MiniDSP... a bedroom affair.

Discussion in 'Speakers' started by ultrabike, Aug 1, 2016.

Tags:
  1. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    So f**k! I'm having a long and hard time equalizing my bedroom.

    I heard many times how frequency response is not everything. How it doesn't paint the whole picture and all that crap. The fact is, frequency response does paint a pretty good picture of what's going on. The deal is, there is not one frequency response. There's many.

    Here are my Mackies measured in different room conditions:

    1) Here are my MR5 MK3 FR measurements in my bedroom (no smoothing) showing clearly reflections, reverberation and shit:

    (I only saved the FR after a very steep crossover at 80 Hz, however w/o the crossover the extension is about 45 to 50 Hz or so, similar-ish to the extension I got at Guitar Center - see plot 3)

    Bed_no_smooth.png

    2) Here are my MR5 MK3 FR measurements in a park (almost anechoic, but not quite) also w/o smoothing:

    Park_no_smooth.png

    3) And here are the ones from Guitar Center (w/o smoothing):

    GC_no_smooth.png

    All that noisiness in my bedroom and GC shows the effects of the room, and they definitively present in the frequency response. An equalizer cannot fight the comb filtering due to bad room acoustics because you move a bit here or there and a new set of comb filtering issues happen. Not to mention nulls are not solvable... shit loads of nulls BTW.

    To make sense out of the things we smooth things out and correct the smoothed out response... which is a frequency response different from the one w/o smoothing.

    Here are the smoothed responses on top of each other (again, the blue is my room w a hard 80Hz crossover in the miniDSP enabled):

    blue - Bedroom (w the 80 Hz LR crossover)
    red - Guitar Center
    green - Park

    Together_1_12.png

    It's still a bitch. As it turns out, the miniDSP has two apps. One affords one the ability to matrix things out to a sub, but only has 2 5-band PEQs for left and right, and 4 sets of 5-band PEQs for the outputs. The other losses the matrix, but affords 6-band PEQs instead. I have both and got confused with how things are going from inputs and outputs. 5-bands and even 6-bands are not that much and one does the best one can with those. I would rather have at least 10-bands on each input and output channel. It is what it is and what it is is good. But it's not that easy. Anyhow, will continue testing and post some on the results later.

    End of ... today here is what I get with the miniDSP + SUB1200 integrated (1/12 smoothing):

    1 meter on-axis
    red -Left
    green - RIght

    sub_corr_1_12.png

    At listening position things slant a bit even if I toe in and bass gets a bit happy, proly because of my bed:

    Bed around where I rest my head (i.e. pillow)
    red -Left
    green - RIght

    sub_corr_1_12_bed.png

    I'm fine with that above. Can't solve problems in every point in a small room.

    However, I may have to move the crossover point up again to 100 Hz, because the damn sub fails to detect signal from time to time.

    Crazy shit note:

    My understanding is that digital PEQs are in general a bank of 2nd order IIR filters (one per band) whose complex pair poles are adjusted towards the unit circle to achieve a peak, and whose zeros are also adjusted towards the unit circle to achieve a null. The closer the pole or zero is to the unit circle, the higher the Q. The phase of the poles corresponds to the frequency. There is also an associated gain. They may be analog as well, with perhaps associated second order cascaded filters. Bi-quad nomenclature for each of the bands may corresponds to the ratio of two (bi) second order (quadratic) polynomials describing the IIR structure.

    By contrast, what is called a "convolution" filter is a pure high order FIR with no poles and a shit load of zeros. An 18000 tap FIR such as the Mega-donkey-dick proly has the around 9000 complex zeros and maybe equivalent to a 9000-band PEQ that can only do nulls. Optimization of such filter to eq stuff (instead of interpolating stuff for DAC-off purposes) may be done using an adaptive algorithm such as the LMS or it's crazy off-springs (RLS, LRLS, NLRLS, PMS...) or whatever.

    Hopefully, this will be supported by REW for the miniDSPHD... or maybe some other awesome-sauce SW, to load into the miniDSPHD.

    Crazy shit note 2:

    Continuing on my frequency response rant. Most measurements are done with tone sweeps which yield harmonic distortion as well. However if we were to have two tones playing at once, one at say 60 Hz and playing hard, and another at say 10 kHz, then if the same driver is playing both tones we would have what @Hrodulf called in another thread "Doppler" effects where the 10 kHz tone source is no longer stationary and would result in the 10 kHz tone to be off in frequency (jitter). This would show on a frequency response plot, but not in one where only one tone at a time is played...Well, not exactly a frequency response plot, but it would show in the frequency domain in a PSD plot (Spectrum Analyzer). If two drivers were used, this Doppler effect may be reduced, but now the source of the tones is not collocated, which results in other issues... One could go coaxial, but then who knows how much control we have at the crossover point... Again, I feel all of these things will show in the frequency domain.

    Anyhow. This is sort of a rant thread (again), with bits and pieces of information (hopefully not too far off from reality) for folks to use as they see fit.

    I may have to re-adjust my miniDSP in which case I will dump some stuff related to it here.
     
    Last edited: Aug 1, 2016
  2. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,594
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Why don't you aim for one of the downward-sloping speaker targets at the listening position? You should at least try something like the B&K target and see what you like better.
    Also, why don't you look at the IR for the two rooms? I feel like one doesn't "hear" nulls, but instead we hear an echo from a certain location at a certain time. Maybe you can reposition your speakers to avoid first reflections as much as possible.

    I think that usually the excursion for drivers is very low at frequencies where it moves very fast and vice versa. In other words, I think that it is hard to get high excursion at a high enough frequency for it to really change the frequency of the tone played. Then again, 20% IMD on a two tone test?
    The other way of looking at it is that the phase of the signal changes as the driver moves, as the distance to the listener varies over time.

    EDIT: Just did my own doppler/frequency/phase modulation distortion tests with my HD800:
    I used a 2Hz tone to modulate a 500Hz tone. I used this website as a tone generator: http://www.szynalski.com/tone-generator/
    I set the levels very high and the distortion was clearly audible, even when I increased the lower frequency. The levels aren't accurate and I had the 2Hz tone a lot louder than the 500Hz tone (500Hz was at 25% vs 100% for the 2Hz tone). It was very loud. The 2Hz tone was probably equivalent to >110db, if it extended that low.
    Doppler distortion torture test3.png
    Would be interesting to try similar things for other headphones, but at more sane levels and a more realistic test (i.e. not 2Hz).
    I should probably look at the waveform.
     
    Last edited: Aug 1, 2016
  3. Hrodulf

    Hrodulf Prohibited from acting as an MOT until year 2050

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    4,276
    Trophy Points:
    113
    Cool post is cool, Ultra!

    About your listening spots - it will be impossible to optimise for both, because two different spots will have different room modes impacting the outcome. The only way is either heavily treat your room or use different output systems. Both are hardly optimum.

    As for your sweetspot curve - it's very decent. Especially with no treatment. I've seen much worse from supposedly treated studios. Here's what I get -

    [​IMG]
     
  4. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    There is a "Psychoacoustic" smoothing mode in REW. I tried it for equalization. It made things quite bass heavy. Particularly at listening location. So I had to redo the whole thing again but with the usual 1/12 smoothing. But you are right. Certain kinds of nulls are heard like room coloration. Because if one moves from one location to the next the nulls move and one's brain fills the gaps. And they are because of room reflections, and sound like so. For nulls that are persistent, because of speaker issues, that may not always be the case. Things may just sound off. Those are the ones equalization may help with... to some extent.

    I think you can do this with a 60 Hz tone and a 1 kHz tone on a driver that is responsible for both frequencies. Set the 60 Hz tone sufficiently high to see some excursion and see how it affects the 1 kHz tone (particularly with a speaker). When a single driver is responsible for 20 or 30 Hz and 10 or 15 kHz the excursion due to the low frequencies may result in issues at the high and mid frequencies. This is probably more of a problem with speakers which had higher excursion to push more air than with headphones.

    I might try doing this as well. But the point is, it seems one can never have perfection. It's just a bunch of trade offs and priorities.

    Depends on the width of the null also and how persistent it is. If a null is due to comb filtering due to some reflection happening long time after the IR peak and settling time, then I think it's more like echo (or bunch of echoes). Room coloration.

    Yup. We do what we can. :-(
     
  5. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    I listened to the setup as is today. It sounds good to me. I think the problem is that I measured below the sweet spot, but I'm too lazy to measure again today. Maybe later.

    So here is what I did:

    These are the two MiniDSP apps that are considered for a 2.1 system along w their top level description:

    1) 2x4 Advanced

    minidspadv.png

    2) 2 Way Advanced 21

    minidsp2waysubadv.png

    There are others that might work but those are the ones I have. For a 2.1 system like the one I have in my bedroom I recommend the 2 Way Advanced 21 because each of the Parametric EQs are 6-band instead of 5-band as is the case for the 2x4 Advanced. One looses the Routing deal, but the IMO is not a big deal. That maybe important if the sub only has one input. Mine has two for left and right channel. You need all the PEQ biquads this thing can give you, and I mean every single one.

    The MR5s do seem to have a bit of hump in the mid-bass area. It also needs to be tamed to sound moar awesome sauce. So first I set the crossover to 80 Hz using a Linkwitz Riley 48 dB/octave HP for the monitors and LP for the sub outputs. Which is why my measurements in room roll off sooner than the others above. The hump needed to go.

    I measured the monitors with REW, applied 1/12 smoothing, and used the EQ application. I set the target to flat. I think I set the range for matching from 80 Hz to 20 kHz, but one may play around with this range depending of what one wants to take care of first. Apply EQ biquad results to Output PEQ 3 (left) and Output PEQ 4 (right) to the respective monitor. REW spits a file that can be read by miniDSP. Use the miniDSP setting in REW for this.

    Then I measured the sub after crossover using left and right channels. One needs extra power there to have enough signal to equalize all the way to low-bass. But not too much so as to make the sub insensitive to input and not auto-turn-on. Apply results to Output PEQ 1 (left) and Output PEQ 2 (right) to the respective subwoofer channel.

    Next we measure the whole thing with 1/12 smoothing. If there are big nulls in the sub range make sure there are no inversions. I had one relative to the right channel due to sub location relative to the monitor. If there are inversions, the signal output to the sub can be inverted. Then run EQ application in REW again and apply EQ settings to Input PEQs for each channel.

    And that's it...

    It works and it works very well. But it does not cure room cancer.
     
  6. Hrodulf

    Hrodulf Prohibited from acting as an MOT until year 2050

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    4,276
    Trophy Points:
    113
    How many measurement points are you using for your sweetspot curve?
     
  7. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    Just one actually, which is not optimal at all. Will redo later.

    More than hitting the sweet spot I was measuring each system on-axis, 1 meter away and correcting in place.

    It is probably better to correct using an average of a cluster of measurements around the sweet spot.

    I also need to characterize the monitors better off axis horizontal and vertical. I think the tweeter design minimizes some dispersion to help with room interaction (regardless of what Mackies marketing says). I say this because the tweeter waveguide looks similar to this one:

    http://www.parts-express.com/dayton...odymium-tweeter-with-waveguide-4-ohm--275-051

    I don't think the tweeter is down that low into the waveguide in the MR5, but the approach seems to actually minimize dispersion and looking at their FR plots off-axis it seems the behavior matches somewhat what I remember getting off-axis in the past with the MR5s (and other similarly designed monitors). This may not be that bad in a small room actually. But we'll see.

    JBL did things differently. They seem to try to maximize dispersion... up to 10 or 11 kHz, at which point things go down hill.

    Like I said. Will redo stuff again with an average from measurements around the sweet spot instead of 1 meter away on-axis with the tweeter.
     
    Last edited: Aug 2, 2016
  8. SineDave

    SineDave Friend

    Pyrate
    Joined:
    Apr 18, 2016
    Likes Received:
    862
    Trophy Points:
    93
    Location:
    Houston, TX
    Home Page:
    Sorry to thread necro, but wondering how this turned out. I wanted to pick up a MiniDSP HD for my computer 2 channel rig, but I was too lazy to search for a case. What did you settle on case wise?

    From my experience with REW and room correction (I used to work with the developer John Mulcahy at HTS) the audyssey measurement approach is pretty effective for this. I like to measure at least 5 positions around the MLP, including 2 at the position of the ears (one on each side), some forward, some back and at varying heights. This usually nets me the best data to program corrections from.
     
  9. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
  10. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    BTW, there is also this:

    https://www.minidsp.com/products/dirac-series/ddrc-24

    it seems like a MiniDSP-HD HW-wise, but Dirac enabled (FW) for a premium price.

    In general, the main advantage of the MiniDSP-HD over the MiniDSP is it's superior DSP (not just the I/O specs). Because of this, unlike the MiniDSP, the MiniDSP-HD supports full FIR equalization. The DDRC-24 integrates their own Dirac FW. It still requires their mic on top of that, probably due to driver support limitations (can't use just any random soundcard + mic).

    I think the MiniDSP-HD is probably a better value, but their Dirac SW may make a big difference. If so they do offer the MiniDSP-HD to DDRC-24 FW upgrade:

    https://www.minidsp.com/products/dirac-series/2x4hd-to-ddrc-24
     
  11. SineDave

    SineDave Friend

    Pyrate
    Joined:
    Apr 18, 2016
    Likes Received:
    862
    Trophy Points:
    93
    Location:
    Houston, TX
    Home Page:
    I am actually in the process of writing a detailed review of the Dirac product for HT use. It's a great tool, but I am not sold. QSC's core 110f can do just as much or more, it's just not as plug and play.

    I already have their mic, since I use it with REW. Less annoying than using a sound card for quick measurements, though not as accurate as my Earthworks EMM8 and preamp

     
  12. Jh4db536

    Jh4db536 Friend

    Pyrate
    Joined:
    May 25, 2016
    Likes Received:
    926
    Trophy Points:
    93
    Location:
    Los Angeles
    Home Page:
    This is a different product than Ultrabike's miniDSP. They sell many different configurations and DIY boards to pretty much cover all your needs. The NanoDigi2x8 has no analog functionality. It is strictly digital (SPDIF optical and coax i/o). It can control up to 8 channels or basically 4 external DACs of your choice.

    I moved rooms. This one is 12' x 10' perfect rectangle and symmetry. The small size makes bass management extremely difficult.
    Nevertheless it is a dedicated listening room and i pretty much have a 5' equilateral triangle (similar to 38% rule) in order to avoid room effects. Sub placement pretty much has to be under the cabinets in order to image correctly. Room is way too small for 4 subs or front and rear position which is the optimal setup according to papers i read.

    BTW, GIK244 really works well for acoustic treatment. in my discussions with GIK, they told me the optimal price performance setup was to hit the side walls and ceiling to absorb the first wave. Since ceiling is not an option for me right now, i opted to put the treatment behind the cabinet which is verified as optimal as i can get for $200 in treatment. Look at the foothill of the red mountain, which decays very evenly and smoothly.
    [​IMG]

    Objectives:
    1) Manage and integrate a 2.2 setup. Little to no analog fukery.
    2) Not PC based and utilize existing gear to the extent possible.
    3) Keeping it "Pure" nothing between the DAC AMP and Mains. I attempted an Active Pre prior to this and i frankly lost all confidence in them.

    This could be considered blasphemy against traditional methodologies but i used a NanoDigi for under $175. This should allow me to preserve the SQ between the two big DACs that i already own (could use a Modi2U for sub amp for all i care). It provides for master (an individual) volume control and remote control and enhanced the capabilities of everything downstream.

    So the chain is as follows:

    VRDS25 (SPDIF) > NanoDigi (SPDIF) > DAC#1 (SPDIF) + DAC#2 (SPDIF)
    DAC#1 > AMP > BK20 BLH
    DAC#2 > AMP > Dayton15 Reference

    What the chain looks like. DAC#2 is on the floor on the left side until i get 15' coax cable. Sub amp is in the closet because it's loud AF.
    [​IMG]

    The NanoDigi is on top and smaller than an O2 amp.
    [​IMG]

    MiniDSP control panel. Don't need to reiterate what Ultrabike has mentioned it works approximately the same way.
    Routing tab routes your channels and dacs (ensure you route them correctly or you could end up with mono instead of stereo...ask me how i know). Output tab is used for correction of individual channels, time alignment, crossover, and digital gain (volume matching) of individual DACS. Input tab is for overall correction (final curve shaping) and master volume control.
    [​IMG]

    Setup of Mains & Subs in Nearfield (smooth/flat) without room effects. As you can see, corrected with IR window applied and farfield (with room effects) is two different things. This is how bad the room messes up your sound and its very difficult to treat. L&R subs and combined in listening position.
    [​IMG]
    Correction of Mains at 1yd from full range cone.
    [​IMG]

    Next you gotta time align the subs and mains. The sound of the subs in my case reaches your head later than the mains therefore you need to add a delay to the mains. REW with acoustic reference gives u this information to plug in exactly to the DSP. The actual number is 7.62ms which you can see on the output tab in my pictures. This just shows that it is reduced to a very low number in measurements after you fix it.
    [​IMG]

    Next crossover setup, you have should optimize phases at crossover range (not exactly a point since there's overlap). You use the overlap to your advantage to fill holes and avoid bumps as necessary. In my case there's 20hz of overlap. This is where you adjust the individual channel gains to match your DACs in the output tab as well.

    [​IMG]
    [​IMG]
    [​IMG]
    This is the full range sweep as a result with IR window.
    [​IMG]
    This is the final finished result. Pretty much full BK curve at listening. The plateau for the subs went too far to 600hz, it should be corrected to 150hz which i can do later in the Input shaping tab (would reduce gain on the whole sub dac and then add a 150hz low shelf for +5). I think you get the idea of what this is capable of.

    This is brute force in digital space and I know the graphs look forced.. There's not actually as much correction as you think. What you measure compensated for room effects and what it would look like uncompensated could be too different things as i showed above. This is my second attempt because the first attempt was a complete failure and i started over from scratch. The first attempt entailed corrected room issues so that i could get the perfect uncompensated FR. I got the FR to measure perfect, but it sounded like amphibian shit and i started over with a different approach..
    [​IMG]
     
    Last edited: Jul 27, 2017
  13. Hekeli

    Hekeli Facebook Friend

    Contributor
    Joined:
    Sep 28, 2015
    Likes Received:
    149
    Trophy Points:
    43
    Location:
    Finland
    I believe nanodigi resamples everything to 96kHz? Would make me sad to know Yggys burritofilter would be kind of wasted.. but perhaps a moot point with all the other dsp stuff going on. End result is still probably very good.
     
    Last edited: Jul 27, 2017
  14. Jh4db536

    Jh4db536 Friend

    Pyrate
    Joined:
    May 25, 2016
    Likes Received:
    926
    Trophy Points:
    93
    Location:
    Los Angeles
    Home Page:
    i will retract my statement on 'no compromise'

    There's always compromises whether it's analog or digital fukry. There's more than 1 way to accomplish objectives and this is not the traditional route obviously. This is kind of a infant stage product almost DIY and like everything else, it will get better and more advanced (including additional filters) in time. That's why this product costs $170...far less than a mutec and about the same price as an eityr.

    This is a lot of correction in analog space. Too much for my taste.
    i could easily use a MiniDSP all in 1 box DSP + multichannel DACs...but it wouldnt be an Yggdrasil. Let's not forget that Yggdrasil is already obsolete...no DSD, not 32 bit, meant for 1990's digital format as is muh 1995 VRDS source. If i was concerned about using all 192khz i wouldve bought a modi2u or equivalent AKM fed by 24 bit HDtrax instead.

    My intent of this thread is to highlight the NanoDigi, which is why i've just called things DAC#1 & 2 / Amp#1 & 2. I wanted the BK20 to work without subs, but basically you can only listen to flamenco guitars/ukuleles or female vocals, anime screaming without a chunk of the music missing. There is bass texture and resolving of the ambiance from where the music was recorded. Even HD800 is extremely light on this ambiance and texture as clean and extended as it measures.

    I definitely want to spend more critical time with it before i talk about what it sounds like, so update to come. Would be difficult to describe the sound of a DSP by nature. Ymmv, because i doubt anyone would have the same setup as this. Reminds me how nice it is to be able to just put headphones on and not have to deal with this extra BS of room correction and alignment of multiple components.

    1) The main concern i had in using this product is what compromises on SQ i would make.
    2) Would the flexibility be worth the compromises? By flexibility, you could set this product to be a digital premp defeating all its other capability. So if you wanted to add remote control functionality and volume control to Yggdrasil, Gungnir Multibit, or whatever DAC you might be using this can be a solution for someone who uses a CDP or LD transport.

    First thing, there's finally thumping BASS! Don't care if the BLH had measured bass below 100Hz - it sucked might as well have not been there. EDM sounded so wrong with BLH. Even jazz or Diana Krall, you could not hear the bass string instruments very well. You hear the faint echo of them, but not the fundamental notes and no thump. Rock sounded hilarious - think grado speakers. Now when i put on some Daft Punk, IM, or even the first 15 seconds of Hotel California the kick drum is like Damn i've been missing out.

    Yggdrasil still sounds like Yggdrasil because it IS Yggdrasil and there's nothing between it and the amp/transducers. The stage is narrow and its in your face; the smaller room actually makes it more acceptable again, because it felt like it was limiting my setup in a larger room where the staging could be a lot wider than what Yggdrasil would allow. I do believe this to still be above USB level because if it's not i'd throw it away in a heartbeat. I loss width because of the room change to a smaller one. Depth still goes out the window and through the front wall. Width was never Yggdrasil's strongpoint. Still has its nice tonality with moffat bass sustain coloration and the tape hiss emphasis of course. Regarding Yggdrasil's strong points i.e. resolution and detail is a different assessment that i need to make.

    Imaging did not suffer as a result of adding DSP to the digital chain. It's still above USB grade IMO.

    There's slight bass bloom which makes it sound slightly veiled and i even measured it. Might be able to correct with DSP, but there was also a small amount which is a 300b thing. I'll know for sure when i integrate the 2a3 amp later.
     
    Last edited: Jul 27, 2017
  15. Hekeli

    Hekeli Facebook Friend

    Contributor
    Joined:
    Sep 28, 2015
    Likes Received:
    149
    Trophy Points:
    43
    Location:
    Finland
    Well I did get it as a free prize.. :p I'll sell it anyway after I get my Gen5 usb upgrade since I'm looking to simplify stuff.

    Was just thinking aloud really, I've looked at nanodigi but I don't like the sampling rate forcing etc out of principle..
     

Share This Page