Calibrating active 3way system with REW

[Ofer]

Hi everybody,

This is my first post. I have long delyed diving into the topic of speaker measurements. I have an active system:
source: emu1212M (patchmix DSP)
Xover: Ashly protea 3.4cl
speakers: based of Andromeda mkII DIY speaker, not all the drivers are the same as the original. The speakers are placed about 1m from each wall.
Three amps driving each set of speakers (base, mid, twitter). The details are in my profile.
I want to use REW to:
1. Set the same SPL to all drivers (this I think I can handle on my own).
2. PEQ the drivers/cabinets to be as flat as possible
3. Choose the best Xover frequencies, filters and slopes
4. Dial the correct delays to the drives (mid and twitter most probably) - this seems very confusing
5. PEQ the listening position.
So there is a lot of questions about how to do it and later how to interpret the plots. For starters:
1. Where should I place the mike (Umik1) to measure the drivers. I was thinking of 30cm from the cone to avoid room response. Is that a good idea?
2. should I aim for completely flat curve on the drivers or some house curve.
3. How to choose Xover frequencies, filters and slopes?
4. How to delay correctly?
I have gained a lot of knowledge from this post:
"Using REW to determine time delays between drivers" (sorry no new to be allowed to post real links)

But it was also very confusing. I will post graphs soon but would appreciate any guidance as to how to go along with this task.
Thanks,
Ofer.

 

[skid00]

I would first measure the main left or right speaker which had the least amount of clutter and reflecting surfaces around it. I would measure at various distances from 3 feet to 6 feet, to see at which distance the measurement curve is flattest. Please note that the tweeter and mid driver will 'beam', so you also need to account for the height of the microphone, and the angle left/right. (You can then use these measurements to decide how to angle the speaker toward the listener).

If you intend to apply EQ to the mid and high range frequencies, you need to understand how the radiation pattern off all sides of the speaker affect what you hear. Your measurement curve won't necessarily reflect these reflections. You can make the sound worse.

I cannot answer any of your other questions...

 

[Ofer]

Hi Skid00, thanks for you reply. So you don't recommend near field measurements? What do you mean by "beam"?

 

[Ofer]

OK. So I started with something straight forward, setting the levels on all 6 drivers. Apparently using REW and EMU DSP isn't that simple. You can't use the Java-wav but have to use ASIO, even after configuring the right ASIO input only one channel produced sound when using the signal generator and one can't use the UMIK1 with asio.
So I plugged in my laptop to the 1212m input. Set the right mid to 75db using pink noise and SPL meter and set off to adjust the rest of the drivers keeping the volume on the preamp constant and only changing the gain for each output channel on the Ashly. The strange thing is that on the twitters I couldn't reach 75db even after adding 12db gain (the maximum in the Ashly)! Is pink noise not suitble for adujsting twitters? or is it that my twitters are faulty or maybe they aren't supposed to be at the same SPL as the rest of the drivers? Any help would be greatly appreciated.

 

[John Mulcahy]

To use ASIO with more than one physical device you need to use an ASIO wrapper like FlexASIO or ASIO4All.

 

[skid00]

Hi Skid00, thanks for you reply. So you don't recommend near field measurements? What do you mean by "beam"?

Sorry I didn't get back to you sooner...

I have Carver Amazing dipole speakers. Think of a 4'x5' sheet of 3/4" plywood, with a ribbon tweeter that is 4.5' long, running from top to bottom, and alongside that are four 12" woofers. These interact with the room in a different way than box speakers. And they are in an oddly shaped great room. After a great many measurement sessions (which always thrill me, John!) I finally realized that if I found the flattest measurement distance, and used a tight window, that I could effectively apply EQ. I *think* that that measurement distance most closely emulates an anechoic chamber.

I *think* that if you also find such a mic placement, that you too will be able to more easily effect EQ.

Of course, I could be dead wrong. But you will need to acquaint yourself with the mic and REW, so you won't be wasting much time.

Drivers in a panel launch the strongest wavefront aligned with the drivers longitudinal axis. (Woofers, with their low frequencies, launch a near-360 degree soundfield). So tweeters sound loudest, and have the highest extension on-axis.

 

[skid00]

The strange thing is that on the twitters I couldn't reach 75db even after adding 12db gain (the maximum in the Ashly)!

Be very careful you don't blow your tweeters out. I can't tell you how loud you can sweep them, maybe someone else has input. But to see you raising the output level worries me.

 

[Sixto]

Hi Skid00, thanks for you reply. So you don't recommend near field measurements? What do you mean by "beam"?

Beaming refers to the off-axis SPL drop you would see if you compared a driver's graph directly on-axis vs 30 degrees off-axis (for example)...it is more noticeable at higher frequencies, and is generally a result of the relationship between frequency and driver diameter, so a .75" tweeter will beam less than a 1.5 tweeter at the same frequency, distance and off-axis angle, meaning the off-axis sound level will be closer to the on-axis level for the smaller of 2 drivers... these are general rules that apply mostly to cone or dome drivers... horns and waveguides often modify a driver's beaming pattern, and ribbons/AMT's have unique patters too.
Six.

 

[Ofer]

So I have made a few measurements over the weekend. An immediate issue that materialized is the fact that my Ashly 3.4CL crossover has only 4 PEQ filters/channel. That really limits my options as both the mid and low drivers produced more PEQ filters. Also there is no Q in the Ashly. All the measurements posted were done and 1m from the cone with a mattress covering the wall closest to the speaker. I post only the right speaker measurements as the left one was pretty identical. Measurements are after SPL leveling of all drivers except twitters. All pictures are after Var smoothing.

Figure 1: right base before EQ.

Figure 2: right base after EQ.

The EQ filters that I chose were:733 Hz Gain -4.0 dB, 116 Hz Gain -3.8 dB, 76.4 Hz Gain -12.8 dB, 84.0 Hz Gain -9.8 dB. As can be seen the EQ didn't change the FR that much. A phenomenon that repeated itself on all drivers.

Figure 3: Mid before EQ.

Figure 4: Mid right 1m after EQ.

EQ filters chosen: 1067 Hz Gain -9.8 dB, 1098 Hz Gain 7.9 dB, 876 Hz Gain -3.0 dB, 567 Hz Gain 2.1 dB
In the Midrange the EQ was more effective (I think).
The twitter naturally was more flat but I haven't measured of axis yet.

Figure 5witter before and after EQ.
The twitter had only 2 EQ filters: 1999 Hz Gain -2.3, 1739 Hz Gain -1.7 dB.
This gave me crucial information as to where is best to Xover the drivers.
Using the posts by Chris A "Using REW to determine time delays between drivers" (can't post links yet) I have decided to enable a rather shallow curve where there is a lot of overlap between drivers.

Figure 6: All drivers after EQ.
Now figure 6 is far from perfect but it helped me choose my Xover filters. I crossed the base/mid at 640hz linkwitz riley 12db/octave and the mid/twiter at 2Khz linkwitz riley 12db/octave for mid and 24db/octave for twitter.
Before punching in the Xover filters I have used the method describe in the miniDSP site to choose delay using impulse response where both drivers are flat: "Time-aligning speaker drivers with Umik-1 and Umik-2" (can't post links yet)

.

Figure 7: Impulse response 1m with microphone height between mid and woofer.
I have calculated the difference at 1.22ms for the twitter mid it was 0.49ms. After imputing the delays (for both speakers) I have punched in the Xover.
Before the last figure in the listening position I would like to say that the sound improved dramatically due to this process. This is far from finished and I have attenuated the twitters in about 2db as they sounded harsh. The best option for applying PEQ seems to be the miniDSP 4x10HD but it is 500$ that I don't have right now, the miniDSP 2x4HD has only 6 PEQ/channel (if I understand correctly) so it to isn't adequate. So any advice as to how to continue from here would be greatly appreciated. Also I realize that this is amateur at best so fire away.

Figure 8: Both speakers at listening position.