Something stirs in the Undergrowth

[IslandPink]

Thanks, that's a very interesting link. I will have a look at that briefly now but also tomorrow . It's interesting because I'll need a low-pass via an R ( in-line ) and C ( to ground ) for this amp anyway . If the inductor will modify the roll-off into some sort of notch on the way down that will be very cool .

Just been listening to 'The musical box' after the Genesis evening last night. There's more coming out that I've not heard before. the precision in the lower mids and bass, and the dynamics , really give you some more insight into the mix. There's some more tone coming out now so running-in is taking place . Superb !

[simon]

Simon - have you run these two drivers yet ?

Sadly not yet. I've made a start on the cabinets for the Alpair 12Ps, which will take me a while to finish because of time pressures, but at least it's a start. Interesting that you prefer the crossover without the 3R3 resistor.

[IslandPink]

I've been thinking about this and maybe the resistor is correct to get it flattish in the crossover area, but audibly perhaps the Fostex with its rising output from 1-4k sounds too bright unless you raise the Supravox a couple of dB to compensate ? Measurements are really called for here. Perhaps I'll change my mind once I roll-off the Fostex to bring in the horn ?

Nick - I couldn't get that website to work so far, just keeps returning an error, but has no hint as to what's wrong. I played around with LTSpice and some varying values of R ( the L and C set the notch freq ) & got a notch but with a rather long slope leading into it from LF to 4k which was not useful . I'll play around with some other configurations ; eg. try getting a 1st-order steepening to 2nd-order low-pass - like with the passives I tried before . Maybe it'll be OK for my purposes with just a 1st at 800Hz, but you never know how vocals could still be adversely affected in the treble by a 4k peak about 15-20db down .

Edit : just playing 'Tres Hombres' by ZZ Top and it sure'nuff sounds excellent . Really is a lot better on drums and bass guitar than anything I've been using lately .

[Toppsy]

I've been thinking about this and maybe the resistor is correct to get it flattish in the crossover area, but audibly perhaps the Fostex with its rising output from 1-4k sounds too bright unless you raise the Supravox a couple of dB to compensate ? Measurements are really called for here.

Now why am I not surprised by this. I too had exactly the same with the new ring dome tweeter on my Norge floorstanders. I and others much preferred a smaller value inline attenuation resistor on the tweeter than the simulations suggest to be the optimum for the relevant efficiencies of the drivers. No doubt down to similar findings as you. Measurement is good to see what the drivers are doing but ones ears usually tell you when the sound is right.

Hope to hear these at the next Owston meet.

[Lynn Olson]

Looks really nice, Mark. Congratulations!!!

The last time I messed around with a fullrange Fostex, the response had several peaks in the 1~5 kHz region, nothing at all like the published curves. If you're going to hit it with a notch filters, I'd find the first (lowest frequency) peak, attenuate by 5 dB or so, and see if that sounds better or not.

If you have a quick way to sweep frequencies (apps on iPhone and Android are available), you can determine by ear where the most audible peaks are, then clip-lead in the notch filter, tuning it by ear for frequency.

But it might just sound dull and flat after the notch is inserted (on the correct frequency); that happens to me about half the time. It seems to depend on the speaker, and the FR and waterfall don't tell the whole story.

[IslandPink]

Thanks. With the work I did with it a few pages back, on the OB with the ribbon ( Basszilla > Pinkzilla ) it appears the Fostex loses most of its aggression when you apply a 3-4db notch centred around 3.5 to 4k . At least that was OK with a ~2kHz low-pass. You can also retain most of the upper mid tone with the limited phase-shifting of that notch . So hopefully for a 800 to 1kHz low-pass I will be OK .The problem really is how to do that in the context of the passive series crossover which is very good and could easily be screwed-up by adding notch elements . I'd prefer to do the notch in the driving amp , but first design efforts have not yielded much . I need to spend a bit more time on it - should be possible .

Latest capacitor news : decided to order a pair of Mundorf M-Cap 'EVO' Oil 100uf/350V caps to improve the current unwieldy collection of Ansars, GE and PIO for the 135uF . Ideally I need to get measuring to determine exactly what's required on that cap, too ... but could do some clip-leading forst & assess by ear .

[Lynn Olson]

Kind of hard to see a good way to do a notch filter in a power amp. Active notch filters usually require a positive feedback loop, which doesn't do any favors to the sound (distortion rises sharply in the positive-feedback region, which is very audible). That's the big sonic problem with textbook Sallen & Key filters; the PF loop degrades the performance of the overall circuit. Not sure, but I think state-variable filters don't require a positive-feedback loop.

I guess you could use inductors as part of a trap filter in a dedicated valve circuit, but they are going to be really big because of the high impedances, and will also need shielding to avoid hum pickup. My general impression is that complex filters using non-feedback valve circuits get difficult and expensive very fast, which is why the old-timers avoided them. It took widespread availability of the NE5532/5534 and the TI TL074 op-amps for active filters to become more popular.

Moving over to high-level crossovers made with passive components, adding a notch filter to a series crossover is a non-obvious exercise. I always use parallel crossovers, partly because I don't want back-EMF currents from one driver to enter another. That's just a prejudice I have, though, and I can't defend it beyond that.

You might try a conventional parallel crossover, just for test purposes, and tune the notch filter by ear (using sweeps or pink-noise stimulus). If the notch filter gives satisfactory results (Fostex doesn't get all dull sounding), then maybe you can puzzle out a series-crossover equivalent.

If it's any consolation, resonant peaks in loudspeakers are usually minimum-phase, as long as the driver hasn't gone into full-blown breakup modes. Minimum-phase peaks can be accurately removed with conventional notch filters with no penalty in phase response, since the poles and zeros cancel each other out, leaving flat frequency and phase response.

But ... the notch filter has no effect on sharp increases in IM distortion associated with mechanical peaking, so the peak can still be indirectly audible as a region of "blur" and confusion. Peaking is also associated with sharp changes in directivity, which in turn changes the ratio of direct sound versus total room reverberation at the listening position. This is why measuring the FR at the listening position can give not-so-good subjective results, since it takes no account of distortion or directivity changes.

Although listening to fast sweeps or pink-noise with an A/B switch while you tune the notch filter sounds crude, it puts the subjective impression first, which is what really matters. I would also listen to massed choral, or something with a dense spectrum, to assess the impact of IM distortion in the 1~5 kHz region.

If the notch filter increases listening fatigue, or creates a sense of congestion or confusion, it's not working correctly. If the speaker sounds more natural, and less "mechanical", the filter is going in the right direction. Ideally, it should disappear and not draw attention to itself.

P.S. I still think your speakers look stunning; the Perspex really is a beautiful material.

[IslandPink]

What this speaker needs is a horn section -

[Greg]

Well you can't do it yet. You are on 99 pages so need some leeway to hit 100. Marvellous thread. Only been running since 12/12/08....over six years! Make sure you get there in the end

[IslandPink]

Thanks - have no fear - will easily hit 100 pages - plenty more to say !
Initail tests with the bi-amping have been interesting ... at first not very good
I just tried the usual 700Hz-ish low pass and high-pass in the two channels . First listening seemed to be lacking tone and with rather blunted transients - lack of detail . I tried swapping the cable polarity to the horn . This was almost the same - couldn't decide
After some thought that evening and overnight, I decided the horn is starting to provide its own roll-off around 700Hz and the phase of the horn section, even without an amp low-pass, is already rotating. I have used the horn without crossover at times in the past.
So, last night I lowered the in-amp high-pass from 700Hz to about 350Hz using a 4000pF(silver mica) /115k coupling to the 300B . This sounded A LOT better . Tone is back and detail much more satisfying . The whole rig sounds more together . Still , it not all that clear which polarity of the connection to the horn is correct - and it should be obvious . I might go down a little lower again with the high-pass frequency .
Now, there's still some occasional harshness on vocals - particularly male vocals . I now need to identify where this is , just by trying the two sections in isolation , first, then with some measurements - at the weekend . I have a few more tricks up my sleeve , anyway ...

[slowmotion]

Sounds similar to my own path. I think Martin in Australia also have similar leanings ....

[IslandPink]

Ah yes - Martin as well ?
Interesting ...
Well, with 175Hz in-amp for the horn, it's better again . My concerns on the male vocals are reducing . Dynamics are getting pretty snappy on 'Brand X' now, just playing .
If I can bodge around with a jig-saw and router a bit more this weekend, I should be able to mount up the Aurum G3 ribbon above the Azurahorn . I'm hoping with the G3 at upper-90's db/w it can add enough to the 8kHz+ end to just boost the Radian 745 usefully . Maybe with a small ( 3 to 6 R ) series-R on the Radian I can get a good balance. It needs a little extra . I'll use a small cap ( 1.5u ? ) & reversed polarity on the G3 , nothing fancy .

[IslandPink]

OK, here's a funny thing ...
I was starting to be not quite happy with the Radian as part of this system, in the last couple of days. Tonally and dynamically it wasn't integrating all that well into the combo.
This evening I just swapped over to the GPA 288H . All of a sudden things sound right - much better integration, more tonal depth in the upper mids. Just sounds really nice - very listenable . Also its efficiency seems about the same as the Radian - should be 3-4db less, but it sounds the same .
It still needs a bit more edge detail, but I think that's just a physical aligment thing ( bearing in mind previous experience with the 288H ) , as the 288H is having to sit higher and a bit further back at present - due to the hardware I made for the Radian - can address that.

However this is the first time I have a full-range system I think I can actually live with as a stereo pair , in the long term .
Result !!

[Lynn Olson]

Good thing I have a pair of GPA 288 Alnico's on hand, as an alternate to the Radian 745P's. I'll follow your lead and try both.

The 288, with its crimped-aluminum tangential surround, has more diaphragm compliance than the Mylar-suspended Radian. This lowers the Fs of the 288 by 100~200 Hz, and results in more output in the 500~800 Hz range.

The phase plug is also completely different too, with the late-model Altec and GPA using the Tangerine radial phase plug, while the Radian uses the traditional circumferential phase plug. This might have a interesting effect on the phase response in the 1~10 kHz region, and the subjective sense of coherency (maybe even tonality).

My dim understanding of circumferential phase plugs is they are "tuned" so the multiple acoustic paths through the phase plug are designed to cancel the first (dominant) mode of the diaphragm, but at the expense of a minimum-phase response in the 5 kHz and up region. (Response is flatter, but group-delay and phase ripples are present.)

By contrast, a radial phase plug transmits the response of the diaphragm as-is, leaving the first and higher modes intact. This would imply that a radial phase plug combined with a Beryllium diaphragm would be ideal, but nobody offers this, as far as I know.

The AH425 is specially designed to preserve phase and impulse response, so these kinds of difference might be more audible than other types of horn.

Also ... the AH425 is directional enough that listening just a few degrees off-axis will roll off the clutter at 10 kHz and above, so there might not be any need for VHF lowpass filtering of the 288 driver.

Thanks for all your experimenting and listening, IslandPink! Much appreciated here in sunny Colorado.

[waveguide]

Good thing I have a pair of GPA 288 Alnico's on hand, as an alternate to the Radian 745P's. I'll follow your lead and try both.

The 288, with its crimped-aluminum tangential surround, has more diaphragm compliance than the Mylar-suspended Radian. This lowers the Fs of the 288 by 100~200 Hz, and results in more output in the 500~800 Hz range.

The phase plug is also completely different too, with the late-model Altec and GPA using the Tangerine radial phase plug, while the Radian uses the traditional circumferential phase plug. This might have a interesting effect on the phase response in the 1~10 kHz region, and the subjective sense of coherency (maybe even tonality).

My dim understanding of circumferential phase plugs is they are "tuned" so the multiple acoustic paths through the phase plug are designed to cancel the first (dominant) mode of the diaphragm, but at the expense of a minimum-phase response in the 5 kHz and up region. (Response is flatter, but group-delay and phase ripples are present.)

By contrast, a radial phase plug transmits the response of the diaphragm as-is, leaving the first and higher modes intact. This would imply that a radial phase plug combined with a Beryllium diaphragm would be ideal, but nobody offers this, as far as I know.

The AH425 is specially designed to preserve phase and impulse response, so these kinds of difference might be more audible than other types of horn.

Also ... the AH425 is directional enough that listening just a few degrees off-axis will roll off the clutter at 10 kHz and above, so there might not be any need for VHF lowpass filtering of the 288 driver.

Thanks for all your experimenting and listening, IslandPink! Much appreciated here in sunny Colorado.

I think we need to thank Lynn for this useful and thought provoking view. Also, while the Be diaphragm is now available in a better metallurgical form i.e wrought dense metal, but clearly much of any advantage is lost because of the diaphragm design/material and wave plug.

Did write more but I will leave it there