Copyright 2010 Troels Gravesen
Greencone Sound The
Modelling Resonanz

SABA PERMADYN 19-200-5298-U8, 5 ohms and tweeter SABA 1670-DU-20 (or -15), 5 ohms

These SABA greencones have true cult status in central parts of Europe and they're not hard to find. They must have been made in 6 or even 7 digit numbers during all of the Fifties and well into the Sixties by SABA, Germany. Other (German) companies made similar products, but the SABA greencones were what Philips 9710 was to Holland, Supravox 215RTF to France and NNNN (?) for the Britts. What is the British equivalent of these 8" extended range drivers? I guess we have to look for Goodmans and Wharfedale speakers to find out. In Denmark we had Peerless, e.g. Orchestra, although I recall my parents' radio actually had a 6" Goodmans unit.

The SABA greencone has a one-piece molded cone including a non-coated corrugated surround. A rugged small dust cap is glued to the cone to protect the voice coil and a light-weight stamped metal chassis holds it all together. A tiny alnico magnet and stamped, cost-efficient metal parts provides the unit a whopping 94-95 dB/2.8V sensitivity. Despite meant for valve amps, Re is set for 5 ohms, but no anaemic 2A3 SET amp will suffer serious trouble from being connected to these speakers - or at least so I thought...

Forget about traditional hifi. Greencones will never make much below ~60 Hz and the tiny voice coil will never generate serious amounts of back EMF, nor require serious amounts of current. Having Fs = 75 Hz we need big cabs to generate something below 100 Hz and the "resonanz" box is most likely the best thing we can do - or use them as a dedicated midrange driver in open baffle systems. The use of resonant boxes is controversial. I know there are people who believe in speakers being "instruments" and I've even had a mail stating a "Steinway isn't made from MDF, right?" Well, I won't argue that these people got it all wrong, because people can do whatever they want and if quality time is a resonant box, who am I to say they're wrong. Don't be surprised if you some day find a thin plywood box for my greencones on this page.

What we're after here is the midrange and minimum phase shift in lower treble range. Generally these greencones display smooth midrange except for the phase shift around 900 Hz, a phase shift much more tolerable than the serious phase shift around 2.5 kHz found in Lowther drivers. Next these drivers have a smooth upper roll-off profile. I wish every midbass driver had a roll-off like this. It makes crossover work so much easier, and not having any serious trouble in all of the important treble range certainly does please the ear. Only the Vifa C20WH can do something similar, although not as extended as the greencones.
What is a general limitation of greencones - and the few other drivers of similar construction - is limited power handling. As long as we don't play too loud, things go well. Cranking up volume, things fall apart, cones break up and we get the "midrange shout", i.e. serious distortion. Using the unit as an extended range driver also calls for beaming of upper mid and loer treble with subsequent loss of detail and perceived level of transparency.

Last, but not least, we're into serious nostalgia when dealing with greencones. We may love the "good old days" and nostalgia is definitely part of the fun, but before rushing to German eBay, we must think about what these drivers can offer - and seriously think about what they cannot. The Supravox 215RTF is a much better driver - but cannot be found for 50 EUR/pair.

Links to greencone sites: go to 26 speaker projects

The drivers
(back to top)

Left: SABA PERMADYN 19-200-5298-U8, 5 ohms. Same unit ID regardless of magnet system.
One of the units to the right was damaged due to poor packing, but could be restored to have the voice coil moving again - and a candidate for experiments.
The dust cap leaves a small space between cone and pole piece acting as a pneumatic break, and I wouldn't be surprised if this was intentional. It
might prevent the cone from excessive movement - and remember - these drivers were meant for being placed in a "radio" = an open baffle, not having the damping
properties of a closed or vented box. As can be seen from the impedance measurements below, removing the dust cap has a significant impact on Qm.

Left: Tweeter SABA 1670-DU-20, 5 ohms. Right: SABA 1670-DU-15, 5 ohms (both bakelite chassis).
These tweeters are some strange creatures having the dust cap fixed to pole piece by a brass screw, thus
a rugged mechanical high-pass filter. The impact on frequency response from this is a broad peak centered around
12-14 kHz, 15 dB above basic level - and some 10 dB above the average response from the 8" driver.
Those who choose to run their tweeter from a single cap alone, usually 2.2 uF, certainly get some upper treble level!

Greencones cabs found on the web
- found from links above.

Resonanz boxen. Enclosures made from thin plywood with minimum bracing and damping.

SABA PERMADYN Tweeter 1670-DU-20 (or -15) measurements
(back to top)

Left: Impedance of 19-200-5298-U8 driver w/wo dust cap (red = wo). Right: Impedance of two 1670-DU-20 (or -15) tweeter units.

The tweeter: SPL/2.8V, 1 meter. Red/green = two units, blue = 4.7 uF in series with tweeter.

SABA PERMADYN 19-200-5298-U8 measurements on IEC baffle

In total I have four Permadyn 19-200-5298-U8 drivers, two in mint condition and two so-and-so, where one was damaged
during transport and had a ripped cone suspension, which was repaired with some flexible glue.
Mint condition drivers = unit I and II
So-and-so condition: Unit III and IV. Both have had the dust cap removed.

Left: Unit I, SPL measured on IEC baffle at o.25 m distance. Input normalised for 2.8V/1 meter. Blue = minimum phase.
Measurements merged at 300 Hz with nearfield measurement.
Right: Same driver front mounted and rear mounted (15 mm baffle). Blue = rear mounted.
The Permadyn drivers were made to be rear mounted, but it certainly doesn't improve measured performance having a short "tube" in front of the driver.

Left: Distortion measured on IEC baffle @ 0.25 meter distance. Input equivalent to ~90 dB/1 meter.
Blue = 2nd, green = 3rd. Equalising the 2-8 kHz range will certainly provide a nice distortion profile.
Right: All four drivers for comparison.

Left: Driver IV (damaged surround) is interesting as the currogated paper surround (integrated part of cone) is quite soft and this seems to have an
impact on the resonance at 900 Hz. This points to some experiments with a very light coating of the surround.
Right: TS data for unit I. Qt = 0.95 suggests an open baffle driver - or a semi-open resonanz box.

Preparing panel for IEC panel measurements.

Routing for front mounting.

Routing for rear mounting, assumed 12 mm baffle.

Measurements on IEC baffle.

Modelling example from web:
(back to top)

Note bass with inverted polarity. Eminence Deltalite-II 2515 used here.

Modelled frequency response with Eminence Deltalite II 2515. Middriver added 12 mm baffle (dZ = -12 mm).

Modelled phase tracking of drivers as suggested by constructor.


When we use 1st order filters we usually have to invert driver polarity and this is how it looks when
we connect bass and tweeter with positive polarity and mid with inverted polarity:

Apparently we get rid of the dip at ~6 kHz, but I wonder how a 10 dB peak at 11 kHz really sounds.
This is a very broad peak ranging from ~6-14 kHz. Some serious upper treble level.


Modelled impedance of system. 15" bass driver impedance from open baffle (OB9).

Building a "resonanz" box
(back to top)

My greencones cabs as they are likely to appear. 15 mm Baltic birch for front, rear, top and bottom bracing. 3.8 mm BB for sides.

Building the cabs

All of this is pretty straight forward. The difficult part is bending the 3.8 mm side panels. Takes a lot of clamps.

Measurements from mounting drivers in resonanz boxen

The basic measurements from midbass SABA PERMADYN 19-200-5298-U8 (left) and tweeter SABA 1670-DU-20 (right).

As predicted from measuring TS data, we're in for a system making 94-95 dB/2.8 volts. The low-ohm voice coil is partly responsible for this high sensitivity. Except for the serious outphasing around 900 Hz the 8" greencone display a fairly linear response. The 2-15 kHz range is some 5 dB above midrange level and calls for equalisation unless we want a much too elevated treble response - because this is going to play treble. The greencone tweeter is very much a pzzzztt-thing adding some airiness to the top octave. What the tweeter does below 6 kHz is pretty much useless and better be filtered out, not interfering with what the 8" driver covers. The useful tweeter level is also some 10 dB above target system sensitivity and may require quite some attenuation.

Before looking further into the possible crossovers, let's take a look at drivers' step response. To the left the tweeter where we have some serious problems. Due to the cone being fastened to the center pole piece we have a kind of mechanical high-pass filter. After the initial positive peak we have a negative peak indicating the cone moving inwards followed by a positive response before settling. Where this happens is hard to say, but using the tweeter only above 6-7 kHz may prove beneficial.
The 19-200 driver to the right looks a bit better although the 900 Hz cone break-up most likely is responsible for the less than ideal decline. From a modern perspective these drivers are not particularly well designed.

The "web" crossover:

Single coil to the tweeter, that's all! The purist's approach.

What's seen above is what can be found on the web. No low-pass filter to the midbass and 2.2 uF to the tweeter. Polarity of tweeter is dubious as wavelengh is short at 7 kHz and whether tweeter should be connected with poisitive or negative polarity doesn't make serious impact on performance. Above to the right measurements at T height (red), between M and T (blue) and at M height (green). At T height we have some outphasing at 6.5 kHz and as listening position is likely to at T height, it may be advantageous to reverse tweeter polarity. From the web it can also be found that a single cap to the tweeter doesn't change the phase... Of coarse it changes the fase.

To the left the response of individual drivers driven from crossover (red and blue) plus summed response (green). Point of crossover around 6 kHz. To the left we have the response @ 1 meter distance and at tweeter height. Some outphasing is seen but having the point of crossover this high, it's hard to hear noticeable difference from positive or negative tweeter polarity.

Left: Adding 15R to the tweeter helps a lot the harsh upper treble is gone. Still, I'd like the tweeter's response at 1 kHz further down.
Right: Again 15R to tweeter and horizontal dispersion at 0, 5, 10, 15 and 20 deg.
Despite having the midbass working up to 6 kHz we have a decent dispersion up to 4-5 kHz. Not bad at all.
Not surprisingly, at 15 deg. off-axis the large cone tweeter starts loosing grip.

Left: Adding 0.27 mH to the tweeter section improves sound considerably. Right: Individual and summed response from #2 crossover.
R1011 increased from 15R to 22R.

What's the impedance of a resonanz box like?
Hmm... these cabs certainly are resonant! Multible peaks in the 40-150 Hz range.
Left: Impedance from #3 crossover. Right: Impedance from #2 crossover (22R to tweeter).

Left: What happens if we turn the speaker up-side down? The 100Hz peak moves some 25 Hz up (red), so the the length of the "tube" and distance to floor impacts performance. Right: Here the two holes (vents) on the narrow rear panel have been blocked and these play a role too.

Final experiment. Sitting behind the speaker I pressed my knees against the bent panels (blue) - and yes,
the resonance modes of the panels certainly have an impact on impedance profile.

Replacing 22R tweeter series resistor with R1031 and R1041 is just another way of doing the same thing. Makes a nicer impedance plot.
10 dB attenuation: R1031 = 3R9, R1041 = 2R7
6 dB attenuation: R1031 = 3R3, R1041 = 6R8
3 dB attenuation: R1031 = 1R8, R1041 = 15R

The Greencone Resonanz Box Sound

Now, resonant boxes are certainly something very different from conventional speakers enclosures, where measures are taken to reduce cabinet resonances as much as possible - or as much money allows. It's not that difficult to find a "resonanz" box in a supermarket "hifi" section. "Cigar-box" speakers have been available since separate speakers became available.
First of all, the usual knuckle test leaves the impression - or expectation - that this just won't work. These cabs are boomy like crazy. I'm sure that if we used a standard 8"/Fs=30 Hz woofer in these cabs they would indeed be way too boomy. But a featherlight paper cone driver, having Fs = 70 Hz is another story. Such a driver won't play much bass in a standard cab anyway - at least not below 70 Hz - despite low electrical damping. But when what it does below 70 Hz makes close to one square meter bent plywood panels resonate - then we have bass. I was surprised to see an FFT plot suggesting bass response down to 40 Hz despite a severe suck-out around 60-65 Hz. This suck-out appeared regardless of room position and I'm afraid it has to do with the open bottom leaving some 130 cm between the front and the rear of the main driver. 1.3 meter is the wavelength of 344/1.3 = 264 Hz. The quarterwave is 66 Hz, halfwave is 132 Hz, which fits nicely with the resonances seen from the impedance plot. The 100 Hz peak may be due to a resonance between top and floor - or the peaks may be derived from the resonating plywood panels. Most likely it's a combination of both. More measurements and tests are needed, and as can be seen from above, damping the side panels seriously alters the impedance profile.
What worries me the most is that once these panels have been energized, they will inevitably return stored/delayed energy to the drivers' cones and smear detail and resolution. In fact, it sounds as this is actually happening when we crank up the volume a bit. The use of bent and light-weight side panels is one way to control this phenomenon and what can also be seen is that the peaks are very sharp, meaning the resonances die out fast, and so they should. In fact, tuning a light-weight paper cone to resonant wood panels may be just as difficult as adding braces to a guitar deck, only here have the energizer (strings) closely linked to the deck - and what we want is sustrain, where the speaker uses air to energize the panels and we want the panels to stop as soon as the speaker cone stops. Air, as we know, is very flexible (easy to compress) and may not be a good energy transmitter here.

The bass response from these greencones/resonanz boxes was better than anticipated. Not boomy or blurred, but lightweight and "short" like from a Voigt horn or similar.
Having an extended range driver doing most of it - here up to 6-7 kHz - sounds like what it is, i.e. due to the cone diameter, beaming starts around 2 kHz and having most of the treble area beamed at you smears detail and reduce perceived level of transparency. This phenomenon has been experienced over and over again and it seems an inevitable trade-off if we really do not want a point of crossover in the 2-6 kHz range. One reason I would like to hear the Altec 755A is due to the cone shape, having a fairly large "domed" voice coil and rather flat cone geometry. This may render beaming less apparent - but some other time should someone ever pop in with a 755 for test.
The "web-crossover", 2.2 uF to the tweeter with no attenuation, is really bad. The upper treble is harsh and forward and well, just too much. Adding the high-pass filter to the 4" cone tweeter makes it do what it's best at: Adding some airiness to the overall presentation. This tweeter, due to its fixed cone, really shouldn't do anything in the 1-6 kHz range. The tweeter diameter will make beaming apparent already at 3-4 kHz and had we started from a blank piece of paper, we would not even consider a large cone like this. A small three-quarter-to-one inch "something" would do much better in the 6-20 kHz range, but this is greencone territory and green it must be.

What surprised me even more was that my 300B amps didn't feel too comfortable with these speakers. I was sure this would be an ideal match, but it wasn't. Comparatively my 300B amp is doing better with my coming OB-11 open panels (15" bass driver + JA8008/TW034). My 20 wpc PSE Audio Mirror amps had no trouble running the greencones, so I wonder if this is really the ideal speakers for "low-wattage 2 wpc DHT SE amps".

To conclude my greencone/resonanz box experiences: These thin-paper greencones have limited soundstage capability and soon start generating serious distortion. Play music with low content of deep bass and they provide an excellent midrange for e.g. the oboe and flute and can be enjoyed free of any point of crossover in the mid and lower treble range. Using the drivers as an extended range driver up to 6 kHz calls for limitations too. Beaming - the enevitable result of having a large diameter speaker doing treble - limits sense of depth of soundstage and treble quality - as such - is not always good if the driver handles bass notes at the same time.
The basic learning I take from these experiments is that a resonant box can make bass - and decent bass too - if not pushed too hard - and finetuning the panels to the actual driver may be an endless occupation.

Believe me: Bending/gluing the side panels was much more trouble than anticipated.


This story turned out to be not quite as closed as planned. Having a spare pair of 19-200 greencones for experimentation I mounted these in the resonanz cabs and I was in for surprise.
These spare 19-200 drivers have had the dust cap removed and the corrugated paper surrounds were treated with a thin flexible coating. As discussed previously, the rugget dust cap forms a cavity to the center pole piece and having a very narrow magnetic gap, the trapped air in this cavity forms a pneumatic brake helping keeping things in place during large cone excursions. Check impedance measurements
above. Coating the surround also makes a smoother frequency response in upper mid/lower treble.

Setting up the speakers with my 300B amp now made a much more full-bodied sound with excellent bass, smoother midrange and the drivers could now handle much higher sound levels with no apparent distortion. I assume the bigger amps are less susceptible to this problem as they simply have better control (damping factor). The 300B really can't manage this problem with its limited 8 wpc. Smoothing the 10 dB dip at 900-1000 Hz certainly improved the overall presentation of acoustic instruments and making an in-room FFT plot now revealed a smooth bass performance down to ~45 Hz with no suck-out in the 60 Hz range as seen previously.

Left: Red = 19-200 driver as-is. Blue: Coated surround and dust cap removed.
Right: The two resonanz boxes at 1 m/2.8V. Disregard dip at 8 kHz. Very much depends on microphone placement.

Left: 1/6 octave FFT displaying adeaquate response down to 40-45 Hz. Right: Moving the microphone around my workshop makes these graphs.
The treble decline is normal behaviour for FFT farfield measurements.

The modded 19-200 drivers. Sorry for the colour tone. They're green.

The cure: Cut off the dust cap with a scalpel. Coat the surround twice with thinned book binders' glue.
Please find out for yourself the brand of glue and how much dilution it needs. The final coating should be slightly sticky without attracting dust.
Glue used for vinyle tiles may be another option. This needs dilution too. Make test coatings with various degree of dilution on cardboard.


Before resonanz boxen is finnally closed an article on "blackcones" will be publised. This from vintage SEAS 21TV-G midbass and SEAS 9TV-LG tweeters.