Page 37 of 58 FirstFirst ... 27353637383947 ... LastLast
Results 541 to 555 of 864

Thread: Project M2 DIY Thread

  1. #541
    Senior Member pos's Avatar
    Join Date
    May 2007
    Location
    France
    Posts
    2,628
    Not completely sure I understand what you are measuring here, but the phase linearization FIR I included in post #538 is not part of the original FIR.
    The original FIR is minimum phase, as is your IIR correction.
    The phase linearization preset can be added to either the FIR or IIR correction and should give the same result in both cases.
    Its aim is to linearize the phase of the acoustical crossover (ie DSP filters+passive components+box/driver/horn).

  2. #542
    Senior Member
    Join Date
    Feb 2018
    Location
    Annapolis, MD
    Posts
    194
    I see. I misunderstood. Yes, just the IIR filter system got the linearization and if that wasn't built into the original FIR, then that would explain the variance.

    Note, when I outputted the LF Crossover FIR filter in rePhase to QSYS, I also removed the delay on the LF, opting to perform that within QSYS.

  3. #543
    Senior Member
    Join Date
    Feb 2018
    Location
    Annapolis, MD
    Posts
    194
    Okay...A quick question. The linearization is for the LF filter?

    What am I listening for, exactly? I've set up the system so I can A/V with/without linearization applied as well as A/B FIR versus IIR versions (or any combination of the above). Unless I'm missing something, the differences are very​ nuanced.

  4. #544
    Senior Member pos's Avatar
    Join Date
    May 2007
    Location
    France
    Posts
    2,628
    Linearization should be applied to both LF and HF part.
    You can apply it before the crossover if that fits your layout better.

    the differences are very​ nuanced.
    Isn't that what HIFI is supposed to be
    Seriously, yes, phase linearization of a properly executed minimum-phase crossover is a very subtle effect.

  5. #545
    Senior Member
    Join Date
    Feb 2018
    Location
    Annapolis, MD
    Posts
    194

    QSC QSYS Voicing Solution for the M2

    So I've been working with various means of getting as close a match as possible for using a QSC QSYS DSP on the M2. I think I've hit upon a very close solution. First, I used the FIR filter solution as provided by pos. Of course, these worked (note, these differ from the Google Docs version in that they are adapted to work with the 4096 tap limitation of the QSYS FIR filter, which necessitates adding an additional PEQ component with two of the filters in that for the LF section). This works fine and probably fine for many people here. Where I ran into issues was the size of the QSYS core I'm running (CORE 110c) and the number of speakers caused me to run out of DSP processing power. 5-sets of FIR based crossovers placed me over 100% processing power before adding anything else! If I were to move to the next larger CORE (510), then those same FIR based crossover would only consume about 12% of the processing power but it would double my cost.

    So I decided to adapt the JBL/Crown settings detailed in the Google Docs paper (links are elsewhere in this thread and in pos' signature) to the QSYS system using their "Voicing" component plus add in an additional PEQ component for the "Input EQ." The results are detailed earlier in this thread. The results were reasonably good as compared to the FIR based solution but did have some variations on the order of .3dB (not too shabby). The biggest difference between QSYS and either RePhase or the Harmon products' DSP interpretation is in the shelving filters. The PEQ filters were a good match.

    I was re-reading various posts here by pos and noting that improvements were made to the FIR filter design to get it to be a better match from the original attempt so I decided to look directly at the rePhase filters and start my QSC voicing with those rather than off of the Google Docs. In addition to some variances, there were additional (4) filters added to both LF and HF sections. So I incorporated all of that.

    And here are the results:
    Attached Images Attached Images    

  6. #546
    Senior Member
    Join Date
    Feb 2018
    Location
    Annapolis, MD
    Posts
    194
    Now it would be possible to combine some of those PEQ filters blocks and even do away with the "Voicing" module and do the crossover as a separate component...etc. but based on how this evolved for me, this was merely building on prior work and ended up with a reasonably clean solution and in an actual QSYS design is reduced to a single container so I'm going to leave it.

    Here is what the the filter results look like:

    Name:  QSYSVoicingFromRephaseHF180721.PNG
Views: 1369
Size:  41.6 KBName:  QSYSVoicingFromRephaseLF180721.PNG
Views: 1359
Size:  47.0 KB

    And here is the comparison of those filters to the FIR filters. Any variances will show up as something other than a flat line. The variance is less than .1dB with most of it to less than .05dB

    Name:  QSYSVoicingfromRePhaseComparedtoFIRLF180721.PNG
Views: 1189
Size:  47.3 KB

    LF comparison. Due to the 36dB/Oct crossover the response analyzer doesn't get enough signal much above the crossover to make a good comparison.

    Name:  QSYSVoicingfromRePhaseComparedtoFIRHF180721.PNG
Views: 1188
Size:  36.9 KB

    HF Comparison. It's pretty tight...it maybe a little off but we are talking about less than .1dB

    That is close enough for me!

    Most importantly...listening tests. I set up my listening system such that I could switch between the FIR solution and this "Voicing" plus additional PEQ solution and to my ears, they are either the same or equivalent. If there is a difference, I'm not picking it up and neither are two other people I had take a listen.

    -Steve

  7. #547
    Senior Member
    Join Date
    Nov 2008
    Location
    Gothenburg, Sweden
    Posts
    761
    Nice !

  8. #548
    Senior Member
    Join Date
    Feb 2018
    Location
    Annapolis, MD
    Posts
    194

    QSYS DSP usage

    So, wrapping up my QSYS story here (at least prior to installation). Here is some potentially useful information for others using QSYS in their M2 futures (naturally, technology may render the limitations I ran into obsolete but it was accurate as of today).

    An empty QSC CORE 110 (the smallest of the Cores) will show that it is consuming 2% of its CPU capacity.

    By adding in a single M2 crossover using FIR components, that will jump to 47%!

    Name:  Screen Shot 2018-07-22 at 3.27.14 PM.jpg
Views: 1208
Size:  77.4 KB

    In my case, doing a theatrical based Atmos system, I'm using 5 M2s. That jumps you up to 86% of your CPU before you put in ANYTHING else like the Atmos receiver components or whatever else your design may need or want.

    Name:  Screen Shot 2018-07-22 at 3.57.48 PM.jpg
Views: 1167
Size:  79.9 KB

    That just wasn't going to work on this core. A CORE 110 can handle up to 128 inputs and 128 outputs and starts you out with 8 analog ins, 8 analog outs plus 8 flex (can be either in or out, individually, by configuration). While the next core up, the 510 would drop that figure to about 12%, it comes at double the cost an starts you with no inputs or outputs (but has the capability of 256 inputs and 256 outputs) so one has to factor those costs into moving to a larger core just to accommodate the FIR filters in the quantity I'm using them. For those that are just wanting to listen to stereo (I suspect a large majority), then your CPU usage is the same as one speaker to just handle the FIR filters of two speakers.

    The probably means, depending on your other uses for the QSYS system, that you have enough CPU to handle two speakers plus those other uses, particularly if it is just I/O switching or some automation/control stuff. Of course, I'm doing Pro-Logic decoding and other processing in my design so I wouldn't be as fortunate.

    So, let's look at things via QSC's "Voicing" component (handles crossover and up to 9 PEQs/HiShelf/LowShelf, per band, LPF, HPF) plus conventional PEQs. Adding in the first speaker puts me up to 6% of the CPU (a far cry from 47%!).

    Name:  Screen Shot 2018-07-22 at 3.19.57 PM.jpg
Views: 1176
Size:  75.5 KB

    If I go up to 5 speakers, it only consumes 10%

    Name:  Screen Shot 2018-07-22 at 3.33.17 PM.jpg
Views: 1189
Size:  87.5 KB

    And, if one uses QSC's DPA-Q or CXD-Q amps, the processing for the "Voicing" module moves to the amplifier so it does not come off of the Core's CPU overhead and your processing drops to a mere 4% for all five speakers with 2% of that just for the the Core to be on. All that is left is the input PEQ plus the LF and HF PEQ additions (4 PEQs each).
    Attached Images Attached Images  

  9. #549
    Member
    Join Date
    Jul 2017
    Location
    Florida
    Posts
    65
    Steve

    A lot of good info there, must admit the only time I ran into CPU utilization issues was in the design phase when I was just throwing stuff into the design and not taking it out. I only have 3 M2's in my final design so the FIR filters are working out pretty well for me. One thing that went through my mind and I will have to try it myself. After setting up the FIR filters for the 5 M2's did you try running the DPA_Q or CXD_Q amp in standalone/slave mode this offloads the processing to the amp rather than having the core do everything. Would be interesting to see what that does to the Core utilization if anything at all.

    Haven't fired up the designer software in a while but I just picked up an I/O frame since I'm looking to do a fully digital DCI compliant audio path from the Altitude through to my CXD_Q amps using a CIAES16 card if I can find one.

  10. #550
    Member
    Join Date
    Jul 2017
    Location
    Florida
    Posts
    65
    And for anyone else thinking of building an M2 clone I can confirm that all of the Parts CD / Waveguide / Woofer are still readily available at Speaker Exchange. Currently 7-10 day lead time. Have 3 of each arriving today.

  11. #551
    Senior Member
    Join Date
    Feb 2018
    Location
    Annapolis, MD
    Posts
    194
    I have not tried to run the amplifier in standalone mode BUT...I didn't think it could take any of the processing other than the voicing module but I'll be sure to give it a try. I'd think that the amp's CPU would be significantly less than a CORE. I had planned on DCA amps on the M2s but am now also considering DPA8.4Qn. In order to do standalone mode, I'd need to switch it to a DPA8.4Q (adds some money) and lose a benefit of the Q series in that it hooks up with just an Ethernet cable rather than running traditional analog cables.

  12. #552
    Member
    Join Date
    Jul 2017
    Location
    Florida
    Posts
    65
    Quote Originally Posted by sguttag View Post
    I have not tried to run the amplifier in standalone mode BUT...I didn't think it could take any of the processing other than the voicing module but I'll be sure to give it a try. I'd think that the amp's CPU would be significantly less than a CORE. I had planned on DCA amps on the M2s but am now also considering DPA8.4Qn. In order to do standalone mode, I'd need to switch it to a DPA8.4Q (adds some money) and lose a benefit of the Q series in that it hooks up with just an Ethernet cable rather than running traditional analog cables.
    Certainly not even 50% sure if that will work, it was just a thought. Your memory on the subject is a lot fresher than mine. The I/O obviously needs to go through the amp rather than be routed through the core to run it in standalone mode so that would rule the Qn's out for sure. It was just an idea I can probably test this out on my system fairly easily since I have 4 of the 4 channel CXD_Q amps therefore I don't think any of the inputs go into the core, no reason not to just throw in a few more FIR filters to see how it responds.

    Saw the CXD8.4Q and DPA8.4Q and n's at infocom last year prior to their release, perfect for surrounds and Atmos speakers. You've got my brain churning again.

  13. #553
    Member
    Join Date
    Jul 2017
    Location
    Florida
    Posts
    65
    I did a bit of testing last night and you were spot on, the Standalone mode works nothing like I thought it did, more like a limp mode since it just allows the amp to boot up and play or continue playing if it looses connection to the core. I didn't bother test it extensively but it was quite clear that no FIR filters are stored in the amp when in that mode, and you cant just tell the amp to go standalone you have to cut the power to the core or disconnect the core from the network otherwise the amps are still using the cores DSP resources.

    What was slightly more interesting is that even though I duplicated the 6 FIR blocks and PEQ's that I use to run my 3 M2's, so I now have 12 FIR blocks logically connected to speakers I couldn't get the utilization up above 60% and that was with the design running on the core not in emulation mode.

  14. #554
    Senior Member
    Join Date
    Feb 2018
    Location
    Annapolis, MD
    Posts
    194
    I'm reasonably certain that a speaker Voicing will work in stand alone mode (makes the DPA-Q amp effectively into a DPA amp. So you can set up what channels do what and put the crossover in there and that is about it.

    As to CPU usage...what Core are you running? I ran into the limit running a 110c, not in emulation mode. My FIR blocks at 4096 coefficients each.

    When I bumped into the ceiling, I was doing other processing and had several Pro Logic blocks in the design (they are, currently, just in the "cinema" version of designer so don't go looking for them or the DPA-Q amps in the inventory.

  15. #555
    Member
    Join Date
    Jul 2017
    Location
    Florida
    Posts
    65
    I’m using the core 110f, most likely the same hardware but sold into different markets. Same as CXD_Q & DPA_Q amps. I’m also using 4096 in the FIR blocks. Could be the Prologic blocks in your design that account for the additional cpu requirements.

    I found the download link, I’ll have to take a look at that version of the designer and see if it runs on my core.

Thread Information

Users Browsing this Thread

There are currently 2 users browsing this thread. (0 members and 2 guests)

Similar Threads

  1. JBL LN3 Project, *Official Thread
    By Nightbrace in forum Lansing Product DIY Forum
    Replies: 58
    Last Post: 07-13-2016, 03:08 PM
  2. Fundraising request; donations to Lansing Heritage Project May loudspeaker project
    By mikebake in forum Lansing Product General Information
    Replies: 7
    Last Post: 04-21-2011, 12:37 PM
  3. Project May, similar privat project?
    By Flodstroem in forum Lansing Product DIY Forum
    Replies: 14
    Last Post: 02-21-2007, 03:42 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •