The best way possible to build an active system.

dallasjustice said:

I use symmetrical 2 octave digital crossovers for 2 subs and my full range speakers. So it's a four channel system for stereo playback. I own and have used both Audiolense and Acourate. I've also used Dirac in the past, but don't use it anymore.

The issues you will run into have to do with hardware/software ASIO interface. I'm not going into it here because it's not important. But if you are really are serious, talk to Uli and you will get very good info about the requirements needed.

Btw, there are differing opinions about whether speakers can be measured in normal rooms quasi anechoic. I think Uli would give you a different opinion than me. But if we are talking "best practices", the actual driver measurements will be lower resolution in a normal room because you will need to window the impulse before the first reflection due to lack of real estate.

Please understand, I'm not talking about the measurement at seated position. That's a different type of measurement and different filters are needed for those types of measurements. I'm just talking about driver linearization. Driver linearization and crossovers have to be done in an optimal way which takes into account the driver polar response over frequency range and cabinet/baffle effects (eg. Baffle step effect and diffraction).

Click to expand...

I went on the Acourate forum and found the only feedback on the Merging Hapi being used with Acourate:

"Hi,

I am using Hapi in my system since this summer. After first few days of
struggle, it has been very stable. I think it gives better sound than
RME AES32.

The music signals (2 to 6ch) from my transports go to Hapi via AES/EBU,
and then to my convolution PC via eathernet/DELL switch/eathernet. After
convolution by Acourate Convolver, the signals (6ch) are sent back to
Hapi via the same route. Then Hapi sends the front two channel signals
to my DAC via AES/EBU, and another four channel signals (two subwoofer
signals and two rear signals) are converted to analog in the DAC card of
Hapi.

I tried Jriver in the above convolution PC, but failed. If you skip AC,
Jriver works fine. If you want convolution, then AC refuses connection
with Hapi. AC does not wait for Hapi's response, but Jriver can wait for
long enough time.

Therefore currently I use Jriver in Mac mini in the same Ravenna
network. Mac mini sends signals to the convolution PC via eathernet as
above, and everything works fine as above."


Doesn't sound like too much of an issue to me. Sound more like a Window's Jriver glitch. Who knows they probably sorted it out by now anyways. A new version of Jriver was released since. Besides Jriver isn't the only media player software out there.

It also seems like a pretty unorthodox method he's using with the Hapi. I wonder what he's using for transports that he's routing through the Hapi? Sounds like he's just trying to utilize other gear he already had. Would probably be better off keeping it simple and just using Jriver as the media player, the Hapi as the DAC, and taking his transports out of the picture.

Keith_W said:

Blizzard, I have been looking at the Acourate website in detail but I can't find an answer. My question: if you are using your PC as the "brains" for a digital crossover, how does it output multi-channel digital for use on your DAC's? If you were to use a DEQX or MiniDSP, the answer is - multiple digital outputs can be found on the back panel. Also, what digital format is sent out by the PC? DSD? 44.1/48kHz 16-bit PCM? The "hardware requirements" does not list any special hardware required.

Click to expand...

All depends on the DAC. If you used an Exasound E28, you just connect by USB cable. If you use any of the Merging DAC's, they use an Audio of IP protocol called Ravenna. So you just connect via Ethernet cable.

Normally DSP can only be applied to PCM signals. That's because DSD is only 1 bit and it can't be altered. Only multibit formats can have DSP applied to them. However HQ player can convert DSD to multibit DSD, and apply DSP to DSD. Only problem is the GUI is pretty clunky. But they are working on integration with Roon, so an awesome GUI can be used with the power of HQplayer. Keep in mind going this route, serious processing power is required.

The easiest way is to just use Jriver. Then you would just generate convolution VST plugins for the Xovers and room correction, and load them into Jriver's convolution engine. Only problem with Jriver is if you want to play your DSD tracks, you will have to convert on the fly to PCM which kills the quality.

Keep in mind you'll need a good measurement microphone and microphone preamp. The better the mic and mic preamp is, the better your results will be. I recommend buying CLIO pocket:

http://www.audiomatica.com/wp/?page_id=1739

There's cheaper options, but they won't be as good. This is a precision system. I've been using CLIO for 15 years and have no complaints. This system just came out last year. I paid $2500 for my original CLIO PCI card based system, so this is a bargain in comparison. You can buy it here in the U.S:

http://www.parts-express.com/audiomatica-clio-pocket-personal-acoustic-measurement-system--390-900

Then there's the Minidsp UMIK-1 as a budget option:

http://www.minidsp.com/products/acoustic-measurement/umik-1

It has a built in preamp and is plug and play with USB. But I wouldn't recommend it for Acourate.

Here's some info on using convolution with Jriver:

http://wiki.jriver.com/index.php/Convolution

Keith_W said:

Thank you for your detailed reply, Blizzard. I saw it a couple of hours ago, and have spent the time reading up on the Merging Hapi and Exasound. I think I will need a Hapi, since I need A-D conversion.

I am somewhat confused by what appears to be multiple RS232 DB25 connectors at the back of the Hapi. Who uses those any more? The pictures are rather grainy and I am not able to read the text. If those are AES/EBU connectors, they look different to what I am used to (the AES/EBU connector I have in mind looks like an XLR port).

I already have a calibrated microphone, calibrated sound card, and use REQW for my measurement setup. For years I tried to tune my system by ear, but this is much faster. A friend bought a new subwoofer and has been trying to tune it for weeks. I turn up, and within an hour I had it all set up for him. This is a small investment that every audiophile should make, IMO!

Click to expand...

No problem. The Hapi uses DB25 connectors because it's a prosound unit and that's a prosound standard. So you'll need the breakout cables. I reccommend the Mogami Golds.

http://www.mogamicable.com/category/products/gold-AES_TD_DB25-XLR.php

But you should also consider that the Hapi with the DA8P DAC board is only $4200, and it uses the same DAC board as the $11500 NADAC. So might be worth it having to deal with the breakout cable.

View attachment 22615


I posted this already, but here's a FAQ on the differences between the Hapi and NADAC:

https://confluence.merging.com/plugins/servlet/mobile#content/view/9338933

And here's the Hapi manual:

http://www.merging.com/uploads/assets/Installers/Horus/Hapi_User_Manual_rev06.pdf


If you decide to go the Jriver route, I'd reccommend ripping your vinyl to DXD (24/352.8) it's a PCM based standard that Merging created as the next best thing to DSD. You have almost DSD quality, but it's multibit so easy to apply DSP.

Rodney Gold said:

Have a look at this speaker , download the white paper
Thats the way to build a DSP speaker
Cheap too foir what you get
80k
http://www.bang-olufsen.com/en/sound/loudspeakers/beolab-90

Click to expand...

Yeah looks pretty good. I was already checking it out.

I read that if connected digitally via USB or WISA, it goes through an TI SRC4392 SRC which either upsamples or downsamples all audio to 24/192 PCM. If analog input is used, a TI PCM4220 A/D converts to 24/192 PCM and skips the SRC. From there 1 channel each goes into dual ADSP-21489 SHARC DSP chips, where DSP is applied to 18 channels. Then from there the SHARC chips send 18 channels of digital 24/192 PCM to 9 separate TI 1798 DAC chips to be converted to analog and sent to Icepower amps for the mids and tweets, and Heliox amps for the woofers. It uses AVB 802.1AS protocol to keep clock sync between the 2 speakers.

I wonder why they didn't use their own amps for the woofers?

I can just imagine where this concept could be taken with Intel Xeon based 64 bit floating point DSP, DSD 256 capability, state of the art DAC's, and Hypex Ncore amplification. But of course the MSRP would be much higher than $80000.

Here's the complete electronics for 1 speaker.

View attachment 22631

Hi Guys, I thought I'd share some pictures of the speakers I built to test DSP electronics. I just used some old Focal "W" cone midwoofers and Raven tweeters I had laying around for years. I built passive Xovers for the top section, as reference. In the passive configuration, the upper midbass is actually running full range. I was able to run these full range, because of special dampening pods on the back of the cones giving the mids a very smooth response up to 9k with a gentle rolloff. The lower midbass uses a single Mundorf Zero ohm inductor in the signal path to gently roll it off at 300hz. This technique is called 2.5 way, and what the lower woofer does is act as baffle step compensation for a flat in room response down to 100hz at the listening position. The tweeter has a 4th order crossover using Mundorf silver in oil caps bypassed with Dueland silver foil and wax/oil bypass caps. Inductors are Mundorf copper foil, and resistors are the incomparable Dueland CAST carbon/silver. I'm using with Hypex NC500 based amps to power the top section, and some Hypex DSP chip based boards going to NC500 based amps to power the lower woofers in the passive config.

The cabinets are constructed from a very high grade of Latvian baltic birch. Viscoelastic dampening glue was used between the layers. Each 18mm 13 ply layer was tuned to a different frequency to cancel out the vibration of the next layer. This technique resulted in extremely dead cabinets. It also made the inner walls have no parallel panels. Theres an enclosure on the back to house electronics once I'm finished my testing. This enclosure will be extremely well dampened with special constrained layer materials to isolate vibration from the electronics eventually. A 3/8" thick black anodized machined aluminum plate fastened with stainless Allen head machine cap screws will be flush mounted to the rear.

I'm still waiting on parts for my state of the art DSP electronics. So far the upper speaker sounds extremely good. By running the upper mid full range, I'm getting perfect phase response with the Raven tweeter's 4th order xover. The soundstage goes as far as 10-15' back and the speakers completely disappear. With the passive version having nothing at all in the signal path of the mids, it will take extremely transparent DSP to match midrange purity I'm getting now. I'm not that impressed with the DSP chip Hypex boards. I firmly believe master clock sync between all DAC's is extremely important.

Here's a couple in room measurements of the passive configuration. No room correction or DSP is applied. Completely passive from 100hz up and a 48db per octave DSP low pass filter on the woofers. No other DSP applied to the woofers. I'm quite happy with the passive results considering the upper mid is running full range, and lower only has a single inductor with almost 0 resistance in the signal path.

One measurement is on axis 1m back, and the other is in the center between the speakers 2m back.

View attachment 22676

anders said:

Blizz have you made any progress with upsampling to DSD256 and native DSD convolution? I've been slowly making my way through the HQPlayer thread on Computer Audiophile and the reports seems to be mixed, seems people are having problems with how much processing is needed with audible glitches.


IMHO that is a very fast processor, seems like you are going to need enormous processing power to accomplish this.

Click to expand...

It's a very processor intensive process. But all depends on how many channels your trying to process at once. Which is why I'll be using a 12 core Intel Xeon processor. I'll also be doing everything on a real time OS.

I'm hoping this will do the job. 1 core dedicated to each channel:

http://ark.intel.com/products/81713/Intel-Xeon-Processor-E5-2690-v3-30M-Cache-2_60-GHz?ui=BIG

anders said:

That quote is from someone not even using it for crossover duty; just room correction and his processor is a 3.4 GHz i7... I'm really not sure if a 2.6 GHz 12 core Xeon would cut it for your much more intensive use? Either way I am glad someone is being the guinea pig and look forward to your results

For your DIY DAC are you going with ESS multichannel, chipless like Miska's Signalyst DSD hardware or something else?

Click to expand...

Dedicating 1 core for each channel it should handle it smoothly. If not I'll have to get something more powerful

@7x the price, I better hope it can outperform the 3 year old 3770S.

Keep in mind each of those 12 cores run at 3.5 GHz in turbo mode. This is a powerhouse processor.

I'm testing a few DAC configs at the moment. Including a new SOTA chip that hasn't been released to the market yet