Patience my ass I'm going to build something too!

[5th element]

What with all the building work going on around here and all the work surrounding coaxials I had to get in on the action too :T

As per usual my design is probably less likely to be built by others as it isn't exactly the easiest of builds due to the driver modification necessary but when have I ever done something simple?

From this thread you may have already seen the Satori with its dustcap removed and had some sort of an idea as to what that was perhaps about - afterall why would you remove the dustcap on such a nice driver except for a very good reason? :B

From my time experimenting around with converting the FST driver into a coaxial I had some Dayton Audio ND16 tweeters left lying around. These had already been slightly modified to allow them to fit into a voice coil of around 30mm inside diameter and as I had the cabinets lying around it seemed like an obvious excuse for buying drivers and building loudspeakers.

I had a look around for some small high performance drivers with a suitably sized voice coil that would fit the bill. Given the wave guide loading that the midbass cone of the coaxially mounted tweeter provides I decided I needed a soft controlled cone if I were to cross for a directivity match. Not only does the ND16 require a steep crossover around 3500Hz to work properly, a cross at about that frequency would be idea for a decent match to ~5" driver.

After having browsed through pages of driver info I narrowed down my choices to either the NE123 from Vifa or the MW13P from SB Acoustics. Both drivers had a large amount of appeal but the Satori won out. If I were using a stock ND16 the Vifa would probably suit better as its voice coil diameter is slightly larger, but the Satori, with its 30.5mm diameter voice coil would fit my already modified ND16s better. That and I've always wanted to use one of the Satori drivers with their class leading motor design.

For those of you wondering here again is the shot of the MW13P with its dustcap removed.



As can be seen the area beneath the dustcap isn't as open as you'd really like it to be. This was expected as I had seen cutaway drawings from the 6" Satori so this was something of a gamble - how high and how wide would the copper rings/cap be? As a point of interest the Vifa also has something on top of the pole that protrudes up a little and can sometimes be seen in shots that show its voice coil peforations so neither were going to be an obvious choice.

It was my hope though that the small motor structure of the ND16 would fit inside the copper ring, sadly this is not the case as the motor on the ND16 is ever so slightly too large to fit snugly inside and this is even with the plastic on the rear side trimmed off. Nevertheless the fit is just about good enough to allow this to work. With the FST the copper cap sits like a veneer on top of the pole and leave plenty of room free, enough so that my modified Aura NT1 old stock could fit. With the Satori that would completely impossible. Thankfully the Aura's ability to crossover at 2500Hz is not needed here. If you were to use a 6" Satori with its larger voice coil the NT1 could probably be made to fit.

Here is a picture with the ND16 in place on the modified MW13.



They fit together quite nicely. As you can see I have flying leads coming forwards from the tweeter as a way of carrying the signal to it. With the final design of these I am going to try feeding the tweeter signal out through the pole vent as it is not needed with the dustcap removed but I am unsure if there will be adequate clearance.

I have measured these and the results look promising. The ND16 behaves much the same was as it did when mounted on the FST with a suckout in the top octave, although slightly worse here due to the half roll termination of the Satori cone instead of the smooth transition found in the flat surround of the B&W driver.

[Steve Manning]

Cool stuff 5th .....

[csmielke]

Matt,
Now I see why you referenced the MW13P in your comments on the modified finalist. Very cool new direction - things just get better and better. I am amazed by what this group is capable of coming up with.
Chris

[ergo]

Very cool project!
Those ND16 are very nice tiny units. I've tried them too in a small speaker marrying them with a 3'' driver. If one makes a precise hole these go so well and snug in it that it looks very nice.

Matt, what do you mean with "my already modified ND16s" - is there something you've done to these small ones in addition to converting them to coax?

[JonMarsh]

Cool, Matt- you're not satisfied just making custom speakers, you want to customize your drivers! Hope this plays out to your expectations... Don't know what you're going to do about the surround influence, though!

[5th element]

Hi guys and thanks. There's not a lot you can do about those surround influences Jon all mid/bass coaxials show them to some degree just look at the C16N001/F from SEAS. Of course with wave guides the situation in the top octave changes significantly with the off axis angle and top octave stuff is usually to be expected. SEAS certainly attempted to do away with most of these issues by creating their new MR18 and C18 drivers, of course these are midrange drivers only, which seems to be a thing, a good coaxial really = midrange driver as you can have a flat surround. KEF, TAD, SEAS, me with the FST, we're all doing it the same way and have seen the same thing.

It is going to be interesting to see how these turn out. For the system that they are going to be used in I have a DSP available so I can fine tune the top octave balance quite easily - if it's needed. Hopefully it wont be.

[5th element]

Speaking of the measurements and unattractiveness here they are!



These have been taken between 5-10degrees off axis. It is usually typical to do this with waveguide/horn or coaxial drivers and use the slight off axis as the main design axis. Unless very well behaved it is usual for the direct on axis to show anomalies that you'd rather were not there but that disappear off axis.

In the above graph the blue trace is the tweeter. As is usual it shows the rising response towards the lower end and issues in the top octave. I am sure that the degree to which the top octave misbehaves is related to the ND16 as much as anything else. It showed similar behaviour in the FST that the Aura NT1 did not.

Below is the ND16 in the FST, as you can see, the same top octave issues.



In the Satori they obviously change in severity with angle as they did in the FST. I am not sure if much can be done about this although I have toyed with the idea of mounting a small phase plug just in front of the dome. That's not easy to do mind you. In the FST it wasn't really audible as the optimum angle to listen to coaxials and wave guide designs is fairly far off axis anyway where the problem goes away/lessens.

That of course leaves the red trace which is the Satori. Nothing surprising there. Just that big peak which will make xover design a little harder than would be optimal for my targeted 3.5kHz xover point.

[sdl2112]

Hey Matt...I applaud your effort...true DIY I tried a small felt phase plug on my D3004/6640 waveguide and it helped. With the grill it was easy to try it. Not sure if it's the right size but maybe you could tack one of the Transducer Labs grills in place and try it.

http://meniscusaudio.com/tranducer-n26grille-p-1528.html

[5th element]

Hey Matt...I applaud your effort...true DIY I tried a small felt phase plug on my D3004/6640 waveguide and it helped. With the grill it was easy to try it. Not sure if it's the right size but maybe you could tack one of the Transducer Labs grills in place and try it.

That's actually a very good idea. I'm on the other side of the pond so ordering something from meniscus would be prohibitive but that's not the point. I've been thinking of ways of attaching something to the front of the tweeter short of using thin rods of metal. A fine mesh/weave metal or plastic gauze, cut into a circle, would be perfect, especially if it was stiff enough to retain its own shape.

[5th element]

So this is the part where you'd figure things would be easy. I have to say I was expecting the same, this is a coax right? But sadly not, the xover design was far more involved and complex than I figured it would be. 8O

The nice thing about coax drivers is that the tweeter and mid are roughly coincident and you'd figure this would equal a nice easy time with the crossover, or more importantly with the phase of the crossover. The trouble with this way of thinking is that crossover filters themselves have phase shifts of their own and you will only have a nice easy time with the xover if the drivers intrinsic phase + xover end up in phase alignment with one another.

At first I thought I may have had something wrong with the measurements so I took them again with no change. It's always worth doing this if you think something should be and it turns out not to be as the hardware can sometimes let us down. Rubbish in, rubbish out and we want to make sure we're not dealing with either.

My target xover for this design was a 4th order LWR acoustic at 3500Hz. This frequency would allow the ND16 to work at its best and provide a nice directivity match to the MW13P. Optimising for perfect LWR slopes was easy enough but with those acoustic slopes dialed in the phase integration was abysmal and to make things right required around 50us of delay on the woofer leg which is a problem. Interestingly though if the MW13P and ND16 weren't coaxial and positioned in the standard way, one above the other, phase integration would have been far better. With the mic on tweeter axis the woofer would then be along the hypotenuse of a right angled triangle formed with the mic, delaying the woofer and bringing into better phase alignment.

Hearing of people using asymmetrical crossovers is nothing new, the basic idea is to play around with different xover slopes/Qs/orders etc until thing start to work and then fine tune from there. This is necessary because you need to apply the correct amount of phase shift in the right places to get the drivers phase integrated. Sometimes you can have the drivers nicely integrated with phase but poorly with frequency response and its the designers job to iterate through different crossovers until both fall into place.

For a quick check of sanity I dialled in a 4th order LWR at 2.5kHz. This is what I had previously used with the ND16/FST/NT1 and with those drivers both phase and frequency response clicked right away. Nicely the same thing happened here with the ND16 and MW13P. This is what I was expecting as the driver configuration is pretty much exactly the same with relative offsets etc.

In this case lowering the woofer xover frequency down helped to bring the drivers into phase but obviously left a hole in the response around the xover frequency. LWR filters use a Q of 0.5, which is fairly shallow and one way to help fill in the hole is to increase the Q of either the tweeter or woofer leg. This helps to fill the hole in as you're increasing the output around the corner of the xover frequency. The trouble with going this route is that it also alters the phase response, so you can sometimes start filling in the hole, but ruining your driver integration. Also in this case increasing the Q of the tweeter filter was less than ideal as it would put more demand on the driver, which is already one of the least capable soft domes on the planet!

In this case I managed to keep the tweeter using a 3.5kHz LWR acoustic target whilst altering the woofer xover to bring them into both phase alignment and with a nice FR. For things to work properly though the woofer required a 4th order 'butterworth' filter, ie a filter with a Q of 0.707 and with a target of 2.7kHz. Quite a bit lower than the 3.5kHz target but the increased Q filled in the hole quite nicely.



Above is the summed frequency response with the tweeter and woofer curves shown in blue and red. As you can see there is deviation away from what could be the ideal target slope on the woofer at the top of the woofers response, where the Satori starts to breakup. This was actually necessary. I had equalised it out with notch filters etc and it affected the phase integration above the xover in a detrimental way. It's just great that in the Satori this breakup is benign, of course that's one of the reasons I chose it in the first place.

Here is the response with the polarity flipped.



It's not perfectly symmetrical but with passive xovers you cannot always have your cake and eat it as asymmetry is required within the xover for things to work properly. In non coaxial drivers this generally alters the shape of the primary lobe or 'point' towards the listener and it's always worth checking on the polar response to make sure that your listening axis is where you want it to be and that you've got enough wiggle room both above and below the main design axis for you within your room. With a coaxial this isn't a concern.



Above is the crossover and impedance. Simplicity is always a bonus with passive xovers. It keeps insertion losses and component tolerance variations lower but you never want to simplify to the detriment of performance. The goal is to still satisfy your design criteria with the least number of components. If you need a complex xover then so be it.

Quite frankly I was surprised that a third order electrical was all that the tweeter required. With the downward hump I was expecting to need a notch filter, thankfully this wasn't the case. On the woofer I tried all sorts of things, first order with bottomless notch and zobel, second order with notch, third order etc. Nothing worked half as well as the good old forth order electrical filter. Going third order simply didn't give the flexibility to shape the breakup region of the Satori, which was absolutely necessary to tune the driver integration above the xover frequency.

Then there's the input impedance. Never ignore the impedance. It is very easy to accidentally create a crossover that is a pain to drive. This is even more true when you're using 4 ohm average drivers than 8 so you need to be especially careful. With this design I tried using an Lpad as it gave more flexibility with shaping the tweeters response but all it did was make the loudspeaker harder to drive. So not only has the design ended up having a reasonably flat input impedance throughout the crossover region the impedance phase (not to be confused with the acoustic phase as they are two COMPLETELY separate things) is also flat and varies only between around -5 degrees to + 25 degrees . No amplifier is going to struggle to drive these. This is also one area where the insertion losses of the series inductors of the woofer can help to make the loudspeaker easier to drive. Yes increasing the DRC will affect the box tuning, so be aware of that, but we're not talking huge amounts of DCR just a little bit here and there and it will also lower the overall system sensitivity. Here I chose to deliberately use inductors of higher than maybe usual DCR for those reasons but also because the cabinets for these are positively tiny and great big hulking inductors were not going to fit, or if they did would reduce the cabinet volume by more than I would like. This was another reason for choosing the 4 ohm Satori rather than the 8 ohm. All things being equal the 4 ohm version will require caps of double the size and inductors of half the size.

For the astute you may notice that very little, if any, baffle step compensation has been applied to these. This has been done on purpose as the intended application is for them to be wall mounted. The system they will be integrated into also has a DSP within the DAC used (a PCM5142), so if any BSC is needed, or if the system changes location at a later date BSC can then be applied as is required.

[5th element]

And then there were crossover components.



Air cores and polypropylene caps from www.intertechnik.de

Thankfully none of the parts required excessively high values. If they had cored inductors and non polar lytics would have been of the order. Those polyprops from intertechnik are very nicely priced though and compact. Most polyprops are rated to 600 volts or so, which is completely overkill for most audio applications and annoys me quite frankly. The only thing that rating the components to this high a voltage does is increase cost and make them bulkier, two things that are unattractive in xover components.

These rectangular caps are only rated to 250volts, what a shame when my amplifier is only rated to 40. In truth though over rating caps can increase their performance (distortion) in applications where you are pushing their limitations, but in typical domestic situations that's unlikely to ever be encountered. The rectangular shape of them is also very nice for stacking caps and gluing the xover together. Which is exactly what I've done below.



Hard wired without any backing board to save on space and hot-melt glued together. The tweeter xovers are on the left and end at the white resistor, with the woofer ones being on the right. They look like they are glued together, but they are in fact completely separate.

[JonMarsh]

This has been coming together pretty nicely Matt! Looks like you're getting much closer to being able to play tunes through these! :T

[5th element]

Thanks Jon I actually have these up and running and have done for a couple of days. Trying to find time to fit in some decent listening is quite difficult as I am sure you can appreciate! I also want to perform some measurements on them.

Now adding to sdl2112's recommendation for the meniscus protector for the Transducer Lab tweeters I went to ebay and looked for some fine weave mesh that could be cut to shape. This happened to land me on things called 'pipe screens'. At first I thought these were for placing in line with water pipes to help stop grit and things, needless to say this was not the intended application :W

These are apparently for use as filters in various devices that one uses for smoking various substances :B

Who am I to judge, they are perfect for my application and readily available, who'd have thought :T

50 fine weave flat discs already cut to 30mm in diameter for about £2.

I am looking forwards to trying this out. If I can smooth out the top octave issues this will help a lot. Even still my quick listening sessions have been very pleasant. This is a polite sounding loudspeaker with excellent imaging and clarity.

[JonMarsh]

well, you know what they say, if the tool fits, use it! And it's certainly appropriately creative to repurpose devices and materials outside their originally intended area of application... sounds like you've got the right fit here! And the price is right, too!

[5th element]

Yeah the price is certainly right and the quantity is too. I can imagine a decent amount of experimentation might be necessary to correct for the top octave issues. That's if they are correctable.

Today I had some time so I moved these to the main system, dialled in some BSC and sat back. Wow is all I can say, these completely disappear. Moving the main speakers out of the way isn't possible really so these were placed on stands in front of them. The listening distance was rather close at just under 2 meters, but set up properly (with enough toe in for the speakers axis to cross in front of the listening position as waveguides/coax's should be) and the speakers just vanish. The main sound stage sitting comfortably behind the loudspeakers and extending beyond them to the sides. Central images always locked rock solid centre stage with no treble bleed through giving away where the tweeters could otherwise be. These are primary traits of what controlling the directivity in some way can do. The FST coax does this and these do too.

The treble balance does need some tweaking. I stayed on the side of caution and kept the balance tilted down as waveguide designs can quickly become too hot sounding, but not so with these. I'm fairly sure this is due to the lack of treble energy within the top octave, the ND16 was the same with the FST, a slightly dull characteristic that swapping to the ND1 solved.

As it stands I am quite happy with these and I'm not going to touch the treble until I've experimented with the pipe filters

[JonMarsh]

well, sounds like you should be feeling pretty satisfied based on the results of that listening test... :T

[5th element]

Yeah Jon indeed. I had listened to them in their intended system and enjoyed what I heard there, but it's not exactly the best position for fine tuning.

Over the years the design process has become a lot easier. The whole mulling things over inside the grey matter, figuring things out from a theoretical perspective, all the research etc. Then on top of that knowing how to make the right measurements, how to interpret them and then what to aim for in the simulator usually gets me pretty good results straight away. Even though I know it's going to sound good it always surprises me every time I hear a new design sing for the first time and the emotions one goes through, well I'm sure you're familiar. It's one of the reasons we do this! Technically speaking I didn't need a new pair of speakers, there was already a formidable pair in place, I just had to build something!

[JonMarsh]

Technically speaking I didn't need a new pair of speakers, there was already a formidable pair in place, I just had to build something!

This part is so, so familiar... :W :B

[5th element]

So sometimes you need to get the turntable out because nothing else will really do.



Before trying to fix the issue it's best to understand exactly how bad the issue is, so full on off axis measurements are needed.

First of all though is a measurement taken similar to the design axis for comparison to the simulation above.



Rather nice which ever way you look at it ignoring the top octave issues.

And here is a polar plot.



Followed by the more typical way of displaying the off axis.



Ignoring the top octave that looks pretty nice. Aside from the far off angles (70, 80 and 90 degrees) The response up to 10kHz is commendably flat and well controlled directivity-wise.

What's interesting though is that even though there are off axis dips there are also off axis peaks at roughly the same frequency but at different angles. This usually helps to even out top octave balance issues so maybe all I need to do is boost treble output to help put the balance into check. Still I'm going to experiment to see if I can improve things with phase plugs/phase stuff in combination with the pipe screens.

[---k---]

Very cool little project.

[5th element]

So I've played around with the filters with varying degrees of success but the biggest issue is trying to fix them in place. Because they are made of steel I had assumed that magnetic attraction would have been enough but this is far from the case.

Nevertheless if you place the speakers on their backs gravity does the job for you so I had a play like this but nothing really appeared to do the job I was hoping for. To a extent this isn't surprising. KEFs waveguides/phase shields are quite intricate and well beyond something a DIYer could hope to implement, especially given how close to the dome KEF place them.

Certainly placing objects in front of the dome changed things in the top octave but that's all really, usually for the worst and with some configurations looking about as good only different.

You can however address the tonal balance issues simply by tilting up the top octave. This is what I had assumed might be the case and doing so reintroduces the 'sparkle' or 'air' that was otherwise missing, so this is the route that I am taking as it's the easiest to accomplish.

[JonMarsh]

Makes sense, and clearly expedient!

[5th element]

Yes indeed, the thought of accidentally damaging something was a very real prospect too. If a fleck of fine weave steel mesh were to come loose and end up within the Satori's magnetic gap it'd be game over...

I suppose this is a nice lesson too and only backs of what Earl Geddes has been saying for a while, that the top octave really matters very little in the grand scheme of things. Sure it adds air and sparkle and if the balance is off it's unpleasant, but that's easy enough to solve with an up or downward tilt. Whereas good control/a uniform directivity below this does have other consequences on the sound, such as stereo perception and tonal balance.