Freq measurements and Hypex Filter Design adjustments

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
This is the Gated Impulse measurement of:
Tweeter Focal Audiom TD5
Mid ATC SM75-150S
Woofer Focal Audiom 15 WX

Crossover with Hypex Filter Design at 500Hz and 2KHz
Crossovers LR4 order all of them
Gain Tweeter Channel 0
Gain Mid Channel 2.5dB
Gain Woofer 10 dB

Measurement only reliable above 300Hz (if I've done correctly)
Doesn't look good to me?
 

Attachments

  • Focusrite19-5-19-256Ksixthsmoothing.jpg
    Focusrite19-5-19-256Ksixthsmoothing.jpg
    568.2 KB · Views: 993
So, this is a measurement of the entire loudspeaker output? Above 700Hz it's not too bad, based on this one particular snap shot as seen by the mic.

Where did you put the mic: how far away? at what height above the ground? Aligned with any particular driver?

How far away from room walls, floor, and ceiling was the microphone?

Please show a pic (if possible) the impulse response and gating window.

Please provide a picture of the loudspeaker itself.

This info is helpful for figuring out what is going on. If you do not do proper measurements on the system and do a good crossover design, it's likely what you are seeing is what cha got... Although sometimes you get lucky, in general you can't apply filters based on the MFG datasheets and expect the phase to be correctly aligned. This can lead to e.g. the large dips like you have at 550Hz, where you have a crossover placed. The funny "bump" up in the response around 200-300Hz might be because the woofer level is much higher than needed, but it looks like you did not correct for baffle step losses, leading to the downsloping response that you see in the woofer response at lower frequencies.

How did you do the crossover design?
 
Measurement is of entire loudspeaker output (only farfield taken). Tweeter, mid and Bass. Mic. Dayton EMM6, interface Focusrite Scarlett Solo and ATRA. Mic 1m away and centred on Mid approx 1.2m from floor. Mic open to outside on left (no wall, floor to ceiling window open), 3m to right wall, 1.2m floor, 1.4m to ceiling. Speakers flush mounted into "wall"

For woofer I should do a near field measurement. Not done yet. The crossovers were via Hypex Filter Design. All Linkwitz Reilly 4th Order (I actually wanted LR8) The crossover freqs I chose looking at manufacturer's graphs and taking the mid pt of the overlap of the flat (horizontal) parts of the graph. The overlap is quite large. I presume I could improve, pushing down the crossover freq so that the smaller drivers deal with more (lower) freqs with perhaps better dispersion?
 

Attachments

  • Focusrite18-5-19.jpg
    Focusrite18-5-19.jpg
    520.6 KB · Views: 919
You don't need steeper slope than LR4 because with DSP you can easily equalize the cone break-ups and LR8 does not help too much with driver excursion on the other side compared to LR4.
With the ATC SM75-150S you can use LR4 at 380Hz like the active ATC speakers.
 
Measurement is of entire loudspeaker output (only farfield taken). Tweeter, mid and Bass. Mic. Dayton EMM6, interface Focusrite Scarlett Solo and ATRA. Mic 1m away and centred on Mid approx 1.2m from floor. Mic open to outside on left (no wall, floor to ceiling window open), 3m to right wall, 1.2m floor, 1.4m to ceiling. Speakers flush mounted into "wall"

For woofer I should do a near field measurement. Not done yet. The crossovers were via Hypex Filter Design. All Linkwitz Reilly 4th Order (I actually wanted LR8) The crossover freqs I chose looking at manufacturer's graphs and taking the mid pt of the overlap of the flat (horizontal) parts of the graph. The overlap is quite large. I presume I could improve, pushing down the crossover freq so that the smaller drivers deal with more (lower) freqs with perhaps better dispersion?

OK, thanks for this info.

Looks like you are limited with mic position as speakers are built into the wall. A nearfield on the woofer would be good. This will relatively accurately show the in-speaker behavior because, being in-wall, there won't be a baffle step. The tricky part is to figure out the relative offset of each driver's acoustic center, since that will influence the delay and therefore phase response of each driver. You can do that with measurements with some care, but you will need a design program to sleuth out the offset. Or you can guesstimate from the depth between the plane of the wall and the dustcap, or where the voice coil sits. For the tweeter and dome, these are probably in very close alignment already, but the woofer AC will be recessed relative to them. You can try to delay the signal to the mid and tweeter by enough such that the delay time, times the speed of sound, gives you the same distance as the amount that the woofer's AC is set back from the baffle. This should bring the timing of the drivers closer together, e.g. time align them. But if you think about it the woofer is lower and you probably listen to the speaker at an elevation that is around the tweeter level. The distance from the woofer to your ear is then even longer by maybe 40% because it is traveling along a path that is like the hypotenuse of a right triangle, where the distance from ear to tweeter is more like one side of the same triangle. So additional delay of mid and tweeter by this amount can be used as well.

Unless you can do many more measurements and then use a crossover design program you will be left with some trial and error, helped by some understanding of what is going on with the sound radiated by the drivers, etc. It's good that you can do SOME measuring as this can be very helpful in ferreting out the problem(s) when it just doesn't sound right. Tweaking the tonal balance by ear is best done when you are already 95% of the way done with the crossover development and might consist of gently tipping of the response and so on.

Try the nearfield measurement and post what you get here.
 
Whats the center to center distance between the woofer and mid? For 500Hz crossover point the best distance is 17cm or less (1/4 wavelength). For 380Hz it's 22,5cm or less which is more realistic with a 15" woofer and a dome mid.

If you want a textbook slope for the woofer, do an ungated measurement at listening position and equalize the response to flat with 1/3 octave smoothing (don't need to equalize every wrinkle) and then apply the 380Hz LR4 filter. You can do the equalization of the mid with the gated measurement and then apply the 380Hz LR4 filter.

Your gated measurement is good down to 300Hz so if you want to check the phase, just flip the polarity of the mid in the Hypex Filter Design and do a gated measurement as you did already. If you have a deep null in the gated response then you have a good phase match with normal polarity. If the null is shallow then adjust the delay until deepens.

Maybe not the most professional method, but it works for me.
 
Last edited:
Your gated measurement is good down to 300Hz so if you want to check the phase, just flip the polarity of the mid in the Hypex Filter Design and do a gated measurement as you did already. If you have a deep null in the gated response then you have a good phase match with normal polarity. If the null is shallow then adjust the delay until deepens.

Maybe not the most professional method, but it works for me.

I think you may be forgetting that "your measurement is valid down to 300Hz" also means that the measurement has 300Hz resolution (at best) at all other higher frequencies as well. As if there is one data point every 300Hz. The algorithm that spits out the data every few Hertz is just interpolating in between the FFT bins at other frequencies to make a smooth data set. This is the very serious limitation that you face when making an indoor measurement in a small space. I don't think you will have much luck seeing a null at 350Hz unless it is VERY broad, like 500Hz or more, which won't happen with an LR4 or higher order.

Also, looking back on your first plot, I forgot about the resolution when looking at it. All of the info below 300-500 Hz should be put out of consideration - it's not "real" data. Thus the woofer nearfield, for which you should be able to use a pretty long gating window and get resolution to better than 50Hz, will tell you what the woofer is doing. Since the measurement looks OK above 800Hz or whatever, I would at this point only concentrate on the woofer-to-midrange crossover at this point.
 
Don't know whether to reply with: Quote, Post Reply, or Quick reply? I'll try Quick reply here, see if gets thro'
Replying to Denibeni: Centre Woofer to centre Dome Mid = 37cm I am considering lowering crossover freq to 380Hz as you suggest. Also raising other crossover to 3.8KHz.
That is from 500Hz and 2KHz First I'm going to do nearfield measurement of Woofer to see if I can get anymore clues. My other problem (for after the measurements) is equalizing with Hypex HFD
Replying to CharlieLaub: Next weekend I want to do the nearfield measurement for the woofer. Will it be as per your fantastic instructions on how to measure freq resp with ARTA? Except for the distance of course. Would 1cm be OK for 15" woofer?
 
With the ATC SM75-150S you can use LR4 at 380Hz like the active ATC speakers.

That would be the acoustical slope. Important to distinguish from the electrical slope generated by the x/over. That could be anything, and most likely it will not be LR4 - but whatever sums up with the driver's natural roll-off to get the acoustical target.

Impossible to tell without measuring.

Troels crosses with electrical 2nd order to achieve 4th order acoustical high-pass: ATC-SM75-150
 
Last edited:
That would be the acoustical slope. Important to distinguish from the electrical slope generated by the x/over. That could be anything, and most likely it will not be LR4 - but whatever sums up with the driver's natural roll-off to get the acoustical target.

Impossible to tell without measuring.

Troels crosses with electrical 2nd order to achieve 4th order acoustical high-pass: ATC-SM75-150

You are right, but if we talking about a DSP solution, as we did here, first you equalize the driver response to flat, then apply the desired filter slope and you get the exact acoustic response, be it BW2, LR4 or whatever.

Of course, you can combine the original acoustic slope of the driver with a DSP slope if you want.
 
Last edited:
You are right, but if we talking about a DSP solution, as we did here, first you equalize the driver response to flat, then apply the desired filter slope and you get the exact acoustic response, be it BW2, LR4 or whatever.

Of course, you can combine the original acoustic slope of the driver with a DSP slope if you want.

EQ-ing low end of a mid driver to flat - a substantial lift at the lower frequencies - and then applying the 4th order HP filter is overdoing it. Yes the DSP can do it but it looks like the reason for doing it is because "it can".

A well behaved unit like the ATC here will (and it does) react well with a simple 2 order HP.

EQ-ing flat(tish) works well for the high end, i.e. for LP filters.

Above IMO of course, many ways to skin a cat.
 
-freq-measurements-hypex-filter-design

OK just done nearfield measurements. Mic 0.5cm from dustcover. Was not able to mute mid and tweeter so some freqs slipped in? ASIO Control panel: sampling rate 48KHz Buffer size 1024
ImpulseGated.JPG

Gated1th.JPG

Gated24th.JPG

ImpulseNotGated.JPG

NotGated1th.JPG

NotGated24th.JPG
I have 1/1 smoothing and 1/24 smoothing. Gated apparently reliable above 100Hz Not gated apparently all reliable?? In all these mesurements the filters were not diasbled so this woofer had Low Pass LR4 at 500Hz
I don't understand these freq responses
 
OK just done nearfield measurements. Mic 0.5cm from dustcover. Was not able to mute mid and tweeter so some freqs slipped in? I have 1/1 smoothing and 1/24 smoothing. Gated apparently reliable above 100Hz Not gated apparently all reliable?? In all these mesurements the filters were not diasbled so this woofer had Low Pass LR4 at 500Hz
I don't understand these freq responses
This is basically all wrong.

Disable all filters. Turn off all drivers except the one you are measuring. Then redo the nearfield measurement on each driver, one at a time. You should be able to use a pretty long window (or no gating at all). Use only 1/24th or 1/12th octave smoothing. Make sure to save BOTH amplitude (SPL) and phase data, minimum phase if possible, using the FRD file format if possible. Go back and check my tutorial if it is helpful.

You really only need to do this for the woofer, but it wouldn't hurt to do it on the midrange as well.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.