Project M2 DIY Thread

[lesliew]

So you are building two more? Adding another listening room?

Thanks for the info.

-Steve

These two are not for me , Actually building 2 Full size M2's then a modified center consisting of just the M2 waveguide/Horn then possibly a separate Sub18.

I've had 3 people ask me to build them a Pair of M2's in the last year but said no, seems like I finally caved in. This will hopefully get me back on the road to building the SCL-2's

[sguttag]

I see. I used the genuine JBL gasket for the M2 waveguide and, I guess, I'm using the genuine gasket for the 2216nd though none is listed in the exploded view. It is a one piece 15" ring so the box it ships in is large.

I would think it would be getting tough to build M2 clone for others since suppliers like Speaker Exchange are clamping down on M2 parts. Are the European distributors having any issues getting the M2 parts? I can say as a JBL Pro reconer, I didn't have problems, per-say getting the parts but there were long-lead times (months) and not everything becomes available at one time. My guess is that because Speaker Exchange was moving parts at a quantity that JBL didn't like for "repair parts" that JBL might have clamped down on them. But that is just a guess.

[lesliew]

I would think it would be getting tough to build M2 clone for others since suppliers like Speaker Exchange are clamping down on M2 parts. Are the European distributors having any issues getting the M2 parts? I can say as a JBL Pro reconer, I didn't have problems, per-say getting the parts but there were long-lead times (months) and not everything becomes available at one time. My guess is that because Speaker Exchange was moving parts at a quantity that JBL didn't like for "repair parts" that JBL might have clamped down on them. But that is just a guess.

I’ve not contacted speaker exchange myself recently, and I won’t be purchasing these parts just having them shipped to me. A guy over in Japan who I sent my CNC files to didn’t have a problem getting parts from SE that was about 5 months ago. Im sure availability varies.

I’m going to need 12 JBL synthesis 8" woofers for the SCL-2’s that I'm planning to build, sounds like I need to talk to you

My son will be up in Annapolis next week; at the Naval Academy.

[lesliew]

I used the genuine JBL gasket for the M2 waveguide and, I guess, I'm using the genuine gasket for the 2216nd though none is listed in the exploded view. It is a one piece 15" ring so the box it ships in is large.

Do you have a link to these for future reference.

[sguttag]

The genuine M2 waveguide gasket IS on the exploded view. It is called "Main Gasket, Horn" and on the exploded view, it is at about 9:30 O'Clock

The LF driver gasket isn't listed (that I could find). I just looked at other speakers that used the differential drive 15" woofer and noted that they all seemed to be using the same gasket...the same way that the 2225, 2226...etc used the same "Spaghetti" gasket. The gasket I'm using is JBL part number 360-70002-01. I have not done a test fit of it yet but I can't imagine why it wouldn't fit.

[lesliew]

Can anyone who has access to the M2 horn and CD measure their overall depth.

I need to know how much space inside the cabinet it takes up.

Thanks

[pos]

Hi Leslie,

I measured 143mm from the back of the horn (the part that sits on the baffle) to the back of the D2.
You might want to use angled spades, as the straight ones I am using do protrude beyond that.

[sguttag]

So I ran into a snag on my Q-SYS implementation of the FIR filters using a CORE 110...I ran out of processing power when trying to do five M2s. One set of FIR filters (high/low plus the LF PEQ supplement) cost 47% of the CORE 110 processing. It stays that way (just 47% processing used) until one puts in the fifth speaker. A problem that few would have, I'm sure. I'm doing a Dolby Atmos system set up for cinema so I'm using 5 speakers. Now, going up to a CORE 510 alleviates the processing power issue (all five only use something like 12%) or using a conventional "voicing" from the jail broken Crown amplifier spec on the CORE 110 also consumes very little processing. The FIR filters, they, apparently, consume a bit of the QSC CORE.

[pos]

FIR typically consumes much more computational power compared to IIR.
That is its only drawback.
One of its advantages we are making use of here compared to IIR is its predictability: any convolution engine should produce the same transfer function for a given FIR.

As the correction here is minimum-phase and pretty simple, it is of course possible to reproduce it with IIR EQ and filters (in fact that is what the BSS/Crown processor does in the first place ), but you will have to measure the output and adjust accordingly to make sure you get the same transfer function as what you get with the FIR.
With IIR you have to deal with different conventions (eg constant Q vs proportional Q EQ, non symmetrical EQ, shelving cutoff frequency, Q vs bandwidth, Q vs slope, etc.) from one processor to another, and then also biquad calculation approximations which will even differ from one internal sampling frequency to another in a given processor!

IIR is a pain, but can be made to work

[sguttag]

Yeah...I'm just having to do some soul-searching on this one. I know that the FIR filter you (pos) works well. I've measured it, listened to it, at length now, with different content. I was only using one speaker during these tests since it was built as a prototype before making the actual five. So I didn't realize the train wreck was coming!

It wasn't until I started to experiment with other processes that will be implemented in the theatre that I got an error because I was using too much CPU. At the time, I had two FIR processing blocks in the system while I was comparing different amplifiers (QSC's DPA8.4Q (audio supplied via QLAN network) versus a QSC DCA 3422 (audio supplied via traditional analog cables). It was just a scratch page so I was merely adding on to it, not removing anything (unless the idea was abandoned).

So, do I spend the money and go for the bigger CORE or work with Crown settings in QSC's "Voicing" module and massage them to a good response? Or, do a hybrid? That is let the primary channels use the FIR filters and Left-Center/Right-Center use IIR/Voicing? I really doubt the later since I really want as close t uniform response/phase as possible channel-to-channel.

I haven't done NEAR the listening tests on the IIR/Voicing as I have on the FIR (once you have something that works well, why look further?). I only did the Voicing blocks because I did them first, before I had the QSC compatible FIR filters via rePhase.

The cost of going to the larger CORE (510c) is significant. Not only is the 510 essentially double the price of the 110, the 510 comes with zero Input/Output modules so each of those, as needed must also be purchased. <sigh>

[pos]

Given your setup it would probably make more sense to use IIR blocks for all your speakers, adjust them till you get a good match (or hire someone to do so), and then possibly use the available FIR power to linearise the phase (LR crossover only, not the ported box) of all your speakers.

What settings did you use in the IIR variant? What do you call "voicing" here?

[sguttag]

The "voicing" is the term that QSC uses for Q-SYS when one applies parameters of a crossover/filter in tabular form.

Last night I was able to do some experiments and also use Q-SYS' "Response Analyzer" component to compare the responses from the FIR filters to the "Voicing" plus "input filters". I was able to massage the voicing/filters to the point that the responses/phase were just about identical. The shelving filters were not a match to JBL/Crowns (big surprise there, not) so a little adjustment was needed on the HF section shelving filter. On the LF section, thus far, I've added an additional PEQ at 150Hz but that may be a matter of working on the shelving filter to get a better match. I did that on the HF section and was able to get it to match pretty well without adding an additional filter though I also had to alter some of the specified PEQ filters to finish the job.

Response is +/- 0.3 from 20-20KHz as compared to the FIR filters (using the response analyzer) with most of it less than 0.1dB variation.

I then set up an A/B listening test. To my ears, they sound the same. I brought in the wife and son to give an A/B listening. They couldn't tell the difference. It is tough for them because at first, they would notice the slight glitch when they were switched (not a noise but you knew that something was switched). I suspect that I'll have an acceptable solution.

[pos]

That is great !
Now if you want to make use of some available FIR power you would have left, you can toss a phase linearization FIR into the mix.
This does not require too much taps with the appropriate centering, 256 taps at 48kHz is enough to address the 750Hz LR 36dB/oct acoustical crossover:

rePhase settings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(copy and load from clipboard in rephase)

[sguttag]

Wow. Thanks! More fun experimentation tonight!

[sguttag]

So I did some experiments. I worked on the QSYS voicing a bit (was able to eliminate the filter I added from the LF crossover section by better fitting the QSYS Shelving filter). I did make a QSYS FIR filter out of the LF phase correction provided by Thomas.

My lack of knowledge of what is going on, phase wise is coming into play. The phase response without the phase linearization FIR filter seemed spot on with the full FIR implementation of the crossover but that isn't the case (according to the response analyzer) when the added FIR filter is applied.

No change was made to the "Input Filter" from the last post so I won't repost here.

Current Voicing (Input filters not shown):



Comparison of LF Response using QSYS Voicing+Input Filter and FIR LF Filter (Ideally, ruler-flat)


Comparison of HF Response using QSYS Voicing+Input Filter and FIR HF Filter.


LF Comparison with Phase Response (Normal)


LF Comparison with Phase Response, with added FIR Phase Linearization to the QSYS Voicing.


No RTA or significant listening tests have been made with/without the Phase Linearization (just a quick 5 minutes)