REW + FBQ2496 and ECM8000 for measuring (home) recording studio

Discussion in 'Official REW (Room EQ Wizard) Support Forum' started by Countcobolt, Dec 31, 2017.

  1. Countcobolt

    Countcobolt New Member
    Thread Starter

    Joined:
    Dec 31, 2017
    Messages:
    8
    Likes Received:
    0
    Hi all,

    I am trying to measure my studio for better sound performance. Let me first describe my setup:
    I am running a focusrite Scarlett 18i20 hooked to a FBQ2496 to 2 ADAM A7X speakers. On the speakers there are small modifiers for sound, but I placed them all to flat.
    I am trying to set up the FBQ so that my frequencies are flat overall (for mixing and mastering purposes). I use an ECM8000.

    Steps I have done:
    1. I have calibrated my sound card (connect output 1 to input 1 for that).
    2. I struggled with a loopback, it appears that the sound coming from out 5 to in 5 on the sound card is really low
    3. I loaded the ECM8000 calibration file

    Until this is appears ok with the exception of 2.
    When I started to measure (1M Log, 8 sweeps) I can adjust the Q/Gain/Freq in the EQ filter to be mostly flat. (I will add pictures later). But the first thing I noticed was that the high is very sharp and low is also too much. When I measured the room again with ARC2 my hunch was confirmed.

    So I started to do a few more things :
    1. Measure the room with the BFQ in bypass
    2. Create new filter
    3 Remeasure the room with BFQ active
    4. Modify the filters from the first measurement to get better results.
    5. repeated 3 and 4 multiple times.

    Although that the predicted line from REW sets that my Mid will be too much, as my high, the remeasurement with the BFQ active seems to be quite flat. Is this normal?
    Also what kind of smoothing do you apply? I currently use 1/6.

    Kind regards

    Steve
     
  2. Wayne A. Pflughaupt

    Staff Member

    Joined:
    May 21, 2017
    Messages:
    129
    Likes Received:
    14
    Location:
    Corpus Christi, TX
    Welcome to the Forum, Steve!

    For starters, it’s difficult to offer help without any graphs or .mdat file to examine.

    That said, a few things:

    You don’t need to generate a calibration file for the 18i20, as I’m certain it has ruler-flat frequency response.

    What calibration file are you using for the mic – hopefully a custom file from a lab such as Cross Spectrum?

    I prefer 1/3-octave smoothing for full-range graphs, because it gets you a graph that looks like what you’re hearing. But 1/6-octave is fine too.


    Yes, you can expect that a measurement would improve after equalization. Seems glaringly obvious, so maybe I don’t understand the question… ?

    Regards,
    Wayne
     
  3. Countcobolt

    Countcobolt New Member
    Thread Starter

    Joined:
    Dec 31, 2017
    Messages:
    8
    Likes Received:
    0
    Hi Wayne

    what surprises me is that when you look at the below graph the dark red is my measurements (FBQ on bypass) . The light red are my correction: One would assume that these setting would give me too much mid and too much high. Yet on the second image you see the measurements after sending these EQ filters to the FBQ. The prediction vs the measuring does not match at all.

    upload_2017-12-31_16-36-20.png


    After EQ measuring

    upload_2017-12-31_16-39-35.png

    So how do I make the prediction match or would I need to keep doing a pre and post analysis until I get a flat?
     

    Attached Files:

  4. Wayne A. Pflughaupt

    Staff Member

    Joined:
    May 21, 2017
    Messages:
    129
    Likes Received:
    14
    Location:
    Corpus Christi, TX
    Are both of those actual measurements, or is one of them something REW predicted?

    BTW, no reason to do screen shots. That little camera icon top left of the graph will save is as a .jpg.


    Regards,
    Wayne
     
  5. Countcobolt

    Countcobolt New Member
    Thread Starter

    Joined:
    Dec 31, 2017
    Messages:
    8
    Likes Received:
    0
    Hey Wayne

    In the first pic the dark red line is the actual measurement. The light red line is the "prediction". The green one in the second pic is the result when I apply the prediction EQ to the FBQ.

    Hope this helps

    Kind regards

    Steve
     
  6. Wayne A. Pflughaupt

    Staff Member

    Joined:
    May 21, 2017
    Messages:
    129
    Likes Received:
    14
    Location:
    Corpus Christi, TX
    Got it.

    Can you show a screenshot of the EQ panel, showing the filter parameters (e.g. frequency setting amount of boost or cut, etc.).

    Regards,
    Wayne
     
  7. Countcobolt

    Countcobolt New Member
    Thread Starter

    Joined:
    Dec 31, 2017
    Messages:
    8
    Likes Received:
    0
    HI Wayne,

    I'll try to perform the following. I'll start from scratch on the measurements. Next I will apply the EQ corrections to go as close to the target line. Then I will remeasure after the correction. I will use 1/3 scaling. After that I will upload the MDAT files and the pictures.

    Kind regards

    Steve
     
  8. Countcobolt

    Countcobolt New Member
    Thread Starter

    Joined:
    Dec 31, 2017
    Messages:
    8
    Likes Received:
    0
    Hi Wayne


    started from scratch today. First measurement (disabled the ECM calbiration as this is supposed to be flat)

    upload_2018-1-2_17-57-57.jpeg

    Then I tried the following as filters

    upload_2018-1-2_17-58-36.png

    Which result in the following predicted curse (That would be awesome)
    upload_2018-1-2_17-59-8.jpeg



    Unfortunately when I measure again, the predicted and the actual have a very high difference :( :( :(
    upload_2018-1-2_18-4-41.jpeg

    Red line is what was supposed to be predicted vs the green, which is the actual measurement :(

    Any ideas why predicted and actual after applying the EQ to the FBQ is so much of? Or is this normal and should I now try to adjust until I get a flat response?

    Kind regards

    Steve
     
  9. Wayne A. Pflughaupt

    Staff Member

    Joined:
    May 21, 2017
    Messages:
    129
    Likes Received:
    14
    Location:
    Corpus Christi, TX
    I don’t know why the EQd result is so far off from the predicted – John will have to answer that one.

    That said, there is no way 18 filters are required to deal with your baseline response. It should take no more than 4-5 filters, based on what I see in your graph. Indeed, I would immediately take out those two filters at 11 kHz and 19 kHz with the huge boost values, before you fry one of your tweeters!

    The goal of full-range equalization is to achieve an audible improvement in sound quality. That is typically realized by smoothing out the worst deviations, not by chasing every little ripple in response (which aren’t audible anyway). Many of the filters in the EQ Panel (such as #1, 2, 5 et. al.) are basically notch filters, which are inaudible or can suck out or exaggerate individual notes (depending on if they are boosted or cut) of the musical scale. You might want to give my article on minimal EQ a read, paying particular attention to the section on "useless filters." It’s written mainly for subwoofers, but the implications are just as useful for full-range EQ. You might also want to peruse this thread, a real-life exercise in full-range equalization.

    Your response from 2-20 kHz, especially between 2-12 kHz, is so bad it cannot be fully addressed with equalization. The best you can hope for is reducing the severity of each peak or depression by 4-5 dB tops. Did you measure each speaker separately or together? If the latter, comb filtering from the mic not being perfectly centered between the speakers could account for some of the problem. You might also experiment with different placement locations, if that’s an option.

    Regards,
    Wayne
     
    #9 Wayne A. Pflughaupt, Jan 2, 2018
    Last edited: Jan 2, 2018
  10. Countcobolt

    Countcobolt New Member
    Thread Starter

    Joined:
    Dec 31, 2017
    Messages:
    8
    Likes Received:
    0
    Hey Wayne,

    I think I get the idea. I am surprised that REW is giving me those results as I used this room for mixing several CD's. I did not measure the speakers separately but combined. You articles did make sense to me a lot. I will try to "hear" the difference as I was really looking at graphs here. I notice I do have heavy low and and lack some mid. My high were actually ok imho. So back to the drawing board.

    Stupid Q maybe: as REW is focussing on 1 single measurement point can it take several locations for measure and combine those? Our mixing desk actually holds 2 chairs next to each other so I now measure in between them.

    Kind regards

    Steve
     
  11. John Mulcahy

    John Mulcahy REW Author

    Joined:
    Apr 3, 2017
    Messages:
    339
    Likes Received:
    53
    Prediction and measurement should correspond very closely, provide the mic is not moved between measurements. To check the measurement is OK try measuring with a single 256k sweep rather than multiple sweeps, in some setups time sync is not maintained when using multiple sweeps which gives invalid measurements.

    Also best using the latest V5.19 beta version (available from the Downloads section) rather than 5.18.

    The filters shown are extreme and fighting against each other (e.g. +15 dB at 19 kHz, -9.5 dB at 20 kHz). It is not a good idea to allow such high filter boosts, the EQ controls allow you to limit that, and it is very bad to use narrow filters at high frequencies. Worth attaching an mdat file for one of your measurements so that we can check it is valid, if you save the file after setting up any filtering you want the filters will be saved with the measurement in the mdat file.
     
  12. Countcobolt

    Countcobolt New Member
    Thread Starter

    Joined:
    Dec 31, 2017
    Messages:
    8
    Likes Received:
    0
    Hi John, Wayne

    thanks for responding: I took a different approach today and eureka!
    1. I think my ECM8000 is broken. I measured with the ARC2 microphone as ARC claims it is a flat response microphone. The measurements seem way more correct.
    2. First I changed the rear speaker settings a lot. I am literally -5db low shelf and -4db high shelf. When measuring in ARC I now only had a few "areas" where there were issues.
    3. I opened up REW but measured with the ARC2 microphone. Response was very similar with REW as ARC 2.
    4. Second found error: When I use match target and send the values to the FBQ the sound is terrible. Measuring afterwards showed again a high boost of the highs, basically the sound was terrible
    5. I reset everything on the FBQ and started to modify manually. I took separate measurement for right. After each modification I remeasured. This actually had a good response/reaction to my modifications.
    6. I repeated this for the left.
    Basically I do have a very flat response now on Left and Right. Due to time I did not do a measurement yet of both.
    I have made an excel file with my filters. To Waynes point, I now can work everything with 7 filters and I assume that some are a bit useless.

    What was an issue is that the measurements send to the FBQ do not match after checking on the FBQ. Honestly, I believe this might have to do with the midi convertor. (They didn't provide Win10 drivers, so had to force win8 drivers on it (it does work for synths etc)).

    Also, but that is my lack of knowledge of the FBQ. I can only use filter 11 to 20. The others are off but I cannot modify them at all. I will read up on the manual of that things.

    I will post my results tomorrow as well as the filters I have applied in the FBQ.
    My next steps would be
    1. Remeasure both channels using ARC2 to check what the results are.
    2. Remeasure both channels using REW
    3. Start changing the filters on the FBQ to have a better result. This will be a challenge as I don't know how the combination of left and right will be.

    Up to know I notice a better clarity in the mixes. I still feel that the bass is a bit too much.
    My mid has definitely improved a lot (FYI: this was a problem area during mixing as my vocals are mainly in that region. I would boost those due to bad room calibration, hence the mix was actually wrong)

    Anyway a very great thanks for all your help already!

    Steve
     
  13. Wayne A. Pflughaupt

    Staff Member

    Joined:
    May 21, 2017
    Messages:
    129
    Likes Received:
    14
    Location:
    Corpus Christi, TX
    It would be best to use matching filters for the left and right from ~300 Hz and above. Using mis-matched filters will do weird things to your imaging. Below 300 Hz it’s okay to use specific filters for the left and right, if needed.

    Regards,
    Wayne
     
  14. Countcobolt

    Countcobolt New Member
    Thread Starter

    Joined:
    Dec 31, 2017
    Messages:
    8
    Likes Received:
    0
    Hi Wayne, John

    I redid all the measuring. I was able to add 10 filters and get to a very close flat response. I did notice that the Midi did not function properly when I send the info to the FBQ. I do believe this is not the software but the Roland Midi USB converter with forced drivers.

    A new thing popped up: to your point Wayne, at 1 central spot in front of the mixing desk, I have an absolute flat range. It sounds perfect. Yet, moving 30 cm to the left or right changes the response heavily. When I measure it, you actually see this resulted. When I put it back to the center position, measurement spot on.

    Now, considering we often mix with 2 people in front of the mixing desk I was wondering, is there an option to "combine" multiple measurements and then correct on that "average"?

    Kind regards

    Steve
     
  15. Wayne A. Pflughaupt

    Staff Member

    Joined:
    May 21, 2017
    Messages:
    129
    Likes Received:
    14
    Location:
    Corpus Christi, TX
    Sure, REW can do averaging, but the problem you’re having is related proximity. No EQ can help that. I assume you were having the same issue before EQ, right?

    I have some good desktop speakers, and the sound doesn’t change much if I move my head one way or the other, other than the soundstage, so I’m having a hard time understanding the problem you’re having.

    The best way to get a handle on what’s happening would be to take some measurements. Measuring a single speaker only, measure at the “dead center” listening position, then over towards that speaker a bit. Probably best to point the mic directly at the speaker for eash measurement.

    Regards,
    Wayne
     

Share This Page