The making of: The Two Towers (a 25 driver Full Range line array)

I have to get back on a topic from quite a while ago. I've always claimed that my arrays ran pretty much full range, even when I added the sub-woofers. This was quite true in the case where my drivers still were the Vifa/Peerless TC9/FD18-08. At the time I did experiment with a more regular crossover to see if it gained me something at mid-frequencies. I never had the impression it did. So I always went back to running them full range and stopped thinking about it. (you know me, I always try something more than once ;))
Fast forward to the Scan Speak 10F/8414G10, from the start it was clear this is a different driver. I've even made a slight comparison here: https://www.vandermill-audio.nl/switching-to-the-10f-8414g10/
I've said that the TC9 was better at low frequencies, something I've always suspected when looking at the German magazine articles about these drivers. At the higher frequencies the 10F is the better driver. Still I didn't use a true regular crossover, the 10F still did some minor lifting down till 50Hz where I did roll it off. Granted, I had rolled off the lifting some 9 Db below the flat response already. Due to the 10F being slightly higher sensitivity and a move towards up-sampling to 96.000 I stopped using the Atom headphone amp as a pré-amplifier quite a while ago. I just never gave it a thought to alter the low end to make the drivers in the array move less.
Recent threads made me rethink that strategy. Seeing the BL curve from the Voice Coil tests made me reconsider it actually.

https%3A%2F%2Faudioxpress.com%2Fassets%2Fupload%2Fimages%2FFigure5Scan-Speak%252010F-8414G10.jpg


Bl curve as measured by Vance Dickason for Voice Coil Magazine, also published by https://audioxpress.com/article/Test-Bench-Scan-Speak-10F-8414G10-Small-3-5-Full-Range-Woofer

Symmetry:

https%3A%2F%2Faudioxpress.com%2Fassets%2Fupload%2Fimages%2FFigure6Scan-Speak%252010F-8414G10.jpg


If you've read the article further, it will show that Kms symmetry is rather good in comparison.

https%3A%2F%2Faudioxpress.com%2Fassets%2Fupload%2Fimages%2FFigure8Scan-Speak%252010F-8414G10.jpg


In the Purify blogs I had read, it was mentioned the distortion due to Kms was less detrimental for sound quality as perceived than the effect of asymmetric Bl.
They might look similar when glanced upon (as noted in a thread here somewhere), but really are two different phenomenons. Reason enough to revisit my bottom end adjustments and tweak it further. I still don't do the regular crossover, though the end result of what I do will look much more like one :). Due to being placed asymmetrically in my room, the left and right arrays have different strong points and weak points due to the room or corner helping out. Plus there are two subs that do most of the lifting. Anyway, I've altered the bottom end and have removed a lot of low frequency energy from both the arrays without losing a smooth bottom end or much output. Having 25 drivers per side already helps a lot to limit any needed movement.
A first listening test: I'll be biased as you know what, but different from my tests with the Vifa/Peerless... The top seems more clear, rounder, less edgy. I've always thought/felt the 10F where more "in your face" than the Vifa's. Literary every part of my DSP is different with these drivers. That edgy character, the "in your face" part isn't a bad thing, but it had bothered me when I was listening when tired. In fact, I started to wonder if I got more tired from listening.
Due to life events etc, there are many more days that I'm tired than I used to be. Never really got back up and running after my heart attack. Walking? Yes, but not quite back to the same level of fitness as I've had before.

Can I ever do a short story? Man!

Anyway, this reduction of heavy lifting of the 10F drivers seems to do the trick. I'll swear I hear more detail :D. Time will tell.

Makes me wonder though, should I also try a tweak to put more impedance before the mid-range? You know, a variation of current drive with a regular voltage amplifier.

https%3A%2F%2Faudioxpress.com%2Fassets%2Fupload%2Fimages%2FFigure9Scan-Speak%252010F-8414G10.jpg


Le curve at different cone positions...

If I put a notch in place, I can raise impedance to have the mid-range act more like current drive, while keeping the bottom end as is for all practical purposes.

notch.jpg


Here's what that does for the frequency curve:

frequency.png

Green is with notch applied, black is without notch.

So there's a 7.5 dB dip that would need to be EQ-ed away, which won't be much of a problem with a 400 watt/8 ohm amplifier. The drivers won't see more power, just more equal current delivery according to the theory behind current drive. Not that I'm calling this current drive (I wouldn't dare) but for practical purposes it comes close enough to see a possible gain.

The impedance plot:

impedance.png

Red is with notch applied, blue is without notch. My impedance correction network is still in place.

Stating again, the drivers won't "see" a difference in power applied. The passive components will burn off the extra power as heat, but there is a possible gain in the current the driver "sees" when it is asked to move due to some lower note playing simultaneously as the mid-range.

I guess I could simply put a resistor in series with the arrays to try first. That resistor would have the same function but simply would drop the SPL everywhere:

frequency2.png


It will do as a first experiment though. Before buying stupid expensive coils.

Will you guys come visit me once they finally come get me? Or might these thoughts be sane enough to try...
Part of me thinks Le is so low already that adding the impedance would not really change anything here.

What do you guys say, anyone willing to enter this mind game?
 

Attachments

  • impedance.png
    impedance.png
    13 KB · Views: 8
I have to admit I have never thought about current drive so I'm not sure what you are driving at and I may be off bass.
I do know you can make a crude current source with a resistor in series with a voltage source, such that the resistor swamps the load impedance. Then the voltage source has to be capable of significantly higher voltages. Inside silicon, its done with mirrors (current mirrors) with more precision.

If you put in a dip in the FR via the impedance trick and then EQ it away, what have you done?
In voltage drive, you apply a constant voltage and the driver power is V*V/Z.
In current drive, you apply a constant current and the driver power is IIZ.
For the same sonic output, IIZ and VV/Z must be the same.

I think if you have a resistor in series with the arrays, you simply make the amps work harder to heat the resistor. That may shift the operating point of the amp to a more linear region, or not. It would certainly solve a minimum Z problem. I don't think it does anything to or for the drivers after you EQ back to the original FR.
 
I believe it does, as it has influence on the Le that is fluctuating with a moving cone. The relative difference of Le depending on cone position changes with that added resistor. The cone will move with the lower frequency tones. While it moves, it may play a higher tone at say ~1 khz. Depending on the position there will be a change in the current that flows trough that voice coil. Current drive keeps that current stable, because the added impedance swamps the change in Le based on cone position.

See: https://www.diyaudio.com/community/...nge-3way-prototype.408882/page-5#post-7603833

The labeling on the graphs leaves a lot to be desired, but sweep 2 actually is H2 distortion, sweep 3 shows the H3 etc. These quite simple experiments show advantages an uneven distortion products. There are many more of these recent experiments using something as simple as a resistor to start seeing improvements. But the advantages will depend on the driver (Le is really low on the Scan 10F to begin with) and how much excursion it's going to have.

But it's different from just heating a resistor. Keep in mind that many frequencies on the array take very little power to move the cone. I won't use a resistor, but as said, the notch I suggested would have a similar function. With the advantage that it won't show worse results at the driver's impedance peak.
 
So its a secondary effect you are looking for? So much for my superficial analysis.

The link to the 3" dome is interesting. It does show a difference in THD. Worst case I saw in the charts was roughly 1% which is many consider inaudible. That is the question - would the improvement be audible? Telling would require a double blind test.

No doubt you've seen this from BrunoP and can or did relate his analysis to your case:
Low distortion filter...
 
Yes, I've seen that :). There is a lot of sense in these theories, and current drive would definitely help some drivers better than others.
I'm thinking with the low Le on the 10F, the differences there may only be marginal.
The biggest step I could make is to reduce it's movement and that has been on my agenda.

I've done a lot of thinking lately, always dangerous, I know. I've had a harder time to get the filtered 10F arrays somewhere really good or even great than I've had with the previous unfiltered arrays. (or I simply forgot how long it took me to get there, that's another possibility) And I think the change in vertical output has something to do with that.

Let me add the comparison again:
shaded-unshaded.gif


It is clear that the floor and ceiling get way less energy distributed to them with these filtered arrays. But if we look at the ~45 degree up which is a reasonable resemblance of the ceiling angle as seen from the listening spot, the unfiltered array is way more even in it's distribution of energy toward the ceiling. In comparison the filtered arrays have pieces of blue sky with green pastures in between. When I first put up damping panels to absorb early energy, I also stole something from the room. The excitement that energy brought was greatly reduced. I've put that back in by adding ambience channels.

My biggest battle isn't the 10F being so different, it actually is that these little nuances, as they seem, the change in off axis patterns make me work harder to regain that sense of a complete room again. As I do fake the room. You don't know what you're missing till it's gone. :) That even vertical pattern of the arrays is worth something.

So I don't need to chase the notch to simulate current drive, I need to work on the room. And that can be done with my ambience speakers. Today's testing has proven that to me. Even though I add a sniff of reverb way down in level from the mains, it has a rather large influence. That edgy or in your face feel I've mentioned a few posts ago could very well be nothing more and nothing less than these little things that stand out when all else is dealt with. At least, it wouldn't surprise me. On listening axis it's way more even and balanced, but even the off-axis balance matters. I guess I'm finding out slowly but surely how big of a change that actually makes.
 
It will be much easier to see if I cut off the graphs to show -45 up to 45 degree angle. The floor will be at an even lesser angle but to get a feel for it.

Unshaded:
25x TC9 FR Unshaded-ABEC Directivity (ver)-45.png


vs Shaded:
25x TC9 FR Shaded 19.0 as build-notches-ABEC-minphase Directivity (ver)-45.png


To talk with the wise words of the Jackson 5: Can you feel it!

So be careful what you wish for, jou just might get it! ;)

It makes me wonder what a more common 2 or 3-way would have if it isn't a coax or similar.
Though one would only notice things like that if you remove the earliest reflections from the room, like in the first ~ 20ms.
 
Don't get me wrong here :). I'm not regretting what I've done in any way. This is just me thinking out loud like I've done several times in this thread.
I've puzzled with the question why it didn't become easier to get everything adjusted to my liking. The answer to that probably is easier than I first thought.
It is similar to an earlier experience. When I removed the first offending reflections, I gained a lot in clarity and detail of imaging. However, something was lost too.
Before I removed the reflections it was like I was in the middle of everything and that was in fact very addictive. It was wilder, yes, but highly entertaining.

I solved that dilemma of the stolen reflections (the added panels) by adding the ambience channels. This time around, I've removed even more of the room effects, but as we can spot, it may have become a bit more restless in the vertical direction. I noticed it in guitar parts and vocals, where some notes seemed to stand out. A while ago I had it covered by changing the settings of mid/side EQ while adding anti cross talk and by adding some low level room reverb using one of Lexicon plugins.

But, as time moved on I returned to mid/side EQ only (removing the anti cross talk construction) and wanting to remove the sniff of reverb from the main arrays.
That too worked out until I changed up the ambience reverb, that now had the room part added. I figured I'd keep the 'early' part of the room reverb and remove the 'late' part as I'd still have the original mixed in Random Hall. So Random Hall was equally edited to remove the 'early' part and keep the 'late' one in. Never gave it much thought but something had changed. I got some kind of listening fatigue, having to reduce the volume etc.

Room-Lexicon.png


Thinking about it even more, I questioned my use of the Scan 10F helping out at the bottom end etc.

Until it hit me: did I leave a gap between that first "room" effect and the Random Hall kicking in? Was this making me listen to that 'robbed energy' from the room, with that slight unevenness I pictured above?

First let me tell you: If I did not have absorbing panels at early reflections, none of this difference in the floor/ceiling reflections would ever have been noticeable. It would have gotten drowned in the overal level of reflections and totally overwhelm any effect it could have had. Add to the fact that if I spot something like this, I'm usually not in a relaxed listening mode and my family members never noticed anything or complained about it. I'm just sensitive like that (lol). Especially when I'm tired.

Yesterday I put back the second part of the Lex Room reverb and all was well again. None of that harsh or uneven tones in vocal parts. Guitars no longer piercing in my (admittedly overly sensitive) ears. Back to a relaxed mode of listening. I could feel the tension falling off of my shoulders. More important: shivers down the spine again! That's always a good sign.

So lesson learned? The more you remove the energy from the room response, the more every possible deviation is going to stand out. It's like putting everything under a microscope. It might also be hinting at the ideal type of speaker: a "coax like" source. Be it a true coax or a Synergy. If done well they have the true even coverage in all directions. It might take a more clever horn than the PA conical type to truly get that kind of coverage. Like an ATH type horn, a big one with the synergy ports added. I don't think a horn + woofer underneath could ever compete unless one would remove floor + ceiling reflections with passive means.

Anyway, I will stick to my concept, it just takes a bit more work to get the most out of it. As far as space savings go, not much can compete. After my session yesterday I'm convinced I'll be back in business very soon, ignore that notch part a few posts back, as I won't be needing it anytime soon.
 
I thought you might enjoy this quote from an Art Welter post in multiway last night

"
I found the TC9FD subjectively to sound cleaner than any compression drivers at high volumes, melted it's voice coil off the former with no sound of distress, while a 3" diaphragm driver well within it's thermal range at the same SPL sounded harsh and IM (Inter Modulation) distorted.
The Purple TC9FD clearly is the winner in output below 600Hz, but with no phase plug, response is erratic above 3kHz."

He compared TC9FD to a few compression drivers, all on the same horn.
 
  • Like
Reactions: 1 user