Semi Active Three Way Speaker Project

[Qwin]

I've never been happy with the three way speakers I have owned in the past, or, baring a couple of exceptions, those that I have heard. I found them sibilant, which I partly put down to the crossover point for mid/ tweeter being generally at 5 to 7kHz. This is at a very sensitive point for reproduction of human voice, as this is where sibilance lies and I was guessing that it might be part of the problem. Two way systems seem to suffer less from sibilance and I attribute this to the crossover point being much lower, at around 3kHz. One afternoon, I tried an experiment and sat an old pair of Tannoy M20 bookshelf speakers on top of my now active Yamaha NS-1000M speakers. I used a 2-way active filter I had, which crossed at 300Hz and fed the 12" Yamaha bass driver the low frequencies and the Tannoys got the higher frequencies, I kept the Tannoys passive crossover in place for mid/upper frequency separation. I was amazed at the results, the Tannoys were great at voices, but now they had a bottom end. This discovery is what kick started this project, my reasoning, when trying to predict what causes the sibilance, could be completely off track, but the results certainly aren't. I started to search for a pair of modern bookshelf speakers that were good with voices and which I could mate with a 12" Woofer or Sub driver. My idea was to use the drivers and crossover from the bookshelf speaker as part of a 3-way Semi Active Project (SAP), in a large sealed enclosure.



After a lot of searching I settled on the Quad S-1 bookshelf speakers shown above, they got good reviews and I liked them when auditioned, preferring the way they handled midrange, to their larger Quad S-2 sibling. I took them to the MiBO meet and I thought they sounded pretty good in the mid range and top end, with an obvious limitation being the 4” mid/bass for the bottom end. They should do the job for mid and top end and I'll cross over to the 12” bass at 310Hz.

The specifications are as follows:

Description:...................2way Ported
Mid/Bass Driver:.............4" (100mm) Woven Kevlar No D-1100
Treble Driver:.................12 x 45mm True Ribbon Model - Fountek NeoX 1.0
Sensitivity:....................84dB (2.83v@1m)
Peak SPL:......................96dB
Amplifier Power:..............25 - 100W
Nominal Impedance:........8 Ohm
Response:.....................58Hz - 20kHz (+/-3dB)
...................................(Tweeter is actually flat to 40kHz on Fountek's response plot)
X-Over Point:.................3.2kHz
Cabinet Volume:..............6.65L Gross 6.0L Net, Completely filled with stuffing
Cabinet Walls:................16mm
Dimensions:...................H285 x W156 x D240mm (Outside)




The plan, is to incorporate the Quad drivers and passive crossover, into a sealed chamber within a much larger cabinet, designed for a 12" Woofer. An active x-over will split the signal at around 300Hz and separate amplifiers will power the Bass Driver and the assembly from the Quads. The Quads are a 6.7L ported design, so the T&S parameters of this small mid/bass driver are needed, to calculate the volume of a suitable sealed chamber.
These results were measured for the two drivers.

..........................A....................... ...............................B
Piston Dia:.........83mm................................. ................83mm
SPL:.................84.78dB 1w/1m.....................................84.99 1w/1m
Re:...................3.389 ohm..........................................3.419 ohm
Fs:...................80.08 Hz.............................................82. 1 Hz
Qts:.................0.583........................ ..........................0.583
Qes:.................0.716........................ ..........................0.737
Qms:................3.154......................... .........................2.805
Le:...................0.175 mH (10k).....................................0.176 mH (10k)
M(ms):..............6.00g......................... .........................5.67g
Vas:..................2.707 L...............................................2. 718 L


Using the figures for "A" and maintaining a Qtc of 0.707 this calculates as a Net sealed box volume of 5.8L (without filling) with an F3 of 97.1Hz. This is more than an Octave below my planed crossover frequency and with the active filter having a steep slope (24dB/Octave), this should work well. A sealed box of 4.0L will give a Qtc of 0.755 and an F3 of 97.6Hz. The higher Qtc of the smaller chamber could possibly benefit the mid range, conversely a well stuffed 4 litres will give a virtual volume of up to 4.8L, so there is a lot of scope for manoeuvre with the design of the sealed chamber.

A screen shot of the impedance sweep + T&S parameters for driver "A", can be seen below.








The impedance plot and T&S parameters for The Quad D-1100 Mid/bass, Driver "A".
Calculated using the Dayton Audio Test System (DATS) software.








The crossover boards revealed fairly ordinary parts. Chinese MKP caps, ceramic resistors and an Iron cored coil in the bass filter. Better quality parts should help here, so I need to draw the circuit schematic and determine what changes will be made. The Capacitor and Resistor values are printed on the parts, I emailed Quad UK and they were very helpful in supplying the Coil values. All parts were measured from the two boards anyway to check how close they matched the stated values, the coils were very close, some caps were and some were not, resistors were also pretty close.



Stock Crossover Schematic
http://www.jkwynn.co.uk/Project_Imag..._crossover.pdf


Modified Crossover Schematic
http://www.jkwynn.co.uk/Project_Imag...ver_layout.pdf







The proposed crossover layout and parts are illustrated above.
The Ceramic resistors have been upgraded to Mills wire wound and from 5W to 12W.
The Chinese MKP caps have been replaced with Ansar Supersound MKP.
The Iron core coil has been replaced by a Jantzen Air Core.
The values are unaltered, but the better quality parts are closer to the required figures.
The Bi wire facility has been removed and instead, a Star Grounding scheme has been applied.
These basic upgrades wont create a night and day difference in sound quality, which was already very good. There was however, a slight coarseness/grain audible in the mids/lower treble, the new parts will hopefully remove this.
I suspect the ceramic resistors and cheap caps in the signal path were the main cause of this.

I already have a pair of 1R5 mills resistors and a pair of 6.8uF Ansar Supersound's from my parts bin, these will help reduce the cost of the upgrade.







The passive crossovers have been completed.
The large coils used were described as 1mH/0.25ohm, I measured them as 1.036mH/0.215ohm.
This worked to my advantage, as when I wound some turns off, to get my desired 0.95mH, it also gave me the exact rdc value I required of 0.2ohm.







This is my proposal for the 3-way layout. The Quad drivers in their sealed chamber are paired with the 12" Scanspeak 30W/4558T00. This driver is classed as a Subwoofer by Scanspeak, yet some dealers list it as a Woofer. This is because it has a usable range up to 1000Hz, this is way beyond what most true sub drivers achieve, they tend to roll off at a much lower frequency. This is also one and a half octaves above my planned crossover frequency, which is pretty much ideal.
My Yamaha NS-1000M have an F3 of 53Hz in a 50L sealed box. The Scanspeak driver calculates (Qtc 0.707) with F3 of 38Hz in a similar sealed volume. If I keep the foot print dimensions the same, I'll be able to use the 300mm stands I made for the Yamaha's with the new speakers.







This is my proposed layout for the active crossover/power amp, with its four channels of amplification using Hypex ucD180 modules (180w/4ohm 120w/8ohm). I have kept Left and Right separated, each side having its own Switch Mode Power Supply. The active "KMTech" crossover board for each side is powered by its SMPS aux unregulated output, so a small adjustable regulator board has been inserted for each crossover, to supply the required +/- 12Vdc. I have made provision for an adjustment pot on the Bass output of the crossover to match the bass driver to the mid/tweeter level. The mid/tweeter are 8ohm and only 84dB/1w/1m sensitivity, were as the Bass driver is 4ohm and 89dB/1w/1m. Using identical amp modules, the bass driver should have more output, but it requires more power than a mid/tweeter anyway, my gut feeling is it will still have too much and will require just a small amount of attenuation to match the less efficient mid/tweeter set up. We will have to see if that works out correctly. AC/DC/Signal lines have been kept apart to reduce the chances of interference.
The whole thing fits into a 2U x 300 19" rack case.







I've completed the lash up from 18mm chipboard, with 18mm MDF baffle. It will test the new crossover and sealed chamber, the chamber is somewhat smaller than the original ported box. I fully stuffed the chamber with fibre glass loft insulation and mounted the crossover on the back for easy access and to duplicate what will happen on the final build. I made a wide baffle to match the width of the design proposed and give a good representation of what it will sound like.







The baffle is also the same width as my Yamaha NS-1000M. With the Yamaha's turned upside down and the lash up placed on top, the drivers are spaced similar to my proposal. I have each of the drivers in my active Yamaha's wired separately to its own set of binding posts, without an internal crossover, so the Yamaha's bass driver is borrowed here for the sound test.







For the first sound test, I'm using my multi channel Nakamichi power Amp. I'm also modifying to suit, my test bed (analogue) active crossover, pictured above. I've swapped out the unregulated KMTech power supply board, for an adjustable regulated PSU from China. I replaced the Chinese caps with Panasonic FC but changed nothing else on the board, these PSU boards are to be used on the final build, so it's a good opportunity to see how they perform.







I made a few changes to the KMTech crossover boards, adding bypass caps on the underside, a 0.1uF Ceramic and a 4.7uF Tantalum in parallel, attached to each power pin on the op amps and run to the ground plane of the board. The Red Wima film/foil caps across the +/- pins were tried, but removed, as they degraded the sound IMO. These boards cross at 310hz and I used 1% metal film resistors and Vishay MKP 1837 1% caps. I also converted the single ended input to balanced, by combining the two halves of the balanced input, using the buffer op amp. I had to cut some of the tracks, add alternative paths point to point and solder a couple of additional parts in place, as can be seen. The boards are single ended, but by creating a balanced input, I can feed them straight from my Pre Amp, which runs balanced outputs, it just simplifies the connectivity and it does run very quite when wired this way.







I used Burson v5 discrete op amps, which I already had, for the Hi and Lo Pass filters, but I changed the buffer to the Sparkos Labs discrete unit pictured above, this mix gives a better tonal balance to my ears. The crossover PCB has pin outs for dual channel op amps throughout, but the buffer makes use of just one channel. The designer did it this way to keep it simple for DIYers and supplies identical op amps for all three locations in their kits, so you can't get it wrong when fitting them. This is a waste of the unused channel at the buffer though and there is a big difference in the price between dual/single channel discrete units. I opted to use a single channel Sparkos unit for the buffer. This complicated things slightly, as the pin outs are different on a single channel unit, so I modified the dip8 socket as shown below to match the single channel op amp to the dual channel pin out of the board. The unused pin locations on the board are not connected to anything, which allowed me to make this change fairly easily.







The modifications to the dip8 sockets are on the underside, so you don't even see them when fitted to the PCB. I had a nagging concern that the unused channel on the stock set up would cause noise, as it is powered up without any input connected, whether it does or not, the mod removes this possibility.

I have around 80 hours on this set up, running it in. There was an initial change at 4 to 5 hours and another major change at around 65 to 70 hours when the mid range fleshed out and the bass got a decent punch. It's sounding pretty damn good now. I'll give it at least 100 hours before giving a final judgement.