TOPIC: JBL M2 crossover with the openDRC

Here is an openDRC preset to act as an active crossover for the JBL M2 loudspeaker
www.dropbox.com/s/8fts4eixivap5t4/OpenDR...%20JBL%20M2.xml?dl=0

One openDRC and one stereo amplifier per speaker would replace the (pricey and noisy) Crown I-Tech DSP-powered amplifiers that are normally meant to be used with the M2.
Using this preset the openDRC should exactly replicate the behavior of the original preset in the Crown.

Please note that this is untested as I do not have these speakers at hand.
Use with care and at your own risk

The crossover, EQ, level and delay alignments are entirely made within the FIR section, leaving the IIR slots available for further tweaking and room EQ.
The FIR was generated using rePhase 0.9.9.
This is a purely minimum-phase correction, and the implied delay is only 1ms (to which the hardware delay of the openDRC has to be added...).

openDRC configuration:
* openDRC preset: www.dropbox.com/s/8fts4eixivap5t4/OpenDR...%20JBL%20M2.xml?dl=0
* One openDRC per loudspeaker (2-way crossover mode)
* Input channel mode must be explicitly set to L and R.
* Output channel 1 is the upper channel (compression driver)
* Output channel 2 is the lower channel (woofer)

crossover settings:
* M2 crossover analysis: docs.google.com/document/d/1bzJyui75ZG_-...?usp=sharing&rm=demo
* rePhase preset for the upper channel: www.dropbox.com/s/oqlmbdjn6dxa5gg/JBL%20...%20high.rephase?dl=0
* rePhase preset for the lower channel: www.dropbox.com/s/pxywfa8pko5xo0v/JBL%20...-%20low.rephase?dl=0
(these presets require rePhase 0.9.9 and up)

Wow... this is neat! How did you get to work on this project? Do you have these nice monitors at home? @ 20k a pop, it's quite a fancy Hifi! heheheh

Thanks again for sharing with the community... Very cool!

Half of that price comes from the Crown amplifiers
Methinks that for home use each of these amps would be advantageously replaced by a miniSHARC and curryman DAC together with a pair of Hypex NC400 amps on a single SMPS600

And... no I don't have these speakers - yet

pos wrote: That does sound like a neat setup! Let's vote on who should contact JBL directly to pitch that idea...

Hehe, I also really like this setup

miniDSP already has some very fruitful cooperations (Dirc, Linkwitz) so maybe JBL is one more candidate

kind regards, Daniel

Bonjour pos,

neat project and somewhat motivating. With two OpenDRC boxes I also want to activate a pair of speakers and provide DRC based on AcourateDRC, and crossover by FIR.

Motivation
Of course, one could easily use the crossover functionality (IIR) implemented in the 2x2 plug-ins. But you get phase distortion produced by higher crossover slopes and you loose many biquads (4 per low-/highpass or subsonic). AcourateDRC (2.0 and up) has an excellent room correction algorithm, using many FIR taps for amplitude correction and the remaining taps (about 20ms, 1000taps) for phase correction and using as many free IIR-biquads as possible for accourate low frequency room correction. Consequently, the phase distortions produced by high crossover slopes will not be corrected adequately.
Solution
Combination of the FIR-linear crossover/subsonic (created by rephase) filter A and the FIR-IIR-AcourateDRC correction filter B (based on the measurement with the FIR-Xo).
Problem
Merging the two FIR filters A and B into one FIR filter C

Can you help me?
Do the files match sufficiently to be correctly convolved (based on 48kHz, 6144 taps)? May HOLMImpulse do the A*B operation, in view of the 6144 taps listed in the .xml file created by rephase and AcourateDRC? Does an Excel/Calc math operation could also do the convolution? What is your experience?

Thank you for your suggestions.

Hello planetti

To "mix" two impulses you will need to convolve one with the other.
That is indeed what HOLM does with the product functionality, but I have never tried to export and use the result.
Sox can also be used for this. It will perform a FFT convolution and the resulting impulse will be the same length as the original one (the one used as the source). Not sure of the result either.

The best solution would probably to do a direct time domain convolution of the two impulse, resulting in an impulse which length will be the sum of the two source impulses.
I don't know if any off the shelve utility exist to do that. I have one in my toolbox that I could distribute if I find some time to package it...

Hi,
in the "convert" page of Align2, you have a function to convolve two impulse responses.
www.ohl.to/audio/downloads/align2.zip
In the same soft, you may then convert it to a bin file and adjust gain of the bin file.

Regards

Hi Jean-Luc

Nice feature.
Is it a frequency or time domain convolution?
What is the size of the final impulse?

Hi pos, hi jlo,

thanks for your suggestions.

Since I am not a programmer, a solution seems to be not obvious for me. None of the programs suits out of the box because of the individual input file specifications needs. Align2 reads .wav data, HOLMimpuls wants .txt in "frequ;dBfs;phase" form.

OpenDRC plug-in provides the data in the manual mode in the form of "b0 = [number]", which can easily copied to a .txt file.
AcourateDRC integrates the data in the Opendrc .xml file as "<subpara row="0">, [hex number]", Rephase creates a .bin file.

Any ideas to provide a convolver tool which may process the .txt file (which seems to be the most compatible with the OpenDRC users)? (This is a cry for help! )

A technical question is: Does any convolver merge the responses always correctly (when the basic parameters match)? It seems that the max pulse in the created Rephase file is perfectly symmetric in the middle at b3072 = 0.986236572265625, whereas the AcourateDRC is not symmetric and has e.g.
b3068 = -0.22571039199829102,
b3069 = 0.00848750676959753,
b3070 = -0.35271111130714417,
b3071 = 0.22794613242149353,
b3072 = -0.24237167835235596,
b3073 = 0.0659007728099823,
b3074 = 0.6056585907936096,
b3075 = 0.13662630319595337.
Is this a problem?

Thank you for help. Perhaps Devteam had already some thoughts about that.

@pos
Is it a frequency or time domain convolution?
What is the size of the final impulse?

I use FFT convolution from octave.sourceforge.net/octave/function/fftconv.html
The final size is the sum of both sizes less 1.

@planetti
I suppose that Acourate has a bin output and rephase also
so a tool that convolves two bin files (f32 files) would be ok ? with gain and length adjustment...

The tool may convolve
- .bin file from Rephase
- fir data out of the .xml from AcourateDRC (yes, the current ver. 2.1.3 creates only the completely loadable .xml file for the Opendrc 2x2Plug-in because of the use of the 14 IIR biquads)
- .txt file from the copy of the fir window within the Opendrc 2x2Plug-in in the form of "b0 = [number]"
(- additionally the common .wav standard?).

Output files respectively....Gain control might also be nice.....The lenght of the filter should fit into the OpenDRC somehow.... That would be GREAT!

In the report www.computeraudiophile.com/content/556-a...ization-walkthrough/,
down in chapter "Apply TD-Function - Convolution of Inverse and XO" (see attachment convolution) there is an indication to convolution in time domain.
I suppose that Dr. Uli Brueggemann, the friendly programmer of (((acourate))), could answer all our questions easily ...

Hello,
I've got all the JBL hardware on order and am looking for a DSP solution.
So long as the learning curve isn't too steep, this might be it.
How many different programs are required to measure the system and use the measurements to make meaningful changes in the DSP?
If it's any more than REW and a minidsp plugin, I'm gonna need a LOT of help.

Hello AaronM,

If you want to exactly reproduce the EQ and filters points of the original Crown DSP amplifiers that come with the M2 then you should look at the first post of this thread.
You will need a module with enough EQ points. The openDRC is a good candidate (one per speaker) and the entire correction can be done in FIR (albeit still minium-phase for now, until we get accurate phase measurements...) which is a guaranty of the lowest possible quantization noises.
If you use the exact same driver complement, same passive protection network and same enclosure volume and port tuning, and same amplifier gain for all channels then you should be good to go without even needing a single measurement, and you should get the exact same response as the original M2+Crown package.

Then if you want to EQ the speaker to tailor it to your room, you will obviously need measurements, and for example use the IIR capabilities of the openDRC to do some additional EQ, or possibly include them in the FIR.

I removed the openDRC and rephase preset links from the first post, because of a bug in rephase 0.9.8 and 0.9.8 that causes a polarity error in the low pass correction (LR implementation bug).
Simple reversing the polarity of the LF section would solve the problem, but I will come back with new preset files when a new rephase version is released.