AVS › AVS Forum › Audio › DIY Speakers and Subs › REW AutoEQ + Behringer DCX2496 --- odd that it's not making a flat FR?
New Posts  All Forums:Forum Nav:

REW AutoEQ + Behringer DCX2496 --- odd that it's not making a flat FR?

post #1 of 15
7/9/12 at 9:35pm
Thread Starter 
So I'm borrowing a buddy's DCX2496 and trying to learn it. I'm a bit confused because I'm not having any luck with the using the REW autoEQ function to provide parametric EQs that actually work? I've used the REW autoEQ function to quick success in the Inuke DSP 3000 at KCNitro's house - but not so far with the DCX2496? I must be doing something wrong but what? Either I'm oversighting something - or the Inuke DSP is more accurate than the DCX2496? It's not much better after applying the REW autoeq suggestions than it is naked FR with no EQ?

If these pictures aren't showing up well you can go directly to the photobucket album

REWDCX2496autoEQfunction.jpg

I tried twice to use the REW AutoEQ function. I even chose DCX2496 to make the filters in REW. I've done quite a bit better than either result just working out the EQ manually on the Inuke DSP 3000. I could just try to manually create the filters and do so in a half hour, but when it works this auto EQ function really rocks and is fast --- so I'm trying to get it to work. As it is -- my own manual EQs are better. rolleyes.gif


All FR captures are unsmoothed:
Blue is raw captivator response in my room.
Orange is Inuke DSP 3000 with me manually plugging in three or four parametric EQs (this is an old capture and the caps were facing sideways for this FR capture -- in all other responses they are toed in facing forward)
Red is REW Auto EQ
Green is MIC2200 manual EQ. One parametric EQ per channel. This is also an old EQ save I pulled up for comparison's purpose.

The Blue and Red were both taken tonight.

I've offset them for easier visibility for each graph.
AutoEQthroughREWnotworkingsohot.jpg

I tried making the tolerance only 1dB instead of 3dB and the resulting FR was actually worse with nine filters than it is with the five filters from the 3dB tolerance shown in the screenprint below. It's weird to me that the auto eq function doesn't seem to be flattening out properly? Anyone see anything obvious that I'm doing wrong on my settings?

Here are the REW AutoEQ settings I've employeed and the DCX2496 filter option radio button in the menus

REWAutoEQsettings.png

DCX2496EQsettings.png

Obviously the omnimic was in the same spot for all of these measurements.
Edited by Archaea - 7/9/12 at 10:47pm
Reply
Reply

AVS Top Picks

post #2 of 15
7/9/12 at 9:48pm
Thread Starter 
Post 102 in this thread shows some previous success of mine with REW autoEQ function and an INuke DSP 3000 amp on a couple Dayton Titanic MK3 drivers in sealed cabs.

http://www.avsforum.com/t/1415823/2-dayton-titanic-mk-iii-15s-or-1-tc-sounds-lms-r-15/100_50#post_22186473

DaytonTitanicMKIIIFROverlay.jpg
Reply
Reply
post #3 of 15
7/9/12 at 10:06pm
How are you setting the "Target Level" inside the EQ section of REW? Did levels change at some point during this process? The red auto-eq'd graph is tons higher than your blue graph but all your filters are cuts so I would assume you changed levels at some point. If that's the case it's hard to tell what your target level should be. In most cases just clicking the "Set Target Level" and letting REW handle it will get you what you want but there are some cases (like the FR stuff you were doing the other day) where it's advantageous to set your own target for one reason or another.
Reply
Reply
post #4 of 15
7/9/12 at 10:17pm
Thread Starter 
On the omnimic screenprint with four FR's the levels are arbitrary..I used the omnimic dB offset feature to seperate them out so they were easier to see. The screenprint with two FR's overlaid is the better one to look at. It too is offset to make them match and show what the REW AutoEQ did to my FR.

I was told initially that I should set the target level dB in REW at or near the bottom of the lowest point so that most everything would be cuts. Do you agree with that? So I set it to 70db which represented close to the very bottom point in the typical subwoofer FR graph - the auto EQ function prompted me that 93% of the EQ function would be cuts --- did I want to continue? I said yes.
Reply
Reply
post #5 of 15
7/9/12 at 10:28pm
That might perhaps be an ok rule of thumb for someone not get themselves in trouble, but in general I don't agree with that, no. Take an extreme example: Say you had a curve that was flat between 20 and 80 Hz at 75dB. Say you set your target curve to 70dB and REW matched it. Say that makes your subwoofer trim 0 when you do your receiver/test tone/whatever calibration. Now, say you set the target level to 80dB instead, so you're shelf-filtering the whole range up by 5dB. However, now when you go to level match your subwoofer trim is going to be -5dB, meaning you're pulling the same exact total power out of the amp and demanding the same excursion from the sub in both cases.

In actuality it's not that simple, as you'd have to gotta watch max power draw and max excursion and etc if you go boosting the natural response down low. However, REW gives you parameters for both individual filters and max overall gains that you can play with. I think you're also well experienced in knowing what your Caps can and can't do so I wouldn't see the problem at letting REW set the target level and go to work. I'm sure one of the folks more knowledgable about REW could comment more effectively in terms of how it decides when things are nulls/dips and can't be EQ'd. There're also times when I've found that if I set the target level too low it seems to not want to reduce the peaks (especially if they spread across a range of frequencies).

Another thing I've seen people do when they're concerned about boost (especially down low) is to start out by matching the target to whatever they're getting at the low end (20Hz or whatever point you want to use). However, since you've got that peak around tuning I don't think that's going to do anything positive for you either.
Reply
Reply
post #6 of 15
7/9/12 at 10:38pm
Also, in my very limited experience I've found that you don't gain anything (except a sweet looking graph to show people I guess) from trying to make REW match to +/- 1dB either. Maybe my ear just isn't good enough to hear it but I had a helper toggle two curves on my Behringer, one was +/- 1dB between about 16 and up to the crossover rolloff. It used 9 filters. The other was +/- 3 dB over the same range using 3 filters. In a random set of material I couldn't tell the difference, and on 3/5 of the pieces I actually preferred the 3 filter solution. What I do now is go into the EQ Filters after I've let REW take a crack at it and see if there're any notch filters or ones with very low boost/cut and then I try to understand what REW was doing and how/if I can get rid of them. It may not be as big of an issue for you with two subs, but notching for me with only 1 sub is extra awful because it means you're doing something really terrible to the frequency response someplace else in the room.
Reply
Reply
post #7 of 15
7/9/12 at 10:54pm
Try bumping up the overall max boost, and setting the target level to midway between the high and low of the graph - or leave max boost at 0 but drop the target level at or below the low point on the graph (so you get all cuts).
Reply
Reply
post #8 of 15
7/10/12 at 5:56am
Thread Starter 
Captivator pair -auto EQ REW and DCX2496.zip 1k .zip file

I'll give your suggestions a try next time I have some time to play. Here are my FRD files before and after applying REW Auto EQ filters to the DCX2496 last night.
Reply
Reply
post #9 of 15
7/10/12 at 7:43am
Did you double check your DCX2496 Sum/Setup and Routing settings to make sure you are EQing both subs? I can see from your screenshot that you are EQing the A input, but I just want to make sure you are using Mono and routed Input A to Output 2. Also go to the crossover tab and make sure you set the HP and LP filters to Off for Outputs 1 & 2.

I set my Speaker Type in REW to Fullrange and the LF cuttoff to 0. I then measure the sub without a crossover and EQ it up to 150-200 Hz. The reason I EQ up higher is to ensure a smoother integration with the mains. Mark S. and Jeff P. both emphasized the importance of this at the Omaha GTG last fall. Here is a quote from Mark Seaton regarding this practice:
Quote:
Originally Posted by Mark Seaton 
EQ will never change the maximum SPL or distortion at a given frequency. It does affect the relative balance vs. other frequencies. One very important detail to remember is that when many say they "matched the frequency response", it generally is not as precisely matched as you might think, and many forget about the out-of-band range.

In the end the biggest difference between various EQ methods or approaches lies in the range above and below what was smoothed. In this specific case, it's the difference of starting with a smooth response out to nearly 200Hz and a smooth fall off into the 2nd order roll off vs. a sharper corner at the top end with an earlier and faster roll off, and possibly a different shape to the low end roll off. While little blips in the response are not very audible unless music or a soundtrack highlights it directly, 1 octave wide variations of 0.5-4.0dB can be readily audible as tonal or quality changes.

I imported your Omnimic frd file into REW and looked at the measurement. After click the EQ icon, I maximize the screen and change the graph axis limits to match the frequency response more. This gives me a better picture of the measurement. I ended up setting the target level at 71.9 dB. I also leave the overall max boost at a positive number. Right now mine is set to 4 dB. I have found that REW optimizes best if it can add some boost, too. This doesn't mean that it is boosting a null, it just means that when optimizing filters sometimes a boost and a cut work together to provide a smoother cut, if that makes sense. However, other times it is just boosting a null. I enter the filters into the DCX2496 and if a filter with a positive gain isn't making a change, then I eliminate the filter since it is just trying to boost a null. If you have a RTA function on Omnimic, you can measure in real time while turning filters on and off.

If you look back at the charts from Subfest 2011, you can see that there were dips that we were able to correct with some boost.

Finally, I sometimes play with the filters in REW and see the affect on its chart. In your case, I might reduce the cut at 20 Hz to make a little bit of a house curve.

338
Reply
Reply
post #10 of 15
7/10/12 at 2:58pm
Can't comment on why you can't get it working properly with the 2496 (maybe by using the filter type "PA", instead of "BP"), but I would like to touch on a few points.

While the AUTO function is rather cool, I still don't think it makes up for the intuition of the operator.

By this, I mean:

1. 9 filters is a bit excessive. I believe there is such a thing as "over EQing". I target no more than 4-5. (As an example: Filter 1 is a -14dB cut at 20hz with a Q of 10. Maybe some massaging of the Q or the center frequency could negate the use of the (IMO, unnecessary) 2.5 boost at 29hz with a "notch-y" Q of 4.)

2. +/- 3dB is a fine goal.


I'd say: 3 filters- Keep filter #1 and raise that sag between (#2) 30-45Hz & (#3) 58-75Hz with a 5dB boost (wide Q) and you should be back on track.
Reply
Reply
post #11 of 15
7/10/12 at 8:08pm
Here's what I came up with, in case it helps:

captivator-eq.PNG

(I think its safe to ignore the EQ #1 rolleyes.gif)
Reply
Reply
post #12 of 15
7/11/12 at 8:48pm
Thread Starter 
I appreciate the responses and Ive tried to implement some of your suggestions tonight.

Overall, I don't think I'm that impressed with the DCX2496. For some reason everytime I power the unit off I loose my settings. I saved them as an internal preset, yet have lost my settings multiple times. The interface is pretty darn kludgy and I keep loosing connection and getting random errors at synch. I connect again and the error doesn't appear etc. rolleyes.gif The product just seems a bit outdated and kludgy to me when compared to the superslick Inuke DSP interface. I do like that I can modify it realtime and didn't realize that until one of you guys said it above. The first few nights I was making adjustments and then pushing them etc. I can definately do RTA on the omnimic so it's nice to also adjust the DCX2496 realtime. (the inuke dsp can definately do realtime too)

I ensured all the EQ was off in the DCX2496 channels like suggested by DesertDome. I put a 48dB Butterworth HP filter on at 20hz for both channels. I ensured that Channel A and Channel B were both working correctly and sounded and measured nearly identical. Output is mono for all six channels (of which I'm only using the first two). I raised the crossover on my Onkyo TX-NR1007 to 120hz for each channel and LFE to ensure the 80hz crossover I was using for the previous graphs a couple nights ago didn't get in the way of this EQ process.

Here's the result.

I've attached my new non smoothed frequency response. This FRD file is averaged over 10 captures - so it may look smoothed, but it is just multiple capture averaged.

Captivator pair naked response toed in facing front 20hz tune - Butterworth 48dB octave filter at 20hz (10 response average).zip 1k .zip file


Using the FRD file I left REW auto EQ settings pretty much as default and let it choose the target spl level. See the screenprint below for all auto settings.
try2dcx2496autoeqREW.png

These are the three EQ filters it chose. Nothing looks extraordinary.
try2dcx2496autoeqREWrecommendations.png

After plugging in these three filters on both channels I get the following blue frequency response. The 48dB butterworth filter and the -12dB PEQ at 21hz really kills my 20hz performance. Not sure about that. You can see the before and after omnimic graphs.
REWDCX2496autoEQfunction--try2.jpg

I then droped the -12dB PEQ at 21hz to just -6dB and came up with the following graph instead to have a bit of a house curve as you recommended.
REWDCX2496autoEQfunction--try2withhousecurve.jpg

I'll listen to this for a bit and see what I think. I guess I just think that it's slightly lame that the AutoEQ isn't more perfect adept. I've spent more time playing around with this than it just takes me to throw some filters together in the Inuke DSP 3000 amp and have a darn near perfectly flat frequency response like on the Inuke 3000 DSP -- without even having any of the filters calculated automatically for me.


There is no smoothing on any of these graphs. I don't like using smoothing for measuring. When you turn on 1/6 octave smoothing the FR looks better obviously.
Edited by Archaea - 7/11/12 at 9:03pm
Reply
Reply
post #13 of 15
7/11/12 at 10:22pm
I dunno, that graph looks like it's about +/- 2dB from the target line using 3 filters which seems pretty good. Did you do your original measurement with the HPF in place? There's no harm in taking the flatness level down and seeing what you get. You could even set up multiple presets, 1 for +/- 3dB with however many filters, and then do a 2dB or 1dB flatness target with however many filters and swap back and forth. If you want a ruler flat matching curve you need to let REW use a little boost though, maybe 6/3 or 9/6 instead of the 3/0 you've got now. If you want to leave more of your 20Hz hump in then you can do that in a number of ways. Telling REW that your LF cutoff is lower (say 15Hz) will make the target curve flatter at 20. You an also tell it to use a lesser LF Slope. If you wanted it to leave it completely alone you could change the "Match Range" under filter tasks. Also, if you want to do anything with a house curve you can do that in REW. Then you're actually matching to that curve with the filter calculations.

But you're right, REW's auto calculation isn't a magic box. Someone who has access to any parametric device can definitely meet or exceed what it's doing in a reasonable amount of time. I think my favorite part about it is that I find it lets me play around with a bunch of different configurations or EQ approaches very quickly, or at least faster than it'd be for me to hand-filter each one. Then when I get something I'm gonna stick with for awhile I can tweak a bit by hand if needed.
Reply
Reply
post #14 of 15
7/11/12 at 10:26pm
I owned a DCX2496 for several years and never had any trouble with saving or loading files. I always used the computer interface and never touched the buttons on the device. I agree that the interface does look outdated.

Try changing your LF Cutoff in REW from 20 to 0. It might allow REW to provide a little more extension on the low end. If you look at your target curve in REW and the results, they match very closely from 30 Hz and down. You can also add the HP filter as a manual filter in REW just to see what it will do to your frequency response. You could probably drop it to a 24 dB filter. Finally, I think the dip at around 30 Hz is the natural rolloff of the sub (from comparing to the chart at JTR's website) before room gain is kicking in. I still think you should change your maximum boost from 0 to 4 dB and rerun the filters.
Reply
Reply
post #15 of 15
7/12/12 at 6:35pm
You will always get better results manually EQ'ing instead of relying on automatic tools. The automation only works when the assumptions behind it are completely true, or don't matter for your scenario - which is usually not the case more often than not.

Also, the way filters are implemented in DSP land is not always the same. I don't know if REW is supposed to compensate for the type of processor or not, but one of the reasons I don't like FIR filters for low frequency is because the frequency response doesn't match that of an equivalent analog filter. IIR filters will.

Somewhere on the internet is an interesting plot showing the frequency response of several different DSP boxes with the same numerical settings, and how they all vary by a few dB depending on where you look. I wouldn't be surprised if that was a factor in your situation.
Reply
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: DIY Speakers and Subs

AVS Top Picks

AVS › AVS Forum › Audio › DIY Speakers and Subs › REW AutoEQ + Behringer DCX2496 --- odd that it's not making a flat FR?