Originally Posted by ansuser View Post
So, what we see here is a low power DSP with limited number of taps and some sophisticated measurement technique (pls. correct me if I wrong)

Hanatsu, can you please clarify some points:
1. Do I need to measure power response in full sphere around the headrest or, for example, in some plane with center at the headrest? Will changing measurement points field alter my results?

2. Windowed measurements in the car are almost useless, and with 200 ms window I don't really see how it can be different from simple pink noise "moving mic" technique

3. What about FR in the listening point using good old REW with single point or averaged? Is it comparable to what you were able to achieve before?

4. FIR filters are great for managing phase. Can you control phase with this device? Are implemented filters minimum phase or linear phase? Square wave response plots would be great to see.

5. Open source FIR filter-generation tools are well developed and tested. What exactly separates APL from them (except hardware, obviously)?
4096 FIR-taps is quite adequate imo. Most car audio DSPs doesn't use FIR based processing at all. Those which use FIR, doesn't really have an accurate method of obtaining a correct power response. Alpine H800 is probably what comes closest but the how the Imprint calculation algorithm performs is questionable.

1. You measure the entire listening space, from the left side window to the right window, from dash height to the roof, all the way from the dash to the seated position. So 'everywhere' in the listening space where the main reflections occur.

2. 200ms is required to get enough resolution in low frequencies, you can choose a much smaller window if you like. I'm fully aware of the limitations and usability of gated measurements in a car. As I'm not the designer of the unit or software I can't give you a well educated answer, I can only observe and review the outcome.

3. I posted graphs from RoomEQ measured using pink noise/RTA, using a "traditional" measuring method in my first post. I can't EQ the response to equal that of the APL. I could get it close but it did not sound or stage the same. I believe it has to do with the summed response. Even if the L/R response is close to that APL outputs the summed response look very different with peaks and dips all over the place. APL does delay/phase corrections as well.

4. Believe it performs minimum phase correction. It definitely "clears up" the the time domain. While I'm not sure, I believe it performs time alignment corrections at the same time as amplitude correction. I leave Raimonds to answer that properly though.

5. I've tried multiple FIR based DSP/computer software able to generate FIR filters. I've messed around with the openDRC miniDSPs with rePhase and a few other programs. They have not impressed me even after extensive experimentation. Perhaps I should not blame the software, it does what it's supposed to do, it's the way you obtain the data. APL uses multiple point sweeps which derives data from an IR. A noise generator coupled with an RTA won't be able to display any data about phase/time domain, which APL does and corrects for. What APL does better, much better than other software I've tried is the way it measures, averages and derives the data from measurements. Again, I don't know the specifics how it works, I can only tell you that it does. It's much more user friendly than most other solutions with similar properties.