New design - A Seas all metal 3-way

[Dave Bullet]

I zoomed in the on L15 merged responses REd = measured nearfield (level adjusted to match farfield), blue = measured gated farfield, black = original blended response (shown in the overall xover above). As you can see the black is a little peaky to the left of the merge where the gated farfield tails off.

The pink is my re blended L15 response.... better matches the baffle step side, but farfield is a little low. As you say - hard to hit it spot on

[Dave Bullet]

Here's my updated SIM using the less peaky more level correct L15 blended response. At least the breakup modes of the woofers have been tamed still (showing seas manufacturing consistency and that these drivers have not wandered in 10 years of shelf storage... PS: they are also broken in). The breakup suppression looks better than reality as we must remember these are on-tweeter axis measurements, so the modes will be down anyway... At least xover component values nail the breakups still.

The 2nd picture zooms in to show the on-axis deviations from flat: Max amplitude is -11.6 dB, min between 60 Hz and 13,000 Hz is -15.1 dB meaning a +/- 1.75 dB speaker. Not bad. The summation error around 1460 Hz is due to no crossover optimisation. I literally plugged in my blended measurements on the original crossover. I feel I am close (apart from crossover simplification is possible).

[Dave Bullet]

Finally - this is just loading my raw gated 27TBFCG and L15 responses in (and removing the L15 Z offset since we are dealing with measured phase. I think Jeff's spreadsheet is on the money? Looks very similar to me.... (ignore the green / woofer to mid section). The only variation is in minor phase summing / discrepancy in the mid-tweeter crossover.

[Dave Bullet]

Now this would be a mega TOWER MONKEY COFFIN if I built one.... Von Schweikert... eat your heart out

[Dave Bullet]

OOOps - I took my eye off the impedance ball... .down to 2.4 ohms in the base region. Not happy about that... back to the drawing board / tweaking!

[5th element]

For the near field measurements, what you're seeing create the 'un-smooth' characteristic is most likely noise. With the lower drive levels used (mic right next to cone) it's easy for noise to creep into the measurement as the mic is picking it up directly from the drivers cone. This becomes more pronounced as the drivers roll off because the signal to noise ratio diminishes.

The way to get around this is to use measurement averaging. ARTA allows you to choose as many averages as you want, I typically use three, but the net effect is it greatly reduces the measurements susceptibility to system noise. I am not sure if speaker workshop (I think that's what you're using?) has that option? Either way what happens is the program performs the measurement several times and can then deduce what's noise and what should actually be signal, the noise being, theoretically, completely random, but the signal not.

[Dave Bullet]

Ok - after a little tinkering with actual measurements, I've arrived at the following.

I've found I can attenuate the 1KHz+ region via just changing the series tweeter L-pad up to 4.7 or 5.6 off the shelf resistors without interfering with phase. Depending on in room brightness, compared with music / recordings.... etc.. this might be helpful.

I've also been playing with running both L15 and L26 in both open and sealed box, with different stuffing (or no stuffing) levels. I realise in room response / modes can vary wildly - based on speaker and listening position but it is helping me come up with "plan b" for tweaking the bass or midrange response without resorting to xover redesign. As I live in NZ - I can't just locally order parts. As it stands, the crossover for this will run into ~ USD$400 for the speaker pair!

[Dave Bullet]

For the near field measurements, what you're seeing create the 'un-smooth' characteristic is most likely noise. With the lower drive levels used (mic right next to cone) it's easy for noise to creep into the measurement as the mic is picking it up directly from the drivers cone. This becomes more pronounced as the drivers roll off because the signal to noise ratio diminishes.

The way to get around this is to use measurement averaging. ARTA allows you to choose as many averages as you want, I typically use three, but the net effect is it greatly reduces the measurements susceptibility to system noise. I am not sure if speaker workshop (I think that's what you're using?) has that option? Either way what happens is the program performs the measurement several times and can then deduce what's noise and what should actually be signal, the noise being, theoretically, completely random, but the signal not.

Thanks. Speaker workshop allows you to take a number of samples I don't know if it averages across these, or just picks the best. I believe an average. I was using 10 as I found any more didn't reduce noise visibly and my patience was being exhausted waiting for tests to finish

[Dave Bullet]

Warning - post contains graphic content. I disembowled the mule....

[Dave Bullet]

I've been playing around with stuffing or not and running open or closed box. The blue and black lines are just 1/3rd octave smoothed, ungated in room at seating position with loudspeaker at intended position.

Is it worth worrying about how to run the box (open / closed) and stuffing levels? or am I just being silly.

Heavily stuffed with an open back is blue, closed box (50L) and moderately stuffed is black and red is the raw nearfield / gated farfield response (purely for level matching and also that is what the crossover is based on).

I've designed the crossover on expected closed box, no room modes, nearfield splice.... so any "extra" dB in the bass region may swamp the rest of the spectrum? Also is running open box - heavily stuffed a done thing? what pros / cons are there?

[Dave Bullet]

I'd appreciate any feedback on both options presented below please.

The top in each picture is the LR4 woofer-mid and LR4 mid-tweeter crossover. The bottom is the LR2 woofer-mid and LR4 mid-tweeter crossover. I am planning to order enough parts to try both.

Pictures in order:
1. On-axis FR
2. On-axis FR reverse null
3. Crossover
4. Impedance

[Juhazi]

How about LR2 at 400Hz? Would give lower midrange distortion. Vertical lobing might be a problem though, but still. Saves expenses too... MT must be LR4 I guess.

[Dave Bullet]

How about LR2 at 400Hz? Would give lower midrange distortion. Vertical lobing might be a problem though, but still. Saves expenses too... MT must be LR4 I guess.

Thanks. That would put me right at the theoretical maximum for driver spacing (1/2 wavelength being 395Hz with a 0.43m spacing between W-M). It would reduce excursion for the L15, but considering it has 5.5 xmax, It should be ok 10dB down at 100Hz. The L26 only has 7mm xmax anyway and my calculations put this as still L26 excursion limited. Of course, we know xmax is one thing, and distortion at xmax is another :-)

The other problem I'm facing with 2nd order acoustic is the top end breakup of the L26. I could shift up another octave, but then I'll probably need another or wider notch on the L26 to end to control it.

[Dave Bullet]

Two things I haven't done and should have....
1. Distortion measurements of the drivers (ensure they are operating in their optimum ranges)
2. Off-axis measurements.

Re #2 - I don't think I'm asking too much of the drivers in terms of frequency response for polar response to be a problem. Of course BDS and other symptoms (built in driver non-linearities) will play a part.

Re #1 - My only concern is rising HD / IMD with a lower xover point and being 2nd order acoustic between W-M.

[5th element]

From my experience there's literally zero difference in sound quality between 2or 4th order networks. Pick which one suits the drivers and the system better. Typically this is the 4th order as it gives better off axis curves and places lower demand on the drivers. At a push the 2nd order seems a tad easier driver and has marginally better phase integration, but that marginal difference will be dominated by component tolerances.

Which one is cheaper for you to build? (I am being serious).

[Dave Bullet]

From my experience there's literally zero difference in sound quality between 2or 4th order networks. Pick which one suits the drivers and the system better. Typically this is the 4th order as it gives better off axis curves and places lower demand on the drivers. At a push the 2nd order seems a tad easier driver and has marginally better phase integration, but that marginal difference will be dominated by component tolerances.

Which one is cheaper for you to build? (I am being serious).

Thanks Matt,

I'm pricing up both now :-)

I had guessed the LR2 would have had a better wave front (lobing) and provide a more seamless transition. The downside is the excursion demanded at the L15 low end. I don't know whether a driver operating near excursion can still provide midrange fidelity. My gut tells me a driver will not have upper frequency finesse, if low frequency excursion is being demanded at the same time. Maybe this is evident in various linear and non-linear distortion measurements at peak SPL.

In any case, I worked out the L15 and L26 would hit xmax at about the same time at both 100Hz and 50Hz levels. The L15 is down 10dB (100Hz) and 15 dB (50Hz) respectively. This puts it on par with my Unibox modeling and xmax at those 2 frequencies for both drivers. So both would let me know i was reaching peak power. Downside is the L15 is moving more air when it doesn't need to (hence my finesse point).

I think I am leaning to LR4 throughout, allowing the L15 to focus on a narrower band.

[Dave Bullet]

Well 6 months have passed and I've ordered the crossover parts.

Findings.

If you can't crank the volume - something is wrong :-)

The Seas 27TBFC/G is ear bleeding. Hated it from the get go. I think the 26KHz harmonic creates something in 3rd or 5th order HD that is just awful.
Seas 27TDFC/G - better. However, straight away, Phil Collins, I wish it would rain down - the opening passage with the soaring guitar was ear bleeding. Awful.

I played the L15 / L26 alone - sounded smooth

I took the L15 out of the picture (wired in a 7 Ohm resistor to replace the load). 27TDFC/G definitely the culprit

The above was with your traditional LR4 1750Hz crossover.

So.. I implemented full 3rd order on the L15 and 27TDFC/G and moved the XO up to 2,500 Hz.

I left in the L15 double notch filters - really killing the 8KHz and 10Khz breakup nodes.

Result - much much better. the 27/TDFCG is not "straining".

Will post revised crossover and simulations.

Caveat - I need to remeasure. And yes, I double checked all wiring.

In terms of L26 balance - lovely. Bass is there and equivalent. I will do some sweeps to see if I can hear any suckouts but happy in that regard.

[Matt M]

Curious.

I remember the 27TBCF/G from Zaph's BAMTMs. I would rather call it easy on the ears, if not a bit boring. And I would call the upper end of the L15 ear bleeding... but since you eliminated this option...

Don't take me wrong: I do not want to question your results, but it is not easily understandable to me based on the data given. What about off-axis behavior?

[Dave Bullet]

Curious.

I remember the 27TBCF/G from Zaph's BAMTMs. I would rather call it easy on the ears, if not a bit boring. And I would call the upper end of the L15 ear bleeding... but since you eliminated this option...

Don't take me wrong: I do not want to question your results, but it is not easily understandable to me based on the data given. What about off-axis behavior?

Thanks - valid feedback. Because others rave about the robustness of the Seas 27Txxx tweeters - it makes me think either:
1. I have a bad tweeter
2. I have wired up things wrong

I need to measure anyway. Totally understand my views so far are subjective (but at the end of the day - we have to like what we listen to. Measurements can provide data to correct an issue - but that is all they are for!)

Off axis is a good question. I noticed standing up from about a 3 - 4 metre distance, things got brighter still.

My listening position puts the tweeter at ear level perfectly. The back of the couch (sofa) is quite tall so that may absorb any backwave from the wall ~ 0.5 metre behind me.

So... I need to do some measurements. I do know however that simply swapping the D for the B tweeter was instantly better. (even though there's only about 1dB throughout the spectrum where the B is more sensitive). A sensitivity mismatch was not the reason. There's some HD or other non-linear distortion going on with the B.....

I'm not sure I can do reliable HD sweeps with my setup. What freeware do people recommend these days? I'll do some looking myself

[Dave Bullet]

Here's some simulations with the new XO (2500Hz 3rd order electrical with L15 notches).

NB: I had to compromise on some values as it is all I have on hand in spares. Some optimisation / tweaking is required. but this sounds much better to me than the "flat" or even "tilted" repsonse (about 3dB fall from 1Khz to 5 Khz) options I had previously - where the tweeter was crossed at a LR4 acoustic slope @ 1760Hz

[Juhazi]

It looks like sounding good! Obviously you had too much stress on tweeter and too wide dispersion around 2kHz. Slowly uprising treble on-axis is good because of the beaming getting stronger above 10hHz. Listening and also measuring "room response" is needed to get voicing right.

How is bass tuned and sounding now?

[Dave Bullet]

Base is good in either 47 or 60 L sealed. I won't bother with ported. Even with only one L26 - there's pleasant and balanced bass to my ears in my lounge.

The cabinets are not properly braced nor damped nor filled yet. I'll keep stuffing until I find the lowest F3 (I think is a general rule of thumb). I'll lean toward the 60 L size (stealing from the midrange enclosure) to get closer to transient perfect (Qtc 0.5). I have a Qtc 0.5 subwoofer and I like the precision and lean-ness of the bass it produces.

I am struggling to get wider / deeper nulls in the crossover regions. I also am failing to ensure a min 3 ohms. I'll keep tweaking but I may need to reset my design... or just stop tweaking once I'm happy with the sound.

With more playing last night, I was really beginning to enjoy what I was hearing. I think things are still a little forward for my liking. Could be a symptom of asking the L15 to go up higher and the breakup (which I tried to combat with a much steeper slope).

Playing the L15 by itself imparted a slight "brashness" to only some small sections of some bright recordings on some frequencies I could hear. but this is drowned out by the tweeter. I don't think the L15 needs further taming. (well I would rather push the L15 up than the 27TDFC/G down based on my first crossover).

[Juhazi]

Perhaps the tweeter should be recessed a little more in relation to L15, to get better phase match? Easy to check by lifting the mid a little up from baffle - insert a sheet of plywood there.

Vertical off-axis measurement at 0.5-1m distance, with 3-4ms IR gating can tell the same, and also the direction of offset! If the null gets deeper upwards, you should lift the tweeter and vice versa, to get perfect match at listening axis.. I am struggling with this with a pair of dsp-active 3-ways when I want the xo around 4kHz, really picky.... That project is on hold now.

[Dave Bullet]

I haven't played with physical acoustic offset. the brightness problem seems to be anywhere in the room / where I stand etc... Listening to Alan Parsons now and smooth as. I might just put some of this down to overly bright recordings.

One thing to me is I need a tilted response as it sounds flat. Any more treble and i would think it is unbalanced to my ears.

The low impedance was doing my head in. So I reverted to LR2 for bass-mid and maintained the Steeper LR4 mid-tweeter xo. Now I have a "healthy" 4.3 ohms. My receiver can drive that happily. I've kept the tilted response as I think this will help the bright recordings I seem to like. The other option is to swap out the metal L15 for a "softer" midrange. But I do like the clarity of these drivers.

Latest trials and tribulations. A little out of phase action between 800 and 1700 Hz. Added network transfer function.

[Dave Bullet]

So far so good. All CDs sound great except the pretenders,. Paul young and a couple of Phil Collins tracks. Very bright / forward. Loving LR2 mid to woofer. Very smooth transition and music sounds full / complete. Amplifier happier too .

[sdl2112]

Hey Dave, I remember your thread now. Nice work! We are in a similar place in tweaking the xover. Looks like we are both using the euro style terminal blocks to link the temporary components...works out quite well. I like your outdoor measurement setup, that's dedication!

[Dave Bullet]

Hey Dave, I remember your thread now. Nice work! We are in a similar place in tweaking the xover. Looks like we are both using the euro style terminal blocks to link the temporary components...works out quite well. I like your outdoor measurement setup, that's dedication!

Thanks. Sadly as mentioned, my mic pre-amp has died so awaiting a replacement to continue.

I've reverted to 2nd order slopes between W-M. Although this sacrifices excursion (more) on the L15 (mid), it provides a much happier bass minimum impedance (4.4 ohms). Although I have a beefy power supply (2 x Cambridge audio 840w amplifiers) and they run class A hot anyway, I want to have the flexibility of powering these on lesser spec'd amps if I want.

I'm also back to conventional 2nd order electrical on the tweeter to get a 4th order LR slope @ ~ 1800Hz. I was running the tweeter a little higher and steeper to tame the brightness but this didn't work. So I'm back to source material being the culprit. Everyone says the 27TxFC family can take a crossover point and slope and I'm sure my woofer is running out of xmax before the tweeter will strain.

The latest attachment will be my revised "flat" on axis response when I can do another measurement and final crossover parts order. I am going to swap the 27TBFC/G back in and (add another 1 ohm of padding to account for the increased sensitivity on the fabric dome version) then really target it's brightness. I might also try a couple of lesser tweeters to see if it is just the particular tracks that bug me. I might even need to have a "brightness contour" that is switchable to satisfy those compressed rock / pop recordings we both like.

I also learned of a way to profile a track's FR via another forum. I'm going to try that on my offending music to see the profile the engineer has mixed in. Might explain the missing link for me.

[BobEllis]

Dave, It may not matter much at speaker levels, but when using stranded wire in those Euro-block terminals an end crimp should be used. I used them in an amp I built for mains and rail connections. I saw a wisp of smoke coming from the amp and shut down. When I opened it up, one of the rail wires had begun arcing to its connector, melting it. I shudder to think what could have happened had a mains connector arced when I wasn't in the room.

[Dave Bullet]

Dave, It may not matter much at speaker levels, but when using stranded wire in those Euro-block terminals an end crimp should be used. I used them in an amp I built for mains and rail connections. I saw a wisp of smoke coming from the amp and shut down. When I opened it up, one of the rail wires had begun arcing to its connector, melting it. I shudder to think what could have happened had a mains connector arced when I wasn't in the room.

Sage advice. I've been very careful to twist strands and overlap within the connector where the block allows (i.e. screwed both sides from both wires). These are only temporary and allow me to securely switch in and out components without risk of a rogue wire shorting out the circuit and causing a min impedance issue for the amplifier.

[Dave Bullet]

Here's my "final" crossover - hahaha. I ended up going 3rd order on the tweeter, even though it only achieves a 4th order LR slope. Why? Better phase tracking in the treble region with the midrange (between 2500 - 3600Hz).

I've also found with simply increasing R6 - I can get a nice tailing treble response without mucking up phase tracking. This will come in handy should I swap in the 27TBFC/G.

I finally got my new mic pre-amp so will give that a try this weekend to see where we fare.

I also am going to go to the "dark side" and apply a PC Equalizer to find out which frequencies are causing the harsh sounds through the tweeter. Here's my front runner candidate:

https://sourceforge.net/projects/equalizerapo/

[Dave Bullet]

So new mic preamp works. Always a good sign. In fact everything does after being in storage for a few months. Surprised the old Lenovo PC I use for measurement and paid USD$16 for still works.. .but it does :-)

Here's the measurements compared to the simulation. This is my "sub optimal" crossover, (not the one in the post immediately above), since I don't have enough parts values to create.

In each case blue = sim, red = actual measurement.

What concerns me is my notch filters on the woofer and midrange to tame the breakups don't seem to be working! I checked they are wired.... will do am impedance sweep next.

[Zvu]

A small cap 0,22uF-0,68uF across the midrange and woofer series coil can tame the breakup better. I have a bunch of those and i almost always have to tweak the value to make it right. Maybe you are having problem with your loudspeakers breakups rather than tweeter ?

[Dave Bullet]

A small cap 0,22uF-0,68uF across the midrange and woofer series coil can tame the breakup better. I have a bunch of those and i almost always have to tweak the value to make it right. Maybe you are having problem with your loudspeakers breakups rather than tweeter ?

I'll try the "series" method of attenuation. However it's almost as if I never wired up any breakup attenuation or my software can't handle it. Having said that the 8khz primary breakup is dealt to the same with very close values as zaph did for his L15 2 way now no longer on his site.

The harshness is definitely the tweeter providing most of the issue. Icpuld tell by just playing each driver individually with its own leg of the crossover.

I've downloaded a 31 band eq. And taking up to 10 dB cut in a parabolic curve between 1.75khz and 5khz definitely "flattens" these bright recordings to my ear.

My solution might be a TBC Q notch over these frequencies that I can switch in or bypass as required. Or a software EQ which would be preferable that I can switch to.'bright mode'.

[Dave Bullet]

Ok - I've found two equalizer settings that work with all sorts of pop and rock music as shown. Next step is to take each driver out of the equation to see if it or its breakup is contributing to the strident / forward / harshness.

It's definitely centered around 3Khz.

Examples of music well suited to the top curve - R.E.M - greatest hits or clapton's guitar on phil colins I wish it would rain down
Bottom curve - most other 80s music

Flat sounds great with vocals, acoustic, pink floyd etc...

[Zvu]

That just looks awful.

[Dave Bullet]

That just looks awful.

It's all information to help pinpoint the troublesome frequency spots. I was listening to some live clapton and Curtis Mayfield and flat the speakers sound wonderful.

[Zvu]

I think that your speakers aren't the problem. It is awful how the recordings are ruined that when you play it through flat speakers, it sounds dreadful.

I'm not sure that there is a way out of that circle at this point. Not in next 20 years at least.

[jkaapu]

I think that your speakers aren't the problem. It is awful how the recordings are ruined that when you play it through flat speakers, it sounds dreadful.

I'm not sure that there is a way out of that circle at this point. Not in next 20 years at least.

Hello,

I suspect that there is a bump in the power response around 3k. Measure the speaker in for example 20 degree steps from zero to 180 degree and upload the data into VituxCAD. Try to adjust the straight frequency response towards more flat power response.

The 2k xo frequency does that with the selected drivers.

J

[Zvu]

Good point.

We'll just have to wait Dave to measure 0-90 degrees to see what's happening.

[Dave Bullet]

Thanks guys. Very good point I have been lazy and must do some off axis measurements.

[Dave Bullet]

Good point.

We'll just have to wait Dave to measure 0-90 degrees to see what's happening.

I will do both "sterile" (gated) measurements plus an in room measurement.

The sterile measurement shows flat on-axis. It is quite possible the off-axis curves show a peak in this sensitive audible band of ~ 2.5 - 4.5Khz and this amplifies in room.

Toed in the speaker is slightly brighter on these tracks but still unlistenable flat (at volume) at angles upto 30 degrees.

Hopefully I get a chance to measure this weekend.

[Zvu]

Look at stereophile measurements of Usher BE718 or Joseph Audio Pulsar. I expect something like that when you start measuring quasi anechoic.

My head is so much into waveguides and coaxes that i totally overlooked the fact that you didn't use it.

[Dave Bullet]

Look at stereophile measurements of Usher BE718 or Joseph Audio Pulsar. I expect something like that when you start measuring quasi anechoic.

My head is so much into waveguides and coaxes that i totally overlooked the fact that you didn't use it.

Cheers. However none of these reviews ever use trashy 80 s pop recordings. They are aloof and only play high end stuff. I suppose they think people who could afford them would never listen to the trash I love ��

[Dave Bullet]

Play brass in pocket by the pretenders and tell me how it sounds on your system

[Dave Bullet]

Supertramp - very best of - sounds great. There's a little bit in Crime of the century with lead guitar which is a bit forward. I could put it down to L15 breakup, but I'm hearing this from the tweeter. There's a certain resonance ~ 2.8KHz the SEAS 27tdfc/g struggles to deal with. Either its the mix or the tweeter or that particular type of guitar and harmonic and mix.