The 2.5 Clone Papers

Transcription

The 2.5 Clone Papers
1
The 2.5 clone papers
by Troels Gravesen
[email protected]
This is a compilation of former 2.5 clone files found at http://members.chello.se/jpo/
•
•
•
•
•
•
2.5 clone measurements and construction, v5, page 2.
2.5 clone without notch filter, page 24.
New tweeter for the 2.5 clone, Scan Speak 9500, page 26.
The final 2.5 clone, the sibilance problem, page 30.
The “new” 8535 drivers from Scan Speak, page 37.
The ProAc sound, page 41.
Only a few changes have been made to the files, thus reflecting the project progression during the last 9 months of
work on the Response 2.5 clone. New are some comments on the ProAc sound at page 41.
Thanks to those who started the project and gathered the basic information needed to get it all going. Thanks to
Paulie, US, for the basic crossover design. And thanks to all who reported their project on the web (a large number
of links can be found at http://members.chello.se/jpo/. Without the inspiration from these people, this would never
have evolved to such a long story.
Thanks to Darryl Nixon, Australia, for all the discussions and constructive criticism. Without the help of Darryl and
his “One Cloner’s Journey” found at http://www.diyaudio.com , we would not have had such a fruitful discussion
on the merits and deficiencies of speakers in general and of the 2.5 clone in particular. There will be different views
on the “right sound” of the clones, but only your ears can tell, what is best for you and your favorite music.
If you have any questions regarding the project, you are welcome to address [email protected]
Please refer to page numbers on specific questions.
Aarhus, 28th September, 2003.
2
2.5 clone measurements and construction, v5
Hello, 2.5 cloners!
Thanks to all for the huge number of mails coming in from Canada, US, Australia, Hungary, Norway, Sweden, UK,
Finland, Russia, Greece, New Zealand, etc. as response to these pages. And I cannot thank ‘JPO’ enough for
lending me the space on his website. Thanks to all who wrote, and commented on the work. Without these mails
the project would have ended another place.
In this 5th (!) version of my file I have added the construction of bass reflex enclosures with final measurements and
comments. Initially the drivers were mounted in transmission line cabinets available, similar in size to the 2.5s,
making reliable measurements.
I have had a lot of mails describing the benefits of adding the LCR circuit to the original design in order to get a
more even frequency response but also with some regrets over loosing some of this immediate appealing
‘technicolor’ sound of the originals.
Some people have been confused over the increasing numbers of crossover designs; they want solutions, not
options. For good reasons, they want to stay faithful to the original design and that’s fair enough. However, we
cannot acquire the original drivers and we will never be able to make an exact copy of the originals. But reports
from people comparing the clones and the originals tell us that the clone can be just as good or even better.
The crossover modifications are fairly simple and the choice is yours! The starting point is the crossover in fig. 1.
This design can be added a LCR notch filter, fig. 14, and you can leave it here. The latest modification I have made
were done to fine-tune frequency and phase response and has - in my opinion – improved midrange response but
has minor impact on the overall perceived sound.
Fig. 1 is the basic crossover (version 1) design I have
been testing.
And here are the MLS measurements from the
Stereophile magazine, as a target for designing the
crossover.
Fig 1. Basic crossover, ‘version 1’.
Fig 2: Stereophile measurements.
I was excited to see if the 2 kHz bump would appear
as predicted from Al.M’s writing at
http://www.geocities.com/diyproac25/.
The bump came out beautifully as seen from the
graphs on page 2!
3
Tweeter polarity
My first comment to the information available is on the discussion on tweeter polarity. It is suggested you try both
options and choose what suits you the best.
With same polarity of woofer and tweeter there will be a major dip in frequency response, which at the same time
can be used to fine-tune the crossover.
TGAudio
MLS - Frequency Response
29-09-02 12.44.59
100.0
CLIO
dBSPL
180.0
Deg
90.0
108.0
80.0
36.0
70.0
-36.0
60.0
-108.0
50.0
200
-180.0
1k
10k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 1.88ms
red=3.3uF
blue=4.5uF
green=5.5uF
Stop 5.14ms
yellow=6.8uF
Hz
20k
FreqLO 306.59Hz
purple=8.3uF
Fig 3: Polarity of tweeter.
The tweeter certainly has to be connected with inverted polarity.
Measurements on LP-section, inductor L2 values:
TGAudio
MLS - Frequency Response
29-09-02 12.18.00
100.0
CLIO
dBSPL
180.0
Deg
90.0
108.0
80.0
36.0
70.0
-36.0
60.0
-108.0
50.0
300
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.01ms
red=0.83mH
blue=0.68mH
green=0.47 mH
Stop 5.12ms
yellow=0.27mH
Hz
20k
FreqLO 322.01Hz
purple=none
Fig 4. Initial measurements of 8535 with 0.83, 0.68, 0.47 and 0.27 mH inductor value of 2nd inductor in LP section.
Not much chance of getting rid of the 2 kHz bump without a LCR circuit. I does strike me however that that lowest
value gives a response closer to the Stereophile measurements. Making the textbook LP crossover the response is
40dB down at 7 kHz where the target seems to more like 40 dB down at 9 kHz.
For the time being I stayed with the 0.83 mH. The 300-1500 Hz response is rather smooth which pleases me a lot.
4
Significance of value of capacitor in LP section.
TGAudio
MLS - Frequency Response
29-09-02 12.27.06
100.0
CLIO
dBSPL
180.0
Deg
90.0
108.0
80.0
36.0
70.0
-36.0
60.0
-108.0
50.0
300
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.01ms
red=6.8uF
blue=7.5uF
green=8.3uF
Stop 5.12ms
yellow=9uF
Hz
20k
FreqLO 322.01Hz
purple=10.1uF
Fig. 5.
The value of this capacitor makes a fine instrument of changing crossover frequency.
Significance of C1 in LP section on full range response.
TGAudio
MLS - Frequency Response
29-09-02 12.36.05
100.0
CLIO
dBSPL
180.0
Deg
90.0
108.0
80.0
36.0
70.0
-36.0
60.0
-108.0
50.0
300
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 1.88ms
red=3.3uF
blue=4.5uF
green=5.5uF
Stop 5.14ms
yellow=6.8uF
Hz
20k
FreqLO 306.59Hz
purple=8.3uF
Fig 6. MLS 0.33 oct. smoothing.
8.3 uF seems to be a too high value, where 5.5 to 6.8 uF looks more appropriate but as seen from the curves this
capacitor plays an important role in determining frequency response.
Series resistor in HP section
I would go for the 5R6 value, giving a quite flat frequency response. Going lower may give you an immediate
appealing sound, but may add to listening fatigue in the long run. But this can be depending on room acoustics and
listening distance.
5
47 ohms resistor in HP section.
In crossover diagrams available, the 47 ohms (R3) resistor is placed at different locations.
1. on tweeter terminals
2. before the 4.7 uF capacitor to ground
The graph above demonstrates the (minor) significance of this resistor placement. I’ve chosen to place the resistor
on tweeter terminals.
TGAudio
MLS - Frequency Response
29-09-02 22.09.35
100.0
CLIO
dBSPL
180.0
Deg
90.0
108.0
80.0
36.0
70.0
-36.0
60.0
-108.0
50.0
400
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 1.62ms
tweeter 47R
red= 47R <4.7uF
Stop 5.57ms
Hz 20k
FreqLO 253.47Hz
blue=47R >4.7uF
Fig 8. Tweeter, MLS, 0.33 oct smoothing.
Construction of notch filter for the 2 KHz bump:
A LCR notch filter was designed to smooth the frequency response between 1500 and 3000 Hz consisting of
1.5mH(0R35)+3.3 uF+10 ohm resistor.
Fig. 9 displays the impact on 8535 response, MLS with no smoothing.
TGAudio
MLS - Frequency Response
06-10-02 18.24.29
100.0
CLIO
dBSPL
180.0
Deg
90.0
108.0
80.0
36.0
70.0
-36.0
60.0
-108.0
50.0
300
-180.0
1k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 2.54ms
10k
Stop 5.61ms
Hz
20k
FreqLO 326.11Hz
Fig.9. LP section +/- LCR and full range response with LCR.
Notice that crossover target is only slightly affected and can easily be adjusted with C1. Fig.5.
6
LCR impact on full range response
TGAudio
MLS - Frequency Response
06-10-02 17.59.22
100.0
CLIO
dBSPL
180.0
Deg
90.0
108.0
80.0
36.0
70.0
-36.0
60.0
-108.0
50.0
400
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.46ms
Stop 5.12ms
Hz 20k
FreqLO 376.47Hz
Fig.10. Full range response +/- LCR, 0.33 oct. smoothing.
TGAudio
MLS - Frequency Response
06-10-02 17.56.08
90.0
CLIO
dBSPL
180.0
Deg
80.0
108.0
70.0
36.0
60.0
-36.0
50.0
-108.0
40.0
200
-180.0
1k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 2.46ms
10k
Stop 5.12ms
Hz
20k
FreqLO 376.47Hz
Fig. 11. Full range response + LCR, same and reverse polarity (no smoothing). 1 meter distance, tweeter height.
7
Significance of LCR on cumulative spectral decay.
TGAudio
Waterfall
06-10-02 17.59.41
0
dB
0.0
-5
ms
0.9
-10
1.8
-15
-20
2.7
400
1k
Hz
10k
20k
CLIO
Cumulative Spectral Decay
Fig.12. 20 dB range CSD, without notch filter.
TGAudio
Waterfall
06-10-02 17.58.05
0
dB
0.0
-5
ms
0.9
-10
1.8
-15
-20
2.7
400
1k
Hz
10k
20k
CLIO
Cumulative Spectral Decay
Fig.13. 20 dB range CSD, with notch filter.
The impact of the notch filter speaks for itself. With the LCR circuit in place an impressive + 1 dB frequency
response is achieved between 300 and 4000 Hz, and the lack of the 2 kHz bump is clearly audible, where especially
female voices gets a natural balance and acoustic guitars which may sound almost too good with the bump, now are
presented with a much more realistic timbre. Listening to pink noise on the 8535 +/- LCR filter strongly suggests
we get rid of the bump. Read Lynn Olson (Ariel) on the use of pink noise! http://www.alohaaudio.com/Arieltxt2.html#top
And best of all, the 8535 does not loose its fresh and crisp presentation. The sound of the 8535 is hard to describe
(isn’t sound always?), but certainly this is a very lucky/clever combination of the right matrix of paper pulp and
carbon fiber, the right cone size and weight (the cone is more flexible that the 8545), voice coil dimensions, magnet
size, all giving a smooth roll off characteristic and simplifying crossover design.
8
Modified crossover, basic design + notch filter (version 2):
Fig 14. Version 2 crossover.
Fine tuning of crossover, version 3:
In order to improve frequency and phase response I have modified the crossover and it looks like this. Red
indicates changes.
Fig 15. Version 3 crossover.
Some people have made complaints on the crossover presentation and here is a more graphic presentation for biwiring. The components in the LCR circuit can be put together in any order,
it does not matter! And be sure to have a decent distance between inductors - like 5 cm, in order to reduce
interaction.
9
Fig.16.
System response with Fig. 14 (version 3) crossover:
TGAudio
MLS - Frequency Response
23-10-02 23.04.34
90.0
CLIO
dBSPL
180.0
Deg
80.0
108.0
70.0
36.0
60.0
-36.0
50.0
-108.0
40.0
300
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.75ms
red=sensitivity, 2.83 V AC, 1 meter
Fig.17. System response, version 3.
Red/blue=left and right speaker.
Stop 5.29ms
blue=min fase
FreqLO 393.85Hz
Hz
20k
10
System impedance
TGAudio
Sinusoidal
29-09-02 14.45.18
50.0
CLIO
180.0
Ohm
Deg
40.0
108.0
30.0
36.0
20.0
-36.0
10.0
-108.0
0.0
10
-180.0
100
1k
CH A Ohm Unsmoothed Stepped
Delay [ms] 0.000
10k Hz
20k
Dist Rise [dB] 30.00
File: imp full range.sini
Fig. 18. Impedance of full range (without notch filter) system.
Not that much different from the Stereophile measurement, although it’s difficult to read the scale on the
Stereophile scanning.
The high damping of the lower impedance peak in the bass is caused by the stuffing of the transmission line and
should be disregarded in this context.
ScanSpeak 18W/8535-00
Finally, here are the TS parameters of my 8535s: The data for the two units are remarkably alike, but the Qt is
significantly higher that the promised 0.4!
ScanSpeak data: Vas=69 litres, Qt= 0.38 and Fs= 26 Hz.
My measurements: Vas = 44 litres, Qts= 0.52 and Fs= 34 Hz.
ScanSpeak is using constant current method at high level, 36 mA, which may account for lack of correlation.
My speaker calculation software says ~ 42 liters from the SS-data, my measurements suggests ~ 67 liters. Another
software tells me this unit is best suited for a closed box!
MANUFACTURER
MODEL
DATE
Fs
Qms
Qes
Qts
VAS
Mms
BL
dBSPL
SD
Re
Fig.19. TS data.
ScanSpeak
18W8535-I
20-09-2002
33.9
2.92
0.63
0.52
43.9
14.4
5.34
86.3
0.0143
5.88
ScanSpeak
18W8535-II
20-09-2002
34.8
2.93
0.63
0.52
43.1
13.9
5.36
86.6
0.0143
5.94
These data were generated with my CLIO measurement system set at 0dB level. Adding another 0.4 ohm resistance
(from inductors) in series with the woofer makes things even worse. The magnet on the 8535 seems to be too small
for a 33-liter cabinet. Object for some tweaking!
11
TS parameters at different measuring level:
measuring level
+10dB
0dB
MANUFACTURER scanspeak
scanspeak
MODEL
18W8535
18W8535
DATE 30-01-2003
30-01-2003
Fs
32.37
35.16
Re
5.86
5.86
Rms
1.15
1.00
Qms
2.64
2.96
Qes
0.54
0.60
Qts
0.45
0.50
Cms
1.62
1.53
Mms
14.96
13.41
BL
5.74
5.36
VAS
46.25
43.71
Fig.21
-10dB
scanspeak
18W8535
30-01-2003
36.65
5.86
1.01
3.47
0.66
0.55
1.24
15.24
5.60
35.40
-20dB
scanspeak
18W8535
30-01-2003
37.35
5.86
0.92
3.72
0.68
0.58
1.24
14.68
5.44
35.40
Series filters
At http://audioclone.free.fr/ two series filters have been proposed and I have tried to wire up the circuits, and here
are my measurements:
Version 1:
Fig. 22. Series filter #1.
First of all a RC circuit has been placed at the crossover terminals which is a strange feature as it generally lowers
the sensitivity of the whole system. I’ll show later the impact of this on frequency response.
12
Frequency response, series filter #1, no RC circuit:
TGAudio
MLS - Frequency Response
07-12-02 23.44.15
100.0
CLIO
108.0
dBSPL
Deg
90.0
36.0
80.0
-36.0
70.0
-108.0
60.0
-180.0
50.0
200
-252.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.68ms
File: 95 cm MLS.mls
red=95 cm hight
Stop 5.57ms
Hz
20k
FreqLO 345.95Hz
blue=min phase
Fig.23. Series filter #1, 1 meter distance, tweeter height. Red=freq.resp. blue=min.phase.
When I first looked at this I thought I’d done a serious mistake and checked the setup several times. Couldn’t fine
anything wrong. Minimum phase indicates serious problems and I tried same polarity of woofer and tweeter:
Frequency response, series filter #1, no RC circuit, same polarity:
TGAudio
MLS - Frequency Response
08-12-02 00.02.17
100.0
CLIO
108.0
dBSPL
Deg
90.0
36.0
80.0
-36.0
70.0
-108.0
60.0
-180.0
50.0
200
-252.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.71ms
File: 95 cm MLS-same polarity.mls
red=95 cm hight
blue=min phase
Stop 5.57ms
Hz
20k
FreqLO 350.68Hz
same polarity
Fig.24. Series filter #1, same pol. Red = freq.resp., blue=min.phase
Well, at least the major dip in frequency response at 2 kHz was gone, but the 2 kHz bump came to life again and
min. phase still isn’t to pretty.
I suppose the series filter was constructed with the intention of keeping inverted polarity of the drivers, so I went
back to this and tried different measuring heights.
Next is the response measured at 1 meter distance, microphone between tweeter and woofer:
13
TGAudio
MLS - Frequency Response
07-12-02 23.45.47
100.0
CLIO
108.0
dBSPL
Deg
90.0
36.0
80.0
-36.0
70.0
-108.0
60.0
-180.0
50.0
200
-252.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.68ms
File: 91 cm MLS.mls
red=91 cm hight
Stop 5.57ms
Hz
20k
FreqLO 345.95Hz
blue=min phase
Fig.25, series filter #1, 1 meter distance, mic. between tweeter and woofer. Red=freq.resp., blue=min.phase.
This turned out even worse, almost an 180o phase shift at 2.5 kHz.
I went back to measuring at tweeter height!
Series filter #2, no RC circuit
In this setup a 1 mH coil is introduced across the tweeter and there are minor modifications to the other
components.
Fig. 26. Series filter #2.
14
Frequency response of series filter #2, no RC circuit:
TGAudio
MLS - Frequency Response
08-12-02 00.12.29
100.0
CLIO
108.0
dBSPL
Deg
90.0
36.0
80.0
-36.0
70.0
-108.0
60.0
-180.0
50.0
200
-252.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.71ms
File: series v2.mls
red= v2
Stop 5.57ms
Hz
20k
FreqLO 350.68Hz
blue=min phase
Fig.27. Series filter #2, no RC. Red= freq.resp., blue=min.phase.
The serious dip in response at 2.3 kHz has been reduced slightly, but this is far from being an acceptable frequency
response. And still there are serious phase problems.
Impact of RC circuit on system response, series filter v.2:
TGAudio
MLS - Frequency Response
07-12-02 23.54.16
100.0
CLIO
108.0
dBSPL
Deg
90.0
36.0
80.0
-36.0
70.0
-108.0
60.0
-180.0
50.0
200
-252.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.68ms
File: plus-minus-RC.mls
inverted polarity
red=tweeter hight, no RC
Stop 5.57ms
Hz
20k
FreqLO 345.95Hz
blue=tweeter hight, +RC
Fig 28, System response, series filter #2, with/without RC circuit. Blue = with RC.
The system response is generally lowered by 2 dB.
The system response of the 2.5 clone is around 83 dB/2.83 V/1meter, which is pretty low. No reason to burn more
energy in the RC circuit.
15
Impedance of system with series filter #2, +/- RC circuit:
TGAudio
Sinusoidal
08-12-02 00.24.24
50.0
CLIO
180.0
Ohm
Deg
40.0
108.0
30.0
36.0
20.0
-36.0
10.0
-108.0
0.0
10
-180.0
100
1k
CH A Ohm Unsmoothed Stepped
File: imp v2+RC.sini
red:v2 no RC
10k Hz
Delay [ms] 0.000
20k
Dist Rise [dB] 30.00
blue:v2 +RC
Fig. 29. Series filter, impedance, +/- RC circuit.
Indeed the impedance is flattened to around 4 ohm above 100 Hz.
My only comment to the RC circuit is that this must be a mistake.
Well, those who might have been annoyed with the 2 kHz bump in the original design certainly eliminate this
problem with the series filter, but this seems to introduce new and more serious problems. Frequency and phase
responses are unacceptable and serious tweaking is necessary to get it right.
The problem with series filters is that it’s difficult to measure the response of the individual drivers. John
Kreskovskij has a method, but I haven’t tried it yet.
http://www.geocities.com/kreskovs/Series-1.html
I have only briefly done listening tests with the series filter but I find the dip at 2 kHz clearly audible and the
tweeter far to loud for my taste.
Replacing the drivers with 8 ohm resistors is not strictly correct but can give us an idea of what is going on.
TGAudio
MLS - Frequency Response
08-12-02 13.13.29
90.0
CLIO
dBSPL
180.0
Deg
80.0
108.0
70.0
36.0
60.0
-36.0
50.0
-108.0
40.0
200
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.75ms
Stop 5.55ms
Hz
20k
FreqLO 358.04Hz
drivers replaced by 8R2 resistors
Fig. 30.
Red= frequency response of system with series filter.
Blue= 8535 woofer with series filter.
Green= 8513 tweeter with series filter.
Purple= with resistor and capacitor for tweeter inverted, which gives a better response because now the capacitor
‘sees’ a much more reasonable impedance.
Yellow = system response with this modification.
16
Fig 31. Part of series filter with inverted C and R for tweeter.
With the suggested design the tweeter reaches down to 1500 Hz giving serious phase problems in the crossover
region.
Inverting the resistor and capacitor helps a lot, but there are months of work to get it right!
Conclusion on series filters:
The series filter has a less than acceptable frequency response, serious phase problems and cannot be
recommended.
And don’t place a RC circuit on top of the whole crossover, this way you will just burn energy and reduce system
efficiency, which is so much needed.
17
Construction of bass reflex cabinets:
A lot of cabinet construction papers have been published and I won’t go into much detail about this.
I have maintained the internal volume but outer dimensions have been changed to 20 x 26.5 x 100 cm and the
bottom plate have been lifted to give room for the crossover to be placed externally. This way it’s easy to make
changes, and the components are not affected by vibration from the driver. The tweeter has it’s own sealed back
chamber in order to reduce vibrations from the woofer. Cross bracing has been added to reduce cabinet resonance.
Cabinets are constructed from 20 mm pre-veneered MDF and front panels are 25 mm (15 mm solid mahogany + 10
mm MDF). Internal bracing is 10 mm MDF.
Fig 32. Cabinets partly assembled.
All walls are damped with 10 mm heavy polyester foam (glued to the panels) and a mixture of polyester and lambs
wool available from Monacor is used for further damping.
Right behind the 8535 several layers of the lambs wool is placed in order to reduce standing waves hitting back on
the membrane.
Deflex damping material is to my knowledge not available in Denmark (?).
Some more pictures:
Fig.33, 2.5-crossover.
Cored 1.8 mH coil and air-cored 0.47 mH, both <0R2.
Polypropylene capacitors and film resistors. (Still waiting for the 47 ohm film resistors).
I don’t think this construction justifies the ultimate crossover components unless you think this is THE speaker of
your life. There are a lot of other variables that have greater influence on the sound.
18
Fig. 34, front plate.
Fig.35, back side of front plate.
Fig.36, damping material.
Fig.37, CO at base plate.
19
At last, the final cabinets with drivers, first play in my workshop; why is it that I after just a couple of months
forget how many hours it takes to build a pair of cabinets!?
Fig 38. First time setup in my workshop.
Crossovers, the never ending story…
- and short presentation of features:
1. The ‘original’ design, version 1 (fig.1)
You are likely to have a major bump at 2 kHz, which sounds very well on certain recordings but makes voices and
violins intolerable. Darryl from Australia calls this the ‘technicolor sound’ and that is just what it is.
2. Original design + LCR, version 2 (fig 14)
You get rid of the 2 kHz bump and can enjoy a wider spectrum of recordings. Enhanced three-dimensionality and
lots of space.
3. Modified crossover + modified LCR, version 3 (fig15)
An even flatter frequency and improved phase response in the critical upper midrange. The choice is yours.
Having finished my bass reflex enclosures I have wired up the three crossover (CO) versions again and was excited
to see whether I could reproduce my measuring results 2-3 months ago! And it didn’t turn out too bad. All
measurements performed at 1 meter distance, tweeter height.
The Stereophile review suggests we have a crossover point of 3200-3300 Hz, where the version 1 gives 3000-3100
Hz, slightly below the original. The version 3 displays a crossover point of 3350 Hz. However, no need to be exited
about +/- 100 Hz. The aim of the v3 crossover was to create a less steep roll-off as seen on the Stereophile
measurements. The 8535 driver is down 40-45dB at 8.5 kHz where we - with the v1 crossover - reach this level
already at 7 kHz. It has been suggested that the OEM-8535 has less carbon fiber and a more flexible membrane
than the ‘DIY’ units. This could mean enhanced high frequency response and a smoother roll-of performance. The
problem with having a notch filter at 2 kHz is that it is so close to the crossover point, that it’s impact is stretched
over the crossover region. I have tried a notch-filter acting exactly in the 1500-2500 Hz region, but this didn’t
perform well.
Fig. 39 and fig. 40 displays the 8535 performance with the v2 and v3 filter +/- notch filter.
Fig. 41 and fig. 42 displays the overall performance of drivers with same and inverted polarity and tells us that we
have a very good phase correlation between drivers in both cases. The v3 has a more symmetrical >20 dB null at
crossover point with same polarity. The 8535 with the v2 CO has a very abrupt, linear and steep roll-off behavior.
20
MLS - Frequency Response
14-02-03 18.35.47
90.0
CLIO
dBSPL
MLS - Frequency Response
180.0
Deg
14-02-03 18.06.30
90.0
CLIO
dBSPL
180.0
Deg
80.0
108.0
80.0
108.0
70.0
36.0
70.0
36.0
60.0
-36.0
60.0
-36.0
50.0
-108.0
50.0
-108.0
-180.0
40.0
200
40.0
300
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.56ms
2.5 crossover v2
1.8mH/7.4uF/0.83mH
Stop 5.53ms
Hz
20k
FreqLO 336.84Hz
-180.0
1k
10k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 2.75ms
3.3uF/10R/1.5mH
2.5 crossover v3
Fig. 39. 2.5 crossover, v2, red: + LCR, blue: no LCR
MLS - Frequency Response
CLIO
dBSPL
Deg
20k
2.2uF/10R/1.5mH
MLS - Frequency Response
180.0
Hz
FreqLO 360.56Hz
Fig. 40. 2.5 crossover, v3, red: + LCR, blue: no LCR
14-02-03 20.06.26
90.0
1.8mH/6.8uF/0.47mH
Stop 5.53ms
14-02-03 20.11.58
90.0
CLIO
dBSPL
180.0
Deg
80.0
108.0
80.0
108.0
70.0
36.0
70.0
36.0
60.0
-36.0
60.0
-36.0
50.0
-108.0
50.0
-108.0
-180.0
40.0
300
40.0
300
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.56ms
crossover v2
LP:1.8mH/6.8uF/0.83mH
LCR:3.3uF/10R/1.5mH
Stop 5.53ms
Hz
20k
FreqLO 336.84Hz
HP:5R6-3.3uF/0.25mH/4.7uF-47R
CO-point=3100Hz
-180.0
1k
10k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 2.56ms
reverse/same pola
crossover v2
LP:1.8mH/6.8uF/0.47mH
LCR:2.2uF/10R/1.5mH
Stop 5.53ms
HP:5R6-3.9uF/0.22mH/4.7uF-47R
Hz
20k
FreqLO 336.84Hz
CO-point=3400Hz
reverse/same pola
Fig. 41. 2.5 crossover, v2, all drivers, polarity
Crossover point = 3100 Hz
0.33 oct. Smoothing. Same polarity, no smoothing
Fig.42. 2.5 crossover, v3, all drivers, polarity
Crossover point = 3350 Hz
0.33 oct. Smoothing. Same polarity, no smoothing
Frequency response of 2.5 with CO v2 and v3
Frequency response of CO v3 at 1 and 2 meters
MLS - Frequency Response
14-02-03 20.32.42
90.0
CLIO
dBSPL
MLS - Frequency Response
180.0
Deg
14-02-03 23.31.35
100.0
CLIO
dBSPL
180.0
Deg
80.0
108.0
90.0
108.0
70.0
36.0
80.0
36.0
60.0
-36.0
70.0
-36.0
50.0
-108.0
60.0
-108.0
-180.0
50.0
300
40.0
300
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.56ms
2.5 crossover v2 and v3
full range response and min. phase
Stop 5.53ms
Hz
20k
FreqLO 336.84Hz
green/purple=v2
red/blue=v3
Fig. 43.Green/purple(min.ph) = v2, Red/blue(min.ph)
= v3
-180.0
1k
10k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 5.74ms
2.5 response tweeter height
red=1 meter
blue=2 meter
Stop 6.91ms
Hz
FreqLO 853.33Hz
purple=same pol. 2 m, 2.5 cm above tweeter height
Fig. 44. Red=1m, blue=2m, purple=same pol.
20k
21
MLS - Frequency Response
15-02-03 21.44.47
100.0
CLIO
dBSPL
180.0
Deg
90.0
108.0
80.0
36.0
70.0
-36.0
60.0
-108.0
50.0
300
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.81ms
blue=left speaker
red=right speaker
Stop 5.45ms
Hz
20k
FreqLO 379.26Hz
For comparison here are fresh Rogers LS3/5a (11
ohm version) frequency response curves, with and
without front grille. A legendary loudspeaker with
phase problems in the crossover region that today
would make any home constructor ashamed of
himself. I keep these shoeboxes to remind myself of
not overemphasizing any single parameter in
loudspeaker construction because they sound so
good.
MLS - Frequency Response
CLIO
dBSPL
180.0
Deg
90.0
108.0
80.0
36.0
70.0
-36.0
60.0
-108.0
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.87ms
red:+front
Stop 5.53ms
Hz
20k
FreqLO 376.47Hz
blue:-front
Fig.46. Rogers LS3/5a, 11 ohm version, 1989.
MLS - Frequency Response
15-02-03 21.51.31
100.0
CLIO
dBSPL
1.400
Increasing the resistor to 12 ohms does do this to
some small extent I guess, but any greater value
begins to re-introduce the upper midrange glare
quite audibly (to my ears, anyway). All things
considered, I still prefer the sound with the 10 ohm
value, i.e. an optimally flat response in the 2 KHz
area.
I did find that reducing the resistor on the tweeter
from 5.6 ohms back down to 5 ohms produced quite
an improvement - I should have tried this before and restored much of that distinctive Proac sound,
more so than I would have expected. (With the
standard (Jacq) crossover + notch filter, I preferred
the 5.6 ohms you recommend.) Increasing the tweeter
output in your latest crossover makes it a lot harder
to choose between the two versions. Even with
increased tweeter level, sibilance is still better
controlled and the midrange sounds more realistic
than it does with the Jacq version + notch filter. I
think I'll stay with your latest version for the time
being, albeit with slightly increased treble. That
change has swung things the other way for me. It
seems a good compromise in my system, and 90 per
cent of the time it sounds wonderful.
ms
90.0
1.000
80.0
0.600
70.0
0.200
60.0
-0.200
50.0
300
I finally got around to experimenting with resistor
value increases in the notch filter, which you
suggested might restore some of the "life" or
"technicolour" sound to your latest (final?) crossover
version.
15-02-03 21.31.53
100.0
LS35A
I am very happy for the evaluation given by Darryl in
Australia and with his permission, here are his
comments:
sensitivity 2.83Vrms
Fig.45. Freq. Response left and right speaker.
Sensitivity at 1 meter, 2.83 Vrms, ~ 83 dB.
50.0
300
Well, whether you choose v2 or v3 crossover you
will get a great speaker in any case.
Be sure to use the right component values in either
case. The most important components are the
capacitor in the LP section, 7.4 uF for v2 and 6.8 uF
for V3, the coil in the HP section, 0.25 mH for v2 and
0.22 mH for v3 and finally the capacitor in the LCR
circuit, 3.3 uF for v2 and 2.2 uF for v3.
-0.600
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.81ms
red=0
blue=10
green=20
Stop 5.66ms
Hz
20k
FreqLO 350.68Hz
purple=30
Fig.
47.
CO
v3
horizontal
0(red),10(blue),20(green)30o(purple)
response,
Incidentally, the listening tests I've been carrying out
have been with a Dynaco PAS3X/Stereo 70 valve
preamp and amp. I also have a Sugden C51/P51
solid state combo, but it's off for repair at the
moment.
22
I was curious to see what would happen with a mid-fi
solid state amp, so borrowed a friend's NAD 1155
preamp and Rotel RB-981 power amp. There is no
doubt in my mind that the Proac is best suited to
valve amplification. With this solid state set-up, there
was a definite "hardness" in the midrange which was
easily provoked by the wrong recording. Sibilance
also became more of a problem. The sound was also
quite "dry" and occasionally harsh, though bass
depth and definition was astoundingly good. (A
Superphon Revelation Basic Preamp did improve
things at the top end.) Nevertheless, the speaker
seemed a lot more tolerant of this set-up with your
latest crossover than with the Jacq + notch filter. My
own Sugden (although no longer young) works far
better, having more valve-like warmth and a much
superior presentation all round, though not as good
as the Dynaco.
To my mind, the Proac Response 2.5 is a seriously
good but extremely fussy speaker, easily provoked
into sounding less than wonderful. The bass-mid is
very transparent to the source, and can easily stray
into hardness with inadequate SS amplification, discreproduction equipment or poor recordings. (I do
wonder whether that hardness is in part due to cone
break-up, albeit at a low level thanks to your notch
filter.)
Another problem is the tweeter, which can easily
stray into excessive sibilance with the wrong
recording, though your latest crossover mods go a
long way towards eliminating this. Yet another
problem is a lack of energy in the lower
midrange/upper bass, which seems to be roomboundary related. Careful positioning and a warmsounding amp can minimise this, but it seems
impossible to cure completely.
I guess I'm being over-critical, given the Proac's
price-point. – and all "high-end" equipment is fussy.
Nevertheless, this is the best speaker I have ever
owned and pretty easy to live with - and it now works
very well in my system. I just wonder how many
people out there are disappointed due to matching
problems in their systems!
Final evaluation
This is probably the most tough part of it all having
to express sonic qualities in a foreign language.
But I’ll give it a shot….
This is probably my 5th floorstander of this design
being bass reflex or transmissionline constructions,
all two-way designs of approx. 20x25x100 cm in size
with SS8545+9500, Vifa PL18+XT25, Vifa
M18WO+D27, etc.
My setup consists of a ROTEL CD modified with
goodies from LCAudio, a CT101 audio buffer from
DanishAudioConnecT (www.dact.com) and a
LCAudio, non-feedback 120 W power amplifier,
Millenium edition (www.lcaudio.dk).
Listening sessions, version 3 crossover.
First disc was Charlie Haden & Pat Metheny:
Missiouri Sky.
It's been a couple of months since I had my initial
transmission lines running and the final setup with
the reflex boxes by far exceeded my expectations.
The bass is significantly better in the reflex boxes and
I had to remove things from my living room that do
not use to rattle with my ‘reference’ system! I can't
believe they go this deep! A tiny 6½" woofer! The
midrange is clear, crisp and transparent and listening
to acoustic music it's very, very good. On track two
the guitar is very closely miked with a lot of low-end
information and the bass attack is most impressive.
The tonal balance seems to favor the highs and
listening to female jazz-singers, strings and big-band
music was not so impressive! ‘S’-sounds and ‘T’sounds are much to pronounced and the tonal balance
of violins are simply not correct compared to my
‘reference’ where voices can be played at loud
volume without distress. The phenomenon is called
sibilance!
Tried to unplug the tweeter and played the 8535 at
loud levels and everything sounds fine except that
you miss the tweeter. The problem doesn’t seem to
come from here.
Another comment from Darryl:
Do you also find a lack of energy in the lower
midrange/upper bass? My own clones sound
beautifully warm provided there are deep lows in the
recording, but if not, they can sound quite "dry".
Yes and no. I can’t say that I have experienced any
lack of energy or level in upper bass register and I
believe actual room acoustics plays an important role
here. On the other hand I like a speaker that has a
very dry sound. I’ve done a lot to reduce vibrations in
my cabinets and this probably helps a lot in providing
a dry sound. I don’t feel any vibrations on sides, front
and back of the cabinet, but strangely enough on the
top plate and I’ll have to ad additional material to
eliminate this.
23
Let’s take another look at the frequency response:
MLS - Frequency Response
16-02-03 18.50.08
90.0
CLIO
dBSPL
180.0
Deg
80.0
108.0
70.0
36.0
60.0
-36.0
50.0
-108.0
40.0
200
It’s a gut feeling, that the 8513 may not be the one
to pick if you want a more true presentation of the
upper register of most instruments and voices. It has
some intrinsic values in terms of speed and
‘sparkling’ sound, and possibly we can turn this
speaker into something that will split constructors
into two groups. One group that wants to stay true to
the original design with its limitations in terms of not
being able to obtain the ‘real’ OEM-drivers and
another group that will take the best of the 8535’s
deep bass capabilities and midrange clarity and
combine it with a tweeter that supplements these
virtues with more fidelity.
-180.0
1k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.70ms
10k
Stop 5.55ms
Hz
20k
FreqLO 350.68Hz
Fig.48. Value of tweeter series resistor
Green=5R6
Yellow=8R2 (sorry for the yellow, hope it’s visible)
Purple=10R
Usually I try to target the BBC-dip curve, giving a ~2
dB dip in the upper midrange/lower highs usually
giving a slightly more distant perspective, but an
overall more balanced sound. The response of the
clones do not exactly meet this criteria. We have a
rather flat midrange response and it better be good
with this level.
Fig. 48 graphs are showing the response at 2.83Vrms
(measured at speaker terminals) and from 0.3-3.5
kHz we have a very flat response of +/- 1 dB. Quite
impressive. But from 4-17 kHz we are least 2 dB
higher and from another construction I learned that
this could make a world of difference. I have worked
a lot with the 8512 tweeter supplementing an ETON
4-300 midrange and this – when properly balanced –
works very well. The 4-300 is a very revealing
midrange driver and matching this driver with a
slightly too highly pitched tweeter makes it
intolerable to listen to.
Maybe the 8512 and –13 isn’t that well suited to
work with the much larger 8535 woofer/midrange
cone. I would like to try the HIQUPHON OW1tweeter without magnetic oil (and probably more
heavily coated) and produced to very close tolerances
with reported low distortion and coloration and an
impressive CSD.
Changing the tweeter series resistor to 8R2 or 10R
seems to correct things and I stayed with 8R2
because with 10R I would have to go through another
fine-tuning of the HP section because it changes the
crossover point to 3.7 kHz and creates a 1.5 dB dip at
~3.5 kHz.
It helped a great deal on the above mentioned
problems although there are still recordings were they
fall short compared to my ‘reference’.
Best regards
Troels Gravesen
[email protected]
PS, 30.03.2003
Have tried the OWI tweeters and except for size and
sensitivity they can immediately replace the 8513,
but the response turns out even flatter than with the
8513s and the sound wasn’t so good. After some
tweaking, I decided this would require a new
crossover and tempting as it was, this is not the time.
The OWIs measures the best I have ever experienced.
Ruler flat from 1 kHz to 20 kHz! Can’t wait to
incorporate these in some future design.
24
The 2.5 clone without notch filter
Sometimes it takes a journey to get back to your
starting point and see what is the real problem in
front of you.
This being the case with the 2.5 clone and the 2 kHz
bump created by the 8535 driver itself and the
crossover topology. One person at diyaudio.com
working with active crossovers for the clone even
predicted that there should be a bump at 2 kHz
derived from the crossover.
In my latest paper at http://members.chello.se/jpo/
(New 2.5 clone tweeter, crossover and speaker
setup) I have constructed a new crossover for
implementing the ScanSpeak D2905-9500 tweeter.
By starting all over again with the crossover it was
obvious to experiment with the Q of the parallel
capacitor in the LP-section. In order to get the target
point of crossover and the target roll-off
characteristic a RC circuit was added. It appeared that
no notch filter was needed.
When this was done it was obvious to try to apply
this approach to the 8535+8513 drivers for those who
want to maintain the 8513 tweeter.
Initial measurements of 8535 driver:
TGAudio
MLS - Frequency Response
07-04-03 19.54.51
90.0
CLIO
dBSPL
180.0
modification refers to the ‘version 3’ crossover
found in ‘2.5 clone measurements and construction,
v5’ at http://members.chello.se/jpo/.
And it implies the use of DAMAR coating as
described in 8535+9500 paper. However, I’m
confident that this modification will also work fine
without the DAMAR coating.
TGAudio
MLS - Frequency Response
07-04-03 20.27.34
90.0
CLIO
dBSPL
180.0
Deg
80.0
108.0
70.0
36.0
60.0
-36.0
50.0
-108.0
40.0
200
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 3.24ms
red= +1.8 mH
blue= +(1.8mH-47R)/(7.4uF)/(0.83mH)
Stop 6.04ms
Hz
20k
FreqLO 358.04Hz
green=+(1.8mH-47R)/(6.8uF+3R3)/(0.47 mH)
Fig.2. Driver response, 0.33 oct. smoothing.
As seen from the graphs the result is a smooth
midrange response, and the level can be adjusted to
personal taste by changing the value of the resistor in
the in LP section, fig. 3.
Deg
TGAudio
80.0
108.0
MLS - Frequency Response
07-04-03 20.43.17
90.0
CLIO
dBSPL
180.0
Deg
70.0
36.0
80.0
108.0
60.0
-36.0
70.0
36.0
50.0
-108.0
60.0
-36.0
50.0
-108.0
40.0
200
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.75ms
red= +1.8 mH
Fig.1. 8535
smoothing.
blue= +1.8mH/7.4uF/0.83mH
driver
SPL
Stop 5.74ms
Hz
20k
FreqLO 334.64Hz
green=+1.8mH/6.8uF+3R3/0.47 mH
response,
0.33
oct.
40.0
200
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 3.38ms
green=2R2
red=3R3
Stop 6.04ms
Hz
20k
FreqLO 376.47Hz
blue=4R7
Blue = 8535 driver with v1 crossover: 1.8 mH//47R +
7.4 uF + 0.83 mH
Red = 8535 driver + 1.8 mH
Green = 8535 driver with new crossover: 1.8mH +
(6.8uF+3R3) + 0.47 mH
Fig.3. Value of R in LP section.
Green = 3R3, red = 4R7, blue = 5R6. I recommend
3R3.
As seen from the graphs it is possible to eliminate the
notch filter by adding a resistor to the capacitor. That
simple!
And the point of crossover can still be adjusted by the
capacitor value (data not shown).
Crossover changes:
I have tried to maintain all component values in the
new design to minimise the cost of the change. This
•
•
•
•
The 47R resistor parallel to the 1.8 mH
inductor removed.
The 6.8 uF capacitor is added a 3.3 ohm
resistor.
LCR notch filter is removed.
No changes to the HP-section.
25
2.5 clone crossover, version 6.
Tweeter level
TGAudio
MLS - Frequency Response
07-04-03 20.48.50
90.0
CLIO
dBSPL
Deg
80.0
108.0
70.0
36.0
60.0
-36.0
50.0
-108.0
40.0
200
-180.0
1k
10k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 3.38ms
red=4R7
Fig. 4. crossover, v6.
(I’m sorry for having to call this version #6, but there
have been a number of designs in between, and I
have to keep track of all changes).
Here’s a graphic presentation for bi-wiring:
Fig.5, crossover, v6
180.0
blue=5R6
Stop 6.04ms
Hz
20k
FreqLO 376.47Hz
green=8R2
Fig.6. Tweeter series resistor, red = 5R6, blue = 6R8,
green = 8R2. I use 8R2.
The choice is yours.
26
New tweeter for the 2.5 clone
After introducing the 2 kHz notch filter to the
original crossover (v2) design and also introducing a
slightly modified filter (v3) in order to smooth
frequency and phase response in the upper midrange,
still people have been complaining about the sibilant
nature of the upper registers. I have defended the
8513 tweeter, being such a proven design, and have
hesitated to make any changes to the tweeter as this
would most certainly for good take us away from the
ProAc Response 2.5 sound with its strengths and
weaknesses.
However, I cannot ignore the fact that a number of
my recordings linger on my CD-shelf as long as the
clones are in place in my living room, this mostly
being records of vocal music.
But first a short story on the tweaks that have been
conducted in order to get to the decision of
introducing a new tweeter.
DAMAR coating
A series of near field measurements of the 8535 were
done in order to localize cone break-up and not
surprisingly the dust cap is responsible for some
serious cone break-ups that create a significant bump
at 2 kHz (fig. 1) (same place as the bump created by
the crossover).
Two layers of DAMAR coating were applied to the
center dome and this to some extent smoothed the
frequency response in the upper midrange (fig. 2) and
also above 10 kHz. Subjectively this had a positive
effect on the overall perceived sound.
Damping the 8535 Dust Cap (diyaudio.com), Darryl
Nixon and Troels Gravesen.
Recent experiments by Troels Gravesen have
demonstrated that there are advantages in applying
damping to the dust cap of the clone's 8535 mid-bass
driver. Troels has been working on the resonance
problems of the 8535 which he found has "a major
intrinsic bump at 3 KHz". In Troels' words, “. . . the
coating seems to remove some edginess in the
midrange with a more smooth performance and
tolerance towards difficult recordings".
The substance used is Damar varnish, which can be
obtained from artists' supply shops. The picture
attached is from Troels and is of Damar as sold in
Denmark. The following is reported with Troels'
permission, together with quotes from his e-mails to
me.
"As a start you may apply a coating until the dust cap
is soaked and leave it there as long as it is not
applied outside the dust cap. The effect should be
there in a couple of hours . . .
"At the beginning of applying the DAMAR the
somewhat porous dust cap readily absorbs the
varnish and I continued to apply DAMAR until the
surface appeared shining. This doesn't mean 'flooded'
with liquid, so 'soaked' may be a little overstated.
Actually the amount of DAMAR applied is moderate.
I should have applied it in mikrolitre quantities to
give recommendations. However, after drying the
application is hardly visible. After 1 hour I repeated
the application with a final coat of 'less than first
time'. After 1 week I don't measure altered
performance, so I guess the treatment is stable over
time. If the coating is to be removed the dust cap is
soaked with turpentine and absorbed with Kleenex
tissue."
The varnish sold under the "Damar" brand name in
my own country is produced by the company Art
Spectrum, and the 100mL bottle I obtained looks
physically different. Also, the consistency of the
substance is obviously thicker than that sold in
Denmark. Applying it as Troels recommended did not
produce the same visual results he described. The
varnish did not really soak into the dust cap as I
applied it, but produced a shiny appearance from the
outset. Nevertheless, I went ahead and applied a
moderate amount. It took several hours to dry,
though it remained slightly sticky in places even 18
hours later. (Mind you, it had been raining here for
several days, so that may explain the drying time.) It
eventually soaked in to a large extent, though there
were still some shiny patches. I reported this to
Troels and he recommended the following:
"If your Damar batch seems to be rather thick I'd
hesitate to apply a second layer of coating. Maybe
one additional layer at the 'center of the centerdome',
like 2 cm diameter. Uneven distribution of coating is
usually a good thing in disturbing resonances."
My listening tests produced similar results to Troels'.
There is a small but definite reduction in midrange
edginess, giving a slightly cleaner sound in what I
consider to be the clone's main problem area. This
benefits "difficult" recordings in particular, so if you
are troubled by the clones' midrange this is a highly
recommended mod. Just don't expect miracles! The
effect is subtle.
The important thing is that you don't apply too much
(though the coating is reasonably easy to remove
with turpentine if you do) – and that you DON'T get
any on the cone itself. (Troels did try damping the
cone with Damar, but the results were very negative.)
27
TGAudio
MLS - Frequency Response
hesitate to use this in some other construction.
This did not – much to my surprise – in any way
change the sibilant nature of the highs! After this
there was only one thing left to do: ‘Thanks to the
8513 tweeter for all the hours we have spent together,
but out you go!’
Having a pair of ScanSpeak 9500s, this was an
obvious choice for a new pair of tweeters.
I have removed the magnetic oil in the voice coil gap
of the 9500s. Otherwise no tweaks.
21-02-03 19.36.42
110.0
180.0
CLIO
dBSPL
Deg
100.0
108.0
90.0
36.0
80.0
-36.0
Construction of a new crossover
70.0
-108.0
60.0
200
-180.0
1k
10k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 1.45ms
Stop 5.04ms
Hz
20k
FreqLO 278.26Hz
File: 8535-1.8mH.mls
Fig.1, red = 8535 in cabinet, no crossover. Blue = 1.8
mH in series with 8535. No smoothing.
TGAudio
MLS - Frequency Response
30-03-03 17.16.19
90.0
CLIO
dBSPL
180.0
LP-section: You can reuse most of your components
from the v3 crossover in this new filter. The 1.8 and
0.47 mH coils are the same. The capacitor has been
raised to 8.3 uF (6.8+1.5) and a 2R2 resistor has been
added to the capacitor giving a smooth roll-off for the
8535. The point of crossover is intended to be around
3 kHz, as I’m now confident that the 8535 will do
well all the way to this point and I want to maintain
the 8535 handling as much of the midrange as
possible.
Deg
80.0
TGAudio
108.0
MLS - Frequency Response
30-03-03 17.42.23
90.0
CLIO
dBSPL
70.0
180.0
Deg
36.0
80.0
108.0
70.0
36.0
60.0
-36.0
50.0
-108.0
60.0
-36.0
-180.0
50.0
-108.0
40.0
200
1k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 2.91ms
10k
Stop 6.04ms
Hz
20k
FreqLO 320.00Hz
Fig.2. Red = 8535 after DAMAR coating, no filter.
Blue = 8535 after DAMAR coating + 1.8 mH.
Identifying the source of sibilance
First the clones were cut off below 100 Hz by a 6 dB
filter and supplemented by a subwoofer in order to
significantly reduce cone movement and ease the
burden put on the 8535 by having to reproduce
everything from 30 Hz to 3 kHz.
This did not in any way reduce the sibilant nature of
the highs. Excessive cone movement does not seem
to be a severe limiting factor for the 8535 in order to
truthfully reproduce the sensitive midrange except
when played at very high level.
Secondly a 3-way construction was tried introducing
a Vifa PL11MH coated midrange at 500–3000 Hz.
This is indeed a very good midrange and I wouldn’t
40.0
200
-180.0
1k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.95ms
10k
Stop 6.23ms
Hz
20k
FreqLO 304.76Hz
Fig. 3. 8535 roll-off with various filters:
Red = 1.8 mH + 8.3 uF + 0.83 mH
Blue = 1.8 mH + 8.3 uF + 0.47 mH
Green = 1.8 mH + (8.3 uF+2R2) + 0.47 mH. All 0.33
oct. smoothing.
The basic 3rd order crossover topology is maintained
in order to give best possible phase response in the
crossover region.
As can be seen, the need for the 2 kHz notch filter is
eliminated by this approach.
28
HP-section:
Not much to say about this part. No problem in
making the 9500 roll off at 3 kHz. See
schematics, fig. 4 and response curves fig. 5 and
6.
TGAudio
MLS - Frequency Response
30-03-03 17.35.07
90.0
CLIO
dBSPL
180.0
Deg
80.0
108.0
70.0
36.0
60.0
-36.0
50.0
-108.0
40.0
200
-180.0
1k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.95ms
10k
Stop 6.23ms
Hz
20k
FreqLO 304.76Hz
Fig. 6. Roll-off of both drivers with new filter.
Graphic presentation of new crossover for bi-wiring:
Fig. 4. Crossover schematics for 8535+9500.
Fig. 5 displays the frequency response from the
drivers with the new filter and with same polarity a
dip is seen at crossover frequency.
TGAudio
MLS - Frequency Response
30-03-03 17.47.34
90.0
CLIO
dBSPL
180.0
Deg
80.0
108.0
70.0
36.0
60.0
-36.0
50.0
-108.0
40.0
200
-180.0
1k
CH A dBSPL 1/3 Octave 51.2kHz 16K Rectangular Start 2.95ms
10k
Stop 6.23ms
Hz
20k
FreqLO 304.76Hz
Fig. 5. Red = frequency response with inverted
polarity, 0.33 oct. smoothing. Green = same polarity,
0.33 oct. smoothing. Blue = same polarity, no
smoothing. All measurements performed at tweeter
height, 1 meter distance.
I’m quite sure tweeter level will be an issue and the
2R2 can be changed from 1–2.2 ohms resistance
without affecting point of crossover.
Fig. 7. Crossover schematic, layout for bi-wiring.
29
Sonic evaluation of modified 2.5 clone
The 9500s have the ability to bring forward the best
qualities in the 8535s and the word that first comes to
my mind is coherence. To my ears the 8535 has a
kind of old-fashioned full-range sound, yet in a
completely other league than the old PHILIPS 9710
‘full’ranger’ or the like.
It has its ‘virtues’ in terms of a rather robust
midrange that is quite demanding on your choice of
recordings. It is rather merciless on poor recordings
and inadequate electronics and will probably always
be so.
With the new tweeter in place the degree of
transparency rises considerably and we know how
much the low end adds to the sense of transparency,
and the 8535 has that ability, so we are close to
getting it all from this modest two-way floorstander.
Quite amazing. The new design appears to give a
slightly more distant perspective and for sure the
sibilant, whizzer sound is gone.
I think that the elimination of the notch filter by
redesigning the LP section does a great deal to
enhance transparency. Notch filters can ‘solve’ acute
problems, but I still have the feeling they can add
some obscure/subtle phasiness to the region affected.
Looking at the CSD data from the region where the
notch filter works, it looks like we have to look over
a hilltop to spot the start of the transient, meaning
that despite having an apparent flat frequency
response it seems as if the energy is slightly delayed
(page 6, latest v5) in the region affected by the notch
filter.
Ideally we want only the forefront of the sound wave
to hit the ear followed by an immediate decay within
the first 0.5 milliseconds.
The 9500s seem to have a slightly recessed high end
(> 10kHz) compared to the 8513s, despite having a
very flat frequency response, and I believe this is a
very common observed phenomenon with most 1”
soft-domes.
2.5 clone with ScanSpeak D2905-9500 tweeter.
30
The Final 2.5 Clone, the “sibilance” problem
1st WARNING:
I have recently (May 2003) acquired my third pair of
18W8535-00 drivers, and much to my surprise these
drivers were heavily coated on the rear of the
membrane.
These drivers were meant for a three-way
construction so they perform as expected, but for
those who buy this new batch of drivers from
ScanSpeak it appears that they will not perform in
accordance with all the material that has been
published until now regarding the 2.5 clone.
Due to the coating they will have an earlier roll-off
characteristic and will require modifications to the
crossover.
2nd WARNING
The tweaks suggested in the following paper deal
with the D2010-8513 tweeter.
You will have to dismantle the driver and –
• remove the ferrofluid
• damp the pole piece
• coat the membrane with DAMAR resin
• if you haven’t coated the dust cap on the
8535 driver you will have to perform this
operation also
• do minor modification to the V6-crossover
If you feel uncomfortable with finer mechanics
you may ruin your 8513 tweeters.
The tweeter is a delicate construction, but with
proper care you can easily dismantle the
construction and perform the tweaks.
The reason for these tweaks is sibilance:
Definition:
Sibilance: “Essy”. Exaggerated “s” and “sh” sounds
in singing, caused by rise in the response around 6–
10 kHz.
See: http://www.linkwitzlab.com/images/graphics/sdqulty.gif
Before proceeding I have to thank Darryl Nixon,
Australia, for an extensive mail exchange on the
phenomenon of sibilance and in particular the less
than appropriate performance of the 8513 tweeter.
The phenomenon characterised by the word
“sibilance” has proven more than difficult to deal
with in the case of the 8513 tweeter. If we stick to the
definition literally, we should be able to solve the
problem by adjusting the response in the critical area.
Various attempts have been tried in order to alter the
frequency response in the 4–10 kHz region by
changing the crossover and introduce notch filters,
etc. But none of these changes gave results worth
pursuing. The sound from the tweeter still sounded
awful on a number of especially vocal recordings.
If you make a search on the web on the word
“sibilance” you get quite a number of hits, mostly
aimed at recording engineers on how to avoid
excessive sibilance by choice of microphones or
electronics. You can even buy a “de-esser” piece of
electronics to solve the problem!
From the work done on the 8513 tweeter, it becomes
apparent that what we perceive as sibilance is not
necessarily only derived from excessive response in
certain areas but also from some intrinsic qualities of
the tweeter. Actually the response is quite flat.
I have done numerous comparative tests with the
CLIO measuring system on various tweeters and
found no apparent poorer performance of the 8513
tweeter, so I will not be able to tell you by
measurements why the 8513 tweeter is inferior as is
or why the suggested tweaks make it sound so much
better.
But I’ll stick my neck out and claim a significant
improvement in performance for those discerning
listeners who like vocals, strings and brass
instruments.
Most likely the 8513 tweeter holds some obscure IM
distortion that on poorer recordings makes you hold
your hands to your ears.
The tweaks will change the performance to a level
not far away from the ribbon tweeters that currently
are my reference for best tweeter performance.
At the same time you will have to adjust tweeter level
to produce a frequency response at +/– 1.5 dB from
400–17,000 Hz. That is with the V6.1 crossover with
8R2 or 9R0 to the tweeter.
However, the tweaks will enhance performance from
whatever crossover you may hold.
I have recently heard the real ProAc Response 2.5
speakers and had the impression that the tweeter was
well balanced to the bass driver but the midrange
hardness and relatively poor midrange/tweeter
resolution was much the same as what characterises
the clone. I had a hard time believing that this tweeter
would have ~ 5 ohm series resistor to the tweeter.
With 5 ohm to the tweeter the clone sounds just
awful.
I don’t care how many reviewers have praised the
successful integration of drivers in the Response 2.5
and apparent midrange smoothness. The speakers had
a slightly smeared midrange with less than proper
resolution and image focusing compared to other
speakers and what can be achieved from the
described tweaks. I’m also sure that some will say
“goodbye ProAc sound”, and except for the bass,
that’s just what it’s all about.
31
Modification of ScanSpeak D2010/8513 tweeter
Disassembling the 8513 tweeter
Release the magnet/voice coil assembly by a
gentle twist of a screwdriver between magnet and
housing. Possibly they will just fall apart.
Fig.4
All driver parts.
Fig.1
With a fine-toothed saw the back plate is released
from the driver housing. Saw blade should not cut
deeper than 2–3 mm all the way round in order not to
damage part of the back plate going into the housing.
Cut where the back plate is melted to the housing.
Fig.5
Carefully lift off diaphragm/voice coil from magnet
gap.
Fig.2
Fig.3
Rubber gasket is removed from magnet with a
screwdriver. The gasket is not glued and is easily
removed. Take care not to damage the wires from the
voice coil.
Fig.6
Voice coil gap and diaphragm assembly.
32
Removing ferrofluid
Fig.7
Remove all ferrofluid with a piece of paper.
Carefully wipe off any ferrofluid from the voice coil.
Adding damping pad to the magnet pole piece
Fig.9
Damped pole piece.
Coating of diaphragm
One layer of DAMAR coating is applied to the
diaphragm.
Use a miniature brush and apply DAMAR coating in
thin and smooth strokes from edge towards center of
the dome.
Fig.10
Fig.8
Here in DK you can buy 17 x 3 mm self-adhesive felt
pads that fit exactly to the pole piece.
Punch or cut an 8 mm hole at the center of the felt
pad and press firmly to the pole piece.
Do not apply DAMAR coating to the cone
suspension!
Leave the coated dome for 1 hour at room
temperature and assemble driver.
Coating will be fully dry after approx. 24 hours.
Assemble driver and tighten back plate to housing
with silicone glue. Add mild pressure until glue has
settled, 6 hours.
33
Listening tests
The listening sessions were divided in two parts. First
the tweeter with no ferrofluid + damped pole piece
was compared to the reference tweeter with no
modifications.
The V6-HP section was used before drivers and a
double switch was attached to the tweeters for quick
change between the two. SPL measurements were
performed to ensure same listening level.
Tweeters were placed aside on a carpet with no front
panels and the average frequency response looked
like this:
TGAudio
MLS - Frequency Response
21-05-03 21.08.53
100.0
CLIO
dBSPL
180.0
Deg
90.0
108.0
80.0
36.0
70.0
-36.0
60.0
-108.0
50.0
20
-180.0
100
1k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 1.31ms
red= no ferro fluid
10k
Stop 3.69ms
Hz
20k
FreqLO 419.67Hz
blue= with ferro fluid
Fig.11
A number of CDs with particularly troublesome
sibilance were used for the test.
Listening to a tweeter – solo – above 3.5 kHz at high
level is no pleasure for sure. Listening to a brass band
crescendo very clearly reveals poor tweeters and
improper recordings.
For comparison, tweeters like the SS9000, 9500 and
HIQUPHON OWI were included, where response
levels were corrected via L-pads.
The coated 8513 immediately appeared to have a
darker sound compared to the uncoated, which was a
surprise as frequency response from the two drivers
was within 1 dB.
Numerous measurements (SPL, FFT, step response,
impulse response, cumulative spectral decay, etc.
have been performed. Tweeter intermods were
measured with a 10 and 11 kHz input, where
intermods are at 9, 11, 12 and 19 kHz with all
intermods down to < 50dB below 90 dB reference
level. Just as good as any other well-designed
tweeter.
In no case was there any significant difference
between the two tweeters.
And yet the sound is very different.
The nature of this “darker” sound was first fully
unfolded after having installed the two drivers in the
2.5 clones with the V6-crossover and 8R2 ohm series
resistor to the tweeter giving a smooth response from
300–18000 Hz +/– 1.5 dB and < +/– 0.5 dB between
speakers.
This “darker” sound now appears in full scale as an
improvement in resolving power, transparency and a
more “full-bodied” sound whether being vocal,
piano, brass, etc.
And still in mono-mode the coated dome appeared to
reduce the midrange “hardness” usually associated
with the 8535 driver.
Frequency response of the tested speakers:
TGAudio
Ferrofluid
The removal of the ferrofluid and damping of the
pole piece did not reveal any significant changes to
the modified tweeter. Same sound from both tweeters
with a wide range of musical sources.
Coating
When you for the first time are comparing coated vs
uncoated it is very clear that you now have two quite
different tweeters!
MLS - Frequency Response
CLIO
dBSPL
180.0
Deg
90.0
108.0
80.0
36.0
70.0
-36.0
60.0
-108.0
50.0
200
-180.0
1k
10k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 2.73ms
File: 8535+8513-un-coat-modV6-10R-rev.mls
The reference un-modified 8513 compared to the
modified and other tweeters appears “flat” and with
poor resolution and depth.
Yes, resolution and depth can be heard in a single
speaker/mono setup.
27-05-03 21.00.55
100.0
red="coated"
Stop 5.61ms
Hz
20k
FreqLO 348.30Hz
blue="uncoated"
Fig.12. Un-smoothed response.
Red = 8535 + 8513 coated
Blue = 8535 + 8513 reference (not modified)
For the sake of good order the bass drivers were
swapped and the same tests were repeated with much
the same results.
34
Getting two bass/mid drivers to sound exactly the
same is not an easy task. You may fine-tune the
crossovers to make the drivers perform within 1 dB
and yet they may sound slightly different.
Thus the need to swap drivers and repeat tests.
Going from the “uncoated” to the “coated”
removes a curtain, however trivial this expression
may be, but it very much describes the transformation
that has taken place.
Coating of second tweeter
Fig.14
After drying for 1 hour the tweeter was assembled
and reinstalled in the 2.5 enclosures.
Frequency response was measured after 1.5 hours:
TGAudio
MLS - Frequency Response
27-05-03 22.11.28
90.0
CLIO
dBSPL
Deg
80.0
108.0
70.0
36.0
60.0
-36.0
50.0
-108.0
40.0
200
-180.0
1k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 2.93ms
10k
Stop 6.04ms
Hz
FreqLO 322.01Hz
Fig.15. Frequency response, no smoothing.
Red = newly coated dome
Blue = 72 hour coated dome
Very much same performance as seen in Fig.12.
(colours reversed)
Fig.13. Tweeter test.
(Gaffa tape covering 9500 rebate)
The intriguing thing about the speaker with the
coated dome is that it now sounds slightly darker, yet
having more presence, which I interpret as a better
midrange-tweeter integration.
From the speaker with the uncoated dome, voices
sound as coming primarily from the 8535 with some
poorly focused highs added from the tweeter, where
the speaker with the coated dome seems to “dress”
the midrange with the proper overtones and presents
a more natural tonal balance.
180.0
20k
35
The 2.5 clone now bear little resemblance to where
it all started with the 2 kHz bump, notch filters,
etc.
To summarise the changes:
• Notch filter is omitted by adding a resistor to
the LP section capacitor
• 47 ohm resistor in parallel with the 1.8 mH
coil is removed
• 8535 drivers have had dust caps 3x coated
with DAMAR resin
• Ferrofluid has been removed from tweeter
voice coil gap
• Damping material has been applied to
tweeter pole piece
• Tweeter dome has been given 1 layer
DAMAR coating
Listening tests on the 2.5 clones with both
tweeters coated:
The speakers fitted with the coated 8513 tweeters
now hold some of the same qualities associated with
the 8535+9500 setup.
And the 8535+8513 holds qualities different from the
8535+9500 combination.
The 9500 setup is a very good all-rounder that
handles almost any choice of music and only rarely
leaves the upper midrange behind in making the
“perfect speaker”, that is from a modest two-way
floor-stander.
The dust-cap coated 8535 + coated 8513 have a
slightly more analytical presentation and enhanced
depth presentation compared to the 8535+9500 setup.
The midrange is improved even compared to the
8535+9500 setup. I cannot account for this
improvement, but it’s significant.
I can now enjoy almost all my CDs, even the poorer
vocal recordings.
I’ll reinstall the 8513 tweeters for good and keep
these speakers and thank all the cloners who
commented on the performance and suggested
changes.
And thanks to JPO for taking in yet another paper.
This is the final.
Fig. 16. 2.5 clone crossover, version 6.1.
Regards
Troels Gravesen
[email protected]
PS. DAMAR coating of the 8535 dust cap:
Apply 3 layers of DAMAR coating to the dust cap
with at least two-hour intervals.
And don’t apply coating outside the dust cap as it
changes the frequency response around 3–4 kHz.
36
Comments to tweeter modifications
1.
First of all your tweeter mod is excellent, the
sibilance has really gone and the sound is much
smoother than before.
Zoltan.
2.
This is quite a delicate operation! The only thing I
found that was not in your instructions was that after
the rubber gasket is removed, it is necessary to tap
the plastic casing on the side to allow the driver parts
to fall free (at least it was in my case). It confused me
for a few minutes until I figured it out. Aside from
that, everything went well. I did both drivers at the
same time and was keen to hear the result, so I've
temporarily attached the back plate to the housing
with electrical tape - don't worry, it's airtight and I'll
redo it properly with silicone tomorrow.
Well, all I can say is that together with the minor
crossover mods and extra Damar coating on the 8535,
this makes the most significant improvement since
your original notch filter. Everything you say in your
paper is 100 per cent correct - the tweeter coating
really does transform the speaker. The last traces of
the original clone's "technicolour" sound are gone, so
I guess it's not going to please everybody, but
sibilants now sound totally natural, even if they are
still a bit more pronounced than in some other
designs. The important thing is that they don't "ring"
and they don't irritate. I also agree with your
description of the treble as "darker", and yet - you're
right - at the same time there's more presence to the
sound! I think we've been unfair to this tweeter; it
had far more potential than we were allowing for.
There is now obviously far better integration between
the two drivers. It's very strange, but I have to agree
with you also that the midrange "hardness" is now
much reduced - or at least much less obvious, and it's
possible to hear depth layering where it didn't exist
before. I have yet to listen to all my "difficult"
recordings, but the few I have tried all showed a
marked reduction in "listener fatigue". I can enjoy
most of them again! I can't say I always find the
"carbon-fibre" midrange to my taste, but I can now
appreciate it for what it does. The clone truly does
now fit the description of a "high-end" speaker. Very
impressive.
Darryl.
3.
I have been briefly comparing Troel's modified 8513
tweeter (ferrofliud removed an applied damar coating
to cone) and it is very nice. I must modify the woofer
crossover to Troel's suggestions before I make any
final conclusions but dropped it into the current
crossover and it sounds more neutral, less sibilant
when CD tracks have too much energy. The same
basic sound character is retained but enough treble
zing is shaved off for those who want this, IMO. At
this point in time I still prefer the standard
unmodified tweeter as there is more of that original
Proac lushness, sweetness and detail, but can be
sibilant on some tracks, which seems to be the tradeoff.
Al.M
4.a
- modifiserte 8513 i går, pluss at jeg satte inn en
10ohms motstand i mitt No1. Filter. Har forresten
bestilt deler til No6. filteret i dag.(gleder meg)
Resultatet av tweakingen er enorm! Må bare takke!
b.
After many more ours of listening to the clones and
the V6 filter, I have come to what will be MY final
filter. Sorry, but I was a little to fast to accept the
filter for the 8535.
The V6 filter sort of shows the true nature of the
8535, a hard sound.......
But with the V1 filter too the 8535, and the latest (in
the link) to the 8513 + the modifications with the
tweeter, the sound is the most acceptable ever. I can
still hear the 2000Hz bump but then mostly when I
listen to TV. The sound on most recordings now is
warmer than with the V6 filter. But now I can for the
first time totally relax when listen to my favourites.
c.
Dear cloners
I really recommend this modification of the 8513
tweeter. I can promise a sweeter treble. And the
metallic sound is totally gone! In fact the whole
speaker can be as fantastic as people said, from
bottom too the top. I have had big problems
accepting the clone`s sound until now. This tweak
together with the v6 filter and I can call this speaker
superior! I agree with Troels that this is the final
version!
d.
There is said many strong words about the sound of
the clone, but I cant understand how so many people
have accepted the sound of the early clones! I can
imagine Mr. Steward Tyler having a good laugh more
than once! But now He he he.........
Regards from Tor Hauge...Norway
37
b. LCR notch filter for the 750 Hz bump = 10 mH
(2 x 4.7 mH, 0.65 ohm) + 10R + 6.8uF.
The new coated 8535 drivers from ScanSpeak
As mentioned in my former file on tweeter coating,
the new 8535 drivers have a coating on the rear of the
cone. This is a deviation from former 8535 drivers
and is easily detected on measuring performance.
TGAudio
MLS - Frequency Response
HP-section :
No changes.
26-06-03 22.00.12
90.0
CLIO
dBSPL
180.0
Deg
80.0
108.0
70.0
36.0
60.0
-36.0
50.0
-108.0
40.0
200
-180.0
1k
10k
CH A dBSPL 1/12 Octave 51.2kHz 16K Rectangular Start 2.66ms
Stop 5.86ms
Hz
20k
FreqLO 312.20Hz
Fig.3. V6.2
Fig.1. New coated 8535 and V6.1 crossover.
What can be seen from this graph is a 3-4 dB bump at
600-900 Hz, and due to an earlier roll-off
characteristic, a dip at 3 kHz.
Here’s the response from the old drivers with
modified tweeter, crossover V6.1:
TGAudio
MLS - Frequency Response
TGAudio
MLS - Frequency Response
19-07-03 22.05.18
90.0
CLIO
dBSPL
180.0
Deg
80.0
108.0
70.0
36.0
60.0
-36.0
50.0
-108.0
27-05-03 22.09.01
100.0
CLIO
dBSPL
180.0
Deg
90.0
108.0
80.0
36.0
40.0
200
-180.0
1k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 2.46ms
10k
Stop 6.02ms
Hz
20k
FreqLO 281.32Hz
File: V6.2-7.8+4R7.mls
70.0
-36.0
60.0
-108.0
50.0
200
Fig.4. Frequency response of new crossover.
LCR components:
-180.0
1k
CH A dBSPL Unsmoothed 51.2kHz 16K Rectangular Start 2.93ms
10k
Stop 6.04ms
Hz
20k
FreqLO 322.01Hz
File: 1st coat 72 h.mls
Fig. 2. Old 8535 driver response. Blue and red are
various measurements during tweeter coating.
It’s my feeling, that these drivers will be best suited
with tweeters like ScanSpeak 9300/9500/9700 and a
target point of crossover at 2-2.5 kHz, possibly 612/18 dB.
Furthermore, a notch filter is needed for the 750 Hz
bump.
To my knowledge, this change in performance has
not been announced by ScanSpeak and can only be
identified by measurements or visual inspection.
I have tried to tweak the V6.1 crossover to adapt to
these new drivers:
LP-section:
a. 1.8 mH-(7.8uF+4R7)-0.47 mH
Fig. 5. mid-LCR.
- And don’t spend a fortune on notch filters unless
you think this is the speaker of your life!
This shouldn’t cost you more than 20 US$ in all for
two.
38
The Scan-Speak 18W/8535-00 has always been
coated both front and back, but as it is an air dried
cone, its absorbency may differ. We use the same
amount of coating glue on each unit so the cone may
sometimes absorb all the coating glue (i.e. the coating
is hardly visible), and sometimes the cone is saturated
after absorbing 80-90% of the glue. The rest of the
glue will then form a somewhat shiny surface,
leaving the coating visible. I hope this answers your
question.
Best regards,
Danish Sound Technology A/S
Carina Sondergaard
Export Assistant
Fig. 6. “New” coated 8535
Fig. 7. “Old” un-coated 8535
Conclusion
Listening to the V6.1 crossover with the new drivers
doesn’t sound that awful.
Compared to the old drivers, we have a midrange
with some more presence and for sure we are more
forgiving to dips (3 kHz) than bumps in the
frequency response (remember the series filter?).
With the V6.2 filter, the dip at 3 kHz is slightly
reduced and you can ad the mid LCR if you think you
have too much midrange presence. With this notch
filter in place, the midrange is much like the old
drivers (and there is not going to be an easy substitute
for this tweak).
I don’t have two pairs of 2.5 clones to compare, but I
would have a hard time pointing out one from the
other.
Or did I hear some more background details (talking)
in the midrange from the new, coated drivers on the
Jazz at the Pawnshop that I haven’t heard before????
13.08.03: NEWS on the “new” drivers from Scan
Speak:
Finally there’s some explanation from Scan Speak
regarding these drivers:
Alain Letendre, Canada, has received this message
from Madisound:
From d-s-t here in Denmark I have received this
message (in Danish):
Hej Troels.
De variationer i udseendet, som dine billeder viser er
ganske normale. Der er ikke lavet tilsigtede
ændringer på denne enhed siden dens "fødsel".
18W/8535-00 er og har altid været coated på både
for- og bagside. Ofte er det dog ikke synligt på
bagsiden, men der kan opstå skinnende pletter, som
dit billede viser.
Membranen er luft-tørret (til forskel fra en kontakttørret membran som anvendes på fx 15W). Den
lufttørrede membran suger/absorbere coate-limen
(som er vand-fortyndet).
Hvorfor så forskellene? Som følge af fremstillingsprocessen forekommer der variationer i den
lufttørrede membrans indre struktur, mens vægten er
let at holde (det er den, der er specificeret med
tolerancer). Det er lufttørringen, som giver det
specielle udseende på forsiden. Nogle gange vil
membranen være lidt mere kompakt end andre gange,
og det påvirker suge-evnen. Mængden af påført
coate-lim altid er den samme (vi kontrollerer desuden
viskositet). Den er doseret til næsten at mætte
membranen. Derfor vil membranen nogle gange
absorbere al limen, så den ikke synligt er coated på
bagsiden (som trænet er det let at se, at enheden på
dit første billede ER coated på bagsiden. Det ville du
også kunne se, hvis du havde en u-coated membran).
Andre gange vil membranen i områder lige nå at
blive mættet, før al coate-limen er påført. Den
resterende mængde coate-lim vil derfor lægge sig
uden på membranen. Jeg håber, det har kastet lidt lys
over sagen. Mvh. Ulrik Schmidt, udviklings-ingeniør
This message says much the same thing as the
message from Madisound.
So, this is production variations and my only
comment to this is:
BEST OF LUCK CLONERS!
Your will probably not be able to determine what
type of crossover will be best suited for your
drivers without measuring equipment.
39
40
New project to come:
“The Point75”: 8535 bass +SEAS W15CY001 di-pole midrange + ribbon tweeter.
Regards
[email protected]
41
Notes on the ProAc sound:
Much has been said on the ProAc sound, and this can
be found in all the reviews seen in HIFI magazines,
etc. If we take a look at the frequency response
curves from a number of ProAc Response designs,
this is what we see:
Response 1S
Response 1SC
Response 1SE
Response 2.5
Response 3.5
Response 3.8
Response 2S
The Danish magazine High Fidelity confirms the
response curves seen from the ProAc 1S and 1SC.
The Response 2S is an exception from this profile
and the Response 3.5 is the worst of them all.
Worst??
Well, first let’s take a look at 10 randomly picked
designs from the Stereophile files:
(A few speakers with very peculiar response profiles
have not been considered)
42
Totem Acoustic Mani2
KEF RDM
Dynaudio Confidence C4
Revel Salon
JMlab, Chorus 706
Joseph RM 33si
Spendor S3-5se
Vandensteen 2ce
43
audience, and the easiest way to do this, is pushing
the “loudness” button = “west-coast-sound”.
“Cheap-trick-low-fi”, that’s what it is.
Thus the need to re-evaluate the crossover design if
you want to listen to acoustic instruments and voices
and have a more natural presentation of an acoustic
event.
KEF Reference 207
The speakers chosen here are from small monitors to
large floorstanders.
Try to zoom out on our PDF reader and watch the
two pages simultaneously.
Few of these nine speakers have some or a slightly
elevated bass response, but none of these have the
persistent increase in bass and treble response as seen
from the ProAc speakers, combined with a generally
recessed midrange. Even the Response 2.5, where we
can see the 2 kHz midrange bump following an 800
Hz bump.
It is apparent, that the Response series was created
with the intention of maintaining a specific response
profile, thus giving the “ProAc” sound to them all,
where the main differences are bass performance,
logically derived from the size of the cabinet and bass
drivers.
This is in no way a new approach and can be heard
from a number of other manufacturers, where the
main differences in “sound”, relates to bass
extension.
The Response 2.5 800 Hz bump, by the way, could
be the explanation for the acceptance of the 2 kHz
bump in the original design.
The “sum” of these two bumps may kind of even out
the overall perceived balance.
(Actually we can now get both of these bumps with
the new coated drivers from SS. Leave out (or
reduce) the series resistor in the LP section, and leave
out the LCR notch filter for removing the 800 Hz
bump + reduce tweeter series resistor to 5 ohm!
How about that? Finally we can recreate the original
ProAc Response 2.5 sound/profile with all the
bumps).
In the history of HIFI, response curves displaying
excessive bass and treble at the expense of a recessed
midrange, was the trademark of the so called “West
Coast” sound. Lots of bass and lots of treble.
Excellent for “Surfin’ USA” by The Beach Boys.
I remember JBL having a 14” bass driver (LE14)
married to a small paper cone tweeter. This was a
“tizz and boom” speaker!
If you have to make a living of producing
loudspeakers, you have to catch the attention of your
The ProAc Response designs are nowhere near the
old “tizz-and boom”/”west-coast-sound”/”boom and
tweet” sound, but displays a specific profile, aimed at
creating an “engaging” sound. The objective of
neutrality appears to be somewhat compromised.
Quote, Martin Colloms:
Subjective effects of first-order errors
“Slight errors in channel balance, either in specific
frequency ranges or in overall level, can subtly
disturb one's opinion of the sharpness of stereo focus.
Statistically well-controlled testing has not only
confirmed the audibility of absolute phase/polarity
but also that of level differences as little as 0.2dB.
These differences may be of octave or several-octave
bandwidth, with a sensitivity of a similar magnitude.
The subjective responses to variations in
amplitude/frequency response are pretty well
documented; the careful reviewer bears these
constantly in mind. For example, less than 0.5dB—
5%—of treble lift in the 3-10kHz range can give rise
to a mildly increased sense of immediacy,
transparency, and liveliness without necessarily being
directly obvious as treble lift. A similar degree of loss
in the 150Hz-400Hz range can make a vocalist
appear lightweight and lacking in power in the
fundamental range, lending a crisper quality to the
sound. This might be preferred on one recording but
disliked on another”.
With the variation in specification of drivers and
crossover components + cabinet construction and
damping, not to mention the front-end consisting of
amps and players, it’s no wonder we get different
reports on sonic performance of the 2.5 clone.
It’s my feeling, that if you choose to target the
original response profile, you’re in for some serious
tweaking on amps, cables, CD players, etc., to get it
right.
If you’re in for the more British school of
loudspeakers, represented by names like Rogers,
Harbeth, KEF, Spendor, etc., and target a more flat
response profile, you will have a speaker, that will
tolerate a larger range of amplifiers and CD players.
And to my ears be more “high-fidelity” compared to
the more immediately appealing ProAc sound.
On the various schools of speaker designs, please
read: http://www.aloha-audio.com/library/speakerdesign1.html