Yes, absolutely. I was hoping someone would ask.
Here is the framework: Start with calibrated hearing threshold results spanning at least from 125Hz to 20kHz for a listener. Test more frequencies in the spectrum, not less. 4 frequencies does not provide enough information. Currently, the software uses template profiles and will move to user test results soon. Next, compare the test results against a reference, which is full spectrum from 0-20kHz (truly nyquist), to derive the users correction curves.
Here is a plot of my test results against the reference curve.
View attachment 355178
Currently, this is completed in a middle/sides optimization stage, not on the left and right channels. I've included two models of this optimization step, which apply different M/S models behind the "character" toggle. This will be fully tunable by the user and allow for additional modes, such as a decoupled stereo mode and a linked stereo mode. There are additional optimization steps governing the application of shared regions of hearing loss and those not shared. This is important for stereo or multichannel listening, as we are now listening in the environment and HRTFs become important considerations.
Next the filters are optimized or deisgned. Generally, we are tying to minimize the error between predicted and target response. Currently, there are both IIR and FIR embodiements with this one being an IIR topology. FIR is next and also presents optionality for things like phase specification. With IIR filters, operating rate can have dramatic effects on sound, so this is another tunable parameter based on what you feed in or if you use the "oversample" mode.
This is a conveince feature. If you really want to experiment with different sample rates, turn it off and try different audio files or resample with something better--HQPlayer or Roon are options.
Apply the correction with the designed filters and then cascade (or remove) the dynamics and tuning sections. These aim to 1) control true peaks to prevent clipping in a musical way (user defines musical way, but we have some suggestions); 2) allow adjustment of the imaging post correction, as applying correction will modify the imaging perceived. These processing stages are optional, should one want to use just the filters. Within the dynamic section you can obviously control how the dynamics are applied, if at all, with the threshold, attack, release, ratio and knee parameters.
Re: Peace Equalizer
I'm not clear on what kind of filters they are using or how they handle freqs above 12.5kHz, as ELC is only defined to this point. It looks quite flexible, however, and seems to have some filter topology optionality, which I'm a fan of. I'm on mac most of the day, but I am going to give this a try. Here's my thought: audition using a common set of test results. If you want to try, let me know and I will share my tabulated results to compare in Peace and in my software--hearing profile #0.
One aim of mine is portability of test results and this is a strong use case.