On a quest to improve my B&W DM640's

[Sixto]

The new USB interface works! - I've run a few measurements in REW of the current in-room response as an updated baseline. It still looks rough with lots of reflections, but is more even than what I got from the old interface. I did get some funny messages about clipping in the reference channel (lowered the output level and that helped) also, I have not figured out whether to use acoustic timing reference or use the loopback feature since this is a 2-channel interface (Is loopback a more consistent or better choice for timing reference - instead of using a 2nd speaker off of ch 2?) I need to use phantom power for the mic, doesn't that mess with the loopback input? I read somewhere that the Scarlett 2i2-4g has the capability to do "Internal" loopback. More research needed...

in the next few days I'll try to move the setup to a larger space and also maybe try an outside measurement in the backyard since temps are expected to climb to the low 40's!

Here are a screenshot and REW file of the new baseline measurement for reference.
Sixto.

 

[steve138]

So how are you getting on now?
I think for an in room response that's not bad. Yes you'll get some swings on the lower end and a gently falling top end. I think thats normal.
But for the speaker, measure at 1m and apply the gated time window ( at 1m and 1W (2.83V) is industry standard to determine the sensitivity and output level). Your output level is too low. When iview your mdat file, i had to apply a gated time window of right = 2,25ms, and the trace disappeared below my normal graph viewing range of 70 - 95dB. I had to change MYscale to 50 - 70dB. and then it was still very wavy. See attached screenshot. if this is getting closer to reality - it's not a good response. and it's very similar to the one I posted before. If its what the speaker is actually doing it definitely needs a new Crossover developing and not surprising given some of the rough responses Danny Richie has fixed on GR-Research Youtube channel.
Your calibration is nice and flat though. I still haven't used my Scarlet and UMM-6 yet but i was going to use an XLR as loopback cable. With REW and my UMIK-1 I haven't used a timing reference and it seems to work fine.
I just managed to buy an Audiomatica CLIO Pocket so I'll be selling the Scarlet and returning the UMM-6 unused

 

[Sixto]

It's slow--going, busy time of year. I'll post progress along the way. Thanks for checking-in!
Sixto

 

[Cristianolo]

Based on the last file you attached, I removed the smoothing and redid the windowing based on two different windows, Flat-Top with a slightly larger window size and Tukey 0.25 based on the standard of traditional guides, that is, at the point of first reflection.
Observing both, it seemed to me that your system has a bit of a baffle step, so I added two proposals for very simple EQ filters that can be implemented: 1) generates a slight decay to a slightly softer top end; 2+3) just an adjustment in the most central region, maintaining the natural brightness of the treble in a flat response.

1)
HS Q Fc 850.0 Hz Gain -4.00 dB Q 1.00

2+3)
PK Fc 1800 Hz Gain -4.50 dB Q 0.700
HS Q Fc 5000 Hz Gain -2.00 dB Q 0.700

 

[Sixto]

Based on the last file you attached, I removed the smoothing and redid the windowing based on two different windows, Flat-Top with a slightly larger window size and Tukey 0.25 based on the standard of traditional guides, that is, at the point of first reflection.
Observing both, it seemed to me that your system has a bit of a baffle step.

Cool!, thanks!!! I need to look into this. I don’t know enough about baffle step, so I’ll be researching that for sure. Impressed at how flat you were able to get the FR graph, with just two or three filters. But first, I want to measure the speaker again in a quasi-anechoic setting (my backyard) and also do near field for the woofers, in two versions , one with the stock XO, and another with just the mid-hi XO. Tomorrow, the temps should get above 50 degrees at 3:00 p.m. so I’ll have a 3-4 hr window of relatively warm daylight to get those measurements.

Sixto

p.s. After a little research I refreshed my memory on baffle reflection and diffraction issues... I will also be making use of some software to model it and visualize it using my actual speaker and driver measurements…

 

[Sixto]

Yes! - I was able to spend some time measuring outdoors yesterday, and this is what I got:

The solid blue line is a fullrange measurement of the speaker with stock XO all around at 1m and amp output somewhere betwee 1v and 1.4v.
The dashed green lines are the twin woofers both at 1m (for delay calculations) and close range.
The red line is the Midrange/Tweeter combo using the top half of the stock XO. (P,S, I attached the raw measurements as separate files below)

First impressions are that my original estimate of crossing from Woofers to Midrange at 400hz may have been premature, since it looks like they could go as high as 800-1000hz. I will test it, but I am still concerned about managing phase when 3 drivers overlap so close to 1Khz.
The midrange 2nd harmonic is 33db below the 78db fundamental at 830hz. and the woofers' distortion jumps up above that starting from 365hz. The midrange also drops about 8db from 800hz to 300hz (baffle step?), but I think I could mitigate that drop by setting my midrange target flat to around 70db at 300hz and boosting the HF amp about 9db to match the LF amp at that frequency.(beauty of bi-amping) (See also Impedance measurements of these drivers at post #14)

I like the rising response of the tweeter on the top end. I think the twin-woofers' interactions are the culprit for the wavy response below 65hz in the 1meter distance measurement, so I'll have to see how that can be managed (reflections/bass traps) after EQing and moving back indoors.

One of my original goals is still to get a punchier low-end frequency response for bass and drums, which is why I want to change the stock (passive) XO to active at the LF/Mid transition and biamp each speaker so the woofers can be better controlled by the power-amps being located very close to each speaker.

My plan over the next few days to a week is to adjust delay and phase, try some simple EQ options, and test XO's between 300Hz and 1000hz to see the impact indoors (weather is turning cold after today). I will try to do all the REW tweaking of Phase, delay and EQ in the two separate FR files (woofers vs mid-tweets), then will see how blending them with different XO configurations on my MiniDSP 2x4HD works. I'm also upgrading to the new REW version, so that might involve some re-learning...Will post again once I get some results.

Helpful suggestions are always appreciated! and remember this is as much about learning as it is about improving my listening experience!
Thanks! - Sixto

 

[Cristianolo]

Well, let's go...
First you must learn to manage measurements better (smoothing, windowing, merging...), avoid working with unprocessed data and just smooth the response.
One thing that worries me is that the measurement times are happening before 0ms, so there is some kind of compensation applied in the acquisition window, exclude this so that you can have the correct sound time of flight.
Given the size of the box, try measurements further away, such as 2m, as the listening distance should not be that small.
The image I bring is the box's response based on what I was able to process from the measurements above.
I personally don't like speakers with this ascending response profile, I would emphasize something more neutral from 200Hz upwards, HOWEVER, if the speaker is used with toe in, it is possible that the response will be more linear at the listening point.
To be sure, off-axis measurements are necessary (15º, 30º, 45º)

 

[Sixto]

Hi Cristianolo, thanks for your response! Ive seen many "standards" for windowing and smoothing, and am still trying to figure out what makes sense to me and why. Consider me a stubborn work in process.

One thing that worries me is that the measurement times are happening before 0ms, so there is some kind of compensation applied in the acquisition window, exclude this so that you can have the correct sound time of flight.

The 0ms point I believe is set by the acoustical reference (fullrange single driver) on Ch2, which was further away from the microphone than the main speaker. I'm mostly interested in the relative difference in delay between the woofer and midrange for the Active XO, when both are measured from 1m away which calculates to be .0715 ms.

I personally don't like speakers with this ascending response profile, I would emphasize something more neutral from 200Hz upwards, HOWEVER, if the speaker is used with toe in, it is possible that the response will be more linear at the listening point.

As I get older, and my high-frequency hearing diminishes, I like boosting the treble a bit, and I do toe-in the speakers when listening to music.

To be sure, off-axis measurements are necessary (15º, 30º, 45º)

Attached, 0d, 15d and 30d. I did not think they told me anything new... basically the tweeters spl begins to drop above 5khz, and the passive crossover is causing the mid/tweet to have an odd directivity pattern between 1,500hz and 4,500hz. Not much I can do about changing directivity of the existing drivers or changing the passive XO between mid and tweet... My NEXT set of speakers will use horns to control dispersion.

I do have a question for the forum: Some people combine (merge?) the woofer and tweeter measurements, and apply EQ process using the combined response and even model the XO in REW. Others keep the measurements separate, EQ them, and test various XO and delay combinations on the separate files using either REW or some XO/DSP Hardware/software. What is the benefit of merging the response files from 2 or more drivers prior to managing phase, delay and applying EQ?

Cheers! Sixto.

 

[steve138]

yes now you're getting somewhere. with a USB mic like a UMIK-1 you dont need a soundcard calibration but just use the mic calibratioin for that mic's serial nr and away you go.
With those traces now you should be able to blend them together well probably with second order filters and get a decent result. we got you there in the end !!

 

[Cristianolo]

I do have a question for the forum: Some people combine (merge?) the woofer and tweeter measurements, and apply EQ process using the combined response and even model the XO in REW. Others keep the measurements separate, EQ them, and test various XO and delay combinations on the separate files using either REW or some XO/DSP Hardware/software. What is the benefit of merging the response files from 2 or more drivers prior to managing phase, delay and applying EQ?

Cheers! Sixto.

Always optimize each speaker first, then the crossover

Will your system be purely DSP based? Or will it be a hybrid, with a passive crossover along with it? What is your miniDSP version?

 

[Sixto]

Always optimize each speaker first, then the crossover

Will your system be purely DSP based? Or will it be a hybrid, with a passive crossover along with it? What is your miniDSP version?

Thanks Cristiano, my system will be a hybrid crossover, The plan to use the existing passive XO for the Midrange to Tweeter transition, and the the MiniDSP 2x4HD for the Woofer to Midrange crossover. Now that I have better measurements, I want to see (hear?) whether he DSP crossover works better at 300hz, 800hz or somewhere in-between. I also plan to test rePhase to flatten the phase and tweak the initial REW filters prior to copying the correction files to my MiniDSP. I will post examples and descriptions of what I do along the way.
Sixto

 

[Sixto]

.With those traces now you should be able to blend them together well probably with second order filters and get a decent result. we got you there in the end !!

Thanks Steve, the fun is just beginning!

 

[Sixto]

Today I measured the Left and Right in-room response with exactly the same REW settings, plus measured the distance to the microphone to be EXACTLY 8' which is also how far the tweeter centers are from each other.

Since I posted the in-room measurement on Post # 26, I have done NO TWEAKING to the speakers other than move them outside to do the outdoor measurements and put them back as close to the original position as possible. That means no EQ, no change in delays or phase, no gain changes other than what I had to do to adjust the audio interface levels for each location.

I found it interesting that there is still a delay discrepancy between the Left and Right (wiring?) (see JPG below, and realized that I might want to EQ the left speaker separate from the right (Right seems to have more reflections) Even though the drivers are identical between L&R speakers when measured at 1m (mid tweet) and close-range (woofer) outside, the room reflections impact the L and R speakers differently. Of course, the summed response from both speakers is about 6db louder than either one of the two individual speakers.

Learning a lot along the way.. found this series of 2 videos (part 1 REW, Part 2 RePhase) about exporting from REW to get Linear phase filters (FIR) via RePhase.
In theory, linear phase filters via FIR are more accurate? but take more processing power... we'll see if my MiniDSP 2x4HD has the power...
Sixto

 

[Sixto]

Today i'm starting the merging and EQ process, so I wanted to document first steps.
I'll be using the outdoor measurements already posted at #31.
I wanted to see how the woofer and mid/tweet align at various XO points so, using the REW Alignment Tool, I tried 4 different combinations that seem plausible based on the FR of these drivers.
First I'll share my steps using a screenshot showing how I aligned the 2 measurements for an XO at 600 hz. (I repeated the same steps for all 4 potential XO points: 300, 600, 900 and 1200hz)

From the "All SPL" tab, I selected "Controls" and then "Alignment Tool" buttons which brought up the dialogue box on the lower left corner.
I placed my cursor at the desired XO frequency (shows in both SPL and Phase Graph)
Then I selected these buttons: First "Level Phase at Cursor" which does Trace A, then "Align Phase at Cursor" which does Trace B and adds any required delay.
I checked to see if Reversing Polarity for either trace gave me a better summed response, and decided that the default polarity was fine.
Finally I selected the "Aligned Sum" button which creates a new measurement of both Woofer and Mid/Tweeter traces combined.

Here is a screenshot of the 4 aligned sums at different cursor points (potential XO's) with dashed lines, and the original traces in Blue (Woofers) and Red (Mid-Tweet)

Note that the aligned sums create a (variable) null much higher than the XO point. The nulls get slightly shallower as the XO points shift higher, also the (negative) delay gets smaller as the XO point shifts higher...

Next, I'll try EQ'ing the aligned sums and see how flat I can get each of them with minimal filters.

Cheers! - Sixto.

 

[Sixto]

I tried EQ'ing the aligned sum of the woofers, mid & tweeters, and got a real mess. Ran multiple scenarios (about 20) and could not get anything that measured better (Indoors) than what I already have (see post #26)

So Im going back to basics, instead of aligning all the drivers into one file before making any corrections, I'll try to work at EQ'ing the individual driver groups (Twin woofers plus Mid-Tweet) and let MiniDSP handle the XO duties.

For now I wanted to share my outdoor Impulse and Step response and see if I could get confirmation that I'm interpreting the readings correclty and which to choose for EQ'ing.

The Mid-Tweet (Green trace) at 1M looks fine to me, and I may not even need to EQ these drivers depending on how the digital XO changes the overall behavior.

The next two are the main source of head-scratching... Nearfield and 1M woofer measurements, plus their associated Impulse and Step response are quite different from each other.

Looking at the 1m (Orange trace) Impulse and Step response you can clearly see the twin woofers are not exactly aligned at this measurement distance, even though I measured the wires to be EXACTLY the same from the binding post to each driver. (they are wired in Parallel). Is it purely the baffle reflections and comb filtering that are responsible for the strange Impulse and Step response at this distance?

The next image is from the Close Range (Blue trace) measurement of the twin woofers outdoors. The mic was vertically positioned exactly midway between the two woofers, and about 1/2" from the face of the cabinet. It shows a much more typical impulse and step response,

A second statement/question regarding the Woofers behavior is this (mostly for the speaker builders on the forum):

1. The distance to the microphone produces quite a difference to the Woofers' Impulse and Step response.
2. The Close Range impulse and Step looks better on the Close Range measurement, so I SHOULD use that measurement as the baseline for EQ and XO right?.
3. Do I just ignore what I see happening on the 1M woofer Impulse and Step graphs, other than using the .07 delay for aligning the timing of the woofer and midrange at XO?

Cheers, Sixto.