Poor Man's Strad    SEAS CA22RNX, MCA15RCY, T27TFFC
Copyright 2006 © Troels Gravesen

CROSSOVER     CABINET     MEASUREMENTS     SIMULATION         

A box of new paper coned drivers from SEAS. The CA22RNX, MCA15RCY and MCA12RC. And soft domes: 27TDC and 27TFFC. Classic paper cones in new spider chassis. A box of opportunities - and many hours of work. It takes five minutes to order a box of drivers and you can spend months and months on learning the strengths and weaknesses of each driver, trying to find the optimum points of crossover and the optimum crossover slopes for producing the best blends of sound that hopefully will integrate the drivers into a coherent and pleasant presentation. The number of options seems infinite.

This speakers should have an Indian name and be called "Comes With a Room" (- like "Dances With Wolves") because this is what wide baffled speakers do. They kind of bring in their own room. There's virtually no edge diffraction due to the wide baffle and the way these speakers recreate music in a room is special and to put it short: I like it! They can play strings, woodwinds and rock'n roll like few speakers of this size and from a 91 dB/2.8V sensitivity they do well with small amps. Read more about wide baffles and what they can do: http://www.troelsgravesen.dk/Acapella_WB.htm

My wife had been attending a three-day course and when she came home and found the Poor Man's Strads in the livingroom she said: "These speakers aren't quite as big as the other wide (baffled) speakers, right?". "Indeed so, these are much, much smaller, actually almost one third the volume of the others" - I replied. So WAF seems relatively high and what helps is the very slim appearance when seen from the side. I have to say that my wife is a darling when it comes to speakers. For many months each year she lives with horrible looking MDF cabs in the livingroom - on their way to the workshop for further fine-tuning.

Builders' response: http://www.troelsgravesen.dk/Builder's_response.htm

Before ordering the drivers I had one particular project in mind: A classical 3-way from an 8" bass driver, a 5" midrange driver and a 1" soft dome. I wanted a 90 dB/2.8V sensitivity, an easy load on the amplifier and I wanted to measure how they would perform on a reasonably sized wide baffle in order to make a Poor Man's Strad.
The midrange is the most important driver in this trio. It has to cover the entire 300 - 3000 Hz range and by the way, what makes a true midrange driver? I've searched the literature for what makes true midrange drivers and not much can be found. However, in the range of PA drivers we find dedicated midrange drivers with high Fs, sometimes with Fs >100 Hz and sometimes with a flat surround as the membrane really doesn't have to move much. Few of these drivers have found their way into hifi and the only thing that differentiates the MCA15RCY from the CA15LRY is a slightly higher Fs for the MCA and most importantly, a short voice coil. You can make a short voice coil as the membrane doesn't have to move much and you get increased sensitivity as more of the voice coil will be in the magnet gap. The MCA15 has a 9.4 mm voice coils (height) compared to 16 mm for the CA15RLY. So the membrane in the MCA15 can only move +/- 1.7 mm and still be fully in the magnet gap.
A classical three-way from CA22 + MCA12 + 27TFFC has been presented,
3-Way Classic, but here were going "wide baffle" to see and hear what can be achieved.

This project has turned out much better than expected and listening to a wide range of recordings, CDs and vinyl, I actually wonder if I should quit speaker building here. There's something appealing about simplicity and this is an easy to build speaker. From all the experiences of the past constructions this one very much fulfils a wide range of demands: High sensitivity, an easy load on the amplifier, a flat frequency response, good phase integration between drivers, deep solid bass and a speedy, transient and non-aggressive midrange. What more can you ask? Well, you may get a slightly more neutral presentation from magnesium or ceramic drivers, but most likely at the expense of sensitivity. However, I would like to try a Accuton 5" driver for midrange, but some other time. Here we go the good old paper way.
Later the EXCEL W22 + W15 will be tried as these drivers fit in nicely with the driver rebating.


Vinyl finish for the sake of photoshoot. Easy way of deciding final cabinet design.
The base plates are certainly not going to look like these.

Simulation
Back to top

The CA22RNX, MCA15RCY and 27TFFC were set up in a small test cab and MDF side panels were added by Gaffa tape to form an approx. 50 cm (W) x 105 cm (H) curved front panel.
Frequency response and impedance files were created to start modelling in LspCAD to hopefully produce a nice summed frequency response with optimum phase tracking. I also wanted to make the crossover from true 2nd order LR filter slopes to get the best possible impulse response and this was what I got:


Très simple, n'est ce pas?
(Thanks, Rodolphe, for correcting my French! Not even this I got right... Hmm.. 35 years since high school)
Bass and mid are true 2nd order and the tweeter needs some more components to produce proper phase tracking
Red are true resistors. "Empty" resistors = DC resistance of coils.


Frequency response and phase tracking of drivers. Points of crossover are 340 Hz and 3800 Hz.
Possibly 3800 Hz for the mid and tweeter is too high, only the ear can tell later.
It doesn't get much better than this, but it's no guarantee the sonic performance will be optimal too.

Frequency response of all drivers on wide curved baffle without crossover. Green = bass, blue = mid, red = tweeter. Something tells me we really are going to get a sensitivity of at least 90 dB/2.8V and the midrange seems to perform remarkably well. Tweeter is basically flat from 1500 Hz to 20 kHz and has the same little wrinkle around 17-18 kHz as seen from the T25 tweeter. Not surprisingly as they share the same type of diaphragm and lack of polepiece damping. The CA22 has some cone break-ups around 3, 4 and 5 kHz and I suspect a somewhat resonant dust cap to be responsible for this. Maybe I should "slice" the dust cap and see what happens.
Let's take a look a the midrange alone:

MCA15RCY: Not an everyday sight. Flat from 200 Hz to 5 kHz and modest wrinkles from 5-10 kHz. Sensitivity = 90 dB/2.8V. This driver is flat over almost 5 octaves! Too good to be true.
Playing music on the midrange alone didn't sound bad at all. The sound appears neutral with no aggressive tendencies.


The Box
Back to top

From these very promising initial experiments a test cabinet was the only thing left to do. But before this we need to think about the proper volume of the cabinet for the bass driver. And we're going to spend more time on this theme than usual.
Initial modelling of a vented box suggest some 40-45 litres to render a reasonably bass extension, but first the CA22RNX TS-data:
(TS data was measured with the CLIO measuring system with the driver driven directly from the board at max. +12dB output)

It's always interesting to see how your TS data measurements correlate with the factory data. The "40Hz/4h" means second measurements after the drivers had been given some heavy 40Hz massage for 4 hours. (Several weeks of use didn't change this figure one single Hz).
Surprisingly the Qms and Qe are in good harmony, but the Fs and Vas.... Actually I don't mind the significantly lower Vas and higher Fs as we can use a smaller cabinet. That's something. I really wonder if the suspension will soften so much it will reduce the Fs by 10 Hz. And subsequently raise the Vas to almost twice the current value. The TS data were done by the added volume method, using a 29.4 litre closed enclosure to measure the increase in Fs.
The 4 hour data were inserted into the LspCAD box function and the usual range of box volumes and vent tunings for SBB4, QB3 and SC4 can be viewed and some 34-36 litres volume is calculated and with the SBB4 alignment a 30.5 Hz vent tuning is suggested with a -3 dB point of 45 Hz is achieved. A little higher than anticipated. From practical experience, driver with a Qt between 0.3 and 0.4 performs the best in accordance with all the nice math. And I would even narrow this down to Qt = 0.33 - 0.38. Drivers with a Qt = 0.25 or Qt = 0.50 really do not fit in.

Doing the same thing with the SEAS data a SBB4 alignment prescribes a 72 litres box and a vent tuning of 29 Hz producing a -3 dB at 32 Hz. Now this is what makes you give up box calculations. I mean, we would probably never put this 8" driver in a 72 litre box. Let's look at what LspCAD predicts:


The calculated response from the SEAS data in a 72 litres SBB4 vented box, Fport = 29 Hz.
Predicted SPL is pretty much as expected: close to 90 dB.


The calculated impulse response from the SEAS data in a 72 litres SBB4 vented box, Fport = 29 Hz


The calculated response from my data in a 36 litres SBB4 vented box, Fport = 31 Hz


The calculated impulse response from my data in a 36 litres SBB4 vented box, Fport = 31 Hz


Comparing the impulse response from the two scenarios. Very little difference. But we needn't worry,
because the filter will ruin these nice graphs anyway.

So how about a 40 litre cabinet with a vent tuning of 35 Hz. Gut feeling:


40 litres volume and Fport = 35 Hz. Minus 3 dB at ~42 Hz.


Next we apply the filter: 4.7 mH in series and 68 uF to ground. Hmm... We get a 2 dB peak at 125 Hz.


And as promised: The spoiled impulse response. And this from a 2nd order LR filter.

And here's the same thing + in-room response from planned placement of bass driver. Minus 3 dB at 41.5 Hz ain't bad at all and there's bass - and then there's bass. All depends. Due to the crossover we lift the response above 100 Hz and room gain makes 1-2 dB below 100-125 Hz and we should get an overall sensitivity around 91-92 dB. We'll see if the mid-driver needs attenuation or not.
The conclusion to all this is that I'll make 40 litres test cabs and I can reduce volume to get the best performance based on measurements and listening tests.

Test cabinets:
Construction pictures




After finishing the test cabs the actual volumes turned out to be 38 litres. A little less than intended, but the damping material will make the virtual volume somewhat bigger.

What's seen above are the only drawings available.


Two vents towards the floor will be placed on each of the “wings” and I'll use PVC tubing of 4.7 cm internal diameter. From a ~38 litres cab and a vent tuning of 38 Hz, the vents should be 80 mm long. I cut them 120 mm to enable some fine-tuning after driver installation.

The tilt of the box is 5 degrees. Running the test cabs it's apparent they need more bracing. In particular the top panel and the panel above the tweeter have resonances. But these are only test cabs.

Thanks to Tim Wilde, here's a link to a Dutch website with lots of pics from constructing a similar type cabinet:
http://www.htforum.nl/yabbse/index.php?topic=41295.50


Mid cab vent

A special section on the mid cab. The information has been here all the time but for some reason I've had the question 100 times about the mid cab vent.

The mid cabs are 4.3 litres net volume and the vent is a 40 x 70 mm PVC tube with an ID of 37 mm. Actually this vent doesn't do much in the traditional way of a vented box. I made it because the mid shouldn't be in a closed box.
I don't dig closed boxes for larger midrange drivers! It kills the midrange. Actually I have a small amount of wool material stuffed in the vent, so its function is more like an acoustic vent. So take your pick.

And the mid cab vent is on the rear panel (I've had that one too :-))


The final PMS version 5 crossover
Back to top


Crossover version 5 for the 27TFFC version. Still - "très simple".
(Actually even more simple than first simulation)

Coil L2031 supplied with the kit has DCR = 2.2 ohm, so you do not need to add any resistor (R2031)


Crossovers have been optimised for various tweeters. Please don't ask which is best.

List of crossover components from: Jantzen Audio, Denmark


The list includes components for all three versions.
Kit including all drivers can be ordered at
contact@jantzen-audio.com and
the extra components for all versions comes at no extra cost (one small coil and 2 small caps).

Crossover construction

Crossover for 9700 version shown. Three additional wires are needed to keep the coils as wide apart as possible. Components connected to ground are soldered to the same spot in the middle of the crossover = star earthing. A short piece of thick copper wire can be used for this purpose. All components are glued to the board by Superfix glue. This is a white sticky mess, but excellent for having all components solidly stuck to the board. If you need to remove the components again, they can be cut off (sliced of) by a thin-bladed hobby knife (Stanley type). Superfix takes a couple of hours to settle and will be firm overnight, much like silicon glue.


Crossover for 9700 version shown.

More layouts:
- sorry for all these different layouts, but making the complete crossover on one board
may pose problems in terms of in-cab placement, so here we go again with two sections:


Note: The orientation of components in the LCR circuit doesn't matter.
It can be LCR, CRL, RCL, etc. Doesn't matter (I have this question all the time).

Despite the crossover being fairly simple, this part of the work was considerably more trouble than expected. Maybe because it’s simple... The bas-mid part was quickly done and it ended up being an almost 12/12 dB giving the best phase and sonic performance. Actually the bas-mid is very close to a true L/R 12 dB crossover for 6 ohms drivers at 375 Hz (35 uF and 5 mH), although I reduce the Q of the mid 4.7 mH coil by a 2R2 resistor.
The mid low-pass section was trouble. Despite being easy to render a smooth 12 dB roll-off at around 3800 Hz, this didn’t fit with 2nd or 3rd order filters for the tweeter. The tweeter's high-pass section ended up being a 4th order to make things match. And the trouble was the 3.8-4.0 kHz point of crossover.
So, time to listen...

This first set-up was including the 27TDC tweeter and despite an apparent high level of transparency, there was something that didn’t sound good. The 27TDC with its rigid, coated fabric diaphragm needs the same cure as the T25 and a felt ring was added to the centre polepiece and the terrible polyester foam plug in the vent was replaced by some wool felt material. This helped somewhat, but before doing more surgery on the poor dome, the 27TDC was replaced by the 27TFFC and things started going in the right direction. Everything started sounding really good except for some vocals. Vocals are the litmus test on any speaker. So back to the LspCAD and some dozens simulations later it was finally decided to take the point of crossover between the mid and tweeter down to 3 kHz.
Despite the ability of the MCA15 to run up to 6 kHz without trouble, it shan’t. 4 kHz is very much the middle treble area and despite a nice performance from measurements, the sound wasn't right. Getting the point of crossover down to 3 kHz was the cure and the phase tracking between mid and tweeter all of a sudden fell into place too.


Response based on simulation (27TFFC before modification).


Measurements
Back to top

Left: Actual measurement of frequency response at 1.0 m distance at MT height and at 2.8V input. Sensitivity in the midrange is around 90-91 dB. Not bad at a amplifier. Watch that volume knob, these speakers can play really loud.
Right: Frequency response at 1 metre distance merged at 350 Hz with nearfield measurement of bass response. Don't put too much into the response between 100 and 350 Hz. What's interesting here is to see the bass roll-off and the CA22 goes down to 60 Hz before the roll-off begins. The vent tuning appear to be around 34-35 Hz. This with the two 4.6 (ID) x 12 cm vents. These vents will stay in place.

Left: Horizontal dispersion: 0, 10, 20 and 30 deg. (Red, blue, green and purple). This ensures a wide listening field with an even sound dispersion.
Right: Vertical dispersion from 1 metre distance. Starting position = between M and T = red graph. Blue = - 10 cm, green = -20 cm. Yellow = + 10 cm, purple = + 20 cm. This is better than expected from a 2nd order filter.

Left: Impedance and phase from V5 crossover. Note the flat phase performance. This is an easy load on the amplifier. The phase at 70 Hz is -49 deg and the impedance here is 10 ohms. This is seen worse and from practical experience this speaker can be played loud - very loud - without my workshop solid state amp - a Rotel RB981 - running even moderately hot.
My friend Darryl in Australia kept raving about the mid and treble from the RB981 power amp and I managed to find one on eBay. This amp has some of the best mid and treble I have heard. The bass is a bit mushy, but is doing excellent on this very transient, paper-coned woofer. A good match.
Right: Impedance again: A true 8 ohms impedance in marketing terms. Many "8 ohms" speakers run down to below 4 ohms in certain areas.

Left: Step response showing negative polarity of tweeter + mid and positive polarity of the bass. Right: Individual and summed response of drivers from the crossover.

Left: Mid and tweeter with opposite polarity. Right: Red = bass and mid with same polarity. Yellow is opposite polarity.


CSD for what it may be worth.

27TFFC tweeter modifications:
This is quite simple. Remove the tweeter faceplate and gently lift it up. The diaphragm is attached to the faceplate. Remove the polyester foam plug in the vented pole piece and add some loose wool felt material instead. Cut a ring of 1-1.5 mm felt material and glue it to the pole piece.
Read here:
http://www.troelsgravesen.dk/AcapellaSEas.htm
And here:
http://www.speakerdesign.net/tweeter_tweaks/pole_piece_felt/tweeter_tweak_ppf_725c003.html