My attempt to a Seas L18-27TDFC MTM

[requiem]

Okay, I’m trying to get the combined frequency response for a couple of seas L18 drivers in a 60L vented box tuned to 34 Hz.

I traced Zaph's infinite baffle FR and Impedance measurements and simulated the infinite baffle with unibox (using a 1000L sealed box). Then simulated the FR for my box (with two L18 drivers mounted in parallel). And finally simulated the baffle diffraction for each one of the L18s.

Now I’m not sure how should I combine all the data because of the 2 drivers I want to simulate.

Would it be correct to subtract the infinite baffle response with a single driver from zaph's measurements, and add my box's response with 2 drivers?

[Amphiprion]

Why do you want to combine the two? Usually I do <200Hz measurements to make sure my bass alignment is right and >200Hz for xo work, assuming you are doing a 2-way.

[requiem]

Well because the predicted box response is quite different if there's 1 driver or two (see attachments).

My goal is to achieve the minimal phase model using FRD tools, and I’m following this guide.
Should I not combine the box SPL (and combine diffraction only) for the crossover design process?

thank you

[Amphiprion]

My bad - I thought you had measurement equipment. I don't know how to use the FRD tools.

When I design a two-way, I simply measure the actual in-cabinet response of all the drivers >200Hz with good resolution. Those measurements are for the xo work, I just make sure not to include more than 4 or 5 dB baffle step comp by calculating baffle step center frequency and looking at the xo filter transfer function. To get the bass right I measure the impedance and then close mic it. Then I just trust the two match up good, because they always do in a 2-way with decent drivers.

But as far as the FRD way goes, I dunno.

[requiem]

has anyone used FRD tools to design a MTM speaker from measured data? is it even possible?

[cjd]

has anyone used FRD tools to design a MTM speaker from measured data? is it even possible?

Yes.

BDS, FRC, are the keys.

C

[cobbpa]

I did not know it was possible to design networks without measuring equipment. Granted, it isn't optimal...but sounds fun

[cjd]

The trick is normalizing your data.

Particularly when you have different sources, perhaps different technique (test baffle, IEC, infinite baffle) - that needs to be normalized (usually to infinite), then baffle diffraction/baffle step summed in, normalized, minimum phase extracted, modeled with appropriate z-axis offset included...

C

[cobbpa]

The trick is normalizing your data.

Particularly when you have different sources, perhaps different technique (test baffle, IEC, infinite baffle) - that needs to be normalized (usually to infinite), then baffle diffraction/baffle step summed in, normalized, minimum phase extracted, modeled with appropriate z-axis offset included...

C

I realize testing on baffle & in place obviously is more accurate because of driver variations (tolerances) and in-room placement, but how close can a non-testing theory get?
Jed designed my towers from raw data, but neither of us expected anything great. They sound alright, but I'm curious as to how close to optimal one can get without testing equipment.

[cjd]

My MTM's were designed this way. I found nothing to change when I subsequently measured them. And they won the budget class at the Chicago DIY event last year.

YMMV.

My 3-way attempts in this way failed miserably.

C

[Jed]

I did not know it was possible to design networks without measuring equipment. Granted, it isn't optimal...but sounds fun

Cobbpa, that's how I designed your speaker/crossover. I normalized the data like cjd commented in his post. I wouldn't be suprised if your speaker's predicted response is VERY close to real world measurements. The FRD tools are very powerful and suprisingly accurate. And yes it isn't hard to simulate an MTM given these tools once you get to know how to use them. I'd start with manufacturer data first though because it is easy to subtract out the IB response and add in the baffle target response. Especially SEAS since they tend to be more precise with their graphs than some other folks.

Jed

[requiem]

The trick is normalizing your data.

Particularly when you have different sources, perhaps different technique (test baffle, IEC, infinite baffle) - that needs to be normalized (usually to infinite), then baffle diffraction/baffle step summed in, normalized, minimum phase extracted, modeled with appropriate z-axis offset included...

C

I've tried to simulate a couple of existing designs using FRD tools and manufacturers data. I found a pretty good correlation with MT speakers, like zaph's speaker #17 for instance.
With MTM's it's a different story. For example, I tried to simulate the Modula MTM (RS180 - 27TDFC) using PartsExpress' data and the predicted response is far from actual measurements...

This is what I’m doing for the woofers section in FRC:

Driver Base: Parts express - RS180 Exported CLIO Frequency Response



Box Measure: Unibox simulated Infinite Baffle (1000 Liters Sealed enclosure) with Parts Express' measured Thiel and Small values. FOR a single RS180 driver.

Box target: Unibox simulated 1 ft^3 box tuned to 34 Hz, with two RS180 drivers in parallel.

(Both responses are calculated at 1 Watt nominal power)

Baffle measure: Baffle diffraction Simulation for an IEC baffle with one 7'' driver (I asked parts express the measurement conditions). Axis distance is 1 meter (39'').




Baffle target: Baffle diffraction simulation for the Modula baffle as specified in the original design. Axis distance: 1 meter (39''). Diffraction for both woofers is the same as they are placed symmetrically on the baffle so I just used the WMTMW arrangement type and exported Axis sum for drivers 3&4 (mid bass drivers).



Room LT/EQ: off

All the data is normalized as suggested in the FRC manual:

driver base: extend top slope - select low slope -12dB
box measure: extend top slope - extend low slope
box target: extend top slope - extend low slope
baffle measure: Extend top value - baffle gain to zero
baffle target: Extend top value - baffle gain to zero

The Response is combined and Minimum Phase model extracted

the tweeter response is calculated in the same way except that the box measure/target are turned off.

The combined response is then loaded in LspCAD with the designed crossover... and what do i get? 8O 8O 8O (See attachment)

The xover components are slightly different, as adjusted by lspCAD. But this is only for testing/learning FRD tools usage and the transfer function is quite similar.

What am I doing wrong?
anyne has RS180 data other than PE's or Jon's measured data?

thanks

[Jed]

When you import the summed RS180 (data reflects 2 in parallel) woofers into LSPCAD you only need 1 woofer represented in the project tree. That's one error I'm seeing. Make sure you adjust the Z axis value for the woofers as well. I think Jon mentioned this value should be set at 40mm if I recall correctly. If you want to have 2 woofers in the project tree you have to have data collected for each woofer, not combined like you simulated originally. That's why you're seeing all those peaks and plateaus.

Jed

[requiem]

When you import the summed RS180 (data reflects 2 in parallel) woofers into LSPCAD you only need 1 woofer represented in the project tree. That's one error I'm seeing. Make sure you adjust the Z axis value for the woofers as well. I think Jon mentioned this value should be set at 40mm if I recall correctly. If you want to have 2 woofers in the project tree you have to have data collected for each woofer, not combined like you simulated originally. That's why you're seeing all those peaks and plateaus.

Jed

Thanks Jed. I didn't think of that. I will use a single driver representing the paralleled two... and the impedance should be then divided by two manually, right?
the Z axis value is the distance between the cone center and the front of the baffle?
finally... is it possible to collect SPL data for each woofer? Adding each driver's SPL will produce different results than simulating both at once.

[Jed]

Thanks Jed. I didn't think of that. I will use a single driver representing the paralleled two... and the impedance should be then divided by two manually, right?

Yes, if you are representing 2 woofers in parallel.

the Z axis value is the distance between the cone center and the front of the baffle?

Yes

finally... is it possible to collect SPL data for each woofer? Adding each driver's SPL will produce different results than simulating both at once.

I think so. Calculate FR for each woofer with BD effects and phase built into the FRD and ZMA files. Then, sum the 2 in LSPCAD by placing both the upper and lower woofer in the project tree and see what you get. I've been using speakerworkshop so maybe someone with LSPCAD can speak from experience on this one.

[requiem]

Thanks for all the help so far, I've been doing some more testing with existing designs and FRD tools and I think the results I’m getting are close enough to begin messing with my own design.

here you can get the seas measurements I’ve been using (thanks Davey, Dennis H).

So here are some results.
Zaph's speaker 17



The gray line (reference) is the traced Zaph's measured response. The tweeters ultrasonic ringing in the >20KHZ region doesn't show because the used base measurements end at 20KHZ and I used the "extend top slope" in FRC normalization.

This is how it would look using seas measurements with Zaph's X-over and NO FRD tools


Quite different huh?


for an MTM example, I simulated the seas odin design and this is what I got adding separate drivers’ response as Jed suggested:



and the predicted response using combined woofers response in unibox/BDS is quite similar indeed



the green line (reference) is the original measured response, I’m particularly curious about where the dip at the xover point comes from. The transfer function looks very attenuated for the woofer in that frequency range however.



As for the parts express drivers goes I gave up until I get better info about the measurement conditions, when I asked parts express all they said is...

"I am pretty sure those were tested on a standard baffle that is 1 meter by 1 meter which is mounted flush directly to a wall"

I thought they mean an IEC baffle but now I just don't know. The published measurements don't look like a typical infinite baffle response either...

Well what do you think. Are these results close enough?

[Amphiprion]

Looks good to me, but that's a LOT of drop in the woofer filter tansfer function from 100Hz-1kHz. More than 6dB baffle step comp and enough series L in the woofer filter to get some ringing at 100Hz (I'm guessing the woofer is ported and has an impedance peak there?).

[requiem]

Looks good to me, but that's a LOT of drop in the woofer filter tansfer function from 100Hz-1kHz. More than 6dB baffle step comp and enough series L in the woofer filter to get some ringing at 100Hz (I'm guessing the woofer is ported and has an impedance peak there?).

If you take a look at the manufacturer's response curves here 6dB seems about right (on axis). This xover was designed by Joseph D'Appolito's and i'm just trying to simulate the measured response using FRD tools, don't know any further details.

[Jed]

the green line (reference) is the original measured response, I’m particularly curious about where the dip at the xover point comes from. The transfer function looks very attenuated for the woofer in that frequency range however.

Try changing the Z offset value and see how it affects the dip at X over point.

Jed

[requiem]

Try changing the Z offset value and see how it affects the dip at X over point.

Jed

I did, as a matter of fact i changed it from -90 to 90mm in 10 mm intervals to see what happens and that's the best response I got, the final setting is at 30mm (in LSPCAD positive values mean away from the listener).

does anyone know any other well documented all-seas MTM design I can try to simulate?