REW+Equalizer APO - ends up muffled-sounding

[thoraxe]

I've just started to play around with REW and Equalizer APO, and I am getting results that sound "opposite" to what my ears expect. I've attached a measurement file for all speakers. This is a 2.1 setup with either a UAD Volt 1 audio interface or a PC soundcard going into a Yamaha RX-375 amplifier, which is plumbed into two Definitive Technology Promonitor 800 and a Definitive Technology subwoofer. A Dayton EMM6 was used for all measurements. I use a standing desk and the subwoofer is under the desk while the monitor speakers are on the desk. I don't have a great listening position as I'm facing into the room with the wall behind me. It's about 7.5ft from the desktop speakers to the wall. The room has no treatment and generally is pretty boomy to start with.

I struggled a fair bit to get measurements that weren't too noisy. I might need to take some measurements at an even louder SPL/sweep so that I can use less gain on the microphone input to reduce some input noise on the EMM6 noise - you tell me. I used the calibration file supplied by Dayton for the EMM6, and I did a soundcard calibration with a TRS-XLR cable going from the headphone output of the Volt to the input. The mic cal file and Volt cal file are also attached. This sweep was done with all speakers active. I tried testing each speaker individually as well but got more or less the same results.

When I try to create filters using this measurement file for Equalizer APO, I get something that ends up filtering out a lot of the highs. I had to attach measurements as text because this forum doesn't allow attaching the mdat file directly, and I don't have permission to post links (to Google Drive). I am targeting a full-range speaker and an SPL of 80 to generate the filter settings using the EQ function of REQ. Here are the filter generation settings I am using:
* Range 20-20k
* Indiv max boost 15
* Overall max 0
* Flatness 3
* Allow both shelves, -6/6
* Allow narrow filters below 200

When matching the response to the target, I get the following filters:
Filter 1: ON PK Fc 270.0 Hz Gain -5.20 dB Q 1.564
Filter 2: ON LS Fc 344.0 Hz Gain 5.90 dB
Filter 3: ON PK Fc 491.0 Hz Gain -4.60 dB Q 3.798
Filter 4: ON PK Fc 620.0 Hz Gain -6.20 dB Q 3.926
Filter 5: ON HS Fc 779.0 Hz Gain -6.00 dB
Filter 6: ON PK Fc 5278 Hz Gain -8.50 dB Q 1.443
Filter 7: ON PK Fc 9164 Hz Gain -4.10 dB Q 1.010
Filter 8: ON PK Fc 13900 Hz Gain -7.30 dB Q 1.745
Filter 9: ON PK Fc 16468 Hz Gain -5.40 dB Q 2.540

As you can see, this shelf boosts everything below 350Hz by nearly 6dB and cuts everything above that. I've attached a screenshot of what Equalizer APO looks like with the filter turned on.

These speakers are normally pretty bright, and this EQ curve does cut down on some of that brightness/harshness, but it's cutting too much, and it gets that muffled/listening through a wet blanket/through a wall tone to it. This suggests to me some kind of user error in my measurements, or that I'm doing something else completely wrong.

Thanks in advance.

 

[Keith_W]

Please read this thread: common mistakes. It is much better to post the .mdat than .txt files. Also, proper labelling of what we are looking at is important.

Anyway, if this is the curve you are trying to correct, it is no wonder that the result will be muffled sounding. It thinks there is a +40dB tilt from 20Hz to 10kHz. This is impossible, there is simply no speaker designer who is that incompetent. Either that or there is something disastrously wrong with your speaker. A tweeter measurement might look like that, but certainly not a whole speaker. Clearly there is an error with your measurement. Here is a brief checklist:

- mic cal loaded? (you said yes)
- mic pointed at the speakers and not at the ceiling?
- used a mic boom tripod and not a mini tripod?
- were you measuring your speakers in situ (i.e. sitting on your desk)?
- care taken to avoid correcting early reflections?

You mentioned you had trouble with ambient noise. To increase the signal to noise ratio, you can either boost the signal or lower the noise.

Lowering noise: close all doors and windows. Choose to measure at a quieter time, e.g. in the evenings, on weekends.

Increasing signal: you can increase the width (time) or height (volume) of the impulse. Choosing to boost the volume risks measuring speaker distortion if played too loud. Always keep an eye on distortion, if your speakers distort it will look like a rising frequency response very similar to what you posted here (I would have checked it for you if you posted an .mdat instead of a .txt file). You can increase the time by choosing a slower sweep, or by taking multiple sweeps, rejecting the bad looking ones, and averaging all the good ones. A 45 second sweep provides about 90dB of noise rejection.

 

[thoraxe]

It is much better to post the .mdat than .txt files.

I didn't notice that zip was an acceptable attachment. After I re-test, I will do that to provide an mdat.

Please read this thread: common mistakes.

Ah... the video I watched did mention a 90-deg calibration, but I didn't have that cal file, and I did have the mic pointing at the ceiling (like in the video I watched), and not at the speaker, so that's probably a big source of the error/problems.

- mic cal loaded? (you said yes)
- mic pointed at the speakers and not at the ceiling?
- used a mic boom tripod and not a mini tripod?
- were you measuring your speakers in situ (i.e. sitting on your desk)?
- care taken to avoid correcting early reflections?

- mic cal loaded (yes)

- mic pointed at ceiling (oops)

- I am not using a specific mic-boom tripod, but I am using a video camera tripod with the EMM-6 mounted to it using the holder that came with the EMM-6 (https://a.co/d/cR8bdpi) - I did not have the "boom" extended on the tripod. How important is that? (see pictures). Note that the picture shows the mic pointed at the speaker (I didn't test like that originally).

- I did have the microphone located roughly where my ears would be, which is approximately 39" from the L speaker (the bookshelf speakers do not sit symmetrically on the desk - see room pictures). The bookshelf speakers were on the desk. The subwoofer was on the floor under the desk but could be relocated to a corner of the room (the LFE cable is quite long) if that would be better. It's unclear if I should measure the subwoofer independently of either L/R (or L+R) and, if I measure it independently, if the microphone should still be pointed at it, given that the sub is under the desk (the mic would point at the top of the desk if it's where my ears go).

- Zero care was taken regarding early reflections because I was trying to measure exactly where my ears would be in the listening situation I would be in (see pictures). I could move the speakers somewhere else, but I don't have stands for them, so I'd have to get creative, and that wouldn't be reflective (pun intended) of how I use the speakers anyway...

You mentioned you had trouble with ambient noise.

Sorry, I didn't mean to write something that suggested that. I mentioned problems with noise due to the gain required to get the levels sufficiently set, but that also is likely due to the mic pointing at the ceiling and not at the speaker. The noise floor, as measured by an inexpensive SPL meter, is ~38dB. The issue was that, even with the SPL meter measuring 80dB at the microphone when playing the test tone, I had to crank the gain on the mic input on the Volt very high to achieve >30dB of signal, so it was really picking up a lot of ambient noise (even though there wasn't much of any). The re-test will probably be much better with the mic pointed at the speaker. I also probably need to use a higher output on the headphone port of the Volt and a lower volume level on the amplifier to reduce noise. I did notice some hiss when nothing was being output by the headphone port of the Volt when preparing to take measurements.

Increasing signal: you can increase the width (time) or height (volume) of the impulse. Choosing to boost the volume risks measuring speaker distortion if played too loud. Always keep an eye on distortion, if your speakers distort it will look like a rising frequency response very similar to what you posted here (I would have checked it for you if you posted an .mdat instead of a .txt file). You can increase the time by choosing a slower sweep, or by taking multiple sweeps, rejecting the bad looking ones, and averaging all the good ones. A 45 second sweep provides about 90dB of noise rejection.

I don't think speaker distortion is an issue in this particular case. I don't think the test tones being played, even at -8 dBFS, are loud enough to distort. I was not doing the much longer sweeps (512k = ~11s x2 repetitions). I see the sub-post from the common mistakes thread regarding averaging, so I will look into that and attempt to do some 45s sweeps when the wife is not home. When I originally tried to use -12 dBFS sweeps and a 75dB calibration level, I got warnings in REW about SNR. When I adjusted the calibration level to 80dB and used -8 dBFS sweeps, those warnings went away. But the lights on the Volt were showing a lot of signal coming into the mic (again, likely because the gain was cranked due to the mic not pointing at the speakers).

FWIW, these desktop speakers have a frequency response of 57-30k, and the subwoofer has a 20-150hz frequency response as indicated by the manufacturer. I was testing L+R with the sub attached in this instance. I suppose I could test just L+Sub (disconnect R speaker). I also probably should set the crossover for the sub in the Yamaha amplifier.

Anyway, thanks so much for taking the time to provide a detailed response and some links. I'll do a re-test and provide an mdat and we'll see where to go from there.

 

[Keith_W]

I will wait for your re-test. In the meantime here are some random observations.

- Wife saw me looking at your pictures when typing the reply. She said "must be a fit guy, has a treadmill when he uses the computer". I asked how do you know it's a guy? She said "look at the racing sim, do you think I would let you do that to our living room?".

- The usual cause of bad sounding corrections after DSP is user error. Either from wrong measurements, or from wrong DSP strategy.

- Wrong measurements: When using DSP to correct the loudspeaker, the usual recommendation is: do speaker+room corrections up to the Schroder frequency and quasi-anechoic speaker ONLY corrections above that. You do NOT want to be correcting short wavelength reflections because what you measure is not what you hear. This means that those reflections need to be windowed out, and if there reflections are early the windowing will be more difficult. At best, you can apply very broad low Q "tone control" type corrections - e.g. you can change the treble tilt. Attempting to chase every curve will degrade the performance of a good loudspeaker. Not my words, Toole said that.

This means two sets of measurements need to be taken and blended together. (1) An in-room measurement for low frequencies, capturing the effect of the room at the listening position. (2) A quasi-anechoic measurement, free of as many reflections as possible for the high frequencies. Or you could simply do (1) and avoid high frequency correction or apply a gentle tone control.

- Re: your camera tripod. The idea of taking measurements for high frequencies is to avoid as many reflective surfaces as possible. Tripods themselves reflect shorter wavelengths and can be seen as small bumps in the frequency response. But you have bigger problems than that, your speakers are sitting on a table.

 

[thoraxe]

- Wife saw me looking at your pictures when typing the reply. She said "must be a fit guy, has a treadmill when he uses the computer". I asked how do you know it's a guy? She said "look at the racing sim, do you think I would let you do that to our living room?".

Your wife sounds smart. Keep her? Fortunately this isn't my living room, but I like to think I'm reasonably fit. Anyway, back to REW.

Attached is an mdat file
* 45s sweeps
* 50-20k, which is the range of the bookshelf speaker
* no sub
* L only
* pointed directly at L speaker
* 38db noise floor
* Pretty sure I don't have any eqing going on in Windows or in the Yamaha receiver. Yamaha receiver EQ mode is set to off, Equalizer APO has no configs loaded.

Before continuing, any thoughts? I could get stands to mount the desktop speakers. In the extreme, I could move them to the far wall and put little shelves up for them. I realize that the desk location is going to cause lots of reflective issues in the highs, and the nature of the room (hard floor, pretty square) is likely to cause boomy issues with the sub.

As far as goals go, I'd just like to get the best sound out of the existing equipment. I do some EDM production, and exclusively using headphones is not always fun. I also do some gaming. This path is likely wild overkill for both, and a little sound treatment and speaker placement improvement will probably pay more dividends.

Thanks again!

 

[Keith_W]

First, let us see if you have taken a good quality measurement.

This is the energy-time curve. I have marked your reflections for you. The first one (at 0.49ms) might be a reflection, might be noise, I am not sure. The way to tell is to repeat the measurement and see if it is still there. IF it is a reflection, 0.49ms is 17cm (6.6") calculated with the speed of sound. I am guessing it is a reflection from your table. The big reflection, marked 3, is at 8.77ms, or 3.08m / 10ft. I am guessing rear wall.

Assuming that arrow 1 is pointing at a reflection at 0.49ms, this means that you will not be able to window out frequencies with a period longer than 0.49ms, or 2049kHz and lower.

This is a zoomed out view of the waterfall so you can easily see the room's noise floor (it is the flat area of the graph). You said you measured your noise floor at 38dB, but it is apparent you did not adjust REW's measurement with an SPL meter because the noise floor was 60dB. Anyway, I adjusted it for you with "Align SPL". It seems to be taken with adequate volume, about 30-35dB above the noise floor, so I will give this a pass. I took a quick look at your distortion measurements to make sure you were not measuring so loud that your speaker distorts, and all the distortion products were below the room's noise floor. So I will give it another pass.

This is a comparison of your new measurement (red) with the old measurement which you supplied as a .txt file (green). IMO your new measurement is much better quality. Especially noticeable is the discrepancy between the two measurements. You can see there is up to a 10dB discrepancy between old and new measurement at the top end.

Finally let's take a look at the frequency response of your speakers. I have applied ERB smoothing to better represent what you will hear. You can see a few things - the response is quite lumpy, especially between 1kHz and 4kHz. And the general trend is an upwards treble tilt (dark blue line). The light blue line is what you should be aiming for.

I note that DT's website says that the frequency response is 57Hz to 20kHz. What a steaming pile of marketing! You can see it rolling off at 180Hz quite steeply, and even Amir's review of the bigger model shows a similar roll-off, although not as steep as yours. This means your subwoofer needs to go up to 200Hz as a minimum, and it definitely be directional at that frequency.

I could not believe that a competent loudspeaker designer would deliberately create a speaker with a rising frequency response like that, so I went and looked for some reviews. I could not find measurements of your particular speaker (Definitive Technology Promonitor 800) but I managed to find Amir's review of the Promonitor 1000 and Erin's review of the BP-9020. Neither of those reviews are particularly impressive, but neither of them show a rising frequency response either.

If this were a proper anechoic measurement of the speaker (and not a flawed in-room measurement contaminated by multiple early reflections which I showed you earlier), I would absolutely trash this speaker and say it is rubbish on the strength of this measurement alone. But I will give it some leeway and say that the jury is still out. Regardless, this is how you are hearing your speaker, reflections and all.

So I suspect I know why you have muffled sound. You have probably become acclimatised to the rising frequency response, which to most people would sound thin and bright, particularly with the lack of bass. An 8dB upwards tilt is A LOT. To compound matters, you were correcting to your earlier measurement which had an even more dramatic upwards tilt. Your software does know any better, I am guessing it must have applied a 20dB cut at upper freqs at the very minimum. So to from a sound that is too bright to a sound with a massive cut would no doubt sound muffled.

This is far from a complete suit of measurements, who knows what other gremlins you might find. Anyway, if you want to keep this speaker, the first thing I would do is to confirm that the rising treble response is real. I would take those speakers outside and elevate them or try my best to get rid of reflections. Or you could try this experimental method that I described. I will criticise myself and say that it is experimental and not backed up by replication from other ASR members, or validation with an anechoic measurement.

I hope I have taught you what to look for to confirm if you have taken a high quality measurement.

 

[thoraxe]

This is the energy-time curve. I have marked your reflections for you. The first one (at 0.49ms) might be a reflection, might be noise, I am not sure. The way to tell is to repeat the measurement and see if it is still there. IF it is a reflection, 0.49ms is 17cm (6.6") calculated with the speed of sound. I am guessing it is a reflection from your table. The big reflection, marked 3, is at 8.77ms, or 3.08m / 10ft. I am guessing rear wall.

10ft is definitely right around where the rear wall is.
6" would make sense for some kind of reflections. There's also a lot of junk on the table that maybe I should also move before the test.

This is a zoomed out view of the waterfall so you can easily see the room's noise floor (it is the flat area of the graph). You said you measured your noise floor at 38dB, but it is apparent you did not adjust REW's measurement with an SPL meter because the noise floor was 60dB. Anyway, I adjusted it for you with "Align SPL". It seems to be taken with adequate volume, about 30-35dB above the noise floor, so I will give this a pass. I took a quick look at your distortion measurements to make sure you were not measuring so loud that your speaker distorts, and all the distortion products were below the room's noise floor. So I will give it another pass.

Yes, I probably did not set that wherever it is supposed to be set. Although looking at the measurement window, there is a checkbox for "capture noise floor" but not an actual setting to enter any values. I did not calibrate the SPL meter after changing the mic orientation though, so that likely explains it.

This is a comparison of your new measurement (red) with the old measurement which you supplied as a .txt file (green). IMO your new measurement is much better quality. Especially noticeable is the discrepancy between the two measurements. You can see there is up to a 10dB discrepancy between old and new measurement at the top end.

Interestingly, the low frequencies are much louder in the old measurement.

Finally let's take a look at the frequency response of your speakers. I have applied ERB smoothing to better represent what you will hear. You can see a few things - the response is quite lumpy, especially between 1kHz and 4kHz. And the general trend is an upwards treble tilt (dark blue line). The light blue line is what you should be aiming for.

I note that DT's website says that the frequency response is 57Hz to 20kHz. What a steaming pile of marketing! You can see it rolling off at 180Hz quite steeply, and even Amir's review of the bigger model shows a similar roll-off, although not as steep as yours. This means your subwoofer needs to go up to 200Hz as a minimum, and it definitely be directional at that frequency.

"Sadly", I have their ProSub 800, which "matches" these speakers and claims a frequency response of 20-150 Hz. Left to be seen/tested.

So I suspect I know why you have muffled sound. You have probably become acclimatised to the rising frequency response, which to most people would sound thin and bright, particularly with the lack of bass. An 8dB upwards tilt is A LOT. To compound matters, you were correcting to your earlier measurement which had an even more dramatic upwards tilt. Your software does know any better, I am guessing it must have applied a 20dB cut at upper freqs at the very minimum. So to from a sound that is too bright to a sound with a massive cut would no doubt sound muffled.

I should not have been applying any correction via EQ APO or the Yamaha when making these measurements, so it's unclear what you mean when you say "correcting to your earlier measurement." When I did the first measurement at a 90deg angle on the mic and then created a filter for that I ended up with some bass boost and a ton of high-freq cut, which made it sound mega muffled.

When I did a filter correction based only on this test and focused on 170Hz-20k, I got a little bit of cut and shelving in the highs (see below). Compared back-to-back, the shelved highs sound much less bright (duh). That makes it seem apparently worse, but this is likely a hugely mental thing where "louder always sounds better." To truly test, I'd have to figure out how to to correct for perceived loudness and blah blah.

However, when coming into the filtered highs fresh, it does sound more balanced and less tinny.

This is far from a complete suit of measurements, who knows what other gremlins you might find. Anyway, if you want to keep this speaker, the first thing I would do is to confirm that the rising treble response is real. I would take those speakers outside and elevate them or try my best to get rid of reflections. Or you could try this experimental method that I described. I will criticise myself and say that it is experimental and not backed up by replication from other ASR members, or validation with an anechoic measurement.

I could do more testing of the speaker in a better environment, but it would unfortunately not really be a zero-cost situation:

* I don't have an easily portable amplifier to use with the speaker
* I would probably need a longer XLR cable to go to the microphone on the tripod, as I expect a "proper" measurement is likely not done at 3'
* I'd have to figure out some kind of "stand" for the speaker
* One of my clothes closets would be decent for anechoic-ish purposes, but there's no outlet in there, so I'd need to use an extension cord for the amplifier (the least problematic problem)

These bookshelf speakers were "free" in the sense that I bought a 5.1 kit of them, but I already had some in-ceiling speakers for the rear channel on the home theater, so the two that would have been used for a rear pair were just sitting around doing nothing. They replaced some really ancient JBLs from a 5.1 kit I had form many, many moons ago. Anecdotally, I have liked the sound of this ProMonitor/ProSub combination for the home theater, but I did no back-to-back testing against the old JBL+Polk subwoofer and have done no testing in that actual room to tell me how good or bad things are.

As far as the computer/office/studio goes, I've also never really felt things sounded bad. The initial YPAO correction on the Yamaha RX-375 made things a little boomy, but after turning all of that off in preparation for these REW tests, I also removed the sub boost, which has leveled things out pretty nicely and removed some of the boominess.

As I've been writing this response post, I have been listening to music with the filter settings that were output from this test run, but I forget what target SPL I used here. I think it was 85, but I don't recall.

Filter 1: ON PK Fc 567.0 Hz Gain -6.10 dB Q 3.208
Filter 2: ON PK Fc 675.0 Hz Gain -5.80 dB Q 4.906
Filter 3: ON PK Fc 1123 Hz Gain 1.60 dB Q 1.001
Filter 4: ON PK Fc 1992 Hz Gain -4.60 dB Q 2.221
Filter 5: ON PK Fc 3128 Hz Gain 1.90 dB Q 3.651
Filter 6: ON PK Fc 4483 Hz Gain -6.50 dB Q 2.933
Filter 7: ON HS Fc 5056 Hz Gain -6.00 dB

I didn't do any back-to-back testing with the unfiltered audio, and things actually sound fine-ish. Maybe the highs are a little dull, but it does sound good/natural. This is very unscientific.

I have been contemplating buying studio monitors to also use for gaming and other purposes, but it's unlikely I'll do that soon unless I find a screaming deal. And, as I do not think things sound particularly bad now, I'm not sure how much an investment in better speakers is going to pay dividends without doing back-to-back testing on something better. In other words, my ignorance is paying dividends right now

So I guess I will end the post with: What do you think the best next steps are? Should I do more testing of various flavors, or just "finish" doing some corrections here and wait to do the studio monitor upgrade? Would some level of room treatment pay huge dividends here (some panels behind me, for example) or does it look like the speaker setup is really the worst offender (again, only given the limited test results so far).

Thanks so much! I also sent you a DM.

 

[Keith_W]

So I guess I will end the post with: What do you think the best next steps are?

My recommendation is to ask yourself whether you like the sound of the upper frequencies. If you are happy, then leave it alone and do bass correction only. Simple!

If you don't like it, or you have the itch, then you will have to go through a lot of trouble. This means putting your speakers in a bigger room for measurement to obtain as much distance from reflective surfaces as possible. You don't need to construct stands for it, you could push your speakers to the edge of a table. Clear out an area of furniture and measure the distance to all the surfaces. You can calculate the lower limit where you will be able obtain a reflection-free measurement with the calculation I posted in this thread. Then after you take the measurement, check the ETC and look at the reflections.

A closet full of clothes is NOT an anechoic environment, it will be full of random very early reflections and it will have selective frequency absorption (some wavelengths absorbed more than others). DO NOT do this.

 

[thoraxe]

My recommendation is to ask yourself whether you like the sound of the upper frequencies. If you are happy, then leave it alone and do bass correction only. Simple!

Without a known better reference it would be hard for me to hard to say. I think it sounds just fine now, and maybe could be even better with some more adjustment. But I've also got a challenging listening environment to begin with, and maybe changing some of those features would pay more dividends. Even something like a speaker stand to get these speakers off the desktop and in line with my ear might make a huge difference and would be a relatively inexpensive test.

If you don't like it, or you have the itch, then you will have to go through a lot of trouble.

But is it really worth the trouble for "free" speakers? My time is pretty valuable (to me, at least), and I have a reasonable budget for upgrades. I was already thinking about something like (powered) studio monitors. Maybe a better bookshelf speaker is a more appropriate upgrade? Who knows. I have many itches but not enough time to scratch them all. I've probably sunk more time than I should have into this particular project in some ways. But it has been a good learning experience (especially with your help!).

Given that I would be willing to spend $400-$600 for a new pair of speakers (assuming the Definitive sub is "good enough" - although I have a Polk sub I can also test), does it make any sense to keep fussing with these or just come up with a plan for better speakers?

As a reminder of the goals, I am doing a mix of EDM production (tech house and adjacent genres) and gaming, with passive music listening during work hours on occasion. I've not been unhappy, but I like to tweak and tinker and suffer from upgrade-itis. I don't listen to things very loud at all compared to most.

 

[Keith_W]

Is it worth it? Who am I to answer that question for you? I can help you take measurements, interpret graphs, etc. but I can't make value judgements for you.

I think it is good for you to acquire the skills anyway - how to measure, how to interpret, how to do corrections. So I would encourage you to persist with it. Not everyone enjoys looking at graphs, or cares about good sound. No judgement on my part, for me it's a hobby and I don't tell anyone else how they should enjoy their hobby.

 

[AnalogSteph]

I would recommend checking the loudspeaker configuration in the RX-V375. Looks like crossover is fairly high, probably 160 Hz - I'd think a 4.5" sat with PR in basically nearfield should be able to handle something lower, try 110 or 80 Hz. You can try running YPAO and see whether it helps any. I think the major dips remaining are a result of placement / reflections, not sure about the weak 200-400 Hz range though. A single position measurement is of limited value, generally speaking you want averages of multiple or go moving microphone method.