Posted 16 March 2010 - 07:59 PM
Here is some reading.
I think most of the questions about the product have been answered here, thanks to someone pasting responses from the now almost dead carsound forum. Thanks for that.
I agree that there haven't been many room correction algorithms that work well, but the one in MS-8 is a good one. There's a huge difference between room EQ in a room and room EQ in a car. Correcting response in a room with real speakers really requires two different kinds of measurements--a near field measurement of each speaker and a correction scheme for that and a second set of spatially averaged measurements for correcting room modes. In a car, we can combine this into one, since the seating positions are fixed and since the speakers are very close to the listeners (compared to a room). There's so little delay between direct sound and reflections (the time and magnitude) are so close that we hear them all as speakers. In rooms, it's important to differentiate between the two.
Can a person do a better job of tuning than MS-8? I can and I'm sure there are a few others who can too, but MS-8 has been designed to make everyone's car sound better--people like my mom who love music but don't know anything about audio--and for installers to implement. I can do a better job than MS-8 but in order to do it I need lots of bands of parametric EQ (currently I have 176 biquads available for eq and crossover), time alignment, phase shifting parametric all pass filters, separate gain control of all speakers, Logic 7, a mic array and multiplexing mic preamp, an RTA with 1/24 octave resolution, an analyzer than can measure impulse responses and phase, and a couple of weeks.
MS-8 includes all these capabilities and does about 90% of the job in about 10 minutes.
Regarding the debate between Car PC and MS-8: A car PC can include many if these tools all kludged together (except Logic 7). A car PC is like a basket full of groceries and MS-8 is like a great meal.
Anyway, the EQ in MS-8 isn't a multi-tap filter (like the Audyssey), nor is it a standard parametric EQ or a graphic EQ (like Cleansweep). It's something far more bizarre. It works great, and with 8 biquads per channel can do more work than 512 taps. It doesn't EQ phase separately from frequency magnitude, but in my experience, that isn't necessary so long as you have a center channel and a matrix or some other center signal extraction method or time alignment. MS-8 has both.
One thing that's important to remember when you're setting crossovers with conventional gear is that what appears to be a gap may, in fact, not be a gap. Here's an example (but without pictures, because I'm lazy today).
Let's say you cross your subs over at 100Hz and your mids over at 200Hz. Both slopes are 12dB/octave. When the output of the subs is precisely the same level as the output of the mids, the subs are down 3dB at 100Hz and the mids are down 3dB at 200 Hz. At 150Hz, both are down 6dB. Now, adjust the input sensitivity of the sub amp, so it sounds like you have bass. Let's say you boost it by 12dB. Now, the sub is up 12dB at maybe 80Hz and below, up 9dB at 100Hz and at 0dB at 200Hz. Now where's your crossover point?
200Hz.
MS-8 avoids this problem by providing one crossover frequency setting for the sub and the midbass, adjusting the final slopes and frequency automatically using the acoustic EQ and then providing a bass shelf filter as a subwoofer level control which is applied to all the channels through the crossover. That way, the midbass and the subwoofer get the appropriateamount of boost at the right frequencies to add bass to the system while maintaining the proper crossover point so the bass doesn't become boomy and direectional. It works great and I have that process running as a VST plug-in in my car now.
Let's make this easy and say a biquad is a filter that can be configured to be a high pass of just about any alignment; low pass of just about any alignment, parametric EQ of nearly any frequency, gain and Q; notch, high shelf; low shelf or phase shift. The MS-8 assigns the filter type and values (frequency, Q and gain) based on the measurements it makes and the algorithm (predefined process or set of instructions for making decisions written as code) that determines how the decision will be made. So, for the purposes of this discussion, MS-8 has 8 opportunities per channel to implement something that does part of the job of fixing the channel's response. The details of how it makes the decisions are proprietary, patented and too difficult for me to try to explain.
Some of it is manual and some is automatic. The user enters the crossover frequencies and assigns the channels. Then the user helps MS-8 make its measurements by placing the microphones and pressing "Go". MS-8 adjusts the EQ and, consequently, optimizes the crossovers and slopes for proper acoustic performance. Then, if you want to make adjustments, you get a 31-band EQ. The 31-band EQ is a separate set of filters that you can use to draw whatever curve you want.
Unlike most 31-band graphic EQs, the response tracks the settings precisely. What many users expect is that if they boost all the sliders by 12dB, that the response should be flat, but boosted by 12dB across the spectrum. This is almost never the case, because making the filter Qs narrow enough to do that makes the response look like a comb. Making the filters wider provides more gain than one would expect when adjacent bands are boosted. Also, adjacent band boosts and cuts are rarely executed by conventional EQs as one would expect. The math used in MS-8's 31-band EQ adjusts adjacent bands automatically so that the curve you draw is the curve you get. This is a big deal, by the way.
For those of you who have an EQ laying around, plug it into your sound card. Make it a loop-back. Generate some pink noise and look at the response as you make adjustments. You may not like what you see and it's one of the reasons that tuning with a conventional 31-band EQ and using a 31-band RTA rarely results in great sound.
The whole point of MS-8 and the point at which it differs most from every other processor that's come to market so far is that it's intended to provide a bunch of tools you can use easily to be successful in making your car sound great. It's not intended to be the tool corral at Home Depot, where almost anything is available, but it's up to you to learn how to use it. If we just took the on-chip library from the TI DSP we're using and added a GUI, this product would have been finished three years ago, but it would have been just like every other DSP EQ/Crossover. There would have been a bunch of people who can pronounce "equalizer" and who have heard the terms "Butterworth", Linkwitz-Riley" and "All-pass filter" raving about the resolution of the available adjustments, but the success rate in making cars sound great and, consequently, the sales rate for the product would have been just as dismal as every one of its predecessors.