DiscoCAT Speaker Design

[S-Man]

Well blow me down with the output of an ATC port ...

Upping L4 to 1.3mH has considerably mellowed the sound of the DCs. On closer inspection the difference in level is more like 0.5dB!
Even Crissie Hynde is not too ear piercing now.

I have a feeling I will end up with a value somewhere in the middle. It might even come down to tuning L4 turn-by-turn to get the right combination of "bite" and tonality.

 

[S-Man]

On the subject of tonality...I took the DCs to a pal's house to try them in another system. I was running V3 XO at the time, which is brighter and more lean than V4 (even with the original L4).

The DCs sounded very bland in his system. His PMC Twenty 23s sounded a lot more lively.
I can't find any published measurements for the PMC, although there are some for a standmount version on everyone's fav website/forum:

https://www.audiosciencereview.com/...pmc-twenty-21-bookshelf-speaker-review.14442/

No wonder my speakers sounded a bit bland in a system that worked quite well with PMCs.

 

[a.palfreyman]

As they say in Nottingham, guz up an daan like Cox's 'osses...

 

[S-Man]

Cone break-up?

Funnily enough I just noticed Troels uses the 15W8434G00 in one of his designs and he comments:

"The choice of the 15W for midrange proved to be right. The fullness of Ingram Washington's voice combined with the crisp presence is there - and a seamless transition to treble range. No upper mid harshness from any rubber resonances, etc."
http://www.troelsgravesen.dk/Discovery-3WC-mkII.htm

Mind you, he would grimace at my lashed up crossovers. One of his fancy caps probably costs more than my entire crossover . And Danny boy certainly wouldn't appreciate the electrical terminal strip for speaker cable connections instead of "toob" connectors .

 

[a.palfreyman]

...as you mentioned difference in phase tracking between L4 values with little difference in amplitude, I wonder if it could be due to a different phase shift between a fundamental of say 1.6kHz (mainly from mid-bass) and third/fifth harmonics at 4.8/8kHz (mainly from tweeter) that changes how 'hard' the sound seems...

 

[S-Man]

Quite possibly! I really don't know how important the phase tracking is and what criteria matter most.
In general people seem to go for a deep and sharp reverse null. Sometimes it's possible to get a shallower but wider cancellation. One octave either side of XO seems to be a reasonable target... but within how many degrees?

It would be interesting to measure some technically excellent commercial speakers like LS50 Metas to see how they have done it. I have access to a pair, but I'm not sure the owner would appreciate them being molested.

I think I may have to crack open the CAOW1s to do more detailed measurements.

 

[James]

I'd prioritise phase tracking over a flat crossover. To my ears, good phase alignment is critical for coherence. By that, I mean the driver pair sounds seamless. So a deep and narrow null with one driver polarity reversed is what I'd look for. Especially with an even order LR cross.

 

[James]

Here is the phase tracking for my E-IXs.

and summed frequency response.

 

[a.palfreyman]

Wish I'd thought of this:

a deep and narrow null with one driver polarity reversed

There you go then @S-Man, rather than pull the CAOW1 speakers apart, measure with both values of L4 and tweeter reversed to compare nulls.

 

[S-Man]

I'd prioritise phase tracking over a flat crossover. To my ears, good phase alignment is critical for coherence. By that, I mean the driver pair sounds seamless. So a deep and narrow null with one driver polarity reversed is what I'd look for. Especially with an even order LR cross.

Maybe my thinking is wrong... but surely the perfect null is deep AND wide? The wideness indicates good phase matching (when not reversed) over a wide range of frequencies. Narrow means that the phase tracking is only accurate at the XO frequency.

Also, the lower the XO order, the wider the range of phase coherency needs to be.

Of course this is all only true at the measurement point! Although adhering to a known target curve e.g. LR2 or LR4 should result in predictable off-axis behaviour.

 

[a.palfreyman]

surely the perfect null is deep AND wide

That depends on the dB/octave roll-off. I'd think LR4 would be maybe +/- 1/2 octave and LR2 to be maybe +/- 1 octave or thereabouts.
In fact, Vance Dickason Cookbook says -4dB at +/- 1 octave for LR2 and -1.5dB at +/- 1 octave for LR4.

 

[S-Man]

Not sure I follow that . Shouldn’t the target be in degrees/phase rather than dB/amplitude?

I have VD’s 6th edition Cookbook. In which section is this info?

 

[a.palfreyman]

What I meant was the width of the 'hole' with tweeter reversed from normal phase. LR2 being 4dB down at +/- 1 octave from cross over point, whilst LR4 is only 1.5dB down at +/- 1 octave.
Chapter 7, passive and active x-over networks, there are graphs for various configs. LR2 magnitude response inc reverse phase graph fig 7.29 and LR4 magnitude response inc reverse phase graph fig 7.58 in 4th edn. Fig 7.29:

 

[S-Man]

Ah yes, I understand now. Thanks! You did well to find that in rather an impenetrable book (for me, at least).

It seems that you and James are right. My idea that deep AND wide is good appears to be wrong. It should be: deep and the correct shape.
That's helpful because it refines what I should be aiming for; previously I was trying to align the phase plots (good) but aiming for the wrong reverse null shape.

FYI the figs are still the same in the 6th edition.


VD presents his LDC6 monitor design in chapter 12. He commentst that "hi-end" speakers often go for 2KHz or lower XO point and therefore LR4 is a good solution. I might have to investigate this approach further.

 

[a.palfreyman]

You need a 81oody good tweeter tho...to cross at 2kHz, and there is the inevitable resonance peak to deal with just below x-over.
With your L4 value change giving you a nicer presentation, I'd investigate the change in null shape with variation in L4. The one that gives you a nice symmetrical null should give better balance...and hopefully confirm your change from 1.05 to 1.3mH

 

[James]

Relative phase is one of the trickiest aspect of loudspeaker design, IMO of course. Two things that matter the most are driver (relative) offset and transfer function. The optimal target is to get driver phase as perfectly aligned as possible e.g. graph of mine in post #48. Sometimes that is not possible, and there is a degree of phase offset, but they track consistently over the crossover region. If the reverse null isn't deep, then it is quite likely you have relative phase offset. Just make sure when you measure the drivers' FRD in-situ, the measurement conditions are identical i.e. same distance (I prefer measuring at 2m) and point in space for the microphone, and toggle each driver to be measured in and out as you capture FRD. @S-Man, do you have a graph showing FRD relative phase of the two drivers?

You need a 81oody good tweeter tho...to cross at 2kHz, and there is the inevitable resonance peak to deal with just below x-over.

I also prefer to cross HF at or below 2kHz. My E-IX, for example, crosses at 2kHz. An earlier design (E-VII) with Scan-speak 2905-9300 and 18W8535-00 (sealed, of course) crosses at 1.8kHz. As long as the fs of the tweeter is at least one, preferably two octaves below, the acoustic crossover will behave predictably.

 

[S-Man]

I think the CAT308 should handle 2KHz. Its Fs is 650Hz.
I have tried to SIM an LR4 at 2K and somehow can't get it quite right. I think the issue is that the diffraction effects of the cabinet and grille frame are very difficult to accommodate. I have some ideas to try to reduce the diffraction issues.

In the meantime I got bored with the sound with L4 = 1.3mH. Oddly enough it sounded fine at low levels and for late evening listening sessions but sounded a bit flat (i.e. the musical opposite of sharp) when having a good blast.
-3 turns results in about 1.21mH and a more engaging sound. I'm really surprised how very small changes can have a significant effect on the sound .

 

[S-Man]

Action against diffraction (10mm thick felt is very difficult to cut!):

Danny-wouldn't-approve crossover:

I made the diffraction felt some time ago and it changed but didn't improve the measurements. Oddly enough, some smaller bits of felt did help the curves and have therefore been in use for a while.
I thought it was worth a go with the full-felt version and it defintiely sounds better and indeed seems to sort the peaky-voice issue. So much so that I will probably revert to 1.05 or maybe 1.1mH for L4 - just feels like the mids need a tiny bit more bite now that the upper midrange glare is gone.
I' have always wondered if the diffraction effects which look so nasty in measurements are actually audible. I think I have the answer now!

I think I'm getting to the point where I can say that the DCs beat the CAOW1s. The clinching factor is the increased resolution, the DCs reveal backing and multitracked voices so much better than the CAOW1s. They also show harmonic structure of many instruments better and reveal more ambient and decay info. However the CAOW1s are kinder to lesser recordings, they have an effect like a really nice valve amp and seem to add a bit of lushness.

I really do wonder if the critical difference is in the cabinet construction?: CLD (or whatever TF it is) versus chipboard/mdf/ply with Blackhole 5 lining. I suppose the only way to find out is to retrofit some CLD in the CAOW1 boxes. When I tried external bracing of the CAOW1 boxes (with quick-clamps) I didn't like the result - so it's pretty obvious that the boxes are contributing something.

I probably ought to get the LS50 Metas over here again and have a comparison/listen. Surely they must be a low diffraction and low box colouration design? (So why can't I live with them ).

 

[a.palfreyman]

Less 'smearing' = better resolution? Could be a combination of good phase tracking btwn drivers, lower box resonances and your rather nice diffraction control which appears to sit flush with the face of the frame.

 

[S-Man]

Yes it seems to a combination of things.

Just got in and plonked RP Rock Mix on for a casual listen. A very busy/manic Santana track came on and I was struck just how many layers of percussion I could hear. Just listening to Beds Are Burning by Midnight Oil and I’ve never noticed the multilayered vocals before. They don’t do this by pulling the music to bits though, quite the opposite.