Welcome, Guest
Username: Password: Remember me
Need help configuring your plug-ins? This is the place to ask the miniDSP community for help. Please read manual first to limit 101 questions.

NOTE: This is a "Community" forum. Please be mindful that community members are here to help as part of a community effort. We therefore appreciate your effort in keeping this forum a happy place!

If you have a specific issue (e.g. hardware, failure) and want help from our support team, please use our tech support portal (Support menu - > Contact Us).
Thanks a lot of your help in making a better community.

TOPIC: Room Curves, Cascaded Shelves and miniDSP

Room Curves, Cascaded Shelves and miniDSP 5 years 11 months ago #16563

  • Jim the Oldbie
  • Jim the Oldbie's Avatar
  • Offline
  • Gold Boarder
  • Posts: 257
  • Thank you received: 78
  • Karma: 29
Howdy Folks,

I've been messing with this for a bit, and it seems like a good topic for discussion. So I'm putting it up here to see how the rest of you are doing this, and if anyone wants to compare notes.

Like many, I've discovered that a loudspeaker that is voiced for flat on-axis response when measured anechoically (quasi- in my case) tends to sound too bright in a normal listening room. I have what I'd call a fairly near-field setup, but a flat EQ still sounds too top-heavy to me.

Among its many other useful features, Room EQ Wizard has its House Curve setting which can be applied to any measurements taken if something other than a flat response is desired. But this approach has a couple of potential issues.

First, the House Curve setting is part of the measurement target, and so becomes somewhat inseparable from the overall EQ settings. You can of course modify your House Curve setting after taking measurements, but (if I understand correctly) this then requires going back and slightly modifying all the individual adjustments that were made for room modes etc, in an attempt to reestablish the modified House Curve.

Second, and perhaps more important, adding a House Curve to your room measurements is only going to be useful for the portion of the spectrum (usually bass) where room measurements are actually used. With my own setup for example, I'm relying mainly on room measurements (and placement of course) in the bass, but sticking mostly with my previously-acquired quasi-anechoic measurements for the mids on up. I tend to agree with the opinion that room treatment is a better choice that EQ in the upper frequencies, so I don't use REW there except perhaps to check on octave-to-octave balance or something like that.

I initially made an admittedly lazy attempt to deal with this top-heavy (and bass-light) sound by stepping the driver levels - subs a bit louder, tweets a bit softer. This helped the original problem somewhat, but of course it created another fairly predictable one: It produced a pair of dips in the perceived (and actual!) response just above the crossover points. I noticed the lower one first - the system sounded more balanced overall, but it just didn't have the right... oomph. But of course - the midbass was being overshadowed by the big shelf I'd made crossing over to the subs. The upper point was more subtle, but the problem was the same - the lack of "presence" I was perceiving was caused by the shelve down at the 2 KHz mid-to-hi crossover.

I then decided to try setting up the system for a more generic flat response. Then, using miniDSP, I could experiment with overlaying various full-range room curves on top of that. And yeah, I know, what we're really talking about here is just a modern high-falutin' version of what we used to call - TONE CONTROLS!! (<gasp!> The Horror!!) But I'm going to assume that anyone reading this has by now traveled far enough down the DSP rabbit-hole to have abandoned that particular audiophile moral dilemma, among others. :o)

First, I returned the tweeter levels to their quasi-anechoic flat settings I'd spent so much time obtaining earlier. Then I EQ'd the (sealed) subs: first for a nearfield Linkwitz transform, then in-room to tame a few big modes here & there. (I now have smooth, _palpable_ bass down to around 16 Hz, whee!) I then adjusted the sub levels with the 1-octave RTA in REW to match them up with the mids, without the big shelf.

Now for some room curves. Back when I had a killer stereo in my car, I wound up with an overall curve that mostly just slanted continuously down from bottom to top at about 1.5-2 dB/octave. This looks a little nuts on paper, but it sounded pretty damn good on the road! I've been Googling this a bit, and I see that some home stereo guys are using a less-extreme version of the same idea, with slopes closer to 1 dB/oct. or less. I decided to start with this idea. To achieve these broad, gently-sloping curves, I'm using cascaded shelving EQs in miniDSP.

I've smacked up a little spreadsheet to help with the calculations. It's attached in both LibreOffice and Excel formats (just remove the .jpg extensions), but I've only tested the LibreOffice version. Here's how it works: Just enter the total shelf depth in dB (remember to use a negative number for a downward low-to-high slope), then the lower & upper endpoints. It will calculate how many filters (1, 2 or 3) are needed to get the job done, and make a crude attempt to adjust the Q value for an acceptable compromise between sharp knees and low ripple. It will then supply the individual filter frequency settings (displaying "n/a" for unused filters) and a single gain and Q setting, which is the same for all filters (as is the HIGH_SHELF setting). I've also attached a screenshot of the resulting curve produced by the intitial spreadsheet settings, taken from the PEQ box on a nanoAVR.

Notes:

1. The curves normally start out at 0 dB and go down from there. If you want to go above unity gain, the easiest thing to do is to change any one of the filters from HIGH_SHELF to LOW_SHELF, then flip the GAIN setting for that filter from negative to positive. This will move the whole curve up by the amount of the GAIN setting, for each filter modified in this manner.

2. These curves are normally applied upstream from any crossover filters, and not to individual drivers. But if you run out of input filters, you can use the output ones. Just remember to apply the same filter to every single individual driver output. This is tedious, but it works. The nanoAVR for example has nothing _but_ output filters, so these curves all have to be applied this way.

3. I'm finding that the trickiest part of this scheme is the first part, which is actually achieving an objectively flat in-room response. Seems to me, the closer you get, the more effective and predictable the overlaid room curve will be. I'm sure I still don't have mine exactly right, but I'm working on it. Meanwhile, this concept is definitely working better than stepped driver levels, to say the least. The midbass in particular has returned, and the music ROCKS when it's supposed to! If you have any interesting curves, please share!

4. Last but not least, if anyone knows of a simpler way to do this sort of thing, I'm all ears! :o)

-- Jim
Attachments:
Last Edit: 5 years 11 months ago by Jim the Oldbie.
The administrator has disabled public write access.

Room Curves, Cascaded Shelves and miniDSP 5 years 11 months ago #16576

  • dreite
  • dreite's Avatar
  • Offline
  • Platinum Boarder
  • Posts: 1416
  • Thank you received: 664
  • Karma: 83
Handy.
This allows to program "tilt"ing equalizations over a wider bandwidth than would be achievable with a single filter.
The 16db limitation doesn't allow EQ's to achieve a full decade coverage, but miniDSP does allow shelving filters of less-than-first-order (Q < 0.5) which sort of approximates it.

Generally, the only time you would need to cascade shelving filters is to create EQ's greater than 16db.

Cheers,

Dave.
The administrator has disabled public write access.
The following user(s) said Thank You: Jim the Oldbie

Room Curves, Cascaded Shelves and miniDSP 5 years 11 months ago #16578

  • Jim the Oldbie
  • Jim the Oldbie's Avatar
  • Offline
  • Gold Boarder
  • Posts: 257
  • Thank you received: 78
  • Karma: 29
dreite wrote:
...miniDSP does allow shelving filters of less-than-first-order (Q < 0.5) which sort of approximates it.

I know the Basic PEQ box won't allow me to enter a Q value lower than 0.5, but I just discovered that the biquad calculator spreadsheet here does indeed allow lower numbers. It seems if we go too far below 0.5, the separation of the poles eventually begins to introduce a fair bit of ripple; but this is still a useful and more economical approach up to that point. I wasn't sure it was even possible.

EDIT: On second look, the ripple there isn't as bad as I thought; it's just magnified by the scaling of the graph.

Generally, the only time you would need to cascade shelving filters is to create EQ's greater than 16db.

Thanks for your reply, Dave. I figured I was at least partially reinventing the wheel here.

-- Jim
Last Edit: 5 years 11 months ago by Jim the Oldbie.
The administrator has disabled public write access.

Room Curves, Cascaded Shelves and miniDSP 5 years 11 months ago #16581

  • dreite
  • dreite's Avatar
  • Offline
  • Platinum Boarder
  • Posts: 1416
  • Thank you received: 664
  • Karma: 83
The 2wayADV plugin allows to enter Q's to 0.3, but some of the other plugins are limited to 0.5 (first-order.) But, as you say, you can utilize the bi-quad calculator and enter coefficients directly to create some interesting curves. Use whatever works to achieve your objective. :)

Dave.
The administrator has disabled public write access.

Room Curves, Cascaded Shelves and miniDSP 8 months 2 weeks ago #46459

  • rup40
  • rup40's Avatar
  • Offline
  • Fresh Boarder
  • Posts: 3
  • Thank you received: 1
  • Karma: 1
Jim, thanks for this calculator. I'm using the same approach of applying the curve on input channels. Were you able to find a better approach? I would like the midpoint of 1d/Octave vurve to be centered at 1K, is it possible?
The administrator has disabled public write access.
The following user(s) said Thank You: Jim the Oldbie

Room Curves, Cascaded Shelves and miniDSP 8 months 2 weeks ago #46460

  • Jim the Oldbie
  • Jim the Oldbie's Avatar
  • Offline
  • Gold Boarder
  • Posts: 257
  • Thank you received: 78
  • Karma: 29
Wow, haven't seen this thread in quite awhile!

Your 1KHz "fulcrum" is a good idea. I'm on the road most of this week but as soon as I can, I'll revisit that old spreadsheet. I'm guessing there's a way to do this.
The administrator has disabled public write access.
The following user(s) said Thank You: rup40

Room Curves, Cascaded Shelves and miniDSP 8 months 2 weeks ago #46470

  • rup40
  • rup40's Avatar
  • Offline
  • Fresh Boarder
  • Posts: 3
  • Thank you received: 1
  • Karma: 1
Meanwhile, I was able to do 1 dB/octave slope with 1K turnaround point

This message has an attachment image.
Please log in or register to see it.

The administrator has disabled public write access.

Room Curves, Cascaded Shelves and miniDSP 8 months 2 weeks ago #46482

  • Jim the Oldbie
  • Jim the Oldbie's Avatar
  • Offline
  • Gold Boarder
  • Posts: 257
  • Thank you received: 78
  • Karma: 29
Your approach above is working of course, but unfortunately uses up a lot of IIR filters to accomplish the slope.

I still haven't had time to look at my spreadsheet, but I did remember why I didn't choose 1 KHz as the "teeter-totter" point: it was to avoid clipping on the low end. Since part of the plan is to establish endpoint frequencies, I decided just to use them as references - the gain at 1 KHz is arbitrary in this case anyway, since it would need to vary downward from 0 dB according to the slope and/or endpoint settings.

Speaking of which, you might want to consider setting your endpoints away from the extreme ends of the spectrum a bit. For example, I found that my room gain adds sufficiently to the extreme low end such that I was able to set the low endpoint of my curve to 32 Hz, and still achieve nice measured response right down to 16 Hz. Of course, my subs use sealed enclosures which helps in this regard (12 dB/oct rolloff vs 24dB for vented enclosures). If you're using ported subs, you'll probably want to stop boosting below the port tuning frequency. I also think that moving the endpoint up might help avoid excess group delay way down there, although I'm not sure about this.

At the top end, I've got my upper endpoint set to 16 KHz. Since my tweeters start to fade at about that point, I figured there's no point in having the house curve doing any further cutting in that top region.

Remember that the closer the endpoints, the fewer cascaded filters are required to form a reasonably ripple-free curve.

Hope this helps.
Last Edit: 8 months 2 weeks ago by Jim the Oldbie.
The administrator has disabled public write access.
The following user(s) said Thank You: rup40

Room Curves, Cascaded Shelves and miniDSP 8 months 2 weeks ago #46485

  • dreite
  • dreite's Avatar
  • Offline
  • Platinum Boarder
  • Posts: 1416
  • Thank you received: 664
  • Karma: 83
Yeah, a lot of filters. :)

This topic has been visited previously. Just three shelving filters is all that's required. (Just adjust the gains and F's to create whatever you want.)
www.minidsp.com/forum/equalization/11325...ccomplish-this#21116

Dave.
The administrator has disabled public write access.

Room Curves, Cascaded Shelves and miniDSP 8 months 2 weeks ago #46486

  • Jim the Oldbie
  • Jim the Oldbie's Avatar
  • Offline
  • Gold Boarder
  • Posts: 257
  • Thank you received: 78
  • Karma: 29
dreite wrote:
...This topic has been visited previously...

Yes, 5 years ago in this very thread in fact... :blink:
The administrator has disabled public write access.
Moderators: devteam