Welcome, Guest
Username: Password: Remember me

TOPIC: The 2-year saga of my system

The 2-year saga of my system 3 years 1 month ago #26633

  • leoman
  • leoman's Avatar
  • Offline
  • Expert Boarder
  • Posts: 92
  • Thank you received: 29
  • Karma: 9
Around two years ago I determined that time had come to use my garage (which is roughly 47'x27'x12'), a big room in which I spend much time, for something besides just a garage. Under the stress of work projects I had ignored audio for a couple of decades. Now retired, I had some time again. Here is a bit of background.

The house stereo has sat essentially unchanged for almost 30 years. The long living room was in fact one reason for buying the house; my eyes bugged at having enough room to put together a system with built-in speakers flanking a fireplace (as seen in a book on speaker enclosures). The woofers were 15" Altecs housed in homemade tuned cabinets, the mids Altec 511 horns fed by Peavey drivers. The tweeters have changed identity a few times due mostly to catastrophe, the latest (maybe even the best) being Eminence drivers feeding small biradial horns. The system is, by the way, entirely solid state (sorry tube fans). It betrays a longstanding belief in efficient drivers, and the three huge power amps, rated from 100 to 250 W/channel, hardly work past idle but are very quiet, critical in direct connection to horn drivers in a living room (back then, generally only powerful amps were also quiet enough to feed horns at idle without hissing you out of the house). It also reflects a belief in steep crossovers, using a (still working!) homebuilt 3rd order Butterworth op-amp xover ('steep' for the 80s at least). Thank you, National Semiconductor Audio/Radio Handbook. This was fed by typical analog sources (plus a CD player) through a Pioneer preamp/selector and a very good Radio Shack(!) graphic EQ plus a few of-the-day analog signal processors, e.g. a Burwen pop/click remover and a homebuilt compander based on the Signetics NE572, adapted from an application diagram for the related NE570 in the Signetics linear book. In other words, it's all typical 70s/80s commercial/DIY components (outside, perhaps, of the speakers).

The carryover point of all this is the opinion that big, efficient drivers deliver the best bang for the buck if they can be tamed properly. The great (and costly) top-tier home systems of the 50s with their large, efficient direct radiators and profusion of horns and compression drivers were such loudspeakers. The advent of electronic crossovers and graphic EQs made taming far easier to do, within the limitations of analog hardware.

The garage begged for a hi-fi even more than the living room, and finally the decision was made to do a few decades of catch-up on audio. Clearly digital was the way to go, offering the promise of linear phase equalization/crossovers, easy driver alignment, less proliferation of component boxes and 'natural' accommodation of source material, today nearly all digital. Most of the catch-up was a pleasant revelation of how far the hobby had come in a few decades (and especially in the past several years).

The progression was humble. Initially, a couple of PA/DJ satellites were used with an old receiver. It worked fine, but stimulated interest. I found the 'pro audio' market not only still kicking but affordable. A slew of (mostly) Behringer eqs, crossovers, amps etc. and a primordial version of the final speaker components began to take over. At some point I wondered what I would do to improve things if I could. Digital seemed the way to go. I designed what I'd want a box to do in software in my head, thinking that if such things existed the cost would be prohibitive. Then one day on the web I stumbled onto the MiniDSP 4x10 which did it all and more, and in the same sequence! That replaced several boxes. But then I discovered FIR filters and SHARCs.

After two years of incremental learning, swapping, testing and building, what was desired is done (subject to continual change of course), an affordable system without obvious compromise and still with upside. It can be duplicated for ~$1700, though a good deal more was spent in the educational/developmental phase (worth every cent). Here is the final (as of now) system and its hows and whats.

The primary source for music is a plain Dell 3646 desktop running Windows 10, already used for a variety of tasks besides audio (and thus not included in the cost) but enhanced a bit with Fidelizer 'audiophile' mode. The music player is foobar2000, through which ripped CDs and web sources provide material. Unamplified natural sources such as symphonic, jazz and big-band vocals are ripped with FLAC 5. Less demanding or more artificial sources (rock etc.) and older recordings with limited fidelity are done using Opus VBR with a decent average bit rate. Resampler-V is employed on output, using SoX at 192k. Output is FIR-suppressed at around 20khz by the resampler with excellent impulse response. WASAPI outputs the stereo signal via USB to a CM6631a-based USB2/spdif converter which, in turn, feeds a modified coax/toslink converter driving a 5mW laser (details of which are available in another MiniDSP forum post). About 40' away, the photdiode of a slightly modified toslink/coax converter reads the signal at 192/24 and feeds the DSP.

The DSP is a pair of miniSHARCs in 'series', details of whose development also are posted on MiniDSP's forum. In capsule summary, the input stage has a DIGI-FP which downsamples the input to 48k. The first sharc acts as an equalizer (including 'room equalizer' if you are a fan of such procedures, which I am not) and then feeds the second sharc which houses the crossovers via spdif. A miniDAC8 sits atop the second sharc, providing all 8 analog outputs. FIR filters programmed via RePhase are used on both sharcs. The crossovers occur at 330hz (192db/oct), 1850hz (256db/oct) and 9000hz (256db/oct). The 2-rackspace box also houses a 5V switching power supply.

The power amps are also DIY'd using very affordable TDA7492 class D amp boards (also shown in the miniDSP forum). These have their endemic issues (one channel is inverted and bridge mode doesn't work) but once the issues are found and compensated (or avoided) they work tirelessly without fans at any power level I can tolerate. Three boards serving the mid, mid-high and high horns share a switching power supply and are housed in a surplus single-rack box. For the woofers, two of the amp boards are used with only one channel employed by each in another single-rack box with another switching power supply. This arrangement is seen on scope (using an Altec 421 as a test load) to handle ~30V peak @200hz without clipping or visible/audible waveform distortion, a power level the Altec, the amp and my ears won't tolerate for long. This is a hefty boost over the amp's stereo power limits, probably due to unloading of common circuitry, and makes the failure of bridge mode superfluous. The thought of such tiny amps driving large speakers to deafening levels seems counterintuitive but makes perfect sense. Monster amp development was largely driven by the sealed bookshelf speaker plague that began in the 60s and into which I was 'born' with my AR-4x and H.H. Scott 222c dorm system. Not to rail against them; the little ARs were a clear improvement over what I'd known before and fit anywhere. But it took time to realize that the "west coast sound" developers of the 50s with their big reflex speakers and horn-loaded HF sections had it right all the time for those who had the space (and money).

Each 4-way speaker in the stereo setup employs the following affordable (and highly bulletproof) drivers:

Bass: Two 15" Seismic 'subwoofers' with TS-matched cabinets are used as woofers. Originally vintage Altec 421s were planned but the Seismic drivers tested so well in their matched cabinets that they persist. A passive 800hz crossover in each cabinet is bypassed. Without it the drivers are capable of well over 800hz linear response, though they are not asked to provide it. Bass horns remain a fantasy, but note that with the current setup the cannon in the 1812 Overture is stunningly loud and undistorted, probably good for stripping loose paint.

Mid: Two Dayton Audio D1075T PA compression drivers are chosen for their low frequency extension, similar to that of more expensive Atlas and EV offerings, along with two classic University Cobreflex horns (modern ABS version, chosen for low cutoff). The 70v transformer in the driver is hardwire-bypassed, else the transformer remains in parallel with the driver even when 'switched out'.

Mid/High: Two Celestion 1446 compression drivers feed two Goldwood(!) CD horns. No horn specs were available, but the photos/dimensions made guesswork easy and they were so cheap as to be riskworthy. They proved brilliant on testing with excellent loading and dispersion, real hidden gems.

High: Two Eminence APT-50 drivers feed two Eminence CD horns. The mike measures excellent dispersion at very high frequencies.

The low and mid drivers exhibit some resonance peaks, located by sweeping a small test signal manually through each and finding voltage peaks across a small in-line resistor. Readings matched up well with published impedance graphs, always a relief. These peaks are compensated (with linear phase) in the respective crossover curves as RePhase allows mixing its 'paragraphic' EQ with the crossover filters. Horn mouths are lined with low-density foam (adhesive weather strip) to minimize diffraction effects, as all are free-standing.

Clearly, the drivers are a challenge to mount in phase. Additionally, they are constrained to be mounted on a shelf (actually the 'roof' of the boiler room) 8' high, making stacking impractical due to the already high positioning. The listener's vertical position varying much less than his horizontal position, stacking would be otherwise preferable to minimize parallax on driver distances when changing listening position. The situation is ameliorated by the abrupt slopes available on the FIR crossovers, minimizing regions of driver interaction.

Several factors were to be considered for phasing:

a. what listening point to choose for phase testing,

b. latency of the FIR crossovers employed, and

c. relative distances of the driver diaphragms.

A ballpark delay was computed in the following fashion:

a. FIR filter delays were known courtesy of RePhase.

b. The front edge of the 'shelf' on which the drivers are mounted was used as a zero-reference point. Then the distances of the voice coils from the front of the shelf were estimated with a tape measure. In the case of the cobreflex horns, which have a 3-way folded path, 30" was gauged as an approximation (it is also the path length published in at least one source).

The amount of delay in milliseconds was computed using 13800"/sec as the speed of sound. To this, the FIR delays (which dwarf physical delays) were added for each driver. From this, the compensating delay to be applied to the second miniSHARC's crossover channels was computed. The driver with the most aggregate delay (the midrange as it happens) was assigned zero applied delay. The difference between its aggregate delay and that of each other driver became that respective driver's applied delay. Finally, a hefty 2ms was added to each delay to provide adjustment wiggle room for the midrange.

Now it was time to fine-tune/confirm. Audacity was used to generate fairly widely-spaced tone bursts at each crossover frequency. Wide burst spacing was of more concern at high crossover points since 330hz has a 3.5-foot wavelength but 9000hz only 1.5". A sine tone was generated over a span several times the desired burst length. Then, editing tools permitting

a. setting the cursor to the nearest zero-crossing,

b. selecting start-to-cursor or cursor-to-end, and

c. silencing the audio over the selected span

allowed creation of clean and useful tone bursts. A high 'project sample rate' is recommended, e.g. 192k. A microphone was placed in a chosen (and necessarily arbitrary) location. I used two symmetrically-spaced positions, one for each speaker system, at a distance of about 15' angled a bit toward the middle. Farther than that and diminishing amplitude plus reflections made scope syncing difficult. The mike ran to a preamp which fed the scope. Delays were fine-tuned at each crossover frequency to result in the correct number of burst peaks and cleanly superimposed signal addition with individual channel level variation. It's a good opportunity to calibrate driver shelving levels also.

The procedure has resulted in smooth driver transition and very good imaging over a wide and long field (near perfect over a large 'sweet spot') and marvelously smooth, clear, detailed sound, the like of which was not heard before on familiar recordings.

And that is that, except of course for a few afterthoughts.

Nothing about the system precludes its being used in a decently sized living room, provided one is willing to build two floor-standing enclosures on a size par with those used in the 50s/60s University/JBL/Altec home systems. A look at University Classic or Dean cutaways reveals a usably small enclosure capable of accommodating the Cobreflex horns and, in fact, a folded bass horn. The exact drivers/enclosures presented of course have a small footprint when stacked, if decor considerations permit. The laser spdif connection of course requires line of sight but can be substituted by cable or (better if your computer is in another room) a 5.8Ghz or 2.4Ghz wireless sender/receiver pair for old composite video (widely available). Regarding such wireless transmission of the spdif signal, the sub-bands (channels) of either frequency will support stereo spdif over video to a 48khz sample rate. Use units with whichever frequency band is not used by your wireless broadband to avoid conflict.

As for the speaker system itself, the choice of Seismic drivers is explained above. The mid horns are marvelous, though they do have two outlets each which some purists may find problematic. No audible effect is heard (especially crossed over at a low 1850hz) but I would love to try Altec 311 horns at some point, based on admiration of the 511s in the house system. Sadly, used 311s are rare and expensive, violating one of my tenets and leaving me to wonder why no ABS clones seem to exist. The choice of mid-high and high horns and drivers was made after testing a number of candidates. Loading and horizontal dispersion are excellent and the drivers are remarkably smooth within their bands. Steep crossover slopes permit use at lower-than-recommended frequencies. Plus, all compression drivers used have proven remarkably tolerant of brutal impulse transients (due generally to tinkering), the likes of which I have seen destroy costly classic drivers (Altec 808-8A, EV 1823 and EV T350 come to mind). Paying $400 (or even $2000) for a driver would have me walking on eggshells. I find tolerably imperfect but correctable and durable $50 drivers vastly preferable to more fragile but (somewhat) more accurate drivers generally costing far more.

Regarding the little amplifiers, note that they replaced my 'professional' power amps whose fan noise became irritating at low levels. The tiny class Ds, shockingly, test as well (within power limits) and sound as good, require no fans and, like their expensive counterparts, lack power on/off transients, a concern for any HF compression driver direct-wired to an amplifier. It is one reason for back-burnering a project to compare and possibly replace them with small class AB amps. These so far have proven touchy (i.e. they thump), which is a no-go. The inductive phase shifting on all class D amps inspired the purist-motivated AB project, but any residual phase shift in the audible band of the class Ds is smooth and relatively minimal, and in any case is essentially identical over all channels, thus having no observable effect on stereo image or driver phasing. In short, the little class Ds sound so good and have proven so durable that I don't see much return on investment in replacing them.

The MiniDSP processors are marvels. The equivalent of what I built is now available in two separate boxes that may be stacked, OpenDRC-DI and the new OpenDRC-DA8. Currently my single-box 'sharc stack' setup uses spdif to interconnect the sharcs (as would be done with the in-a-box options mentioned), though the added option of I2S board interconnection exists with the common enclosure.

Except for the MiniSHARC plugins ($10 each) and Fidelizer (~$40 IIRC) all s/w used is free including (e.g.) foobar2000 and its plugins, plus Audacity and of course RePhase. Amazing.

Transmitted sample rate, despite the 48k limitation of the DSPs, is kept at 192k partially because of an ulterior motive, that being the testing of my laser setup, which has hundreds of hours on it now at 192/24. The system as a whole, in fact, has been left on (including in my absence) for over 6 hours at high volume regularly and without issue, including in ambient near-90-degree temperatures (the garage is heated but alas not AC'd).

One final note: for fans of signal enhancement, even inexpensive (or free) s/w like DFX and Floorfish (usable with available VST plugin adapters for foobar) can provide loads of fun decompressing vintage recordings (especially vocals), though I tend generally not to use them for various (totally arbitrary) reasons. And hundreds more plugins exist for the so-inclined.

I hope someone finds part or all of this somehow helpful. Comments, questions and corrections welcome of course.
The administrator has disabled public write access.
The following user(s) said Thank You: devteam

The 2-year saga of my system 3 years 1 month ago #26652

  • pos
  • pos's Avatar
  • Offline
  • Platinum Boarder
  • Posts: 345
  • Thank you received: 196
  • Karma: 39
Very nice system and approach leoman!
I like the fact that you are using reasonably priced and carefully selected/tested drivers, and pushing them to the max (and beyond!).
It is indeed much more rewarding than using top-of-the-line super expensive drivers that are perfect out of the box, plus it is also more fun to play with as you don't have to freak out as much if (when...) a problem occurs (filter error, on/off thumb, etc.).
download the last version of rephase here: rephase.org
Last Edit: 3 years 1 month ago by pos.
The administrator has disabled public write access.
The following user(s) said Thank You: devteam, leoman

The 2-year saga of my system 3 years 1 month ago #26688

  • devteam
  • devteam's Avatar
  • Online
  • Administrator
  • Posts: 8459
  • Thank you received: 1037
  • Karma: 156
Such a nice and detailed explanation of how you came to build your system. Obviously something that wasn't born in a day! Thanks so much for your post! We'd love to see some pics, we can push them together to the Online Gallery. :- )

ohh and we couldn't agree more that rephase rocks! ;-)

DevTeam
MiniDSP, building a DIY DSP community one board at a time.
The administrator has disabled public write access.

The 2-year saga of my system 3 years 1 month ago #26703

  • leoman
  • leoman's Avatar
  • Offline
  • Expert Boarder
  • Posts: 92
  • Thank you received: 29
  • Karma: 9
Thanks, POS and devteam! I'd like to add some omitted details. The mid-hi horns were Goldwood GT-400PB, the hi horns Eminence APT-150S. Plus the conspicuously absent photographs.

I don't suppose anyone needs to see a Dell 3646 running foobar2000, so no pic. But, in sequence from that source, here are:

1. The little usb/spdif converter, on the 'roof' of the office,



2. The laser driver stuff (modified converter to the right of the power supply), fed by the usb/spdif converter,



3. The mounted laser unit (seen at extreme right of prior pic),



4. The laser receiver (slightly modified toslink/coax converter; unmodified will work too),



5. The front panels of the DSP/amps (Ultramatch not in the path, just in the picture),



6. The speakers,




And the rest on the next comment; I've run out of 'add files'.
Attachments:
Last Edit: 3 years 1 month ago by leoman. Reason: fix bad photo
The administrator has disabled public write access.
The following user(s) said Thank You: devteam

The 2-year saga of my system 3 years 1 month ago #26704

  • leoman
  • leoman's Avatar
  • Offline
  • Expert Boarder
  • Posts: 92
  • Thank you received: 29
  • Karma: 9
And finally some details of some of the above.

7. Inside shot of the 'sharc stak' box,



8. Its rear panel,



9. Inside shot of horn amplifier,



10. Rear panel of bass amp (inside looks roughly like horn amp; differences were cited),



11. Test version of converter modified into laser driver, top of board (different 74hc04 inverter shown in socket),



12. And the addition of 2 more inverters to the fan out/in on the bottom,



and that ought to do it.
Attachments:
Last Edit: 3 years 1 month ago by leoman.
The administrator has disabled public write access.
The following user(s) said Thank You: devteam

The 2-year saga of my system 3 years 1 month ago #26755

  • devteam
  • devteam's Avatar
  • Online
  • Administrator
  • Posts: 8459
  • Thank you received: 1037
  • Karma: 156
Gotta have a sticky for all this goodness! :-)
We'll make sure to post few pics in our gallery shortly.

Thanks again for sharing!
MiniDSP, building a DIY DSP community one board at a time.
The administrator has disabled public write access.

The 2-year saga of my system 3 years 1 month ago #26814

  • david914
  • david914's Avatar
  • Offline
  • Fresh Boarder
  • Posts: 15
  • Thank you received: 3
  • Karma: 0
Very interesting setup, and well written write up! I have a soft spot for the horns, having owned a pair of original Cobraflex horns (aluminum), and currently own a pair of Altec Voice Of The Theater speakers (Valencia cabinets) and a pair of 511b horns (in storage). Perhaps someday I'll try bi-amping the Valencias using a miniDSP!
The administrator has disabled public write access.

The 2-year saga of my system 3 years 1 month ago #26815

  • leoman
  • leoman's Avatar
  • Offline
  • Expert Boarder
  • Posts: 92
  • Thank you received: 29
  • Karma: 9
I am jealous, David! Valencias are beautiful. Years ago, I found a pair of Sevilles (that look like mini-Valencias) in Cambridge for (I think) $150 for the pair. The woman was replacing them with a pair of little bookshelf speakers, to my delight/horror. I installed them in my parents' living room.
The administrator has disabled public write access.

The 2-year saga of my system 3 years 3 weeks ago #27322

  • leoman
  • leoman's Avatar
  • Offline
  • Expert Boarder
  • Posts: 92
  • Thank you received: 29
  • Karma: 9
As usual, nothing stays the same around here so I wanted to post an update. Several changes have been made.

1. Grounding was improved on all boxes (always a good idea).

2. I was not satisfied with tone burst response I saw on my midrange and thought to try something else, but not something costly. I replaced the Dayton PA compression drivers with Atlas PD-5vh. They are quite remarkable drivers that feature Alnico magnets, amazing low range that permitted me to drop my low-mid xover point to 300hz without issue, and gave me improved tone burst response. They sound subjectively great, and are about $140 each.

3. Similarly, while happy with the deep low end of my Seismic 15" woofers in their matched sub cabs, I didn't care for the tone burst behavior. A pair of Altec 421A (original, not recones) were substituted, and proved much better at tone bursts. Their efficiency is remarkable, approaching that of the horns! Their surrounds and cones showed their age (though not audibly); that was cured with a rag and a bottle of Armor All!

4. Despite their marvelous clarity, the Altecs lack the excursion capacity of the Seismics and a bit of the bottom end dropped off. To this end, I ran the optical output of the Digi-FP on my sharc stack (from sharc 1, which is used full-range) to a new 2x4HD. It is FIR programmed for an 80hz lowpass (as well as RePhase can handle the low filter, which is plenty good enough).

5. Now I needed another amp. I (sadly) cannibalized a power supply box and a fully wired box for a class AB project that I had shelved, placing a single class D board into the box running a single channel, and downconverting the PS from +/- 24V to just +24V. I would have stacked my two supplies but the amp alas will not tolerate 48V. Not that I'd ever use it.

6. Seismic's subs may not match the Altecs in 'tightness', but they have excellent excursion and are unrivaled AFAICT in the affordable subwoofer category, coming in matched tuned cabs. A downfiring 18" cab (for a little over $200) was placed (with much effort) between my woofer cabs. Phasing was done, and the under-300hz region now has Altec clarity on the high bass to low mid, along with powerful response down to around 30hz (or lower) in the below-80hz region, with added durability for the unloaded Altecs.

A couple of photos: First, a look at the speakers, as before from this angle but with the sub now amidships and two lovely 421A's staring at you from the woofer cabs.



This is one Atlas PD-5vh driver mounted on a cobreflex.



This is the processing stack now, only the Ultramatch unused (except in testing). Note the PS box and the sub amp, the top two boxes. Also note the little 2x4HD atop the pile. Clearly some aesthetics are in order, but it's a garage after all. The laser receiver still sits to the left atop its lab jack.

This message has attachments images.
Please log in or register to see it.

Last Edit: 3 years 3 weeks ago by leoman.
The administrator has disabled public write access.
The following user(s) said Thank You: devteam
Moderators: devteam