Averaging Listening Positions when doing multi-channel equalization. (1 Viewer)

[Larry Chanin]

Hi All:
In trying to figure out what's going on with my processor filtering the test signal, I just recalled reading the following instructions prior to taking room measurements in Lucasfilm's Home THX Audio System Room Equalization Manual:

Switch your Home THX Controller to the “Dolby Pro Logic Surround” mode. The Home THX Cinema mode must be switched off for this procedure. Note: For Controllers featuring Dolby AC-3 Decoders, there is no easy method to insert broad band pink noise into this signal path. Equalization should be done through the Dolby Pro Logic mode on these controllers as well.

Although I performed the measurements with my main speaker set to large and the subwoofer off, I was using my default surround setting on my Lexicon processor, which was Logic 7. Tonight I'll try another measurement with Dolby Pro Logic and see what happens.
Larry

 

[brucek]

Larry,

Was the rs.cal the same as we discussed above with the low frequency correction included?

You should not be using any DSP modes at all, never use a suround mode - you should be using bypass mode or mono mode or stereo mode, depending on your processor.

brucek

 

[Larry Chanin]

Hi Bruce:

Confirm EQ
• Reconnect System
• Confirm splice of Subwoofer with Center Channel by playing Center pink noise in Dolby Pro Logic mode
• Confirm splice of Subwoofer with Left and Right Channels by playing Left & Right Channel Pink Noise in Stereo or Bypass mode

Larry

 

[brucek]

larry,

The reason they are specifying prologic is because if you're going to do an EQ of more than just your mains (left and right) and your subwoofer, then you need a method to get sound to the surrounds and center. Prologic would likely be the least offensive in that regard, in their opinion.

Myself, I would get more creative about it than that and never use a surround mode.
For the individual sub, left and right speaker I would use bypass, mono or stereo mode (in that order,with bypass being the most desirable). I would EQ each of these speakers separately and then when I wanted to check them in concert for interactions, I would use simple stereo mode with the two mains and sub playing at the same time. If you are going to use the small setting for your mains in the end, then I would set that up at that time when I was checking for interactions.

Then to EQ the center and rears individually, I would use bypass or mono mode and wire up the center channel to the left main channel speaker connection for example and proceed with the EQ on it.
When finished with the center, then I would wire up each rear speaker to the left main channel and EQ them individually. This would remove any DSP or surround mode bias that might be introduced.

Then when I was all done that I would do a full system check in proplogic so that all speakers where on to see any final interactions.
It would be nice if your processor had a 5 channel stereo mode to do this final check...

brucek

 

[Larry Chanin]

Hi Bruce:
Thanks for the advice.
I think I'm finally starting to circle for a landing.

The Party effect allows unprocessed stereo signals to be played over all speakers for background music or for maximum acoustical output of the system.

Since this sends a signal to all eight channels, I could just turn off the amplifier feeding the surround back channels.
Thanks again.
Larry

 

[Larry Chanin]

Hi All:
Well, I think I'm out of the woods! At least for now, that is.
I reran the measurements connected to the processor and running it in 2-channel mode. The results are almost the same as the Direct to the amplifier measurements.
Here's the latest frequency response curve. (Bruce: I went back to the original correction file with the low frequency values.)
Frequency Response Curve through Processor in 2-Channel Mode
We're looking at an 18dB drop from the peak which isn't great, but at least its not 55dB.
Based on this I think I can resume taking measurements and seeing whether the Bijou can help smooth out the peaks.
Larry

 

[brucek]

Larry,

Now that's a normal curve that I would expect to see. The two channel mode is perfect.
I don't know how much I would trust the RS meter above 15Khz. You can see how it drops off rapidly. Don't know if that's the speaker or the meter response.....

Looks good.

brucek

 

[Greg Monfort]

My take is something's still not right, 18dB is still too much. Even if the speaker has max baffle step compensation built in it would only be 6dB, and there's no way it's boundary/room gain, otherwise there wouldn't be so much roll off below the LF peak.

GM

 

[brucek]

Hey Greg,

Do you not think that the room could be causing that peak? I know I have to EQ out about a 15dB peak in my sub with my EQ.

brucek

 

[Greg Monfort]

But it's a fairly high Q (narrow BW) peak, right?

According to his room mode chart, the first axial mode, which will be the highest in amplitude, is at 24Hz. On his FR plot, it's ~ +3dB WRT the higher BW average, so how can the room be making an extremely low Q (wide BW) +18dB peak?

Again, going back to using your ears, does the speaker sound extremely bass heavy or boomy?

GM

 

[Larry Chanin]

Hi Bruce:

Thanks for the encouragement. Believe me I need it!

Larry,

Now that's a normal curve that I would expect to see. The two channel mode is perfect.
I don't know how much I would trust the RS meter above 15Khz. You can see how it drops off rapidly. Don't know if that's the speaker or the meter response.....

Looks good.

brucek

According to the THX manual we shouldn't bother trying to equalize anything over 6,000Hz. An old test report by Stereo magazine puts the -5dB point on a newly manufactured speaker (Polk Audio SDA-1C) at 18,000Hz. Mine's about 12 years old now if that makes a difference.

Larry

 

[Larry Chanin]

Hi Greg and Bruce:

Hey Greg,
Do you not think that the room could be causing that peak? I know I have to EQ out about a 15dB peak in my sub with my EQ.
brucek

With all the miscues I’ve had in coming up the learning curve, it’s of course still quite possible that I’ve done something wrong. Below is my first attempt at equalization. Maybe these results can shed some light on the subject.
First, let me say it takes the patience of a Saint and the fingers of a safe-cracker to get any results worth looking at. I’ve spent about a hour fiddling with the sliders to get these initial results. I think its possible to improve them with more effort.
Improvement of Frequency Response by Equalization
Of course I didn't attempt to do anything with the steep roll-off occurring after 17,000Hz. However, with an hour of effort I was able to get all the values within plus or minus 4dB (within 8dB of the maximum)for this channel.
Larry

 

[Pete Mazz]

Hey, now you're talking! Looks good.
I'm not sure I'd get too crazy about the FR. It is, after all, just a SPL meter taking the measurements. Try looking at the FR with 1/6 or 1/12 smoothing. That might show some specific room modes.
Pete

 

[Larry Chanin]

Hi Pete:
Thanks.
Yeah, I did look at the 1/6 octave curve to see more detail, but the sliders on the main speakers are only 1/3 octave. (11-frequency bands between 80Hz-800Hz) The sliders on the two subwoofer channels go to 1/6 octave (12-frequency bands between 22.4Hz-80Hz).
With all the false starts I've had I share Greg's concern about the low frequency peak on the "Before" frequency response perhaps not being the "real deal". Naturally a concern about the low frequency response of the Radio Shack meter comes to mind. Right now I'm torn. I'm not sure that I want to go out and buy a precision microphone and preamp just yet.
I wasn't kidding about needing to have fingers like a safe-cracker to make adjustments. The ETF5 software plots hundreds, maybe thousands, of data points. So, for better or worse, its letting you see a lot more precision than any manual charting method (or one of those expensive RTAs with 31 LEDs). The result of this is subtle adjustments of the slider controls show significant changes on the frequency response curve. To make it more interesting there's an interaction between the frequency being adjusted and other frequencies.
[Start Sidebar Discussion]AudioControl goes to a lot of trouble to explain that the Bijou is a constant-Q equalizer, meaning that adjustments in one frequency band has little affect on adjacent frequency bands. While it may be true that adjacent frequencies are not greatly effected, I observe changes in other frequencies hundred or thousands of Hz away. I figure that if I'm surgically attenuating or boosting a particular frequency, I'm also attenuating or boosting its harmonics. So depending on whether those harmonics are in or out of phase with the direct sounds at those higher frequencies, I see a boost or cut in frequencies distant from the point where I've made the actual adjustment.
[End Sidebar Discussion]
So,…the net effect of all this is while the movement of a particular slider has a predicable effect on the frequency being adjusted, it has an unpredictable effect on higher frequencies nowhere near the adjustment point. I figure this is how I was able to use trial and error to flatten the frequencies thousands of Hz up the spectrum when my controls only go up to 800Hz. [rant]Have I told you how much I hate trial and error, especially the error part??!! . I love the equalizer manual, you see a 3dB peak, you cut it by 3dB. Simple right? I don't think so![/rant]
Anyway, all this greatly challenges one's patience, and I suspect that this exercise could actually be harmful to those of us who are incurably anal retentive.
Larry

 

[brucek]

So,…the net effect of all this is while the movement of a particular slider has a predicable effect on the frequency being adjusted, it has an unpredictable effect on higher frequencies nowhere near the adjustment point

Yep, and that's the rub with equalizers. They not only alter the amplitude of the signal, they affect the phase. This can make for some unpredicatble results. There is always trial and error I believe. Myself, I only use a sub equalizer, but I know exactly what your talking about..

Wayne P uses analog equalization on his mains I think, he might be helpful here to give you some tips....

brucek

 

[Greg Monfort]

>AudioControl goes to a lot of trouble to explain that the Bijou is a constant-Q equalizer, meaning that adjustments in one frequency band has little affect on adjacent frequency bands.
====
This would be solely determined by whatever Q they used. If it is high Q then this would be true, but the fewer the center points, the lower in Q it is, and since the octave scale is logarithmic, each octave band slider covers a wider BW. This is why parametric EQs are the Hot Ticket.

Still, I wouldn't sweat this too much as it will sound fine as long as the FR is +/- 3dB in-room with no glaringly wide broadband null/peaks in the 1k-5k BW since our hearing acuity isn't really all that good above/below, and gets progressively worse with decreasing/increasing frequency.

I've EQ'd the same room with both 1/3 and 1/6 using 31/62 band RTA/pink noise, and I could barely tell the difference with music. With a parametric, it was subjectively quite a bit smoother, but for movies I don't see the point unless you're a perfectionist.

GM

 

[Larry Chanin]

Hi Bruce:

Yep, and that's the rub with equalizers. They not only alter the amplitude of the signal, they affect the phase. This can make for some unpredicatble results. There is always trial and error I believe. Myself, I only use a sub equalizer, but I know exactly what your talking about..

Well, based on my brief taste of equalization, it looks like I'm in for a protracted series of head-banging sessions.
What I particularly don't like about a graphic equalizer with mechanical sliders is that you can't "save" various settings and recall them from memory. If you go too far on a manual adjustment you can start to migrate away from a good curve by trialing and ERRORING in the wrong direction! [rant]Have I told you how much I hate trial and error??!![/rant]
Well, I better stop ranting or I'm going rationalize myself into thinking that I can't live without a $9,000 AudioControl Diva.
Larry

 

[Larry Chanin]

Hi Greg:

Still, I wouldn't sweat this too much as it will sound fine as long as the FR is +/- 3dB in-room with no glaringly wide broadband null/peaks in the 1k-5k BW since our hearing acuity isn't really all that good above/below, and gets progressively worse with decreasing/increasing frequency.

If we believe the corrected ETF5 frequency response curve is telling us the truth, I believe I'm at the +/- 4dB point with my corrections. Do you agree? As I continue to attempt to improve this response is there anything on the curve I should turn my attention to?
Thanks for your continuing interest.
Larry

 

[Greg Monfort]

Hmm, I haven't kept up with most consumer components for many years now, but there used to be some fairly affordable ones out there. If I were to go buy something today it would be a Beringer Ultracurve and use the auto FR correction feature to get a baseline, add EQ above/below it's basic range, let it redo the FR and forgetaboutit. +/-0.5dB 20-20kHz at the listening position in a typical room for ~$900 including calibrated mic (last street pricing I saw) and an hour or so of time while you're still reading up on someone's software..... Considering your room, it would be cheap at twice the price. Sorry, but I couldn't keep quiet any longer. :frowning:
Anyway, working with what you're committed to, WRT the FR, obviously I don't know how it sounds, but the dips are too broadband looking for my hearing, plus you've EQ'd away much of the speaker's dynamic headroom, so will it even play at reference level without clipping the amp, and at how high a distortion level?
At this point, if this were my situation I would take one of the speakers (with a preamp/amp or basic receiver to power it) outside somewhere there's no boundaries in any direction for at least 50ft (to get accurate LF response) and measure it at 2.83V/m to see how it compares to a factory FR plot (often 1 octave to make it look good). Until you know your measuring equipment's valid, you're blind.
If it turns out that it's working fine, then I would attenuate/boost the FR using a 2.83V/m input until it at least approximates the manufacturer's FR plot at the primary listening position.
GM

 

[Larry Chanin]

Hi Greg:

At this point, if this were my situation I would take one of the speakers (with a preamp/amp or basic receiver to power it) outside somewhere there's no boundaries in any direction for at least 50ft (to get accurate LF response) and measure it at 2.83V/m to see how it compares to a factory FR plot (often 1 octave to make it look good). Until you know your measuring equipment's valid, you're blind.
If it turns out that it's working fine, then I would attenuate/boost the FR using a 2.83V/m input until it at least approximates the manufacturer's FR plot at the primary listening position.

Logistically this will be quite difficult for me. However, by setting the gate time of the software short enough so that its measuring just the direct sounds it is possible to approximate a quasi-anechoic frequency response in which the effects of the room are totally ignored, i.e. the speaker response. The problem is that as the gate time is reduced, so is the minimum frequency that the software is capable of measuring. Since my speakers are literally up against the wall (and I guess so am I )when I try to measure the direct sound a meter from the speaker, the first reflections are about 1ms, which is lower than the software can handle.
An approach with less logistical demands, but similar to your suggestion, would be to attempt to get the speaker in the middle of the room with the microphone close and the gate time at say 10ms. This might yield a good approximation of the speaker response. My speakers are Polk Aduio SDA-1Cs and they are about 12 years old. Frankly I don't think that the manufacturer response curves are available, but I will contact Polk Audio. They are ususally very accommodating.
As we have already mentioned, I think another approach to somewhat verify the ETF5 measurements situation, is to use other less sophisticated measurements to see if "we're in the ballpark". I'll try to do that before locking into new equalizer settings.
Thanks again for your helpful and candid response.
Larry