This just in! There's a new "king" in the land of bang-for-the-buck
drivers. I recently received a shipment of Usher drivers and have been
evaluating their 8945W 7" woofer. I had figured that it would take me
longer to form a clear impression of this driver's performance, especially
after all the hooplah it raised on the Madisound discussion board. After
listening to these drivers for only a short time, however, has made it clear
-- they rock. You can skip the details and just go to
Euphase Audio and order up a few matched pairs, or you can read on for a
little bit more on why I'm so impressed.
Usher 8945W 7" woofer
Ok, so I received two matched pairs of the woofers and installed them in my
Project X enclosures, just for starters. I tested their TS parameters in
Clio and found that they were very close to the manufacturers specs. I
ran a few box simulations in Soundeasy and compared their frequency response
with the Vifa PL18's and found that the Usher woofers are about as close to a
dead-on "drop-in" replacement for the PL18's as I could have asked for.
Since I have 5 different PX prototype enclosures, I chose one with dual .82 ft3
(net) internal volume for each woofer and a tuning frequency of 42 hz.
Why not choose a lower tuning frequency? Simple -- you need to evaluate
mechanical power handling, transient response, distortion, and many factors
when choosing a low-frequency alignment. Just looking at the lowest
possible tuning frequency for a woofer in a vented enclosure might be fine if
you're only going to pump a watt or two into your enclosures, but I want to be
able to play it LOUD (or at least a reasonable volume at a 3m listening
position). In the enclosures I've chosen, these woofers will reach 115db
output before their mechanical excursion limit is reached. From what
I've heard so far, they'll do it with exceptional grace and clarity. A
nice, alternate enclosure for those building systems for small listening rooms
would be 1 ft3, tuned to 35hz -- you'll get very deep, clean bass at the
expense of power handling (you'll be limited to about 30W of power input,
before excusion is reached.... which isn't bad at all, but I'd prefer to have
at least 50W of power handling -- if you want to get more than that in the
<35hz region, you'll need a dedicated subwoofer).
Based on my initial measurements, there has no need to modify the PX crossover
for the Usher drivers. This, however, relies on the 450hz crossover
frequency for the woofers in the PX. I haven't studied whether the the
Usher could be substituted for the PL18's in the Eros without modification,
but I suspect that the results would be just fine -- the frequency response of
the Usher drivers and PL18's is very similar throughout the woofer's passband
in the Eros.
Finally, what's it sound like? Clean. Clear. Perhaps a bit
cold (like the Scan Speak 8545's that these drivers resemble). Some of
you might remember the discussion that erupted on the Madisound board over
these drivers -- some people claim they're "clones" of the Scan Speak 8545's,
due to the use of the same cone (which is, in fact, purchased from the same
supplier that makes the cones for the 8545's). Are they clones?
Who cares? They're not identical to the 8545's, and they're priced close
to the Vifa PL18's -- so, to me, the proper comparison for the Usher drivers
are the Peerless, Vifa, and Audax drivers in the $60-$80 price range, not the
terribly expensive Scan Speak drivers. And guess what? The Usher
drivers beat everything else in their price range hands down. The
midrange is excellent (not quite up to the Scan Speak, but is in a league of
its own in comparison to similarly priced drivers.) The bass is on a par
with any other 7" driver out there -- deep, well resolved, and clean.
How's this for a simple review: after listening to the Usher woofers, I
can say with complete confidence that there is simply no better 7" woofer on
the market for less than $100 and most people aren't likely to notice much
difference between the the Usher drivers and driver costing under $200 (the
8545's have slightly better midrange and the Seas 7" excel has moderately
better midrange at the expense of some bass response). The only driver I
have that beats the Usher, flat out, on a pure performance basis (i.e., cost
notwithstanding) is the Scan Speak 7" Revelator at $225/each.
After a few days of listening without the felt treatment on the baffle, I
decided to put it back on. The smoother response with the felt does
sound better. Clearly better. I then decided to measure the on and
off-axis response of the system with and without the felt, to see if I could
see what my ears heard. I did. As shown below, the off-axis
performance of the PX system without the baffle treatment shows peaking in the
2.5khz - 3khz region. There's also some peaking in the 6khz - 9khz area
in the 30 degree and higher off-axis. These effects are audible as some
inconsistencies in the soundstage (i.e., a loss of focus and breadth of
presentation, I think) as well as simple audible prominence of those frequency
regions.
With the felt treatment reapplied to the baffle, I measured again. The
response below shows nearly textbook perfect off-axis response.
Sonically, the speaker is better balanced and, most importantly, the
soundstage and image are broader and in better focus. Conclusion?
I'm sold on the use of felt. It allows me to keep the drivers centered
on the baffle and time-aligned, without having to employ extensive
equalization in the crossover to eliminate peaking in the response. The
felt also appears to function independent of the listening axis, making it a
more overall effective technique than trying to optimize off-axis response in
the crossover (although I used crossover frequencies that we appropriate for
the driver's natural abilities and baffle separation, to avoid any undue
problems). For the $20-$30 you might spend on felt for any project, it's
probably worth it. Also, for those of you without measuring and design
equipment, I can't imagine a better tweak than to apply felt to the baffle --
any baffle.
All that aside, I redesigned the highpass for the PX to improve the impedance
profile (you will see below that there was a 2.2'ish ohm minimum around 10khz
that I didn't like). The new topology represents an interesting tradeoff
that some will argue is irrelevant. I'm not sure.
As you see, I've replaced the series notch filter with a CR network (sometimes
called " high boost" or "shelving"network, as it will rolloff a
highpass through the capacitor until the cap's cutoff frequency is reached,
then it begins rolling off through both the resistor and capacitor, until
finally the load is purely resistive at 2 ohms. For those of you
wrestling with "LR" networks used for baffle step compensation, think of a CR
network as it's highpass counterpart. ) This allows me to obtain the
target Linkwitz-Riley 2nd order acoustic rolloff with the same precision as
the network shown in my earlier update, but (as shown in the graph) while
maintaining a flat impedance profile. The supposed tradeoff is in having
multiple caps in series with the driver. I'm not finding any negative
effects from this, although I'm using Solen capacitors for the highpass in the
prototype. They seem a bit less veiled than the Dayton caps with this
tweeter (maybe it takes a tweeter as revealing of this to begin considering
more expensive caps? Perhaps.)
The work continues.
I'm sure you've all heard the story of the boy who grows up, moves to a far
away place, and after many years meets the woman of his dreams who turns out
the be the girl he grew up living next door to. I felt that way this
week as I racked my brain in the hunt for the "perfect" tweeter for my
"project x" speakers (they don't have a name yet, hence the generic
designation). Anyway, I began eyeing those fancy new
Scan Speak phenolic ring radiators that cost a small fortune (about $450
each). They're used in the latest megabucks Krell speaks and sound quite
impressive (whether they're worth the price, or not, isn't even worth
discussing).
So, I got to thinking. Before plunking down an amount of money that
would put Project X out of reach for most DIY'ers (and would thoroughly piss
off my wife), I wondered what made them so much better than the
Vifa XT25. You remember the XT25, right? They're that
neat-looking "nipple" tweeter that was all the rage for about a week and a
half last year. Everyone was dying to get them to the DIY market and
when we got them, everyone raved over their exceptionally flat frequency
response and the fact that they didn't use ferrofluid gave us great promise of
highly detailed sound (many folks believe that the use of ferrofluid robs a
tweeter of the ability to reproduce very fine detail -- and the fact that some
phenomenal tweeters like the Scan Speak 9700 and Revelator don't use
ferrofluid only confirms this belief).
(The Vifa XT25)
Then, reality reared its ugly head and nobody could do much with the XT25.
Despite its' nice, flat response, it wasn't very cooperative and quickly fell
from grace. I felt somewhat responsible since I hounded Larry at PE to
talk to the folks at Vifa to see if these could be brought to the DIY market.
I even found examples of OEM's selling these via retail backchannels (e.g. as
"replacement" units -- except that people were just calling and buying them
and not replacing anything). Based on this, the folks at Vifa agreed to
sell the XT25 to any of the DIY parts suppliers in the U.S. that wanted to
carry them. Madisound and PE began carrying them, no one bought them,
I didn't do a damn thing with the pairs I had, and all that was left was a few
scattered reports that they weren't all that hot for their $60 price tag.
Personally, I found that everyone I showed them to thought they looked really
cool, so I decided to use them in a design at some point in the future.
The only problem was that I was somewhat prejudiced against them, based on
other people's reports. For that reason, I couldn't justify putting a
$60 with any inexpensive woofers (i.e., in the $30-$50 range) and figured it
would be a waste to match them up with much more expensive woofers.
That brings me back to project X. I still needed a tweeter with higher
sensitivity than my beloved Scan Speak 9500's, but wanted something with
high-resolution and that could be tamed to produce acceptable frequency
response. I tried many units, but ribbons were unacceptable due to their
poor vertical dispersion and sharp rolloff (project X uses very low-order
crossover slopes, so extended response is a must, and this is a WMTMW design,
so vertical dispersion in the tweeter is quite important). None of the
"high-efficiency" metal domes pleased me. I'm not against metal domes,
I've just never heard one I really liked. Among other speakers (Fostex,
Cabasse, etc.) the resolution just didn't impress me. Then I remembered
that the XT25 has excellent efficiency (it's rated at 91.5db, but it could
legitimately be rated at 92db, which is just barely what I was looking for).
I also realized that I had a few pairs sitting on the shelf among that
bazillion other drivers I own that will be collecting dust over the next few
years. Oddly enough, the faceplate on the XT25 allows it to fit
perfectly in the cutout for the 9500. GAH! Could it have been
meant to be? Could this be my "girl next door"? Only running some
tests would tell me for sure.
So, the next evening, despite being dead-dog tired after one of those days
that made me wish I had decided to be a professional mattress-tester for a
living, I tossed the XT25 into the PX (I'm tired of writing out "Project X")
enclosures and ran some measurements. Here's what the response looked
like at 1 meter.
"Hmm", I said. It does have nice, flat response. But the comments
made by a few folks I trust were still nagging away at me. Nevertheless,
I modeled the response in Calsod and optimized a 2nd-order acoustic (LR)
rolloff at 3khz, checked the impedance and summed frequency response, and
decided to give it a go. I built a pair of "add-on" boards to go onto
the PX prototype crossovers (shown in a recent update) and hooked these things
up and gave them a listen. I was impressed. I was still skeptical,
but impressed nonetheless. Over a few days, the sound of the units
opened up quite a bit (I guess the diaphragms were a little stiff after a year
on the shelf). I found the XT25's to have the detailed response I would
expect from a high-quality unit without ferrofluid. In fact, I couldn't
find anything wrong with them. My only remaining concern was the
imaging, as that was the aspect that a few folks had commented on in a
somewhat negative manner. Oddly, though, I found them to actually image
better than the 9500's in the PX prototypes. Now that it's been a few
days, I'm starting to wonder whether these are truly a 9500-beater. Of
course, everyone compares tweeters to the SS 9500 and claims to have the same
level of performance for less money -- it's the "in" thing to do when you like
a tweeter. Try it sometime, it's fun. Just take any old tweeter,
claim it's as good as the 9500 but only costs "1/Xth" as much and go around
the message boards telling the story over and over again. People will
believe you. I won't. Others will. I'm not going to claim
that the XT25 is a 9500 beater. To me, it doesn't have to be. It
appears to be a great tweeter for $60 and it's better suited to my current
application than the 9500. Is it better in a 2-way? Dunno.
Don't care. Next time I build a 2-way, I'll try them both and see which
one works better. Remember -- despite the flaming horsecrap you read on
the message boards, the "best" tweeter is the one that gives you the best
performance in your application, not what some dork in East Pancake, Iowa
tried and liked (and failed to mention that it was the only tweeter he tried
-- but why let little things like facts get in the way of making really
persuasive recommendations?)
(click to enlarge)
So, after deciding that the XT25 was going to be the way to go in PX, I
decided to do a little measuring and pass along a few more details. PX
isn't all about the drivers. No, no, no. The drivers are just a
means to an end. PX is about making a great set of speakers by
evaluating various crossover slopes, sensitivities, baffle configurations and
treatments, etc. No one detail should define a project and I'm
not one of those "designers" (say it with a snarl when you see it in quotes,
please) that slaps some drivers in a Madisound Woodstyle box, spits a
crossover out of some program, and calls it a speaker. I'll leave the
mundane aspect of providing people with computer optimized crossovers up to
other guys (not that some of these guys don't do it well, they're just not
"designers" in my books, they're technicians -- and some of them are lousy at
it because they use ill-conceived combinations of drivers and do uninspired
things with them. Sure, like the proverbial million monkeys typing on a
million typewriters for a million years, they're bound to produce something
that is passable, but their contribution to the art is purely trivial).
The PX, on the other hand, is intended to be a speaker that's been considered
from all-angles. The box, the baffle, the drivers, the crossover, the
impedance, and a lot of other factors that I won't even begin to discuss until
the final write-up.
As you see in the picture above, the felt-treatment on the baffle had been
removed for comparative listening tests and measurements. Here's what
the response looks like with and without the felt and an overlay of the to
graphs (presented in that respective order):
Quite obviously, the felt has a significant effect on the on-axis response.
Off-axis, however, the effects are less obvious (not shown). In
listening tests, I can clearly hear the difference, but I'm not sure whether
the increased dip at 3kh (without the felt) isn't actually preferable.
That's why these things take months. I'll go back and forth for a while
and, most likely, publish the final project showing both measurements and let
any of you who choose to build this project offer your own opinions.
Additionally, here's the system response (with felt treatment on the baffle)
showing the individual driver responses (note that this is only valid down to
about 350hz, which is the limit for 2m response in my testing room right now).
Note that the total system efficiency is now about 92.5db @ 1m/2.83v.
Finally, there must be a catch right? This is all too
good to be true, right? There is and there isn't. The impedance
profile of the current design is a solid, nominal 4 ohms, but it does have a
dip down to 2.25 ohms in the 10-15khz region. Is this a major problem?
No. Most high-quality (high-current) solid state amps will handle this
just fine. My 300b tube amps seem to just love these things (the
impedance phase angle is quite friendly to them). Cheap mass-market
receivers may not be as happy with this load, though (but you should get some
good quality amps, before investing of speakers of this caliber anyway -- snag
a Bryston 60wpc 2-channel amp off eBay for around $200 and you'll be happy as
a clam. Better yet, snag a pair of Bryston 4B SST's and run them bridged
in mono -- but that will cost around $5500 if you buy them new).
(impedance with and without crossover)
So what does the crossover look like? Glad you asked. Please don't
construe this to be the "final" version, but here's what the current prototype
looks like. Please don't try to go run out and build this, unless you
are very eager and willing to do some work on your own. At the present
time, I will not be posting enclosure construction information (although if
you want the internal volume and tuning information, I can do that). The
enclosures for PX have an internal enclosure for the mids and tweeter and a
time-aligned baffle. This is not for beginners. Also, I'm simply
not done yet. It may be a few months before I settle on the final
design.
Last thing for tonight, then it's sack time. A few folks have written
and asked about the DIY cat-5 speaker cables that I built and have mentioned
that I'm very pleased with. I will not be posting detailed information
on how to build these. Check out
Chris Venhaus' site and direct any questions you have to him. I'll
be happy to share my impressions of these cables and the various methods
suggested for making them (twisting vs. braiding, etc.) but simply do not have
the time or intimate knowledge of the project to assist other folks if they
want to build a set. Nevertheless, here are a few pictures of the pair I
use.
(diy cat5 speaker cables with WBT banana plugs,
shown with 14 awg flatwire cable as reference)
As always, stay tuned...
