Can anyone help with Alignment?

[Flaesh]

There are xt1464 in them. And nd1480 and 15nd930.
But it is not known whether the crossover there is good enough.

 

[Keith_W]

@Flaesh is right. You don't have to learn it all at once. DSP is an incremental process that you will revisit over and over. You will learn to take better measurements, develop a philosophy on what you want to do / avoid. We all start off this way. You may eventually disagree with some of the things i've said, and that's fine. I have my way of doing things, it does not have to be your way. Anyway, let's look at your measurements.

First, I should commend you on the quality of your measurements. I only looked at the left measurements, and this is the left woofer. I removed the 1 cycle FDW and examined the ETC for the first reflection. Earl Geddes said that you should not try to absorb reflections when taking these types of measurements, you want them to stand out like a sore thumb so that you know where to apply your windowing. With this measurement, you should be able to obtain a clean reflection-free measurement down to 110Hz. That is excellent for somebody who is measuring at home. I can't get that low, and I sort of know what I am doing!

I also looked at the distortion measurement (to make sure you're not measuring too loud) and the waterfall (to check the SNR). These were both excellent as well. Great job! Every measurement you take needs to be thoroughly examined like this, and rejected and repeated if necessary. I'm not going to do it for you. I'll take a quick look to make sure it's OK, but this is something you need to do while you are taking your measurements, not after. Because it's annoying to have to set everything up again to repeat a measurement.

Regardless, I don't think you need to obtain a reflection-free measurement below 110Hz, because you will correct any freq lower than this together with the room later. So for your purpose, this is good enough.

Same can't be said about your midrange measurement, though. Don't know what happened there. Regardless, this is below your XO point so you won't be correcting it. This measurement is not as good as your woofer, but it's good enough.

This is a comparison of your native woofer measurement (red) and after the amplitude linearisation you applied (green), compared to the manufacturer's data sheet on the right. There are two problems here: (1) the measurement of your woofer does not match the manufacturer, and (2) your post-EQ measurement is not flat.

I don't know what measurement distance you used, but one reason for the discrepancy is the baffle step. Manufacturers measure their drivers on test baffles, so it is normal for your measurement to differ to published measurements. You need to measure far enough so that you can see the baffle step and compensate for it. The baffle step is a volume loss of about 6dB over 4 octaves, and its centre frequency can be calculated with F = 109500/d (where d = baffle width in mm), or F = 4311/d (where d = baffle width in inches). This is why you measure 2x baffle width away from the speaker as a minimum.

Staying with the left woofer, and assuming that you have actually measured the baffle step ... here are 3 curves. From top to bottom: purple the measured response of your woofer with 1 cycle FDW applied, minimum-phase copy; green the EQ you generated, blue the EQ I generated. You can see that my EQ is flat. You do this by setting a flat target in REW's EQ, then move the target to an appropriate SPL, then get REW to match response to target.

BTW, you still haven't told us what DSP you will be using, so I don't know how many biquads you have available for correction. If you have fewer biquads, your result won't look as nice.

There is no point examining the results of the next step in detail (i.e. after application of the crossover) because this step is not completed. All your curves need to look like this before you apply the crossover. But I did take a quick look:

The faint curves are your woofer, mid, and tweeter with linearisation and XO applied. You can see you have a major problem at 2kHz. Without knowing what XO you used, I don't know why you have this cancellation. This is why I asked you (twice) to include the XO in the measurements. The PLAIN XO that you generated in REW.

This is what you need to do:

1. For each measurement, apply FDW of 1 and generate a minimum-phase copy. For measurements of your tweeter and midrange, you can probably go higher than a FDW of 1, but you need to calculate what the ideal FDW should be.
2. Working on the minphase copy, linearise each driver so that it is flat.
3. Generate an XO in REW and INCLUDE THIS IN YOUR MDAT. Before you convolve this XO with your measurement, check that it sums to flat (REW Trace Arithmetic A+B).
4. Convolve the XO with the measurement with REW Trace Arithmetic A*B.
5. Sum all the drivers again and check the result. Consider inverting the polarity of one driver to get it to sum to flat (SPL&Phase tab, right click, Invert).

 

[Flaesh]

what DSP you will be using

https://www.audiosciencereview.com/...help-with-alignment.66272/page-6#post-2414927

 

[Keith_W]

https://www.audiosciencereview.com/...help-with-alignment.66272/page-6#post-2414927

D'oh! There was someone else using that same DSP unit which I commented on, can't remember if it's the same guy.

 

[3ll3d00d]

First, I should commend you on the quality of your measurements. I only looked at the left measurements, and this is the left woofer. I removed the 1 cycle FDW and examined the ETC for the first reflection. Earl Geddes said that you should not try to absorb reflections when taking these types of measurements, you want them to stand out like a sore thumb so that you know where to apply your windowing. With this measurement, you should be able to obtain a clean reflection-free measurement down to 110Hz. That is excellent for somebody who is measuring at home. I can't get that low, and I sort of know what I am doing!

there are reflections much earlier than this, the one you pointed at is certainly larger but there is a bunch of it before then though using an FDW is probably good enough for the woofer if you're just roughly drawing a crossover

all of these measurements have ref time = 0 though so I assume no timing reference was involved? basically unusable for xo design and the amount of contamination in the measurement (+ the lack of off axis) means one can't expect too much of the resulting speaker. I think OP really needs to learn how to measure... vituixcad has a good guide for rew https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_Measurement_REW.pdf and there's a thread here too https://www.audiosciencereview.com/...ents-spinoramas-with-rew-and-vituixcad.21860/ which goes a bit more step by step

 

[3ll3d00d]

if you zoom in the woofer IR, it's more obvious how messy it is with respect to reflections

part of that is because it has a rather extended frequency response (can't imagine who needs a woofer to play to 17kHz, it must be like a laser beam by then!) as usually the high frequency noise is naturally filtered out by the speaker itself.

 

[Keith_W]

Given how early they are, do you think they are internal cabinet reflections?

 

[3ll3d00d]

I didn't see a cabinet pic but I would think not, 4ms is typical sort of floor bounce distance unless you lift it high up

 

[Robbie010]

here is a brief overview of the measuring setup:

The speaker is brought in to the middle of the room and placed on a table, the rear, front and side walls are all at least 1.2m away and I used cushions and throws to cover reflective surfaces (TV, leather sofa etc).

There is a rather funky ceiling light that is only 42cm and 76cm from the drivers respectively, so this will be the first reflective item:

 

[Robbie010]

Using an online refection calculator I think the first reflections will be:

Midrange - Ceiling Light 0.96ms/519.49hz - Ceiling 3.97ms/125.88hz

Woofer - Ceiling Light 2.83ms/176.69hz - Floor 4.57ms/109.51hz

Tweeter - Ceiling Light 1.72ms/290.56hz - Ceiling 5.24ms/95.47hz

 

[Robbie010]

Using the details above, I applied the windows to each native sweep and I get this:

 

[Robbie010]

Playing around with the EQ in REW, in order to flatten the entire response of each driver:

Tweeter between 1khz - 20khz, Target Level to 70db, +/-3db, it gives me 3 EQ filters:

Midrange between 1khz - 20khz, the Target Level is 74db +/-3db, it gives me 3 EQ filters:

Woofer between 400hz - 20khz, the Target Level is 72db +/-3db, it gives me 6 EQ filters:

The first issue here is that I only have 6 EQ filters on each output of the Crossover / DSP unit........

 

[Robbie010]

I assume its an issue that the window applied to the woofer cuts off the frequency response below 400hz?

 

[3ll3d00d]

There is a rather funky ceiling light that is only 42cm and 76cm from the drivers respectively, so this will be the first reflective item:

that's like a torture test for this sort of measurement

I assume its an issue that the window applied to the woofer cuts off the frequency response below 400hz?

largely irrelevant as that's in room tuning territory though the woofer is used to anchor the overall level of the response so it's good to get that right. Measuring outside, if possible, would be a good idea.

Woofer between 400hz - 20khz, the Target Level is 72db +/-3db, it gives me 6 EQ filters:

if you're crossing at ~1kHz then most of those filters are irrelevant, frankly the measurements aren't really good enough for designing a speaker, just too much contamination for me

 

[Flaesh]

It should look something like this [(very) schematically]:

6 EQ\channel are quite enough.
I wrote the rest yesterday))

 

[Flaesh]

Calculated 12-250tc frequency response in 2pi on axis and acoustic power in 4pi (gray):

44 litres sealed.

The Fane speaker is full-range, so you can listen to them just as they are, without all this fuss with measurements and crossovers [this is mostly a joke (but the audiophiliastic magic of full-range speakers is easy available to you))]

 

[Robbie010]

Ok chaps, I think I took a few small steps in the right direction today, a went through the process as described by @Keith_W and tried to flatten each response and apply the crossovers again, here is where I ended up:

As you can see, when the overall response is measured, there is a large ~+6db bump at the first crossover point and ~ + 4db where the tweeter and woofer meet at ~3.5kh. The filters used were 12db Linkwitz, do you think the crossover filters need to be steeper?


Edit: I didn't apply any EQ to the Woofer past 2khz.

 

[3ll3d00d]

I tried to view the response with the 1 cycle FDW setting applied but something strange happens to the tweeter response. ]

why do you keep applying a 1 cycle fdw?

fwiw I drew this by hand/eye (there's also a gain offset of -6 on T and -7 on M) just using your earlier raw (no eq) measurements using xo points roughly around where I thought you were aiming, of course the sum is nonsense as the measurements don't have correct timing and I have no idea where to really aim (as idk the directivity of each driver). Nevertheless it's in vituixcad and is intended to try to illustrate that a) it's not so hard to make on axis flat (+/- a few dB) and b) use the right tool for the job (REW is not a speaker design tool)

 

[Robbie010]

Apologies, I realise now that the suggestion to use FDW 1 cycle was to help clean up the impulse response.

The crossover points I have chosen are ~1.2khz and 6.5khz.

 

[Salt]

You may correct me, but the only crossover I can see ist between Woofer an Tweeter at around 3+ khz? and the mid is running free....