Page 1 of 3 123 LastLast
Results 1 to 10 of 26

Thread: Tuning from scratch

  1. #1
    Tester Extraordinaire ErinH's Avatar
    Join Date
    Aug 2010
    Posts
    1,247

    Tuning from scratch

    I hesitate to call this a tutorial, but it is a documentation of my next tuning foray. I've been talking about doing a video of the RTA process and it'll probably be rolled up in to this 'tutorial'. I will not be taking the time to explain how to build an RTA. There are numerous threads on that already and frankly, I just don't have the time. If I tried to address that, I'd never get this thread finished.

    I have decided to start from scratch with my tune. The system sounds pretty darn good as it sits but I want to play around with some things and starting from scratch will be easiest to see what I need/want to see. The reason why is because I often think of things I need to look in to or try and, as most know, once you make an adjustment to the system, it's kind of hard to really know what is affected and that rabbit hole often leads you to an ending where you're left scratching your head, trying to figure out how you got there. So, with that said, I'm going to try to put some method to the madness this time and log my methods and results as I go. I don't expect this to be a full up tutorial on how to tune; I've not yet gotten to the point yet where I feel comfortable doing that. However, I will do my best to make not of things and post results as I go.


    First off: figuring out where to start.

    Usually we just jump in and start messing with gains or T/A. I've not ever had much luck there and I think I know why...
    Over the years I've learned that T/A is VERY hard to set unless you have left and right side frequency response pretty well aligned in terms of level matching. The reason is pretty simple: I have adjusted time alignment in increments way beyond what I should have adjusted and still aspects of a soundstage pull to one side or the other (or behind you). As an example: If I want to center up the vocalist, usually we just apply some t/a and levels... but if there is an annoying band or two in the frequency response that overrides another, I can delay the left side all I want, but the left side still takes precedent. So, then I lower the gain... but something still pulls to the left. Often I find it's either a reflection based issue or an issue in the FR itself (due to the speaker or the install) and the localization of one particular frequency/band overrides everything else I do in a means to align the stage. Therefore, it's a worthless attempt to center the stage because I'm having to hear 'through' the problem frequency(s).

    So, I usually start off with response matching; which includes level matching to get a broadband response close and then tweaking with the EQ to level out any major peaks in response. Taking it a bit further, though, I will first focus on modal issues. Because, modal issues are a function of the cabin's pressurization, they will not change no matter what kind of level matching or t/a you do. Plain and simple: they suck. And until they're really gone, it's very hard to get a good sound in the car. I spent a long time working to tame them and had to give up a lot to do it because I only had a 1/3 octave eq; a parametric really is needed because the Q of modes can be as high (or higher) than 10.0, whereas a G-EQ only can adjust in Q of about 4.0.



    Below is a summary of the above which will help me in this thread and help you guys understand where I'm at in the tuning process. It'll change a bit as I go, but this gives me a good footing to start from and is better than my notebook. ;)

    1. Taming Modal Issues
      • Using the RTA to determine what is a modal resonance and what is just ol' phase related issues. Long sweeps, used purposely to ignite the room and cause reverb.
      • Using the Parametric EQ to tame these modes
    2. Level Matching with the mic and amp gains
      • Noting noise floor
      • Careful to not to clip the output signal (badly)

      • Broadband (overall levels to get within some tolerance)
      • Specific Band (L/R response matching using the graphic eq to nail down blips)
    3. Crossover Setting (HD, polar, and phase)
    4. Time Alignment
    5. Revisiting L/R EQ
    6. Cabin gain and the Target Curve
    7. Fine Tuning (by ear, by mic)




    I'll probably think of more to add later and as I go, the details will come with it.

    Not sure how long this will take. Doing it on my own, I can fully tune the car in a day. But, stopping to take photos, video, and document the steps is going to make it take a bit longer. Hopefully it'll help others and get some ideas going on further tuning your own system.



    - Erin
    Last edited by ErinH; 06-18-2012 at 09:26 PM.
    Your ears: The best tools you have... and they're free, too!

  2. #2
    Senior Member cvjoint's Avatar
    Join Date
    Mar 2011
    Location
    Northern California
    Posts
    801
    Real Name
    George
    cool, waiting for the pics!

  3. #3
    Senior Member
    Join Date
    Oct 2010
    Posts
    169
    Erin, you have measurement gear, and you're good at using it and interpreting the data. So why on earth are you setting your T/A by ear? I think it's the first thing that should be set because it's the only independent variable. For those who don't have measurement tools, I usually recommend at least approximating it with distance measurements and calculations, because altering it after the fact has an impact on every other adjustment. So doing T/A first basically saves you an iteration.

    I also blend steps 2 and 3 on your list. I identify modes and phase issues initially like you do, but then I try to adjust the crossover parameters to try to account for them without having to rely on the EQ. You can move offending frequencies into different bands, or center the crossover near the problem areas so that you can play with Q and slope to try to take care of them. Within reason, of course.

  4. #4
    Founding Member OSN's Avatar
    Join Date
    Aug 2010
    Location
    Philly
    Posts
    252
    Subscribed!

  5. #5
    Tester Extraordinaire ErinH's Avatar
    Join Date
    Aug 2010
    Posts
    1,247
    Quote Originally Posted by MarkZ View Post
    Erin, you have measurement gear, and you're good at using it and interpreting the data. So why on earth are you setting your T/A by ear? I think it's the first thing that should be set because it's the only independent variable. For those who don't have measurement tools, I usually recommend at least approximating it with distance measurements and calculations, because altering it after the fact has an impact on every other adjustment. So doing T/A first basically saves you an iteration.

    I also blend steps 2 and 3 on your list. I identify modes and phase issues initially like you do, but then I try to adjust the crossover parameters to try to account for them without having to rely on the EQ. You can move offending frequencies into different bands, or center the crossover near the problem areas so that you can play with Q and slope to try to take care of them. Within reason, of course.

    I've tuned via mic many different ways and setting T/A with it a lot. The problem, getting back to my OP, is the 'ear' part. When I get in to listen, it's hard to hear past the level issues. So, that's why I'm a fan of leveling everything first. Then it's easier to adjust T/A (in any method) and understand if it's helping or not. I tune as an iterative process. I used to go in blind and just tune away via RTA but I've since learned that's not very wise. So, I'll do some things and then jump in to make sure nothing is off.

    FWIW, the last time I did alignment with the mic (about 2 months ago), I tried to see if doing a spatial average changed anything. I measured each speaker's impulse in 3 different positions: left, center, and right. All were about 2-3" apart. The results were pretty much identical for every measurement. I thought that was kind of interesting.


    As far as crossover to avoid resonance, I've not had much luck doing this. It seems to imply that you can just shift the crossover up or down to kill the mode, but if that's the case, then you may also be verging on creating distortion via a lower high pass. Do what works best for you, and we all are pretty competent here. I'm just thinking of the 'other side' of the ball in this scenario. Like you said... within reason. I have found that leaving a gap in certain areas (namely, the 200-300hz range in my car) helps to absolve some nasties. It's interesting to me that so many try to cross a midrange very low, yet that's often where most cars are very boisterous and provide their own output. It makes sense to me to just leave a gap there (as long as there's no glaring reason why you shouldn't) and let the car do it's thing while you enjoy the benefit of crossing a bit higher and lowering distortion.
    Last edited by ErinH; 06-19-2012 at 06:57 PM.
    Your ears: The best tools you have... and they're free, too!

  6. #6
    Senior Member
    Join Date
    Oct 2010
    Posts
    169
    Quote Originally Posted by ErinH View Post
    I've tuned via mic many different ways and setting T/A with it a lot. The problem, getting back to my OP, is the 'ear' part. When I get in to listen, it's hard to hear past the level issues. So, that's why I'm a fan of leveling everything first. Then it's easier to adjust T/A (in any method) and understand if it's helping or not.

    FWIW, the last time I did alignment with the mic (about 2 months ago), I tried to see if doing a spatial average changed anything. I measured each speaker's impulse in 3 different positions: left, center, and right. All were about 2-3" apart. The results were pretty much identical for every measurement. I thought that was kind of interesting.
    I guess it will depend on the properties of the software you're using to do it. That's going to have an impact on precision. Three inches should give you a delay of about 200us, right?

    In the thread I started the other day, you can see how reproducible the impulse responses are in my recordings (and this is without a mic stand and with the neighbor mowing his lawn, btw). Have you been able to verify reproducibility across repeated measurements? When the measurement noise is small, you can align recorded waveforms and get accuracy down to about a single sample fairly easily (less if you use interpolation). My distributions of delays end up pretty narrow, so that my delay resolution ends up somewhere in the ballpark of 100us. Which isn't great, but probably good enough. Under ideal conditions, the human brain is capable of performing ITD discriminations at 10us, so that should be the target resolution, although I think you'd have a hard time hearing the difference in the car.


    As far as crossover to avoid resonance, I've not had much luck doing this. It seems to imply that you can just shift the crossover up or down to kill the mode, but if that's the case, then you may also be verging on creating distortion via a lower high pass. Do what works best for you, and we all are pretty competent here. I'm just thinking of the 'other side' of the ball in this scenario. Like you said... within reason. I have found that leaving a gap in certain areas (namely, the 200-300hz range in my car) helps to absolve some nasties.
    Yeah. I was just pointing out that the crossover is basically an EQ (in fact, my crossover VST is the same as my EQ VST...), so you can kill two birds with one stone. It seems silly to have a crossover at, say, 100Hz and then an EQ dip or boost at 150Hz, when a lot of times you can just move the crossover point a little bit, and adjust the slope and Q in a way to maintain the phase coherence between drivers while dealing with the issue at 150Hz. This is where processing power comes in handy.

    I'm a firm believer in individual crossover parameters for each speaker, and there are four of them: level, f, Q, order. The first one is obvious; the second one is accepted by most; but some people don't like to adjust the last two individually because they're afraid that the transition will be different between drivers and that phase coherence can become a problem. Other people (like Andy and you?) don't think that phase coherence is an issue if the amplitude response is cool. I haven't read enough or experimented enough to know how audible phase coherence is in the absence of amplitude differences, but I think it can be corrected sometimes with a phase knob anyway.

  7. #7
    Tester Extraordinaire ErinH's Avatar
    Join Date
    Aug 2010
    Posts
    1,247
    I've not really tested repeatability per se... but I have ran multiple measurements just to make sure I wasn't crazy. ;)

    Last time, I started with the furthest driver, measured it with aRTA and saved it as my reference. I then played the other speakers and adjusted the delay until it lined up with the furthest speaker (which, IIRC, is my right midbass). Made it all go pretty easy. That's one plus for having a computer GUI for the dsp (or a carPC... cheater).


    I'm not really sure I don't buy in to phase coherance as an issue or not. I'm just more of a proponent on getting crossover points set pretty well within reason and putting importance on the magnitude side of things. It's the same means to an end, I suppose. I just think that people are now getting caught up in measuring 'phase' when, at least in the car, response is stupid and measuring/correcting phase is futile at higher bands. I do certainly use the crossover points and measure them to make sure I'm not creating nulls (especially in the sub/midbass/midrange drivers area) but I don't set out to specifically measure phase. If that even makes sense.
    Your ears: The best tools you have... and they're free, too!

  8. #8
    Senior Member
    Join Date
    Oct 2010
    Posts
    169
    Yeah, my program is a little different in that it computes delay directly, and doesn't rely on the user to line things up, which improves resolution. You play a set of alternating impulses (recorded as a wav on the L & R channels -- so it requires that you can map L & R to any arbitrary pair of speakers). And then the program collects the impulse responses at a period of 2 * (T +/- delay), assumes that every other impulse belongs to speaker 1 and speaker 2, respectively, and then aligns the waveforms and forms a distribution of delays based on the time intervals between pulses. Make sense? [The alignment part is actually the tough part to code, at least when the S/N is high]

    If I can find a VST impulse generator that will allow me to route alternating impulses to 6 channels at a time, I could bang it out all at once. For now, it's a chore having to do this a bunch of times to T/A the whole system.
    Last edited by MarkZ; 06-19-2012 at 08:19 PM.

  9. #9
    Founding Member pocket5s's Avatar
    Join Date
    Jan 2012
    Location
    Oklahoma
    Posts
    35
    Real Name
    Robert
    When you say level matching with the mic, are you looking for a specific rta output or db matching?
    Last edited by pocket5s; 06-19-2012 at 10:39 PM. Reason: Grammar

  10. #10
    Tester Extraordinaire ErinH's Avatar
    Join Date
    Aug 2010
    Posts
    1,247
    just matching the left and right response. no real target, per se, but I only make attenuation adjustments. no boosting
    Your ears: The best tools you have... and they're free, too!

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •