Three-way floor stander DIY crossover design– help welcome

[Nik76]

It’s been a while (more than 25 years) since my last attempt at DIY speaker and crossover. Been reading a lot in the past few years regarding home DIY speaker building and I am really amazed at what one can do today with a little cash and a lot of enthusiasm .
Anyway, I am building a 3-way floor stander, with a rather strange combination of speaker, but, for one reason or another, I am “stuck” with the speakers that I have and want to make the most of them.
As mentioned earlier, I am a newbie in today’s world of DIY speaker building, so I am open to any suggestions or tips on how to improve the crossover design. I do have some knowledge in acoustics as it’s been my hobby for as long as remember, so I am not a complete beginner.

Drivers used:
tweeter: HiVi RT2h-a
midrange: Faital Pro 5PR160 link
woofer: Heco TC300 (used, re-foamed speaker)

I started with what I had, Heco TC300 and HiVi RT2h-a. Was shopping around for a midrange that could fit well, as a base point I wanted to retain high speaker sensitivity so was searching for something with higher sensitivity, break up that is easy to handle and low distortion. Finally, decided on Faital Pro 5PR160, based on the manufacturer published response and simulation of box and baffle. But in hindsight, I should have gone with a speaker with independent testing. Well, you live and learn.

I have made speaker measurements using REW, and EMM-6 microphone. All measurements were done in an actual speaker box, app 70 lit net for the bass and 3 lit for the midrange. Both sealed. Heco is an acoustic suspension speaker (one reason why I like it), so sealed enclosure only. Vituix CAD was used for crossover simulations. I have followed instructions from Vituix cad and REW, so measurements shouldn’t be much off. I couldn’t measure outside or use a larger room, so the measurements are not perfect, especially the off axis ones, with more than 80deg (it was clear to me that the room walls are not far enough), so please don't take much consideration of the results above +-90 degrees off axis.

Also, while measuring the impedance of the bass in last iteration (after test front baffle mounting), I realized that the impedance curve below 100Hz was “rough” and the resonance spike was not smooth, which was due to poor seal of the front baffle in a few places. I measured impedance before with adequate seal, and was very similar to the one presented here, but with a nice smooth line, so nothing wrong with the speaker.

To cut the long story short, I wasn’t successful at reaching the goal, which was among others, a smooth frequency response and nice dispersion, with not too complicated crossover. My greatest issue was the response of the midrange (which I was hoping to be an easy speaker to work with). Break up at 6kHz was much worse than expected, that forced me to use a notch filter and 3rd order crossover to handle it (maybe I have overdone it ) . Below is the response on the actual baffle, measured as stated above.

Speaker box size is below, together with the image of the test speaker (before baffle edge chamfering):

The box is made of 20mm mdf with double front plywood plate.
Measurements were taken at 1m distance, Fairfield and nearfield, and merged, for each speaker (except tweeter offcourse).
Crossover simulation was performed for listening axes at tweeter level. All speakers are measured flush with front baffle.

I have many versions of crossover, 2nd order, 3rd order, combinations, different slopes and cutting frequencies and for now settled on the final 3 presented here. I am not satisfied with any of them, but I am unable to do better for now. All three versions below (V1, V4 and V6) have some degree of inadequacy. My intention for now is to go with R6 version and make a test crossover (with cheaper parts) and measure the assembled speaker, and if that goes well to buy higher quality crossover components and finalize the design/speaker. Being that the streamer will probably be WIIM, I will use PEQ to address the largest peaks and dips especially in the midrange response after which I think the speaker should be OK.

I am aware that the crossover network is fairly complex, but I simply couldn’t find a way to handle two large break ups (bass and midrange at 2kHz and 6kHz) and provide for a decent phase coherence at the transition frequency between speakers (app at 330Hz and 2.1 kHz). I am fairly happy with horizontal dispersion at crossover frequencies. Inverting the midrange polarity (all three are of the same polarity) shows significant cancelation up to -30dB, for on axis and off axis, which I think is good.

Impedance curve is not very amp friendly, dipping below 3ohm, but I don't think this will be an issue, amp is rated at 2 ohms (Hypex Ncore NCX252MP based power amp).

I had to pull the tweeter all the way down to 2kHz, which is not recommended by the factory, but a steep curve of 4th order filter should solve this problem.

So, if you have any recommendation, observations, or items that you think I have missed or completely sc***ewed up , please let me know, it will be greatly appreciated.


V1, normal and midrange reversed

V4, normal and midrange reversed

V6, normal and reversed

 

[a4eaudio]

Your Y-offsets should be relative to the tweeter so the tweeter's own Y-offset should be zero.

 

[hex168]

I hate to add even more complexity, but the remaining midrange bump at about 1.1K may still be a problem. It also shows up in Faital's impedance curve. I'd consider putting a notch filter on it.

 

[Nik76]

Your Y-offsets should be relative to the tweeter so the tweeter's own Y-offset should be zero.

Your Y-offsets should be relative to the tweeter so the tweeter's own Y-offset should be zero.

Actually I made a mistake in my original post, listening axis is 100mm above the tweeter, so its ok., But thanks for the correction, spot on.

 

[Nik76]

I hate to add even more complexity, but the remaining midrange bump at about 1.1K may still be a problem. It also shows up in Faital's impedance curve. I'd consider putting a notch filter on it.

I have done that in V4 crossover, but had to "pull" both the tweeter and the woofer down a few decibels, so I was thinking that its better to adjust it with PEQ on the streamer than to add a large capacitor, resistor and coil in series with the midrange (expensive and in the signal path). Since the hump is not caused by the filter or the brake up, and its there on axis and off axis, I was thinking that it is best to adjust it in the streamer EQ.

 

[eddantes]

Actually I made a mistake in my original post, listening axis is 100mm above the tweeter, so its ok., But thanks for the correction, spot on.

I don't see the vertical directivity graphs - but I doubt one should stray more than 10 degrees vertical from tweeter axis. Instead of setting the measurement axis 10cm above the tweeter - consider tilting the speaker to be on tweeter axis, but I don't know this tweeter, perhaps you are right to be above it.

 

[Nik76]

I don't see the vertical directivity graphs - but I doubt one should stray more than 10 degrees vertical from tweeter axis. Instead of setting the measurement axis 10cm above the tweeter - consider tilting the speaker to be on tweeter axis, but I don't know this tweeter, perhaps you are right to be above it.

I didn't post vertical directivity graphs, I had problems during measurements of vertical response, so I thought its best not to provide any info, than potentially inaccurate ones. I was considering this option, of tilting the tweeter, and tried this in the simulations, but didn't provide any significant difference (with the results that I have). Adjusting axis to the level of the tweeter doesn't significantly change the overall response, major issues remain the same, so I was thinking it is not that important.

 

[audio_tony]

I have done that in V4 crossover, but had to "pull" both the tweeter and the woofer down a few decibels, so I was thinking that its better to adjust it with PEQ on the streamer than to add a large capacitor, resistor and coil in series with the midrange (expensive and in the signal path). Since the hump is not caused by the filter or the brake up, and its there on axis and off axis, I was thinking that it is best to adjust it in the streamer EQ.

A notch filter is usually in parallel. A rough starting point for values would be 5.6R and15uF in series to negative.

This would be right on the driver.

Excuse my crude drawing. (The 'C' looks like and 'L' oops)

 

[hex168]

I have done that in V4 crossover, but had to "pull" both the tweeter and the woofer down a few decibels, so I was thinking that its better to adjust it with PEQ on the streamer than to add a large capacitor, resistor and coil in series with the midrange (expensive and in the signal path). Since the hump is not caused by the filter or the brake up, and its there on axis and off axis, I was thinking that it is best to adjust it in the streamer EQ.

I agree with your approach and considered suggesting EQ as an alternative but did not know if that was acceptable to you. I would not worry about components in the signal path; there is a Purifi paper that I think applies to this case that shows a series notch filter could actually reduce distortion compared to using EQ.

 

[Nik76]

A notch filter is usually in parallel. A rough starting point for values would be 5.6R and15uF in series to negative.

This would be right on the driver.

Excuse my crude drawing. (The 'C' looks like and 'L' oops)

View attachment 520478

Thanks for the input, and I have tried that, but just couldn't make it work, it would adjust the peak, but made the dip even worse. The only notch filter that worked was the one in series with the speaker (parallel notch filter), which did the job, but lowered overall speaker sensitivity by 2dB, so I thought adjusting it with a EQ would be better.

 

[a4eaudio]

Also, keep an eye on the minimum impedance. You have a minimum impedance of 2 ohms in some variations. It is probably safe to target a minimum of 3.2 ohms (or more).

 

[Nik76]

Also, keep an eye on the minimum impedance. You have a minimum impedance of 2 ohms in some variations. It is probably safe to target a minimum of 3.2 ohms (or more).

Yes, that is true, but V6 crossover version dips to 2.3 ohm min (below 3.2 ohms from 170-320 Hz only). Being that the amplifier will be Hypex Ncore NCX252MP based, which is rated at 2 ohms (2x180W @2 ohms), I was hopping that it wont be a problem being that impedance curve is mostly above 4 ohms. Do you think it may not be enough, that it impedance should be higher (in this case only, not in general)?

 

[a4eaudio]

I think you are "probably" okay with the Hypex. BUT, if you play around with the components you can probably get the impedance higher.

1) Try larger inductors and smaller capacitor
2) Consider putting in actual DCR of inductors and see what a higher DCR air coil inductor does compared to lower DCR laminate core
3) You do not want a resistor in series on your woofer.

 

[Nik76]

Thanks for all the inputs.
@a4eaudio have tried varying combination of capacitor/inductors, but couldn't achieve adequate impedance without the series resistor. So i started from scratch for the woofer and midrange, added a series notch filter on the woofer to flatten the resonance peak (instead of a resistor) which increased impedance, and a parallel notch for the midrange peak at 1kHz. This lowered the response by app. 2-3 dB but I think I am finally there! Impedance is now 3.1 ohm minimum (dip at 55 Hz, the rest is mostly above 4 ohm). Shouldn't be a problem for the amplifier.
Frequency response is now more even than before (there is a small dip at 1.5kHz and at 4kHz, but nothing significant. Roll off above 10 kHz (I cant hear anything above 12-13kHz anyway ).
Midrange and tweeter appear to be properly phase aligned (see reversed midrange response graph).
I have added the vertical directivity graph (there were a few problems during measurements, but probably they are not far off), not very nice, but probably to be expected for this type of planar tweeter (it would have probably been better to raise the cutoff frequency from 2.3 kHz, but just couldn't make it work...)

If anyone has any input/advices they are very welcome!

 

[MEGB1262]

hi there,

the real problem is the difference in the distances between the points of sound origin of the two critical sources—the isostatic tweeter and the cone midrange driver—relative to one another

i have simulated a lot with the Visaton BoxSim software - an EN version is available - try similar Visaton chassis for the simulation and you will know exactly what i am talking about

they have very accurate on axis and off axis measurements for their own drive units - 180 degree in 5 degree steps - 19 measurements for each driver

you will never get a decent phase in the crossover region of about 2000Hz - 3000 Hz for more than 1 very close listening angle - the contrary of constant directivity let's say

without a physical "time alignment" changing the baffle itself only a active crossover with a delay for the tweeter will do it may be

with a passive crossover you will have to move the midrange driver forward with a ring
recessing the wider tweeter horn won’t work

i have one example how it looks - optically it could be a challenge to like it

may be a custom 3D waveguide with more distance between throat and mouth for the isostatic tweeter will look much better, but i am completely lost with 3D printing, may be other forum members here can help

looking very close at the construction drawings of the midrange drive unit and the tweeter with the horn probably the tweeter - midrange the difference in distance between the points of sound origin of the two sources is about 20 - 25 mm but this a inaccurate guess

the HiVi waveguide horn construction is peculiar - only about 25 mm waveguide horn length for the quite big waveguide mouth - no load of the horn of the waveguide below 3000 Hz i guess

oh, and there are probably other effects caused by diffraction at the edges of the baffle going off axis with the mic

you will have to take with a good measurements system a look at the step response of the midrange - high range combination
at minimum you have to take 0 degree on axis and 15 and 30 degree off axis for all this - even better are 5 degree steps

hope it helps, Stefano

 

[Nik76]

Hi, thanks for taking the time to look at my design. I understand in general what you are saying, but I am not sure if I should align the midrange and the tweeter as you suggested, but I might be wrong.
Looking at the vertical cross section of the speaker box, I think that the speakers sound origin points relative to one another is close (radiating surfaces in red), max diference is I think no more than 8mm (comparing the dustcap of the midrange and the isostatic).

My understanding is that if I have measured all 3 speakers in horizontal, one measurement for each speaker 10 deg apart (full 180 deg and vertical 20 deg apart), exactly 1000 mm from each speaker to the microphone, with speakers mounted on the actual baffle/speaker box(es) and merged with nearfield measurements (for midrange and woofer), speaker crossover simulation (in this case in Vituix cad) should show if speaker horizontal offset should be applied (for example by moving the midrange forward as you suggested) to achieve proper time alignment and phase, based on location of listening axis location/height.
Looking at Vituix Cad GD & Phase graph at midrange/tweeter cutoff frequency (approximately at 2.2 kHz) I think that the phase of the speakers are in line (circled in red, green and blue lines - representing midrange and tweeter phase lines), so phase should be ok.

Looking at horizontal gray and blue lines (normal group delay and excess group delay) at the cutoff frequency, there are no humps or similar, the lines are almost horizontal in this zone. Turning off tweeter/midrange (one off, the other on) to check individual time delay, there is a difference in the cutoof zone of approximately 0.14ms between them (0.32ms vs 0.46ms respectively), while group delay is 0.4ms app.

I am not sure if this time delay difference (0.14ms) should be corrected by moving the midrange forward, or is it maybe acceptable. Any thoughts?

 

[MEGB1262]

Hi there,

the real quick calculation about the difference of 140uS in air is at 20 degree / 68F is 48 mm, in the crossover region between 2000 and 3000 Hz this is non neglectable, would be easy to verify in the simulation introducing the option for an active delay to correct it

hope it helps, Stefano

 

[Paweł L]

Don't worry about the woofer tracking the mid perfectly since the wavelengths are large comparable to tw/mid. And 90% of the time the sims don't exactly come the same as measurements, but they're close. Now if you used minimum phase during the measurement process, the phase will reflect correct delay for the drivers at 1m from the baffle at the Axis reference point. If you changed the microphone position so each driver was measured at 1000mm then you have add additional difference to the driver's position, and in case of mid/tw section these 5mm make a difference. Most of the time the TW serves as reference and you would move the mid on the Z axis, w/o the waveguide by adding 20-30mm to the mid. But as point midway in the design process I run quick measurements to check if my phases in the sims follow reality. Just slap some parts together and measure in phase and out of phase, moving the mic up and down from the reference axis to check if your antiphase dip follows what's on simulation, adjust the Z accordingly. Remember also that the offset will be slightly smaller at 3m-4m from the baffle which is mostly the listening distance in the larger room. In the end I always end up tunning by ear on the mid/two side mostly because directivity difference between these two, distortions patterns, some combinations play better than others. Hope it helps

 

[Nik76]

Don't worry about the woofer tracking the mid perfectly since the wavelengths are large comparable to tw/mid. And 90% of the time the sims don't exactly come the same as measurements, but they're close. Now if you used minimum phase during the measurement process, the phase will reflect correct delay for the drivers at 1m from the baffle at the Axis reference point. If you changed the microphone position so each driver was measured at 1000mm then you have add additional difference to the driver's position, and in case of mid/tw section these 5mm make a difference. Most of the time the TW serves as reference and you would move the mid on the Z axis, w/o the waveguide by adding 20-30mm to the mid. But as point midway in the design process I run quick measurements to check if my phases in the sims follow reality. Just slap some parts together and measure in phase and out of phase, moving the mic up and down from the reference axis to check if your antiphase dip follows what's on simulation, adjust the Z accordingly. Remember also that the offset will be slightly smaller at 3m-4m from the baffle which is mostly the listening distance in the larger room. In the end I always end up tunning by ear on the mid/two side mostly because directivity difference between these two, distortions patterns, some combinations play better than others. Hope it helps

I have measured the response of each driver at exactly 1000m from the actual baffle (in line with each speaker center/axis), and simulated the crossover for an expected location of the listeners (axis direction and height in relation to position of each driver on the front baffle), so I assume that phase indicated in the simulation should be correct for that listening position and horizontal/vertical position of each driver, or am I missing something?
I will for sure do test measurements when I complete the test crossover, as you suggested.

 

[Paweł L]

I have measured the response of each driver at exactly 1000m from the actual baffle (in line with each speaker center/axis), and simulated the crossover for an expected location of the listeners (axis direction and height in relation to position of each driver on the front baffle), so I assume that phase indicated in the simulation should be correct for that listening position and horizontal/vertical position of each driver, or am I missing something?
I will for sure do test measurements when I complete the test crossover, as you suggested.

You should just leave the microphone in one position for the reference axis say 1000mm mid point of tw and mid axis and use minimum phase function if you measuring program has such feature, then every next measurement will correspond to the reference axis point and phase will have the correct delay included with increase or decrease in distance (to the reference point).
Remember there's only one valid Reference Point for summation of drivers output on Reference Listening Axis, and for correct simulation 'all drivers phases' have to correspond to the distance from acoustic center of each driver to that point. So while you tweeter might be 1000mm, the midwoofer or mid won't have the same distance unless the tweeters waveguide somehow will move tweeters acoustic center to match That of the Midrange. The correct minimum phase with Delays Included is very important in passive x-o design, if not you can kind of throw all your sims in the garbage. Unless you treat the design like coax driver.
Hopefully if you could assign each driver the height location, and use simple trigonometry to extract the extra difference, the Vituix will also compensate for the delay change.
I usually measure my driver's at mid point between tw and mid, good reference point even if the listening distance/height changes a little.