Copyright 2008 © Troels Gravesen
inversely proportional to the moving mass
The number of (triode) valve amplifier producers today compared to two decades ago is phenomenal and there's a steadily increasing interest in high-efficiency speakers, thus this page on my experiences of high-efficiency speakers and some comments to what I've picked up from the web.
Why bother about 95-100 dB/1 watt
speakers when we have almost unlimited power
available from solid state or digital amps? Why do people
build two watt single ended triode valve amps with
subsequent need for extremely efficient speakers? Is it
pure nostalgia or do these systems have something to
offer conventional hifi does not?
There are not a whole lot of places to go and hear low-wattage SET amplification combined with good high-efficiency speakers. And even if we do find such a place, there's a fair chance we may leave, saying: "Is that it?" I've heard 50,000 USD triode systems that really didn't impress me. Silver trafos and everything to make the best of sounds and yet it didn't convince me at all. And I heard a system consisting of a cheap Chinese triode amp connected to pair of mock-up HES speakers that forever changed my ways of thinking hifi.
Art Dudley/Stereophile has these comments:
"The punch and the
drama" - is what we're always looking for
and sometimes SET + HES can deliver this kind of aural
nirvana. I've been sitting next to people experiencing
SET + HES for the first time and saying: "Holy crap,
this sounds like music!" I've had people in
listening to the OB7 and saying:
" - whauu...this sounds like a live
performance!" So, what do these people do after
returning home? They start yet another project based on
an 85 dB heavy alu cone driver. Either the prospect of
bringing two 100-200 litre cabs into their living room is
too scary or the last 30 years of hifi-tradition is so
engraved into our way of thinking hifi that we resign and
turn to well-known technology.
A lot of things can be found
on the web on high-efficiency speakers and what these
speakers can do - and not so much about what they cannot
Aiming at really, really
high-efficiency, there are prices to be paid
like in any extreme direction. 2-3 watts from a 2A3 SET
triode is not a whole lot of power. Some manufacturers
claim their 2 watt SET amps to perform well with 90 dB
(minimum) speakers. Yes, yes, if you never ever play at
anything more than - to my volume setting - well below
average listening levels. I know what it's like to sit
5th or 10th row in a club and listen to a jazz quartet.
It's damned loud and I want my ultimate speaker to be
able to recreate some of that drama. This doesn't mean I
drive my speakers to 95 dB average level - far from - but
I may play my speakers a bit louder compared to the
average listener judged from the visitors I have. Playing
the 15" Tannoys at realistic levels I remember one
visitor who almost fell off the chair from hearing a
snare drum being hit hard. I mean, this is what it's like
at a concert....
Looking for high efficiency
speakers on the web, we find the single-driver
aficionados. These people use wide range drivers with
very light-weight cones, reaching well up above 10 kHz,
making the need for tweeters questionable. Sometimes only
acoustic equalisation of the drivers are permitted.
Crossover components are definitely no-no! Purity and
simplicity at any cost?
If we increase driver size to
8" diameter, treble may be limited unless
we use a whizzer cone and the whizzer cone introduces
some new problems. I still have to see (and hear) a
decent response profile from a whizzer cone driver.
People play all sorts of games to tame the uneven
response from these drivers and the often seen rise in
amplitude at 3-8 kHz (e.g. Lowther).
There are a few exceptions
to this. The vintage Altec 755C may be worth a look. http://members.myactv.net/~je245/755survey.htm
I think the French PHY-PH drivers look interesting. High efficiency drivers with lightweight paper cones - and no whizzers! These drivers are aimed at an upper limit of some 8-10 kHz and an unusual tweeter is supplied; a small 70 milligrams paper cone driven by a piezo electric device. It doesn't say much below 10 kHz. I would love to do an open baffle 3-way from their 12" drivers, the 8" for mid and possibly this piezo "thing" for treble. But this would be something like 25-30,000 DKK or ~5-6,000 US $. Quite some money. Let's take a closer look at one of these PHY-PH drivers, the H 21 LB 15. I recently found a French website showing a magazine test on drivers in this category. But first the data from the PHY-PH website (left image) - and some LspCAD simulation/crosscalc. Even with a few data we can try calculating if what the company claims can be right. Quite often we find major deviations from claimed data, usually too high sensitivity, too low Fs and too low Qt.
Usually Re and membrane area is correct. Moving mass may be measured before assembling the driver. These are the easy parameters. Rms (the mechanical resistance of a driver's suspension (lossiness) in N·s/m) and Cms (compliance of the driver's suspension, in metres per newton (the reciprocal of its stiffness)) are some of the tricky ones as they are dependent on the signal applied during measurement. All drivers are highly un-linear devices and what we get is dependent on how we measure and there's no fixed rules for doing so. By manipulating cone mass, compliance, etc. in LspCAD we can usually by cross-calculation get a set of data that looks plausible. And I've often found these data to better correlate to what I measure for e.g. sensitivity. The claimed efficiency on the H 21 driver seems to be quite optimistic. 98 dB, although the website doesn't tell whether this is sensitivity or efficiency. Either way, I suggest being a bit sceptical when someone claims 98 dB from an 8" driver. At least don't expect much bass from such a driver.
Click image to the right: This is what a French magazine found. Here we read "94 dB" - and again we are not told what it is. 2.8 volts or 1 W? It does make a difference when we're dealing with a driver having an Re = 12.9 ohms. Fs is 10 Hz higher than claimed. I'm not surprised. This was what I found for the Supravox drivers as well. The frequency response doesn't quite look like what is shown at the PHY-PH website. Looks like we have a 5-6 dB peak at possibly 5-6 kHz. Hard to read the scan. The Qt is much higher than claimed. 0.83 rather than 0.59. Anyway, we have a driver that is born to be used on an open baffle. The interesting review can be found here (in French): http://petoindominique.fr/php/planhp.php
One important thing: What
we often hear reported from single driver set-ups, is the
revelation of minute details in the recording, things we
have never heard before. BUT, if we increase a narrow
frequency band some 3-4 dB - and from fullrange drivers
with whizzer cones we often experience major peaks in
frequency response, much more than 3-4 dB - we will
always hear things we never heard before.
One single driver set-up I think looks interesting is found at Decware: http://www.decware.com/newsite/mainmenu.htm. A tweaked Fostex driver and some interesting vented designs aimed at reducing standing waves inside the cabinet. But Steve Deckert, please show us high-resolution impedance curves from your designs.
An interesting comparison of single drivers can be found here: http://www.tnt-audio.com/shows/etf03_e.html; I wish I'd been there..
I recently had an opportunity
to test the all time full-range classics: Lowthers. Read
2-3-way high-efficiency systems
Next on the web we find people aiming at high efficiency but from multiple drivers and yes, we may even find crossover components. The Dick Olsher website is an example, http://www.blackdahlia.com/html/tip_56.html. Large 12-15" bass drivers in large vented cabs - or on open baffles - and a speedy 8" (Lowther, Fostex or even greencones) for mid + e.g. ribbon tweeters for highs. Gut feeling tells these are best driven by bi-amping and the bass section may like a sturdy solid state amp.
An interesting TQWT variant is the double Voigt horn, where all drivers are placed in conical horns (TQWT) of possible different lengths and firing into the same common middle section. The woofers can be connected in parallel or series depending on impedance of actual drivers.
The two bass drivers are fed through a huge coil, the mid no-whizzer-Lowther is having a series LCR to smooth the upper midrange/lower treble and the tweeter runs from a single cap. A fairly simple design, but finding the right drivers and doing the necessary driver tweaks has probably taken quite some time. I've heard a system like this fitted with a no-whizzer Lowther and a ribbon tweeter and it wasn't good. I brought my torture female vocals and it couldn't handle it. Serious distortion in upper midrange.
It's remarkable that so few
impedance plots can be found on high-efficiency
commercial designs. Actually I don't know of any. How
does a Lowther impedance profile look from an exponential
horn? What about the Living Voice d'Appolitos? How
about Beauhorns? And the Horning Hybrids? A few frequency
response graphs have shown up lately, but impedance?
Nobody apparently wants to speak about it - or
manufacturers don't want to show it. Unfortunately few
horn speakers show up in Stereophile to be dissected by
John Atkinson. Impedance plots show an awful lot about
how the drivers are performing in the actual cabinet, and
higher up in frequency a lot about intrinsic driver
performance, e.g. resonances in cones and suspensions.
No matter what we do, we have to balance
these three basic parameters: The trade-offs in
So, once upon a time -
If size is not a limiting factor, we can get really low bass and very high efficiency. If size is restricted, we may have to sacrifice efficiency - or low-end extension. If an F3 of 60 Hz is fine, we can get efficient speakers from quite small speakers. Somehow I don't like the idea of supplementing a HES speaker with a mega-powered inefficient subwoofer as seen from some of the Avangarde-Acoustic speakers. 12" woofers in small cabs usually have very low efficiency and need a lot of equalisation to get the bottom end going and they are slow compared to the frontloaded midhorn they have to play up to. I'd expect the same kind of trouble here as seen from hybrid electrostats.
I had a mail from a guy who had just heard a pair of vintage Pioneer speakers, actually the HPM-100. Below to the left:
The questions from the writer were
these: Why does this speaker provide such a
pleasant sound? Why is it so easy to listen to? What does
it have that most modern speakers do not have?
A 12" driver usually has a cone area of a little more than 500 cm^2. This is more than 3 times a ScanSpeak 18W/8531G00 driver! Imagine 3 x 6½ drivers in the same cab. A 12" based system can do things none of today's 6" skinny runts can do. It seems like most diy'ers start with a 6" 2-way making deep bass from 24-33 litre and an efficiency rarely exceeding 85 dB. We're all impressed from the bass it can make - but after some time we find out it really can't play very loud and we try another 6" driver or we make a 2½-way from 2 x 6" drivers. And we're never really getting anywhere. So, if you can handle a 60-70 litre cab, find a 12" driver, mount it on the side if the speaker must be skinny and use a 6" for midrange with a point of crossover around 250-350 Hz - and now hear what happens. With a 12" making 90-92 dB and a large middriver that no longer also has to handle the bass - we have a much better starting point in making an easy-to-listen-to speaker. Generally people have way to high expectations from a 24 litre floorstander. And they're often mislead from much too optimistic specs. Very often you can subtract 2-3 dB from claimed sensitivity and 5-15 Hz from claimed low-end extension.
The speakers of the Sixties
and into the Seventies often had high-efficiency
paper-cone drivers - where the British constructors were
heading towards heavily coated bextrene cones with an
all-time low sensitivity of 78-80 dB/2.8V and became
famous for their neutral sounding speakers, e.g. Rogers,
B&W, Spendor, etc.
By the way: Read interview with Joachim Gerhard (Audio Physic) on the secrets of old drivers: http://www.speakerbuilding.com/content/1039/ but don't think high mechanical Q is the only thing that makes a good driver. It must be balanced against a lot of other parameters.
2006 has brought the C17, Classic-3W and Poor Man's Strad, all producing
more the 90 dB/2.8V and the C17 has had another
round of tweaking, now running with a Monacor DT300
tweeter + waveguide. Read C17-II.
What I wanted was an 8" driver with paper cone, foam surround and sensitivity around 91-92 dB/2.8V and this CA21REX comes close - except for the foam. The CA21REX has a short voice coil compared to the new CA22RNX and it is not made for excessively large cone excursions and may benefit from some acoustic loading. An additional magnet will provide another single dB of sensitivity. More on this later.
Last year I ordered new foam
surrounds for a pair of Dynaudio 21W54 drivers
from Simply Speakers in the US. The
surrounds didn't fit the Dynaudios at all - they said it
didn't matter if the outer diameter was too small; my
goodness, of course it matters - but these foam surrounds
make a perfect match for the SEAS CA21REX drivers,
although they are rather thick and may provide a higher
Fs than intended. We'll see. The foam surrounds will
replace the rubber surrounds of the CA21REX. For midrange
I think surrounds made from foam sometimes perform better
than rubber, but it depends on a lot of things. Foam has
very low weight and for lightweight paper cones we may
not need a heavy load of rubber (mass) at the end of a
thin "sheet" of paper.
Surfing the web I also came across Audio Note and here we find an 8" drivers with a claimed "high sensitivity", presumably around 91-92 dB. The SPKR04 looks interesting. Large magnet, long fibre paper cone, foam surround and a coated dust cap. The chassis is a non-vented old SEAS version, the same as used for the CA21REX. Silver voice coil. There's only one obstacle to this driver, the price: 310 £/ea. Serious money.
Following the Audio Note path, I also came across the review of the Audionote AN-E Lexus Signature and this is interesting reading: http://www.stereophile.com/standloudspeakers/506an/index.html. First of all the claimed sensitivity of 98 dB (!) turns out to be 92.5 dB as from John Atkinson's measurements. Some difference I should say. I'm sure the answer from Audio Note would be that this speaker should be placed in corners, giving the additional low-end decibels, at least in the bass. But it won't raise the response in the midrange and treble by 6 dB, that's for sure. However, this clearly shows there's no limit to how much producers will overrate their products. The point of crossover is claimed to be around 2 kHz and it turns out to be close to 3 kHz. See graph below.
This speaker is also available in an all-silver version for the humble amount of 69,500 £. Three quarter of a million Danish Kroner. Well, well, well......
An 8" driver running to 3
kHz? Well, all depends. Next the frequency
response from this AN-E speaker is claimed to be 17 Hz to
23 kHz. Now that's something! Qvortrup suggests that the
woofer's theoretical maximum sensitivity is in the
neighborhood of 100dBand the overall system
sensitivity is stated as a very high 98dB....(read here)
"theoretical maximum sensitivity"..... what a
load of rubbish! 98 dB sensitivity and only 6 dB down at
17 Hz. Dear friends, seek no more. Go buy the AN-E and
all you need is 2-3 SET watts from a good 2A3 valve.
The Tonegen tweeter in
this kit seems a bit odd. Peaking severely at 10 kHz and
falling some 8-10 dB from 10 kHz to 20 kHz. It's really
not a problem getting a 92 dB tweeter these days that
will perform better. I don't say the sound of the Tonegen
tweeter isn't good, but....
The notch filter is interesting
as it works exactly at 1 kHz and the two 1 ohm resistors
in parallel may not only flatten the impedance notch but
also take the impedance close to zero at this frequency.
If we have to go higher in sensitivity, we need a much lighter membrane like seen on the Supravox 215GMF where we are down to around 12 grams moving mass. This makes some 96 dB sensitivity, but there are other problems when we do this. I've recently tried the 215GMF and I was very impressed by its sensitivity, but not so much from the serious cone break-ups at 1-2 kHz, right where it was least needed. This driver apparently needs a horn or a high-sensitivity middriver from around 500-800 Hz.
Anyway, this wasn't going to be about Supravox, rather a conventional 2-way from an 8" midbass and a dome tweeter. Back to the -
SEAS CA21REX + Dynaudio Esotar.
These drivers were set up in a 60 litres cabinet and a vent tuning of approx. 30 Hz was provided. The crossover is this:
Fairly simple crossover. Basically 1st order on the electrical side and drivers has same polarity as the tweeter being in front of the bass driver. Same as Audio Note. The CA21REX doesn't behave too well at 800-1000 Hz and a minor bump had to be removed by a notch filter.
By the way: The "DD" of the SEAS CA21REX refers to "two high conducting rings placed at each end of the motor coil. These act as eddy current brakes, firmly halting additional cone excursion as the rings enter the magnetic field". Thanks to Martin Colloms/High Performance Loudspeakers, page 79, 6th edition.
So what does it sound like? Well, better than expected I have to say. The Esotar is an excellent tweeter but it's on a tough job running from a single 6.8 uF capacitor. Placing a finger on the dome it can easily be felt that it's moving a lot due to the shallow slope of the crossover. What struck me the most when I first connected the speakers was how lean it was. The bass goes really deep from the 60 litres cab volume and you surely don't miss any low-end information. The bass is fast and transient and maybe lacking in some upper bass/lower midrange warmth. This cabinet is reasonably braced and I guess the severe resonances of the low-damped AN-E plywood box may add to the overall sound and possibly give a warmer presentation. The midrange is far better than I would have thought and dispersion doesn't seem to be a major problem and I'm sure the tweeter's 1st order filter helps here. Would I listen only to my latest vinyl - the Anthony Wilson trio/our gang - and similar stuff - then I could easily live happy ever after with this speaker. Transparency is fine and listening to a drum kit is very enjoyable. It is physically a big speaker, and it sounds like a big speaker.
The CA21REX is well above 90 dB for 2.8 V, but had we used the 1 kHz rule it would have been around 94 dB. Actually the PD from SEAS says 93 dB, which is not very useful. The Esotar performs exceptionally well but is does have a very early roll-off compared to many other 1" domes. Really designed for 1st order filters. And I'm sure the Dynaudio people would tell us not to worry, this tweeter can take a lot of heat.
Changing the rubber surrounds of the CA21REX to foam surround.
Changing the surround was easier than anticipated. The rubber can usually be peeled off from the metal chassis and with a scalpel the rubber was cut close to the cone. Slicing off soft rubber from a soft paper cone with a scalpel is not funny at all. However, in this case it was possible to also peel off the rubber remains from the cone - very slowly and it only ripped off a very small amount of paper pulp material. So all in all, a nice job could be done.
New cabs for this experiment:
A pair of test cabs were made resembling the AN drawing above. Approx. 55 litres of volume and I chose MDF for front and rear panel, chipboard for top, bottom and sides. Mixing various materials may not be such a bad idea. The modest bracing of the AN cab was followed closely and also modest damping material was added. I admit this cab appear less resonant than I had expected. No bitumen pads were used.
TS data CA21REX-foam/magnets:
The Fs really went up
from this foam surround. I had hoped for some 40 Hz, but
47 Hz is a bit on the high side. From the Supravox
drivers I learned not to worry too much about the Fs as
these drivers have an Fs around 65 Hz! Yet, they provide
excellent bass from 80 litre cabinets.
The modified SEAS CA21REX
to the left, 1 metre distance and 2.8 V. Red = SPL, blue
= min. phase. I now call the driver CA21REX-FDM. Foam and
double magnet. Now this SPL response is a surprise. The
bass response is very dependant on from where you merge
the nearfield response; here at 400 Hz, so don't put too
much into the bass response. It very much looks like this
from 400 Hz and down, but the absolute level may vary
relative to the mid and upper midrange.
To the right the SPL response from the DT300 tweeter with waveguide. Placing the DT300 + waveguide on a relatively wide baffle further increases the response in the 1000-1500 Hz range. Compare to the narrow baffle of the C17-II. The DT300 + waveguide is indeed very suitable for speakers with an average sensitivity in the 95 dB/2.8V range.
To the left both drivers on the same graph. To the right, the impedance and phase profiles of the drivers. With a 72 (ID) x 50 mm vent, a box tuning of 40 Hz is achieved. It seems possible to achieve relatively high overall impedance and hopefully an easy load on low-wattage valve amps.
My initial approach to the crossover was a first order topology. Like the C17-II. Why not? Same tweeter and a big 8" bass rather than two smaller 6" drivers. Modelling seemed easy but soon after starting hardwiring the crossover everything went wrong. A coil for the bass and a cap to the tweeter and LCRs to control the roll-off at point of crossover. No problem at all, BUT the phase tracking was horrible - and it really didn't sound too good either. Same or reverse polarity of the tweeter gave almost the same results despite nice performance in the LspCAD. Now, in LspCAD you have to correct the relative distance from microphone to drivers to render the phase tracking right and this can sometimes be trouble. Where the C17 drivers and the DT300 seem to emerge from almost the same plane, this was definitely not the case with the CA21REX-FDM and the DT300. The step response displays the bass driver more delayed compared to the tweeter. Actually the tweeter had to de added 10 mm distance to the mic, with the bass at zero. Hmm....
To control the peak at 2600 Hz from the bass driver, a notch filter is needed with a first order filter and it performed reasonably well. The DT300 was no problem, but I guess the steep roll-off from the large 8" drivers was what caused the trouble and the C17 approach simply didn't work. To cut a long story short, a 2nd order filter was tried and this immediately made everything fall into place. Flat response, good phase tracking and the nulling effect from reverse tweeter polarity was deep as should be.
So, speakers connected to the amp, volume tuned up one third, CD in the drawer, a drummer hitting his snare drum very hard at the intro - and I almost got a heart attack. 95 dB sensitivity is immediately heard - and felt. Talk about a fast speaker. Some CDs later this tuning turned out to be just too much. Basically flat SPL tuned up to around 95 dB, the bass a little retracted but very fast and articulate. But voicing this speakers for at 95 dB sensitivity produced too much presence and "right-in-face". Details in spades, but vocal and sax to energetic. Time to revise the crossover once more.
The V5 crossover is a bit annoying as it came very close to the AN-E crossover and I didn't want to make a clone of this speaker, but after very little modelling and very much cut'n try this emerged:
A simple 2nd order filter and same polarity of drivers.
Abbreviations used: foam - driver fitted with foam surround. 2 mgn - driver added additional magnet. WG = wave guide. NO = no magnetic oil (oil removed).
CA21REX + DT300-WG-NO, version 5 - as seen above
- has been playing for some time in our living room and
I've gotten used to the sound. So, time to evaluate.
It's difficult to review
your own speakers and in this case I feel like what I
read in magazine reviews sometimes. Scepticism at the
beginning, but after some time you get accustomed to the
sound and start enjoying what this speaker is good at.
Because very few speakers are good at everything.
The CA21REX/DT300WG is not an idiosyncratic speaker. It's easy to listen to; gives major insight into the music, transparent midrange, undistorted treble and the overall transient attack is simply great. It's a revealing speaker and may leave a few poor recordings out, but not to a degree where I would call it idiosyncratic. I've pulled out - what I thought was the most difficult recordings - and have been surprised to find them performing well where a few recordings I thought wouldn't cause trouble, wasn't so good.
few weeks later:
Looking at extinct Snell E/II and /III speakers, I came to think about the old Vifa C20, 8" driver. The C20 is found in numerous versions like the C17. The C20 has a lightweight paper cone, a narrow foam surround and the "H" provides a 90 mm magnet for a 1" voice coil, probably giving decent sensitivity. Googling the C20WH, I found a single shop where the C20 was available, probably some old stock and this is what I found in seller's storage.
Really nice looking C20 drivers. Fresh foam surrounds and untinted cones. An additional 90 magnet was added to increase sensitivity. When you add an extra magnet it has be to pushed away from the speaker's magnet. If the extra magnet is sucked in, it will decrease magnetic flux in the gap. Strangely enough, if it's very easy to get the extra magnet in place, it usually only means a small increase in sensitivity. If you really have take special measures to keep the magnet in place, there's usually an significant increase in sensitity. I keep the extra magnet in place by tape after adding a small amount of silicon glue between the two magnets, and add a clamp overnight.
The usual frequency and phase measurements
were done at 1 metre distance to front panel and files
were imported into LspCAD and modelling could start.
turned out to quite a long article and without
summarising all key points above, it all ended at the
point where I had my own high-efficiency, extended range
driver made by SEAS/Norway. This driver turned out very
suitable for a wide range of applications and the results
can be seen here: