warning - 80Hz crossover frequency means that the bandwidth of the subs is very small, i.e. the group delay and risk of poor transient response becomes very high (especially with sub designs of order higher than 2) . Acoustic filter design for the subs is something that makes all the difference in terms of LF perception. I suggest leaving that topic for a later discussion.
Your LF response should now be sorted. Shift focus to high frequencies:
Proceed investigating the effects of reflections on the transfer function at higher frequencies. Using ETC and trigonometry you can identify where strong reflections come from (simple in your case - dominant ones come from the ceiling with minor ones from the floor and rear). All important is the level difference between the first and subsequent peaks. Your ETC shouldn't have any dominant early reflections higher than -15dB compared to the initial peak. The ceiling one may require diffusion to reduce its level - it doesn't strongly contribute to spaciousness but may mess with the coloration of the reverberant signal (I assume you got a suspended plasterboard ceiling which absorbs LF). Floor reflections should be mitigated using heavy felt underlay and carpet.
I would then proceed to investigate the FR differences between the two speakers at the LR and use targeted EQ above say 300Hz to minimise these. You therewith increase the coherence between the two channels - you will find that phasing due to head movements at the LP is significantly reduced. Take care when interpreting your FRs - change smoothing from 1/20 to 1/6 to 1 oct to tackle large trends first and narrow deviations second. I know many people object EQs in the signal chain for audiophile reasons, but the gains outweigh potential signal degradation.
Your low reverberation room design should now show an ETC with no dominant reflections, waterfalls without dominant ringing, and FR that is essentially flat up to 14kHz. Bear in mind that 3dB sound pressure level difference corresponds to halving the acoustic power.
Given your low reverberation time and consequent absorption of all ER, I would investigate the IACC / ASW / LER to check whether you have enough reverberant energy left in the sound field to achieve the all important perception: immersion and source widening! Experiment with 1m2 reflector panels at the lateral reflection points and see how this affects your perception. And sit down in the room with the missus and a bottle of shiraz and talk with no background music. When you both are comfortable with the acoustic quality of the conversation (intelligibility, ease of communication, apparent liveliness, lack of background noise level) than your room design is good. If you feel slightly uncomfortable, you need to re-work.
Edited by svenr, 18 December 2011 - 10:06 PM.
