PDA

View Full Version : JBL M2 measuring curve to DSP filter setup.



sebackman
02-08-2015, 03:28 AM
I知 wondering if anyone has seen an SW that can calculate active filter setting for a given measuring curve given certain filter (EQ&PEQ) parameters.

When JBL decided the parameters for the JBL SDEC (aka BSS BLU160) they use for active filter and EQ they must have had a program determining the first set of filter parameters and from that done additional listening and measuring.

What I知 looking for is a program where I can feed a measured curve and then give the program calculation boundaries in form of what EQ痴 and how many are available.

Par example;
high shelving slope 6db
high shelving slope 12db
low shelving slope 6db
low shelving slope 12db
PEQ Q=1-10
+/- 6db for all filters
Resolution 0,1db
max 14 filters in series used (like in JBL M2 HF below) or less

http://www.audioheritage.org/vbulletin/showthread.php?33972-JBL-Master-Reference-Monitor&p=354293&viewfull=1#post354293
Post #500 (coutesy of member 4313B)

It should not be so complicated programming a Monte Carlo simulation, or similar, to run 20.000 iterations (or more) using 0,1 db increments with all filter combinations to determine what combination of 14 filters (or number of choise) would result in a reasonable flat curve. From there more listening and measuring can be done.

To do that by hand would take ages with 14 filters.

Any thoughts

Kind regards
//RoB

macaroonie
02-08-2015, 05:39 AM
I suspect that you need to have real word measurements of the intended driver set sans networks.

As has been mentioned JBL set out to provide as maximally flat response as is reasonably possible out of the box. Whilst the DSP units can be doped to adjust for room respomse there is no way that the OEM can factor this in > They need to supply a constant.

Earl K
02-08-2015, 07:02 AM
I’m wondering if anyone has seen an SW that can calculate active filter setting for a given measuring curve given certain filter (EQ&PEQ) parameters.

When JBL decided the parameters for the JBL SDEC (aka BSS BLU160) they use for active filter and EQ they must have had a program determining the first set of filter parameters and from that done additional listening and measuring.

What I’m looking for is a program where I can feed a measured curve and then give the program calculation boundaries in form of what EQ’s and how many are available.

Par example;
high shelving slope 6db
high shelving slope 12db
low shelving slope 6db
low shelving slope 12db
PEQ Q=1-10
+/- 6db for all filters
Resolution 0,1db
max 14 filters in series used (like in JBL M2 HF below) or less

http://www.audioheritage.org/vbulletin/showthread.php?33972-JBL-Master-Reference-Monitor&p=354293&viewfull=1#post354293
Post #500 (coutesy of member 4313B)

It should not be so complicated programming a Monte Carlo simulation, or similar, to run 20.000 iterations (or more) using 0,1 db increments with all filter combinations to determine what combination of 14 filters (or number of choise) would result in a reasonable flat curve. From there more listening and measuring can be done.

To do that by hand would take ages with 14 filters.

Any thoughts

Kind regards
//RoB

The Auto-EQ window within REW (http://www.hometheatershack.com/forums/downloads-area/47460-v5-01-beta-downloads-asio-support.html#post1020106) can spit out EQ suggestions ( & those filters become manually tweak-able from within REW's filter window ) . The user always needs to establish a target curve, that REW then auto-EQs against .

Obviously you'll need to learn how to make meaningful measurements with REW ( & how to use whatever hardware you decide to buy to capture the sweeps / not exactly trivial endeavors ) .


:)

pos
02-08-2015, 07:37 AM
Doing 14 manual EQ points is not that hard when the measurement is of good quality and the analysis is good.
What is hard, as Earl noted, is to get an accurate and meaningful set of measurements, and to know what shall be EQed and what shall not.
JBL uses an averaged measurement over a 30ー/10ー listening window, in anechoic conditions (direct sound).
When a loudspeaker is EQed using this data (like the M2 is) the only thing that the user needs to adjust is the room interaction down low, and possibly up high (target cuve).

sebackman
02-08-2015, 10:46 AM
Hi,

Thank you for the answers.

The idea is to take real measurements on the elements without passive networks as input, ie in anechoic conditions.

I already do have good measurement SW and calibrated card/mic (Earthworks), so quality measuring is not a big problem. I have access to a large hall and/or can measure outdoors.

My plan is to copy the M2 topology with an separate EQ section for the speaker and a section for room correction in series. That way I know that the speaker is reasonable flat and the rest is the room. I can then change the room but keeping the anechoic setup intact.

I use the same XO as M2, BSS BLU800, albeit I use digital input.

Are there any alternatives to REW? I had a peek and it does some of what I知 looking for.

More ideas?

//RoB

pos
02-08-2015, 11:11 AM
If your unit can load arbitrary FIR coefficient then there are software out there that do automated correction based on inpulse inversion like DRC-FIR, and other that can calculate biquads and output a FIR like PORC. Align2 will let you both, and help you with the analysis.

That said, If you look at the M2 EQ you will see that is it pretty "simple" and easy to build manually (even it it of course requires multiple iterations with listening sessions to validate the choices, but same would go for automated solutions, with less control...).

sebackman
02-11-2015, 11:00 AM
Hi

Thank you for the replies.

My unit can indeed to FIR and maybe that is the way to go. I need to do some homework on that.

What I was looking for was really IRR filters optimizers. I will try REW but would also welcome other alternatives.

Kind regards and thank you
//RoB