My System
This is a description of the evolution of my system. The oldest iterations are at the bottom, and the newest at the top.
Much of the text is written at the time the changes were made. It reflects my thoughts and impressions at the time, and may not necessarily be what I think is a good idea now.
2021
As mentioned, in 2018 and 2019 I spent much of my time finishing the Horn Book, so I didn't have much time to work on my system. During 2020 I did various experiments building 15" woofers for the baffles, and also tried various tweeter solutions. More on this later.
But early 2021 I decided to go ahead and build a bass horn. I haven't had a bass horn since i moved from Øyfjell in 2010, so I was really looking forward to some proper horn bass again.
The design and construction is described in the article on the Big BenD Bass Horn, so I won't go into detail here. The setup is shown below.
The bass horns behave a bit differently from the open baffles, since they are monopoles. Because of this, the first test of the horns in the room was disappointing, since the horns were fighting the room modes. The room is about twice as long as it is wide, causing two modes quite close to each other at ~50Hz, and only a single mode below that at 25Hz. In between there is a deep dip. I hope that a well placed Helmholtz resonator can help.
The tweeters are modified Pioneer PT-406, which have received neodymium magnets from Celestion CDX07-1075. This doubled the gap flux and increased sensitivity markedly over the quite sad original state.
But this isn't the only change. This summer I aquired a BSS Blu100 DSP through a friend in Norway. The BSS has significantly better capacity than the Behringer DCX, and, it turns out, also sounds a lot better out of the box than my modified DCX.
The downside with the BSS is that it doesn't have a digital input, and I have several digital sources. I therefore bought a second hand Musical Fidelity M1 DAC. It has the required number of inputs and balanced outputs, and one was available to pickup within driving distance.
The BSS also has a few other issues that had to be addressed. The input and output connectors are screw terminals, it has two small, noisy fans, the output level is high and that makes it noisy when driving horns, and I need an input selector and volume control.
The connector issue was solved by placing it in a rack cabinet, where I fitted XLR and RCA connectors to the rear panel. The BSS has 12 analog inputs (2x6 stereo channels), and 8 (2x4) analog outputs, all fully balanced. The input channels have up to 48dB gain and phantom power, so I can actually easily connect a microphone to one of them and use my system as a small PA. Which will probably never happen, but the gain can be useful, as described below.
With 2x6 inputs I basically have a full preamp. The BSS has a set of control inputs at the rear, where switches or control voltages can be used to control certain parameters. I set up one input to control the input selector, another to control the output volume, and a third to select crossover type. Now the volume control isn't ideal, as it limits the resolution of the DACs, but I plan to add a microcontroller controlled relay attenuator to it later, as I did with the DCX.
The noise issue was solved by attenuating the LF channel using a 20dB pad at the output, and by removing one amplifier stage in the midrange and HF amplifiers, effectively reducing the gain of these by about 20dB (I had previously added an amplifier stage to the SE809 HF amp, but it proved a bit noisy anyway).
With the high available gain, I looked into making a digial RIAA implementation. The amplifiers seem to be quite low noise, even when driven by a Denon DL103. But the RIAA correction proved to be difficult to achieve using the available equalizer settings. The shelving filters implemented in the BSS just don't match the shelving required for RIAA correction. I was considering and outboard solution when I found the files for a RIAA FIR filter at Linear Audio. This worked very well, and the whole setup now works without any issues and sounds at least as good as my RTP RIAA.
The final modification was to implement delay-derived subtractive crossovers (DDSC) in the BSS. I couldn't do this with the DCX, and it's also not available in most other DSPs unless it's a type that can be freely configured. (The only one I have found which provides it out of the box is the Linea Research ASC48. They call it a Linear Impulse Response (LIR), but the resulting highpass slope looks very similar to what a DDSC would produce.)
The DDSC filter is based the paper A Family of Linear-Phase Crossover Networks of High Slope Derived by Time Delay by Stanley P. Lipshitz and John Vanderkooy from 1983. I came across this paper when working on the Horn Book, and had wanted to try it since. It basically creates the highpass filter by subtracting the lowpass filter output from a delayed version of the input signal. The delay compensates for the group delay through the lowpass filter, giving up to third order highpass slopes (as opposed to the first order slopes with a peak you get without the delay), and the sum of the highpass and lowpass sections is of course a perfect representation of the input signal.
I set up the BSS so that I could switch between the standard 4th order Linkwitz-Riley (LKR) filter and the DDSC using one of the control inputs, enabling me to easily compare the two options. I use this configuration for crossing over between the LF and MF channels. The DDSC sums slightly different to the LKR, but I have tried to match their responses as close as possible. Does it make a difference? Yes. There is a subtle but audible difference in the clarity of voices, for instance, which makes the DDSC a small but worthwile improvement over the standard LKR.
2018
I write about this update to the system nearly two years after the fact. There are two reasons for that. First, that it involves the Celestion Axi2050, which was not available except as OEM at the time of writing, and Celestion was already getting requests after the driver had been shown at ETF17. I wrote about the drivers here. So they asked me to not cause any more "trouble". Second, the writing of the Horn Book took a lot of my time in 2018 and 2019, so updating my website took the back seat. Now the Axi driver is available from Lean Business, so I feel it's safe to write about it.
So, back to the system upgrade. At the time (2018) Celestion was still not having a full volume production of the Axi2050, and the units that were produced were going to important OEM customers. But we (yes, I work at Celestion - but don't ask me to get cheap drivers for you!) had a couple of prototypes on the shelf.
The drivers are quite big, and with the 2" exit they didn't fit the midrange horn directly. For ETF I had had a new throat section made, so these were put into service in my own system as well.
So if you ever wondered what the mysterious box at the ETF-17 horns was, it was just a frame for holding the 3D-printed throat section for the Axi205s.
The Axi2050 prototypes have been in my system for about two years now (2020), and I'm still very happy with them. The wide band seamless midrange is addictive (and maybe part of the magic in Western Electric systems). The drivers are also low distortion and very transparent, without the high-frequency hashness of typical large-format compression drivers. That was actually the first thing I noticed about the Axi2050 when it was demonstrated to me at my first visit to Celestion.
I have used them both with and without extra tweeters, and I cross them at 300Hz. They are a bit rough in the top octave (not harsh, just not as smooth), and they don't go all the way to 20kHz, so the system sounds a bit smoother with an extra tweeter. But it doesn't break it to not have the tweeter. I have usually crossed over at 8-10kHz somewhere when using tweeters (this is also where the midrange horns start to lose horizontal directivity control). Unfortunately my SE 809 amps don't have quite enough gain to really keep up with the rest of the system, so before I do more tweeter experimenation, I need to modify the amps.
Summer 2017 (2)
Moving again... The first place I lived in England was not that great, and the living room was a bit small for a horn system. So this summer I found a better place, with some more space.
The horns are set up again, and not showing in the picture are the tweeters I have been trying. First a pair of Pioneer PT103, which sound nice but are very directive. Then PT410, which have wider directivity. The tweeters are driven by the SE 809 amps, the PP 300B amps drive the midrange horns, which still run Altec 288Bs.
Summer 2017
This summer I decided to get my horns back up. I bought some materials and built racks for the horns that mounted on the back of the baffles for the LC1Bs.
 |
 |
In addition, I switched to Altec 515-8Gs in the baffles. They have more bass and more efficiency. And although it says "Low Frequency Horn Loudspeaker" on the back of the magnet, they work pretty well in the baffles too.
2016
In November I moved to England, to start in a new job. My first house was not very big, so I didn't set up the horn system at once. Instead I put up a pair of open baffles with RCA LC1B in them.
First things first. As you can see from the picture, I'm in the process of packing out my things, but the system is already set up. The baffles were made by Thomas Dunker for ETF 2016. I took them to England afterwards.
The LC1 is a nice driver. It's a full range unit, but it is two-way, with a small tweeter cone in the center. It has quite wide dispersion too, and sounds pretty good.
2013-2016
The time I lived in my apartment in Trondheim the system did not change much. I experimented with different tweeters and different placements of them, and I finished building the 300B PP amps. They were built into Tektronix 501 oscilloscope cabinets, as shown in my post about ETF 2015.
Authum 2016 my time as a PhD student was over. I packed down the system and most of my other things at the end of August, and then lived without my horns again for a while. I also sold the Altec bins. After the PhD defense in October, I moved to England.
2013
In August this year I moved into a new apartment. And finally I had enough space to set up my system properly. The living room is not huge, about 4.5 by 3.7 meters, but that's enough if you really want a horn system up running...
When I moved, I also brought the amp racks, and installed everything: the transistor amp runs the 816A bins, and the 809 amps run the AH425 horns. After playing on it this way for a few weeks, I hoked up a pair of PP 300B monoblock prototypes on the horns. You can see them on the middle shelves in the outer racks. The bottom shelves hold the power supplies. The amps are heavily inspired by Lynn Olson's Karna amps, but the output stage is closer to the one used in Western Electric 86A, and the input and driver tubes are European ones.
When I moved, I also brought the big 200Hz midrange horns. After getting around to building stands for them, where they are suspended in heavy rubber bands, I put them into the system. The 300B amps drive the midrange horns, and the 809 amps drive the tweeters, which are modified PAudio WN-D34 tweeters. They have been modified to a 0.5" exit, and play through a pair of University 4401 Reversed Flare horns.
The tweeters, which I also used with the AH425s for a short time, provide much more treble energy into the room that the AH425s managed. Both because the AHs beams quite a bit in the top end, and also because the Altec 299-8A drivers are slightly rolled off in the top end.
Having the big horns back in the system was very satisfying. The directivity of the system is more constant than with the AHs, which means that the sweet spot is much larger, and I also like to let a single horn play more of the voice range.
2012
For some time now, I have used my AH425 horns with Altec 299-8A drivers with my 816A bins (which now have Altec 515-8G drivers). I use my modified Behringer DCX2496 as crossover and preamp, with the big transistor amp driving the bass bins, and the SE 807 amps driving the midrange horns.
I have also acquired a Sony TTS3000 turntable, which is now fitted with a Denon DL103 MC pickup and runs through the RTP RIAA.
This setup works pretty well, but the space is really too limited for setting it up properly. But at least it keeps the "horn withdrawal" at bay... And it shows, for those doubting it, that a horn system can be set up in a 4 x 2.5 m room.
2011
Fall 2010 I moved to Trondheim to study. First I used a pair of Jordan JX92S in vented boxes, but that was not very satisfying for a horn enthusiast. I obtained a pair of Altec 816A bins with 511B horns, which is what I use now. That works...
2010
For several years now, I have been planning, simulating, and building new midrange horns, designed to cover the lower midrange using compression drivers. The horn is designed for modified Altec 288B drivers, with a cutoff frequency of approximately 200Hz.
The performance of these horns is the best I have heard so far. Improved horizontal directivity control gives a much wider sweet spot than the Le Cléac'h horns. They merge better with the bass horns at the crossover frequency (about 300Hz), it is almost seamless. Having most of the voice range in a single channel (300-5000Hz) results in a very good and intimate reproduction of voices.
The tweeters are Fostex FT17 units, they are a bit better matched than my pair of 2402s, but are not as efficient. Also, they are not as good as the manufacturer's data indicate...
This was the last change I made to my system. A month after completing and installing the horns, I had to pack it all down, as I was moving. I'm now living in a quite small apartment where there is no place for a large horn system.
2009
During the fall of 2008, I collaborated with Lynn Olson and Martin Seddon (of Azurahorn.) in the design of the AH425 Le Cléac'h horns. My part in that project was mainly simulating the horn using Boundary Element Method (BEM) SW, and to be a discussion partner. This horn was specifically designed to match the Altec 288 series of compression drivers.
I decided to buy a pair of these horns, both because I was curious about how they sounded, and because I had just acquired a pair of 288C drivers. (There is a slight difference between 288B and 288C. The C version has better HF response, and is usually the better choice. The B version has a larger distance between the diaphragm and the phasing plug, which gives it better low frequency headroom, but poorer HF response. Taking this into account, it is odd that the B version is more expensive, since the HF response is usually what is most important in the systems where this driver is used).
The 288C on the AH425 sounded really good. But even the 288C doesn't go all the way up, so I crossed in the 2402 tweeters at about 7kHz. The performance was superior to anything I've had to that point. Better imaging, more details.
In addition to new horns, I also built new racks for the power amplifiers. They are constructed along the same lines as the preamp/source rack, using threaded rods and wood boards. Ikea had resized their butcher blocks, so they are now too small for rack shelf use.
And finally, I also finished the new balanced RIAA stage that I have been working on for a long time. It's basically an implementation of Allen Wright's RTP5 preamp, without source selector or volume control. It sounds good, and is very quiet. The stage lineup is first a hybrid differential cascode stage using 2SK170 and E88CC. Then an ordinary differential E88CC stage, followed by an E88CC/IRF840 SLCF. The output goes into the modified Behringer DCX2496 preamp/crossover.
The RIAA is built on the chassis of a Tektronix plug-in module, which plugs into a power supply that is used with the modules when they are operated as stand-alone units. (I can't remember the type no.). Both the module and the PSU have of course been completely rebuilt to fit the requirements of the RIAA. With this plug-in system, I can try different RIAA solutions without having to build a new PSU for each.
2008
At a junk yard, I found a pair of Philips EL6400 amps. Since 20W of class B power is not what works best with high efficiency horns, I decided to keep the output and power transformers, and build a pair of new amps on the old chassis. The amps look pretty cool, and have plenty of space when the microphone inputs are not required. The circuit is a differential stage with E88CC driving a pair of triode connected EL84s, for about 3-4W output. I use this amp for my tweeters.
The tweeters are now JBL 2402. This was a great step up from University Mid-T drivers, which are actually very poor. The horns are very interesting designs, but they need good drivers. The original University drivers have too low a mass corner frequency, and are not well suited for tweeter use.
The tweeters and midrange were aligned for best possible impulse response. This required not reversing one of the drivers, but rather separating the crossover frequencies to avoid the dip at the crossover frequency. Reversing the tweeter surely made the frequency response look better, but the impulse response got worse.
This summer I took a pair of Ikea wooden blocks and made a new plinth for my L75. Much better than the original flimsy particle board plinth.
This year I have also upgraded my DCX2496 with Frank Oettle's input module, an upgrade which greatly enhanced the sound, and removed some odd behaviour at the digital input.
2007
This amp is designed in a way a class A2 amp should not be designed: the output tube does not draw grid current through the whole cycle. That is not practical with 809, which is a 25W transmitting triode with a µ of 50. But with a low impedance driver, in this case triode-connected D3a DC-coupled to the 809 grid, it works well. The input stage was originally a 6SN7 differential stage, but it had too little gain, and was changed to 6SL7. Operating point and anode load resistors were changed too. This amp sounds very good, I think it is my best amp so far.
2006
During the winter 2005/2006 and spring 2006, not much happened, except that I removed the JLH amp from midbass duty and installed the big transistor amp there instead. The subwoofer is now driven by a DIY 60Wpc transistor amp with build-in cross-over and parametric equalizer. And I also built a small single-ended 807 amp with Hammond iron. The 807 is operated in tetrode mode with feedback from plate to grid, and is driven by an E83F.
In October 2006 there was a big change in the speaker setup. During the summer a friend and I had built 4 big basshorns for sound reinforcement, two for sale and two for personal use. The horns are 103cm W x 64cm H x 81cm D, or about half a cubic metre. Since we needed a place to store the horns when not in use (which will not be very often), I took them home.
The drivers are JBL 2205B (alnico). The horn itself is a 1.8m long folded hypex horn with a 43Hz cutoff, designed for floor/wall placement. It has good dynamics in the midbass, delivering tight, powerful bass, and meets the subwoofer much better than my previous horns. It may seem strange, but I have more floor space now than with the old midbass horns. The new ones are only 81 cm deep compared to 130 cm for the old ones.
In addition the old tractrix prototype horns have been replaced by a kind of constant directivity horn, using parts from my old Trinity horns. They are NOT conical horns. The first part of the horn is exponential, two coupled segments of different flare, terminated in a conical mouth part that controls the directivity. The throat impedance is very smooth down to about 500Hz, as is the power factor.
Tweeters are University 4408 with MID-T drivers.
2005
As all the horns in the system have different lengths, and as I plan new horns that require different crossover frequencies, an adjustable active crossover with time delay would be nice. The Behringer DCX2496 is a cheap solution, with lots of flexibility. Correcting the delay differences between midbass, midrange and tweeter made the system sound much more coherent and "real". The DCX can barely be seen at the bottom of the picture, under the old preamp, now in use as phono stage. The new rack (right) can also be seen. It is made from butcher blocks and all-threads, and is not as wobbly as the former rack (still in use, to the left).
Pr July 2005, the 6AS7G amp drives the midrange, the JLH amp drives the midbass, and the 3.5W SE UL amp drives the tweeters.
I must mention that I was very sceptical about introducing a DSP device into the signal path, but the sonic degradations (I lost a little dynamics and definition) are small, and just time correcting the horns was a great step forward. But anyway, there is a lot of hacking to do in the analog stages, and by running the input at a constant high level and having the gain controls at the output, I expect great improvements.
2004
Summer: Horns again!
June 2004 I visited Thomas Dunker, and listened to his horn system. Getting back home, I realised that even if my system sounded good tonally, it lacked dynamics. As I had just bought a pair of JBL 2470 compression drivers on eBay, I hooked them up on a pair of unfinished Edgar-style tractrix horns. And the dynamics came back! I switched back to the 3.5W SE UL amp, and the 150Wpc transistor amp is now back in subwoofer service.
The 2470s do lack the ultimate top end, even with titanium diaphragms, so I had to add tweeters. First I added the Seas 6x9" fullrange drivers I used long ago. They are quite a bit less sensitive than the horns, so they needed an extra amp. The result was a 6Wpc PP amp using 6AS7G in the output stage.
February 2005 I bought a pair of University 4401-horns with Mid-T drivers. They are much better than the Seas drivers, and are now in use in the system, as shown in the picture above.
March
Much has changed... I still use the subhorn and midbass horns, but from 500Hz and up, I now use a pair of Jordan JX92s, driven by the Slone 150W amp (the 3.5W SE amp is too small). To drive the subhorn I use an old Sammic PA amplifier using only one channel and both subhorn drivers in parallel.
Trying the Jordans was a great step forward. They are extremely detailed, and have the most natural voice reproduction I have heard. They have a bit too much in the top end though, so I use a small line level correction network to level them out.
The preamplifier PSU also got a little upgrade, a shunt regulator similar to Allen Wright's SuperReg, but without the current source. This improved low level details and perceived noise floor.
he Sony player is also upgraded with new clock, extra bypass capacitors around the VC24 filter, and a new analog stage. Right now I use an analog stage similar to Allen Wright Level 1 upgrade, but the plan is to make a balanced tube based analog stage.
A somewhat controversial improvment was to use thin solid core silver plated speaker wire. Just good quality wire-wrap wire. It sounds very good, with improved imaging and better "intertransient silence". I tried it first on a friend's system. When we first listened to it, we just looked at each other and laughed; the improvement was obvious and big. And it's cheap too!
2003
This is the setup at fall 2003. The CD-player has changed from a modified Marantz CD63to a Sony DVP-S9000ES (Why does Sony have to use such long complicated names?), the JLH class A amp is used for the midbass, the new SE amp is used for the Trinity speakers. I use the big transistor amp for the subwoofer, which has changed to a horn subwoofer.
Here is the new SE amp. It has two stages of E80CC driving 4654 in UL mode. 4654 is an old European tube with P-socket. Maximum power at clipping is about 3.5W, which is OK for normal to fairly loud listening on the Trinity horns. Distortion @ 1W with 10dB feedback is 0.12%.
The horn subwoofer.
The new, big transistor amp. It is a design from Randy Slone's book "The Audiophile's Project Sourcebook", and works very well. Power is about 150W/ch with very low distortion. Distortion is also very low at low power. I've had it measured with a Rohde & Schwarz Audio Analyzer. IIRC, THD was down at 0.0008% @ 1W from 20-20k. Pretty good! In addition, it sounds very good. Far better than Electrocompaniet AW60, more dynamic than the Zapsolute amp. One reason for the good sound is that it uses 2-pole compensation, which reduces HF distortion significantly. Without it, this amp is nothing special.
2002
The listening room in the barn is cold in the winter. But around Easter (2002) things started to happen again. I moved the system to one of the long walls, and that really improved the LF response. The bass was more even, without the dips that had previously made the sound somewhat thin.
I decided to use the mid compression drivers I had used in the styropor horns together with the Sammi SM-110s in the Unity inspired design, with a passive crossover to make a nice mid/tweet combo. I call this speaker Trinity (3-driver unity summation), and it sounds quite good. I switched from the OTL in the HF range to the old John Linsley-Hood's class A design. This is an amazingly good amplifier, which makes you wonder what audio engineers have been doing since 1969. It's the gray box at the top of the left shelf. Take a look inside. It runs quite hot, so I will paint the heatsinks black.
Another new thing is the preamp. I constructed a completely new pre with separate PSU and built-in active crossover. The PSU has tube rectification (PY88) and choke input without electrolytics. The linestage is the same topology as the old one, but with ECC82 instead of 6SN7. For the 500Hz crossover I use PCC88 as cathode followers (they can drive a 10k load with 0.03% thd), and op-amps for the sub XO, 85Hz 4th order L-R.
The preamp has a 64 step relay controlled attenuator with nixie tube indication, and all inputs are relay controlled. It also has provisions for a remote control, which is not finished yet. Right now I use a cable from switches on my chair to control the volume.
2001
After some work, I had a brand new listening room. 4.3 x 6 m, sloping roof: nice. The system was moved in, and the music played again! But there were improvements to come. I added 20cm to the midbass horns to make them go deeper, and later I installed an Eminence Delta 10 woofer in each. This increased efficiency quite a lot. I also made new midrange/tweeter horns. They are constant directivity horns, and look quite like the Lambda Unities (which were the model for my design). At the moment I use only the tweeter, without any midrange drivers. The tweeter is a Sammi SM-110, and it sounds very good to my ears. I have also made a version of the TNT FleXy rack, but using thinner rods and 12mm wood fibre plates. It supports the record player, CD-player, pre-amp and tuner. It was a real improvement, the record player does not jump anymore, which proves that it decouples effectively from the floor. The midbass horns are also placed on spikes, and the amp rack moved to where the record player stands in the picture.
2000
Summer
There is always room for improvement. In the summer of 2000 my midbass horns were born. 80Hz exponential, designed for 10" drivers. But I had a pair of 8" drivers I wanted to try, and they worked pretty well. Midrange and treble as before. The subwoofer is painted in the same color as the midbass horns. For the sub I still use the Boyer 100W amp. The midbass runs on the first prototype of the Surtamp, adjusted for about 3.5W, the OTL runs about 0.1W. Efficient, or what?? But how is the sound? The most important thing in a system is not the look, but the sound. Well, I can say this: This must be the most price-effective system around. Compared to performance, it is incredible. Not counting the sub, all the drivers are at least 20 years old. And besides that: It delivers music as easy and unburdened as only horns can do. I love it.
Spring
Then it happened: The Great Horn Experience! (have a look at my Why Horns page for more on that.) With some advice and inspiration from Timo Christ, I made some nice, blue styropor horns, using drivers I bought from him. The best sound so far, my old Monitors sounded dead and lifeless by comparison. Here are dynamics, life and MUSIC! The speakers at the top are 6x9" tweeters (!) used from about 3.5kHz and up. They only go to 13kHz.
January
I moved into grandma's house with my system. Later I also built a new preamp. The old subwoofer is placed in the middle, driven actively by the amp next to it. This is a Boyer 100W mono PA amp. Next to my preamp/CD-setup is my turntable. The table where it stands is not very stable, so I have to be careful when I walk around.
1999
This is when my DIY hifi adventure started for real. For years I had built electronic stuff, occasionally some speakers and amps, but mostly radios. For some years my system had consisted of a NAD 501, a Marantz amp, and Monitor Audio 201 speakers. In the summer of 1998 I made an unsuccessful attempt to build an OTL amplifier. After a year away from home, I tried again, and succeeded. It was the best amp I had owned to that date. A true inspiration that initiated the line of evolution chronicled on this page.
This picture shows my system as it was in the winter of 1999. The amp is my dear OTL, a DIY project. The preamp is also DIY, so are the speaker stands too (surprise!) The speakers are Monitor Audio 201, the CD-players are a NAD 501 and a Marantz CD-38. Not shown is my turntable: an ITT L-75, a CC-SUB subwoofer, 4m Audioquest speaker cable and some 4mm^4 cable for the subwoofer.
