AOS Studio 24

The AOS Studio 24 BE (S24) is a closed box MTM loudspeaker (Photo 1). The S24 is sold as a kit by Axel Oberhage Starnberg (AOS) in Germany. The original AOS design is a “bookshelf” format box. However, free-standing operation of the S24 on a stand is acoustically better than stowing it away in a bookshelf. The S24 uses high-quality driver units made by ScanSpeak.


Photo 1: My AOS Studio 24 BE, which I re-desinged as a full-blown floor-standing loudspeaker

A unique feature of the S24 is that the woofers protrude by 18 mm from the front baffle, whereas the tweeter sits flush with the baffle. Most speaker designs lack this offset, which, due to the recessed woofer cone(s), causes a lag of the sound wave(s) emitted by the woofers relative to that of the tweeter. The additional offset used in the S24 therefore avoids the typical misalignment of the different drivers and therefore allows time-coherent sound emission from the woofers and the tweeter.


Photo 2: Close up of the tweeter and the midwoofers, which are mounted on aluminum rings to achieve an offset relative to the front baffle

I decided to modify the original AOS design a little when I built my S24. In particular, I used a full-height floor-standing enclosure for the S24, revised the crossover filters a bit, and added an active subwoofer. A detailed description of my tests, modifications, and extensions to the S24 are available as a separate PDF document.

Linnenberg UDC1 DAC

Note: I also published this review on Computeraudiophile. udc1_inside2I was looking for a DAC to complete my new stereo system. I use an Apple MacBook with Decibel to play digital music via the DAC. The rest of the system consists of DIY preamplifier (Thel VX-D, battery powered class-A MosFet output), two Hifiakadmie PowerAmps, and AOS Studio 24 BE speakers (top-of-the-line ScanSpeak drivers). Some more specific requirements for my new the DAC were:

  • High-Res (192/24) asynchronoous USB or FireWire input
  • Discrete analog stage would be nice
  • No volume control or other gimmicks required
  • Down-to-earth price! It’s just very wrong to pay $5000 for a $15 DAC chip, a bunch of $0.05 opamps and a $2.50 «audiophile» face plate made of the same material as kitchen aluminium foil. I really don’t want to support this sick part of the audio industry.

I looked at many different DAC offerings, and I found a very nice list of different DACs with asynchronous USB and FireWire inputs. One of the least flashy DACs on this list was the new Linnenberg UDC1, which seemed like a perfect fit to my requirements. Also, Udo Linnenberg of Linnenberg Audio claims that he’s not making a living from his audio «business», so maybe the price-to-value ratio of the UDC1 is better than with other DACs. I gave the UDC1 a shot in in December 2011, when price was hot. Udo Linnenberg also offered a full return if I don’t like the UDC1, so I couldn’t really go wrong. The UDC1 has a built-in M2-Tech USB interface, which accepts up to 192kHz/24Bit (or 32Bit, I’m not sure). The USB interface has its own dedicated power supply, so it doesn’t rely on the noisy USB power from the computer. The DAC chip is a good old Burr Brown PCM1794A, which drives a discreete analog stage. The analog signal is output to symmetric XLR jacks only, so there are no RCA. I therefore had to use XLR-to-RCA adapters to use the UDC1 in my system. No big deal, but RCA jacks would’ve been nice. The power supply is completely built into the box. The lack of a wall wart is nice because I hate wall warts cluttering up the power outlets. But I also like wall warts because they can be replaced by something better (e.g. a DIY power supply), which is often a very easy tweak to considerably improve the performance of audio electronics. The USB interface built into the UDC1 requires a special driver software, even for current Macs. This is a little unusual, and I am a bit afraid of what’s going to happen if the drivers stops working with a newer version of Mac OS X and M2-Tech stops supporting the software for the UDC1. It also means that it might be difficult to make the UDC1 work with Linux, because there is (currently) no Linux driver available (shame on M2-Tech!). Another slight annoyance is the power indicator LED, which is waaaay too bright. The UCD1 could almost light our entire living room with its uncomfortably bright blue LED! Udo Linnenberg instructed me how to reduce the LED brightness by replacing a resistor. I’d be comfortable doing this, as I know where to hold a soldering iron. Others might not. I solved the bright LED issue simply by stuffing the UDC1 away behind my audio rack, somewhere in the middle between the computer and the preamplifier. So, how does it sound? I compared the UDC1 to my Apogee Mini FireWire DAC (hot rodded with a hefty DIY power supply, which replaces the original wall wart) in my system as described above. With the UDC1, the sound was more detached from the speakers. The sound was «juicier» and livelier, whereas the Apogee sounded dryer and calmer. The bass was better resolved with the UDC1 and there was more air in between the instruments and vocals. I’d say the overall score of my digital chain with the UDC1 is about the same as the vinyl chain (although they certainly don’t sound the same). Really good! Although I was not able to compare the UDC1 to the usual $5000+ «reference» DACs with $0.05 opamps behind the face plate, I am pretty sure the sound quality is up there. I therefore don’t know if the UDC1 is a giant killer, but it might well be… But can the UCD1 be improved? Yes. First of all, the LED brightness needs to be reduced, it’s just ridiculously bright. And then I am a big believer in clean supply power. That’s why I use battery power in my phono and line preamps, and an elaborate outboard power supply with a large C-L-C filter for the Apogee Mini Firewire DAC instead of the cheap wall wart. However, doing this with the UDC1 is not trivial, because the power supply is built into the box and sits on the mainboard. And, although the USB interface works very well, I’d really appreciate the freedom to be able to use it without a proprietary driver that might stop working in the future or does not exist for some operating systems.


Note: an earlier version of this text was first published as «The DDDAC1794 is no (ordinary) DAC!» on Computeraudiophile.


Why the heck would anyone buy and build a DDDAC1794? This thing seems very much out of place in the arena USB and FireWire DACs out there, it costs a lot of money, and it requires an intimate relationshop with a soldering iron, too. However, I have been in the DIY world long enough to know that nothing beats a good DIY system.

I had many and very different DACs in the past. My beloved Stokes DIY Tube DAC was restricted to S/PDIF and red-book 44/16 audio. I plunged into computer audio with a not-so-great Headroom USB DAC. Then I hot rodded an Apogee Mini FireWire DAC with a hefty DIY power supply. I use an Audiolab M-DAC in our living-room system (and it tends to break from time to time). My main system had a Linnenberg UDC1. And I’ve listened to the Weiss and many other cost-a-lot stuff. However, while some of those DACs sound pretty good compared to others, they all screw up the music in the same way. And I don’t mean the painful “S” sounds and similar boorishness from crummy DACs. Even the «good» DACs take away the life, flesh and breath from the music, very much in contrast my trusted vinyl rig (Scheu platter and bearing, Teres motor, and Schröder arm). The Stokes Tube DAC and the Audiolab M-DAC both allow using different built-in digital filters with different characteristics. The different filters usually sound slightly different, and some sound «better» than others, but they never get rid of the artificial sound completely. But, maybe unfortunately, the filters cannot be turned off completely.

When I «stumbled» over Doede Douma’s description of his DDDAC1794 that does away with digital filters I knew I had to try a non-oversampling (NOS) DAC sooner or later. Why not just skip the digital oversampling/filter, if it affects the sound by inventing new sound data that never existed in the first place?

Doedes technical description and documentation is very comprehensive and makes a lot of sense. My only hesitation was that I didn’t want to start yet another DIY project that I’d never finish, because time is limited (there’s a family, work, and too many other hobbies). But Doede sells completely assembled and tested DACs modules, power supplies, and USB interfaces. He even gave me a copy of the files needed to order a very nice custom-made chassis for the DDDAC1794 at Schaeffer AG. And when I asked him about the specifics of the additional bits and pieces needed to build a complete DAC, he simply included these in the package. For example, when I asked about which power switch would fit in the chassis, Doede just put the switch in the package (three switches, to be precise. Just in case I’d break the first and loose the second one). All this allowed me to build the DDDAC1794 in no time. The only gripe was when the Schaeffer chassis was a wee to too small to fit the power-supply heat sinks, but there was an easy fix (just a little side note to illustrate how smooth building the DDDAC1794 was: before I found the DDDAC1794 on the net, I asked the local Bryston distributor if I could borrow one of their DACs to give it a try. They keep promising I could have one once they receive one. In the meantime, I am playing my music using the DDDAC1794).

How does it sound? Spectacular? Phantastic? Superb? Damn good? Fucking great? Yes, all of this. But that’s all secondary with the DDDAC1794. The really important thing is that the DDDAC1794 doesen’t sound like a DAC at all. It’s a bit like a vinyl rig on steroids, but without the pops, clicks, and rumble (and I don’t mean the old record player your dad had when he was a boy, but the freaky good 2013 stuff). The music and all the little details are just there in a very relaxed way. Ry Cooder is having a party in my house, Phil Collins’ (yes!) drumsticks are flying in front of me, Sophie Hunger has moved to my house (was close anyway), Willy DeVille has risen from the dead, no more doubts about No Doubt, Timber Timbre is timber timbered, Mark Knopfler is in Dire Straits, Marianne Faithfull finally confessed her love for Bruce Springsteen, Jeff’s Wine is as Lilac as it gets, Glen Hansard got a shave, Depeche Mode are Exit(er)ing, Lou Reed made me a Perfect Day, and Giant Sand and Marla Glen just called to be the next acts in my listening room. In short: I hear the music, not a DAC. In contrast to oversampling and digial filters, the NOS concept not only works, but also sounds good!

As a final and very important comment, I’d like to congratulate Doede not only for designing the DDDAC1794, but also for documenting everything in full detail. The deep insight into how the DDDAC1794 works provided a lot of confidence that convinced me to try Doede’s design and to buy his stuff. One can only guess why others don’t do that.

Update 11.3.2013: Doede sent me two Sowter 1298 transformers, which he designed as an alternative to the standard coupling capacitor in the analog out line. Apart from avoiding the coupling capacitor in the signal, the transformers also allow using the inverted output of the DAC chips, thus cancelling out even-order distortion. I immediately noticed the sound improvement with the transformers. The music sounded as if the musicians just got a pay increase! The transformers were expensive, but the money was well spent in my case.

MATAA (Mat’s Audio Analyzer)

mataa_screenshotMATAA is an extremely flexible and versatile audio analysis system. Similarly to many other computer-based audio analysis systems, MATAA applies a test signal to the device under test and simultaneously records the response signal for subsequent analysis. MATAA uses the computers built-in soundcard (or an external audio module) to play and record test signals to and from a device under test.

MATAA runs on all sorts of computer platforms because it uses Matlab or GNU Octave as a base (these are powerful number crunching programs which provide a huge toolbox of routines for data analysis and processing). This explains why MATAA is so flexible and easily extendible with custom test signals, data analyses, plotting procedures, and scrips to automate routine measurements.

MATAA is distributed as free software under the GNU General Public License.

For further information read the MATAA article in the audioXpress journal (I also provide the original version of this article, where the command examples are easier to read).

Obtaining and installing MATAA

  1. Make sure you’ve got Matlab or GNU Octave installed on your computer. I tested MATAA on Matlab version 6 and 7 as well as Octave 2.1, 2.9, 3.x, and 4 (I highly recommend Octave 3.0 or newer). Matlab is a commercial and proprietary software, whereas Octave is free and open source. See below to find out how to obtian and install Octave for your computer platform.
  2. Download MATAA. There are two methods to obtain MATAA:
    • Download a recent version of MATAA as a zip archive from the MATAA project website at GitHub. Unpack the archive (if your browser didn’t do so already). If you later need to update to a more current version, download the most current package file, expand it and replace your previous version with the new one.
    • Use subversion to download and track the most current version of MATAA with the following commands:
      • Download MATAA: svn checkout
      • Then rename trunk to mataa (on Linux or Mac OS X run: mv trunk mataa)
      • If you later need to update to the current version: svn update
  3. Follow the instructions in the manual included with the files you downloaded in the previous step to install MATAA in your Matlab/Octave environment.

Further information

Further information on MATAA is available at the MATAA project website at GitHub. In particular, you can access and subscribe to the MATAA mailing list, where all sorts of things about MATAA can be asked and discussed. Using the mailing list is preferred to contacting me privately by email because:

  • Other users subscribed to the mailing list can respond and participate in the discussion.
  • Mailing list contributions will archived and will always be publicly accessible through archive.

If for some reason you cannot use the mailing list, you can still try contacting me directly by email.

The picture at the top of this page shows a screenshot of MATAA in action (with Octave 2.1 on Mac OS X).

MATAA needs your help!

  • Complain if something does not work as expected! Chances are you found a bug, or the documentation is unclear. These things can only be fixed if you report them.
  • Suggest new functions and tools for MATAA.
  • Linux is fully supported. MATAA supports sound input and output on Mac OS X and Windows (thanks to Shu Sang [] for enabling sound I/O on Windows using ASIO!), but not yet on other operating systems (e.g. Linux). Someone needs to enable sound input and output for MATAA on Linux (and other operating systems). The program(s) that handle the sound input and output from the soundcard are based on PortAudio, which provides a platform-independent environment for audio programming. Compiling the MATAA programs for sound in/out for different Linux distributions should be straightforward. I can’t do this, however, because I do not have a Linux computer.

If you think you can help in any way, please get in touch through the MATAA mailing list (see above).

Where to get Octave and how to install it

There are different ways to get and install Octave:

  • The official Octave site has instructions on how to obtain and install Octave.
  • Most Linux distributions have recent versions of Octave in their repositories. Use the package manager for your flavour of Linux to install it.

Passive pre-amplifier

preampThis “passive” pre-amplifier is nothing more than an input-selector switch and a transformer attenuator to adjust the signal volume. In comparison to using a potentiometer (or a switchable array of discrete resistors), which attenuates the signal voltage by dissipating some (or most) of the signal power, a transformer attenuator has the advantage of passing the full power to the amplifier. With the transformer attenuator, the source can therefore drive the line stage and the power amplifier much better.

Sheldon Stokes DAC


The electronics of this DAC were designed by Sheldon Stokes. It uses the old (but good) PCM63K D/A converters, which are directly coupled to an output stage with 6DJ8 / 6922 tubes. The DAC is housed in a custom-made enclosure made of solid beech wood and aluminum plates. The wooden parts are internally shielded with copper foil to avoid hum and noise due to electromagnetic interference.

THEL phono pre-amplifier

thel_phonoThis phono stage is based on the phono modules by THEL. The discrete phono stage is powered from batteries, which are recharged using the built-in charger. Due to a clever combination with the Teres, the phono stage is automatically disconnected from the mains supply during music playback, whereas the batteries are recharged while there’s no record spinning. During music playback, the phono stage is therefore completely isolated from the mains AC and its noise. This separation allows the cleanest possilbe amplification of the very low signals produced by the phono cartridge.This phono stage is based on the phono modules by THEL. The discrete phono stage is powered from batteries, which are recharged using the built-in charger. Due to a clever combination with the Teres, the phono stage is automatically disconnected from the mains supply during music playback, whereas the batteries are recharged while there’s no record spinning. During music playback, the phono stage is therefore completely isolated from the mains AC and its noise. This separation allows the cleanest possilbe amplification of the very low signals produced by the phono cartridge.

Scheu Turntable

plattenspielerThis turntable uses inverted bearing and 5-cm thick acrylic platter made by Thomas Scheu. The base is made of birch ply. The tone arm seen on the picture is the venerable Schröder with the magnetic bearing and a Pertinax wand holding a MusicMaker Mk III cartridge. The platter is driven by a battery-powered Teres Signature II DC Motor, which is based on a high-quality Maxon motor.

Cheap Trick 154

cheap_trick_xyA small two-way floorstanding speaker, developed by Bernd Timmermanns (then the editor of the german DIY-HiFi magazine Klang & Ton). This speaker uses the a Morel MDT 29 soft-dome tweeter and a Monacor SPH 135C 13-cm midwoofer.


scanspeakerI designed the ScanSpaker as a mid-sized but high-end two-way speaker using Scan-Speak drivers. This floorstanding speaker uses the Scan-Speak D2905/990000, a 1″ soft-dome tweeter, and the Scan-Speak 18W/8545-00 mid-woofer, which has a 17-cm diaphragm made out of carbon fibre and paper.

1. Drivers

As the ScanSpeaker name suggests, all the drivers are made by ScanSpeak. I used the 8545 woofer (17cm diameter) and the 9900 tweeter (aka «The Revelator», 28mm dome). These are very good, but expensive drivers. The 8545 midwoofer has the reputation of very smooth reproduction of voices. This wasn’t true in my speakers at all until I added impedance-compensation for the impedance-rise caused by the voice-coil. The 8545 has a strange “bump” at about 600–700 Hz. It doesen’t look like a membrane-resonance at the waterfall-diagrams so it can be corrected with an LCR-network. The 9900 tweeter has a special frontplate which is supposed to improve lobing behaviour. I heard and read rumors suggesting that this special frontplate design causes a slight resonance somewhere between 10–20 kHz, but I couldn’t see it on my measurements. The dome diameter is 28mm and the resonance-frequency is at 530 Hz, so the Revelator goes lower than most other tweeters. The 9900 has a reputation to acts a bit like a «diva» when designing the crossover, which I attribute to the very pronounced impedance peak at resonance.

2. Enclosure


Drawing of the ScanSpeaker enclosure

I used a bass-reflex construction for the ScanSpeaker. I put two bass-reflex channels in the bottom of the case, «aiming» at the floor. This works because the speaker stands on spikes at about 2 cm above the floor. The internal volume of the case is about 18 liters. The front is 18.8 cm wide which is about the minimum possible with the 8545 woofer. An ideal case does not vibrate at all – the driver’s membranes are the only thing that should move. To get a «dead» case I built the case from 18 mm particle-board (inside) and 10 mm MDF (outside). Enclosure vibrations are further dampened by a layer of ceramic «bathroom» tiles and a layer of «Hawaphon» on the insides of the enclosure walls. A layer 10 mm felt and some wool  dampen internal sound resonances and eflections in the case. In addition, I installed a piece of cardboard at the top of the case with some foam glued to it. The cardboard is angled at about 30° relative to the face plate, which helps to avoid standing waves along the long axis of the enclosure.

3. Crossover network

The final crossover filter networks for the midwoofer and the tweeter are shown in the figures below. The initial design started out by calculating theoretical parts values, which were then optimized by acoustic measurements and by ear. As mentioned above, the 8545 woofer frequency response shows a slight «bump» at about 600–700 Hz. This is compensated using an LCR network. Also, for proper operation of the tweeter filter, the impedance peak of the 9900 needs to be compensated. This is achived by an L-pad in between the tweeter and the filter network. The L-pad is also matches the tweeter SPL to that of the woofer.


Midwoofer crossover filter network


Tweeter crossover filter network

The parts values for the midwoofer filter are:

  • L1 = 1.8 mH
  • L2 = 15 mH
  • C1 = 10 µF
  • C2 = 15 µF
  • C3 = 3.9 µF
  • R1 = 5.6 Ohm
  • R2 = 4.4 Ohm (tot. resistance of LCR including L2)
  • R3 = 1 Ohm

The parts values for the tweeter filter are:

  • C4 = 3.9 µF
  • L3 = 0.82 mH
  • R5 = 10 Ohm
  • R6 = 8.9 Oh