JBL 4429 biamp and dsp correction

[bubbleboy76]

I have sold my 4365:s. They were too large and expensive. I have now downgraded to 3 pieces of 4429 instead as L-C-R speakers in my living-room cinema.

I will try this procedure, that Greg Timbers used, for the fun of it, and see where it takes me.


"I haven’t gotten back into working with the DBX or with any other speaker models. I am using an AllDSP model PLP226 Signature. It is basically 2 in and 6 out which works for my Tri-amp system I am not using the UHF at all. I actually did my design with my speakers in place in my living room. I put in a 4 th order LR crossover target (750 Theoretical) and made 4 mic measurements of each of my woofers and HF units. Averaged the 4 positions for each transducer and used those 6 curves as the transducer measurements. I then applied the necessary crossover slopes and frequencies to get the acoustic measurement to match the electrical target as closely as possible. It doesn’t matter what electronic slope and frequency is used so long as the actual acoustic response is close.

Next I applied the necessary EQ to get the crossover slopes tuned in. Once done, I turned everything back on and made a Left (or Right) channel measurement using the same 4 mic positions averaged. The summing came out nearly perfect, as it should. I did the time delay by choosing one of the mic positions that gave the smoothest curve through crossover and inverting the polarity on the HF. I adjusted the delay for the deepest and most symmetrical notch at crossover. Flip the phase and all goes flat again. Finally, I applied some global EQ to the entire system to clean up any remaining room of summing issues. It is very simple to do, but it takes forever due to the multiple measurements and repetitive nature of things. My results were awesome.

The DBX Driveline 260 can do the job nicely although it is a little limited in PEQ filters. You have to be pretty efficient. I am currently looking at other High End solutions which incorporate built-in Preamp functions so I can ditch my analog preamp. Among the contenders are the Ground Sound DCN28 and the DEQX HDP-4. All are pretty pricey. The DBX is cheap at about $1000."

The equipment I will use:
Crown CTs-1200 amplifiers
JBL 4429 speakers
REW measuring software
Calibrated mic
Focusrite 2i2 sound interface
DBX 4820 dsp

Comments, help and tips are welcome down the road :)

[bubbleboy76]

If anyone has access to 4429 unechoic measurements or spinorama diagrams, now is the time to post them!
4313b, you had it for the 4365. Do you have it for 4429 as well?

[bubbleboy76]

To get all the measurement gear up and running, I started with some delay tryouts.
My experience from the 4365 was that finding the perfect delay makes a big improvemement. But that it also is hard to get clear readings when measuring this from listening positions due to my room interferences.
I tried different delay settings around the approx. 10cm physical offset between the 1200fe-8 and the 175Nd-3 drivers on the 4429.
I measured 1m from the speaker on-axis horizontally, with the mic in the middle between center of woofer and horn vertically.
0.375ms delay on LF gave me the smoothest and deepest cancellation around the crossover frequency when inverting phase on HF.
1/6 smoothing. Single-position measurement.
Speaker position in left corner, where I usually have it. I did not pull out the speaker in the middle of the room. I interpreted Greg post that he had the speakers in their normal position, and that he measured with mic in different listening positions. But this is a guess from my side.

Subjectively, it sounds better. It sounds less like an speaker playing. The sound is more released from the speaker.
Note, this opinion is not from blind-test, so it might very well be imagination from my side! But I can turn on and off the delay on-the-fly in the dbx, and the difference is there every time.

[pos]

Magnus, delay can only correctly be set after the final high pass and low pass filters are in place.

[bubbleboy76]

Magnus, delay can only correctly be set after the final high pass and low pass filters are in place.

Yes I know. But maybe I choose only to do delay-correction, and no other EQ. I suspect that will be the end-result of this.
This was for me to get up-and-running with the measurement gear. If I do other dsp-corrections to get the slopes right, then I will redo the delay measurements.

I have measured 4 mic-positions now, for both speakers, both LF and HF+UHF separately. Now the analysis starts. What could possibly go wrong?! :)

[bubbleboy76]

Mic position 1a: Left speaker. Mic 1.5m on-axis. 85cm to floor. 155cm to roof.
Mic position 1b: Right speaker. Mic 1.5m on-axis. 85cm to floor. 155cm to roof.

Mic position 2a: Left speaker. Mic 5m on-axis. 85cm to floor. On back left corner of sofa. 155cm to roof.
Mic position 2b: Right speaker. Mic 5m on-axis. 85cm to floor. On back right corner of sofa. 155cm to roof.

Mic position 3: On middle of corner sofa seat. 85cm down to sofa seat. 1.09cm to roof.

Mic position 4: On middle of main listening position sofa seat. 67cm down to sofa seat. 1.26cm to roof.

[bubbleboy76]

I wonder if 750Hz LR24 is the theoretical target for 4429? Looks more like 700Hz on my single-position delay test. But I know single-position measurements sucks...
I will post the 4-positions averaged measurements for every driver soon.

[bubbleboy76]

Left and Right 4-position averages.

[bubbleboy76]

Then I guess Greg averaged the averages for L and R, and did the same settings for both speakers. Because his end-result were one set of settings.

This is the curves I should use to generate EQ for getting a more perfect acoustical LR24 slope on, if I understod Greg's procedure right.

[bubbleboy76]

I do not see how to add a target LR24-slope into REW, and generate correcting EQ to my slopes.
REW seems to be too limited. I can only see some kind of generic 24dB target slope, that does not look like Linkwitz-Riley.

Can I do this step with rephase, Thomas?

I am thinking the LF-bump at 1500Hz should be reduced. Maybe an woofer breakup?

The big suckout at 150-250Hz is a room-node, or a speaker boundary interference, I guess. Both speakers are positioned deep into corners, angled 45 degrees into the room.

[bubbleboy76]

JBL TS of 1200fe-8 shows no such breakups at 1500Hz.
But it might be something else causing this maybe, like a resonance in the cabinet, a backwave, or something like that?

[grumpy]

port resonant freq?

[bubbleboy76]

port resonant freq?

Yeah, maybe!

[bubbleboy76]

It's cool how these useless in-room single-spot measurements togheter averages into something seemingly useful/correct.
The lesson of the day.

[pos]

JBL TS of 1200fe-8 shows no such breakups at 1500Hz.
But it might be something else causing this maybe, like a resonance in the cabinet, a backwave, or something like that?

The dip around 1.2kHz and bump around 1.3kHz are visible in the anechoic measurement.
You should probably not worry too much about them as they are some 20dB down already.

[pos]

I do not see how to add a target LR24-slope into REW, and generate correcting EQ to my slopes.
REW seems to be too limited. I can only see some kind of generic 24dB target slope, that does not look like Linkwitz-Riley.

Can I do this step with rephase, Thomas?
Yes, but I have heard that REW also has filtering simulation features.

[bubbleboy76]

Yes, but I have heard that REW also has filtering simulation features.

It would be possible to do in REW, I think, if I had 750Hz LR24 target curves to load in REW as "house curve". I found some excel-sheet that can generate LR24-curves. But I do not have Excel at the moment in my pc, and I do not want it either!

I tried to generate those LR24 target curves in rephase, but I only found how to generrate IR.

I also tried to generate IR from REW to import in rephase, but I could not do that for the averages.

[bubbleboy76]

I will try this tool:
http://audio.claub.net/software/ACD/ACD.html

Which smoothing is appropriate to use for generating EQ? 1/24?

[bubbleboy76]

I found an excel-sheet that could generate 750Hz LR24 slopes to make an "house curve file", so I now can use REW for generating EQ.

As you can see, the crossover-slopes in my system are pretty far from perfect (if this measurement procedure is to be trusted).

[bubbleboy76]

First try of making EQ-settings in REW. This was the best I could achieve while using 6 PEQs per driver at the most (the limitation in my dbx4820).
A 10dB boost from 130Hz and down, I generally use to like.
I worked on 1/6 smoothing granularity.
Comments?

[bubbleboy76]

I entered this settings in my dbx, and did new delay tryouts from the main listening position.
0.354ms delay on LF gave the deepest notch this time. And the notch actually showed up at the right place, 750Hz, this time!

[bubbleboy76]

Any opinions, do you think it will sound better or worse with this EQ and delay applied?

I have tested it now (not blind-test)... :)

I will try to do a blind-test on my friend with the golden audiophile ears.

It became "much more" EQ than I expected to be the outcome of this. I was hoping it would only be some small adjustements.

[johanwholst]

In my experience, you should be cautious not including too much room correction in your EQ. I would concentrate on getting good measurements of the midrange/tweeter. Gated and measured free standing in the middle of the room.
For the bass, I would do near field measurement, to do some correction of the driver itself, and just apply a shelf in the bottom (if needed). No more EQ in the bass except for obvious room modes below 200hz.
Then I would consentrate on getting good crossover slopes. The target curves you are using is a great tool.
When this is all dialed in, you must time align the speaker.

[bubbleboy76]

In my experience, you should be cautious not including too much room correction in your EQ. I would concentrate on getting good measurements of the midrange/tweeter. Gated and measured free standing in the middle of the room.
For the bass, I would do near field measurement, to do some correction of the driver itself, and just apply a shelf in the bottom (if needed). No more EQ in the bass except for obvious room modes below 200hz.
Then I would consentrate on getting good crossover slopes. The target curves you are using is a great tool.
When this is all dialed in, you must time align the speaker.

Thanks for your input!
Yes, if I would try to do fully active settings, then I would do like you suggest. But now I wanted to test this G.T-procedure, for only fine-tuning the passive speaker a bit. But it is still unknown to me how G.T placed the speakers and the mic when he did it. And his living-room setup is probably acoustically much better than mine.

[grumpy]

his living-room setup is probably acoustically much better than mine.

I think you might be surprised. Had a very nice visit not terribly long ago... Yes, the system at the time
sounded great... no, the room did not appear particularly treated in any special way (although the
speakers placement did seem to be a priority in the room... nice to have that option :)).

I'm curious as to how your process will affect how you enjoy your speakers. Looks fun
and educational, regardless.

[bubbleboy76]

I think you might be surprised. Had a very nice visit not terribly long ago... Yes, the system at the time
sounded great... no, the room did not appear particularly treated in any special way (although the
speakers placement did seem to be a priority in the room... nice to have that option :)).

I'm curious as to how your process will affect how you enjoy your speakers. Looks fun
and educational, regardless.

Aha :)

I will keep you posted.

It would be fun if the "main man" would comment here eventually.

[srm51555]

I will keep you posted.



Please do.

[bubbleboy76]

EQ Version 2.
Now, I tried to adapt the whole spectrum to my wish/target, not only the crossover slopes.
I choose to use the 12 available Pre-PEQs in my dsp for the HF+UHF. So I have 18 PEQs there. And 6 PEQs for the woofer.
I choose to not EQ anything below 220Hz, where the big suckout starts. I think this dip is due to speaker positioning in the room.
I used 1/24 smoothing for horn, and 1/6 for woofer.

[bubbleboy76]

My target curves vs. my measured data.

[bubbleboy76]

When I open the dbx, I see it has only 9 pre-peq, not 12.

I have to redo it.

:eek:

[bubbleboy76]

I can not get the EQ2-plan working with the limited amount of PEQs in my dsp. I will stick to the first version, only fine-tuning the crossover-region.
If I would have had around 20 PEQs per driver, then I could do it nicely for the whole spectra.

[bubbleboy76]

REW-files on my dropbox, if anyone wants to play with it.

https://www.dropbox.com/s/nou9noqn3vn1w5c/Measurements%202016-09-09%20JBL%204429%20biamp.zip?dl=0

[bubbleboy76]

I have listened now for a time, and I think definately it is an improvement.
The midrange is clearer / less muddy.

[bubbleboy76]

My longterm vision is to build these guys into a baffle wall, go active on the LF, and add 2269h subs.

[bubbleboy76]

Which frequency and slope would you use for crossover to a potent subwoofer helping the 4429?
Like 4645c, a DIY box with 2269h, or something like that.
Thomas?

[pos]

If I were in your situation (HC setup) I would probably try to restrict the sub to LFE duty only :dont-know:

[bubbleboy76]

OK.
No, I would like the subs to play for music as well.
My vision is to use my small LSR310s subs as balancing subs, the Earl Geddes way. Together with one potent 18" sub like 4645.

[bubbleboy76]

Sold the 4429:s today. New adventures ahead!
Something large.
Something JBL
Something DIY.
The theme is "If I could do it all over again...".

[srm51555]

Sold the 4429:s today. New adventures ahead!
Something large.
Something JBL
Something DIY.
The theme is "If I could do it all over again...".
Any components you can hint us with?

[bubbleboy76]

There will be a separate build thread soon :)

[johnlcnm]

Sold the 4429:s today. New adventures ahead!
Something large.
Something JBL
Something DIY.
The theme is "If I could do it all over again...".
Dang Magnus. I was getting ready to join you on this 4429 venture! Waiting for my copies now!

[bubbleboy76]

Dang Magnus. I was getting ready to join you on this 4429 venture! Waiting for my copies now!

Congrats, 4429 is fantastic! And some extra percents fantasticality can be squeezed out with this kind of EQ I did. So please continue this venture!

[bubbleboy76]

Found this today, regarding mic positions.


REW 5 is the way to go. Be sure to get the Beta version that corrects the curve averaging method. For Mac, it is Beta 20. I use a Scarlett i212 as an external sound card. Most PC's have lousy frequency response so if you want to use your internal card, just do a loop back with the REW software to be sure things aren't too bad. I have a B & K mic, so I cannot comment on the suggestions others have made for mics. Since you don't have the ability to calibrate your mic, I would suggest sticking with something that is good quality or at least a known entity. If your mic has a non-flat characteristic (smoothed over the full frequency range) you will end up with less than desirable results. You need to be able to correct the Octave to Octave balance of your mic to less than 1/2 dB from 100 Hz to 10000 Hz. I plan to provide steps for in-room measurements and EQ soon, but not today. For now, consider REW 5 to be your software. It is excellent. Get a good mic, or at least one with flat (or known) response. REW has the ability to do a loop back correction for your electronics and it also takes a Txt file for mic calibration.

You will need to set up a mic array (one measurement at a time is just fine) around your listening position. 3 to 5 positions will be sufficient. If you try to include too large an area, you will cause more problems than you fix. I started with 5 positions and found that what I measured and what I heard did not agree. I found 3 positions that much more closely agreed measurement and sound wise. Just luck. I would start with 5 positions. one at your listening point, and the other 4 as a square or rectangle around the first mic. I typically space the mics about 2 feet from each other. One possibility is like the number 5 on a die (or dice?) Each of the 4 perimeter mics would be 4 ish feet from each other with the center mic in the middle. You should choose the same height for all mics and that height should be your ear height in your actual listening posture. REW can simply measure each mic position individually and then you can average the 5 into a single measurement. REW only has 30 bins so you need to throw away the individual measurements after the Average has been done (to save space).

Your M2/Crown setup will come with the recommended crossover and EQ already configured. The configuration is for flat response in a chamber and is not correct for a listening room without some adjustment, particularly in the 400 Hz and below region. There are several views regarding EQ above 400 Hz or so. I have had very good results correcting things above 400 Hz if you know something about your loudspeakers directivity behavior. I have a suggested in-room target shape which is not flat. I will publish my suggestion anotherday as I don't have the curves with me today.

This should be a fun project and you will end up with spectacular results.