TOPIC:

I've posted about this in the newbie section, but thought I'd post here also. I've completed a project whose purpose was to give me FIR filtering with maximized tap numbers for both an input stage (2 full-band FIRs, one per channel) and an output stage (6 FIRs serving as crossovers plus driver EQs). The resulting DSP is meant to receive source input and emit analog output for six drivers, 2 lo, 2 mid and 2 hi.

This required two minisharcs, one feeding the other. The first receives input and processes it through opendrc2x2 s/w with 6144 taps available per channel. Its input and output are via digi-fp. I cannot make the recommended spdif connection between minisharcs internally as the opendrc2x2 has no routing matrix and spdif-tx reads channels 7 & 8. Thus, the processed signal exits through digi-fp.

The second minisharc receives the output of the first via loopback cable. An RCA jack feeds a 2-pin motherboard connector. The second minisharc is part of a stack with minidac8, which happens to have a 2-pin connector for spdif that level-converts to TTL and passes that to the minisharc below. That sharc runs the minisharc4x8 plugin to accommodate 6 output channels. The 2 woofer and 2 midrange outputs get 2048 taps each (max), while 600 or so are left over for each tweeter output, more than sufficient at high frequencies.

Each minisharc is equipped with a vol-fp and usb (the first via digi-fp, the second via simple cable and jack). 25W of absolutely flat 5V power is supplied via a marvelously cheap supply seen in the photos attached.

RePhase is used to generate the filters. The crossovers are in 192-db LR format, the lower at 380hz, the higher at 4700hz. 3 Behringer power amps are currently used almost at idle to drive three very efficient drivers for each channel. At this writing, the LF speakers are inexpensive (but quite tunable) Seismic Baby Tremor cabs fitted with old but fabulous Altec 421-8H drivers. The mids are the modern form of the classic University Cobreflex horns, using Dayton PA compression drivers which not only sound great (the 70v xformer circuitry has been bypassed fully via rewiring) but also are expendably affordable. The tweeters, also very good and very affordable, are Eminence CD horns and drivers. All drivers are capable of high power consumption, yet I doubt even the woofers ever use more than 20W on peaks. The system sounds great at quiet through deafening (not kidding; one test killed a deer fly) levels.

Since I had to use digi-fp for input (to provide output for the second sharc), an ADC was required. I'm using the also affordable Behringer SRC2496, which not only gives a choice of formats (I'm feeding the digi-fp through toslink and using aes/ebu format) but also permits switching via front panel between digital input and analog input. A variety of parameters may be varied for best match.

Results are startlingly good. I believe this greatly affordable system rivals any audiophile system costing possibly many thousands more in subjective quality, not to mention acoustic power. I have always believed that commercial/reinforcement drivers provide unparalleled potential, needing only to be 'tamed', and now I can tame them without introducing phase anomalies to possibly interfere with imaging.

In the front panel photo, the unlabeled box sits atop a 4x10hd and beneath the SRC2496.

Thanks for documenting your project Leoman!
Sounds like a lot of work went through to make that custom enclosure and your dual sharc!

DevTeam

Cool project!

Just an addendum to the project to bring it up to date.

The capabilities of rePhase and the openDRC2x2/minisharc4x8 combination are permitting me to perform experiments on driver/xover combos essentially undoable with less capability.

For example, as I type I'm listening to the local FM symphonic station sound more lifelike than ever before owing to a new configuration on the minisharc, whose tap capacity is used now solely for crossover duty, any eq being done on the 2x2 with its 6144 taps/channel.

I am running just under 1800 taps/channel for my woofers and midranges, and 600/channel for my tweeters and supertweeters (yes, 4-way, more on that below).

This gives me the capability to cross over the Altec woofers at 380hz, 192db/octave (without way more taps this is as steep as I care to go as rePhase begins to fail to trace the curve perfectly, though still maintains linear phasing). Thus, after more than a decade of bandwidth, the woofers pass the signal to the Cobreflex mid horns, which operate for slightly under a decade, rolling off at 3 khz (but at 512db/oct!).

Here the rolloff capability at higher frequencies of the FIR filters, in conjunction with the digital delays built into the 4x8, permit me to do things classically considered heresy. It is nearly as difficult to find horns/drivers that handle everything past 3khz very well as it is to find one that operates from 380 to the ends of audibility. The continuity of a single driver is always subverted by some non-optimal behavior somewhere, rolloff, distortion, dispersion, peaking, whatever. I've gotten around this by implementing a 4-way system (always unthinkable to me) using remarkably cheap horns and drivers (also once unthinkable to me). The crossover is so abrupt that once shelved and phased properly the transition becomes transparent. Thus, limited horns and drivers are used only within their optimal ranges (true also of the big Cobreflexes).

Caution purists: you may gag at first. A small Pyle PDS221 compression driver loaded by a Dayton H6512 waveguide runs from 3khz only to 7.5khz, less than 1 1/2 octaves. It passes the signal, again at 512 db/oct., to an Eminence APT-50 supertweeter driver on an Eminence APT150S horn, which (with the proper curve applied) disperses the highest frequencies via its CD properties, and again handling less than 2 octaves.

These four smaller horns/drivers cost less than $100 total. I plan to keep playing around, but I am doubtful I will find anything so superior that I would pay the price. Plus they are bulletproof, something a guy who's lost expensive classic Altec, EV and JBL drivers in the past (without ever knowing what killed them) appreciates. And the 'sharc stack' is making the current set sound like nothing I've heard. Voila.

Next, priority shifts to seeing whether I can build a 6-channel class D power amp in a single small chassis to power the mid/high/super horns. Power should be easy; I don't believe that together at high volume these horns ever could demand more than 50 or 60 watts peak cumulatively. It would free up a lot of space, but if it compromises quality audibly I won't use it. I'll report on that when I know something.

Thanks for the update @ Leoman!

Lots of interesting opportunities indeed opened with the FIR filters. Keep us in the loop of your findings, we all love hearing of real life implementation and discoveries.

DevTeam

i want to do the same thing ,a 2in 6 out Fir x-over/EQ and would like to implement a set up like yours. do you use rephase to eq as well as x-over...for example once the x-over is generated for say for the high pass/ low pass of the mid range driver- then take a measurement? import it into rephase? do an eq with the slider adjustments? and import the corrected curve for the driver into the same dsp block that holds the x-over? .........what are you using the open drc for?an EQ for both L/R channels together-room correction?.......you clearing this up for me is much appreciated thanks

Hi Jacko,

I dabble with room correction but frankly have never found it a really useful tool even when properly executed, though I'm all set up for it if I want. My purpose was always to have full-band EQ (as with a plain old graphic or parametric equalizer box) feeding my crossover, but the best EQ/Xover I could get.

I use the FIR output filters on the minisharc4x8 plugin on my second sharc, which is stacked with the miniDAC8, to generate the crossovers using rePhase, which has linear-phase LR bandpass filters with whatever slope your allotted number of taps (and chosen frequencies) can handle. I can generate 512db/oct linear filters at higher frequencies and 192db/oct linear woofer xover at 380hz. Not bad.

I use the opendrc2x2 plugin on the first minisharc, the one that accepts input from the outside world via a DIGI-FP and sends it again via the DIGI-FP (on an external cable) directly to the second sharc. I use the 2x2 plugin on the first (input) sharc to get 6144 taps/channel for full-band FIR input EQ (that's my 'graphic equalizer'). RePhase is again used, this time with the paragraphic equalizer drawing my 'room curve'. If you wanted to generate room EQ, that would be where you'd do it. I just use the old manual method and my ears. The second sharc is well-enough occupied handling the crossovers.

Unless I miss my guess, in this modern world (i.e. the last month or so) you can do what I did minus all the box cutting etc. by sending your source material to an openDRC-DI and cabling (optical becomes an option) its output to the new openDRC-DA8. With the exception of the second sharc already having a DIGI-FP front end, I believe the combination would be exactly what I've written about here.

In my case, I hooked the first sharc via its DIGI-FP's coax spdif-out to the second sharc by using a small port on the miniDAC8 (on top of the second sharc) that does coax-to-TTL conversion and then feeds spdif-in on the sharc underneath. But I suppose if I had the openDRC-DI and the openDRC-DA8 I'd take advantage of the DIGI-FP available also on the second sharc and use the optical interface to connect them.

hello leoman......

that makes it clear that you are using the first drc to EQ the entire speaker (as a whole) after the x-over. and that when you implemented this set-up,the drc da8 was not yet available as the 2nd shark and that you are not taking measurements to EQ each individual drivers response 'flat' . ......thanks.

Yes, just one thing, the 2x2 occurs *before* the xovers, of course. But correct, not doing any eq added onto the crossovers. You could do it, but it's harder, so why not do it on the full-range input side instead? Plus I've found it easier and more precise to correct for things like CD horns on the 2x2 input eq. It leaves the limited per-channel taps of the 4x8 to deal only with making my xovers as steep as practical.

yes, the limited taps, and the fact that resources of the dsp are not being used efficiently i,e-4 inputs ......who uses 4 inputs

Hi Leoman!

Congratulations and my due respect for completion of this innovative FIR project!

My own long-term project is a 4-Way Speaker as well, and because of 4-Way I've been considering miniSHARC for some time. Although, I'm not using horn drivers, but "conventional" membrane, dome and AMT drivers. Also, I'm not a sophisticated electronics cabinet builder...

Well, to get to decisions here, I have a few questions which I hope you will be so kind to give answers on from your experience or judgement:

1. After finding 96 ks/s being the optimum resolution for my ears I's like to ask for your listening experience regarding the ASRC miniDSP uses. I'm asking this because e.g. DEQX does not do ASRC, but has dedicated filter sets for different sampling rates which "feels" technically better to me.

2. Do I understand correctly, that your "Sharc Stack" could be built using an OpenDRC-DI and an OpenDRC-DA8 plugged "in series" by SPDIF cable, yielding the same result or functionality you have built yourself?

3. When programming the two miniSHARCs, can you connect each board to one USB PC port each and the plug ins find "their" unit? Or is it necessary to plug/unplug a unit at a time with one USB cable?

4. Maybe I missed something or got something wrong... but I'm wondering about your usage of SPDIF to connect the miniSHARCs. So far as I understand, both could be connected via I²S, one as master, the other as slave, with the advantages of only using one clock source, avoiding the potentially "jittery" SPDIF coding/decoding and so on. What were your reaons to not go down the I²S connection route?

Sincere thanks for some insight!
Best Regards,
Winfried

Hi Winfried,

First thank you for your original suggestion!

1. I run the 48khz plugin on the input sharc, but my ADC is set to 96khz. It just seemed subjectively a bit clearer, though there may be no difference. I also use the optical connection and the ADC is set to AES-EBU, which seems to work fine. SPDIF would too, I suppose.

2. Far as I know, that is correct. Plus with the openDRC-DA8 as the 'output sharc' you'd also have the luxury of another DIGI-FP, so you could use the optical 'inter-sharc' connection if you should prefer it.

3. Each of my sharcs needs a USB port. The first uses the port on its DIGI-FP. The second uses a cheap port/cable that connects directly to the sharc. Since both are on the back I usually just keep two USB cables plugged in (and labeled!).

4. I expect you could use a I2S connection. My memory of this is a bit fuzzy now, but I had planned I2S originally. Devteam steered me toward the spdif. Whether that was due to suspicion of my digital h/w proficiency (not totally unwarranted, btw), just better faith in the spdif and ASRC functionality of the sharc, or perhaps suspected glitches in nonstandard clocking modes I don't know, but it does work quite well with spdif. Of course, with the two separate boxes you'd have no practical choice anyway I suppose. With my second sharc having miniDAC8 mounted to it, I was glad to avoid some possibly messy wiring anyway (not knowing at that point if anything would work, and looking for the most conventional path), but if you should try a single box with I2S I'd love to know how it works for you!

Regards,
John

Hey John,

thanks for your prompt and useful answers! Much appreciated! I'm really intreagued by this miniSHARC Stack

While seting up the "stack" using OpenDRC-DI and OpenDRC-DA8 looks quite simple as a solution, at the same time there is redundancy added/payed by having two VOL-FPs and two DIGI-FPs in the chain... So, maybe putting the two miniSHARC Stack, one VOL-FP, one DIGI-FP and the miniDAC8 board into a miniDSP 2x8 Box or something similar would probably make more sense and may even provide the space to include a decent powersupply in the box.

I've just read your several topics/posts on this project which gave me pretty good insight and direction. But a few pictures of your device would come in handy, as, unfortunately, the posted pictures are gone...

Thanks and Regards,
Winfried

Hmmmm......

Just did some budgeting which makes me really wonder IF building the SharcStack as DIY is worth the effort:

To me it has become very tempting to take advantage of not having to make the hardware work, mount boards in the case, do the wiring, and so on. Being able to focus on the Speaker alignment right from the start has a lot to it for me.

So, what would be the significant advantages of having all pieces in one box, i.e. doing this DIY? Maybe the Devteam can also make a suggestion here, as it also seems not advisable to connect the miniSHARCs wit I²S anyway

Thanks for helping to decide.

Winfried

Hi Winfried, I agree with your math. Also, at least as delivered to me, the minisharc/DAC8 stack wouldn't fit in a single-rack box, so you'd cut more holes than anticipated.

You'd see all that in the pictures. They seem to be there but are impossible to bring up (on either of my posts with pics). Perhaps victims of posting during a long site upgrade? Perhaps the devteam can report on this issue? If it persists I'll send a mail. Other posts with attachments seem ok, though they've been there longer. There's no economic incentive to discourage a sharc stack, as you've pointed out