First attempts at measuring KEF R3 and HD6xx with REW and Umik-2

[Gruesome]

First steps with REW and EQ APO. I know there are help files and Wikis and FAQs, so these questions are more meant as a sanity check. If anything looks off, I'd appreciate a comment.

The main issue for me so far (besides just figuring out the relevant REW settings) is the input and output device selection in REW, and how REW and Equalizer APO work together. Is REW keeping two lists of all audio devices it has ever seen, or seen since the last restart? Why are there multiple copies of the same device, beyond the EXCLusive access qualifier? Is this REW, Windows, or user error?

It seems when I start up REW with the Umik-2 already plugged in, it is NOT selected as input device for measurements, even if REW was shut down with that device selected. The devices in the REW input list with Umik-2 in the name also seem NOT to include the real Umik-2 currently connected to the computer, and selecting one of them does NOT select the real Umik-2 for use in REW measurements. It seems I need to unplug it and plug it back in, and then select it. And also, the calibration gets forgotten. Does that seem right/expected behavior?

And when I configure EQ APO with an equalization curve generated in REW, I have to make sure to NOT select the exclusive version of the output device, otherwise the APO will not be employed; is that correct as well? And windows does NOT need to be rebooted for APO setting changes (gain, eq settings) to take effect? Since I see an effect in the first graph below, I think I got it right.

In any case, I got the Umik-2 with the hope that the lower noise floor would also allow distortion measurements.
Here are some screenshots from trying to measure a KEF R3 in my office, a harmonic distortion spectrum for that speaker at 75dB, and an attempt to measure the frequency response of my Sennheiser HD6xx, using a foam baffle (with mic hole) to seal off the headphone.

KEF R3 frequency response (at 1m), and equalization attempt

Does this look reasonable? The room is 11' by 10', and 8' high, with a 3' by 3' doorway extension. The green curve is measured after loading the equalizer settings into the EQ APO, and it seems to behave as intended. I assume for real room equalization I should measure both speakers at the same time, and generate one equalization curve for both?

Example of R3 distortion spectrum

HD6xx frequency response
And here is an attempt to measure the frequency response of the HD6xx; without the baffle/seal the response has a lot more peaks and dips, and it is very difficult to get something reproducible. With the baffle there still remains a dependence of the relative response above two kHz on the exact distance of the microphone capsule from the headphone membrane. Putting the mic in contact with the inner guard screen of the headphone seems to give reasonably reproducible results (top three curves in graph below):

A picture of the setup, with the Umik-2 sticking through the baffle into the headphone ear cup:

 

[HarmonicTHD]

First steps with REW and EQ APO. I know there are help files and Wikis and FAQs, so these questions are more meant as a sanity check. If anything looks off, I'd appreciate a comment.

The main issue for me so far (besides just figuring out the relevant REW settings) is the input and output device selection in REW, and how REW and Equalizer APO work together. Is REW keeping two lists of all audio devices it has ever seen, or seen since the last restart? Why are there multiple copies of the same device, beyond the EXCLusive access qualifier? Is this REW, Windows, or user error?

It seems when I start up REW with the Umik-2 already plugged in, it is NOT selected as input device for measurements, even if REW was shut down with that device selected. The devices in the REW input list with Umik-2 in the name also seem NOT to include the real Umik-2 currently connected to the computer, and selecting one of them does NOT select the real Umik-2 for use in REW measurements. It seems I need to unplug it and plug it back in, and then select it. And also, the calibration gets forgotten. Does that seem right/expected behavior?

And when I configure EQ APO with an equalization curve generated in REW, I have to make sure to NOT select the exclusive version of the output device, otherwise the APO will not be employed; is that correct as well? And windows does NOT need to be rebooted for APO setting changes (gain, eq settings) to take effect? Since I see an effect in the first graph below, I think I got it right.

In any case, I got the Umik-2 with the hope that the lower noise floor would also allow distortion measurements.
Here are some screenshots from trying to measure a KEF R3 in my office, a harmonic distortion spectrum for that speaker at 75dB, and an attempt to measure the frequency response of my Sennheiser HD6xx, using a foam baffle (with mic hole) to seal off the headphone.

KEF R3 frequency response (at 1m), and equalization attempt
View attachment 285069
Does this look reasonable? The room is 11' by 10', and 8' high, with a 3' by 3' doorway extension. The green curve is measured after loading the equalizer settings into the EQ APO, and it seems to behave as intended. I assume for real room equalization I should measure both speakers at the same time, and generate one equalization curve for both?

Example of R3 distortion spectrum
View attachment 285070

HD6xx frequency response
And here is an attempt to measure the frequency response of the HD6xx; without the baffle/seal the response has a lot more peaks and dips, and it is very difficult to get something reproducible. With the baffle there still remains a dependence of the relative response above two kHz on the exact distance of the microphone capsule from the headphone membrane. Putting the mic in contact with the inner guard screen of the headphone seems to give reasonably reproducible results (top three curves in graph below):
View attachment 285072
A picture of the setup, with the Umik-2 sticking through the baffle into the headphone ear cup:
View attachment 285075

Looks good at first glance.

 

[Gruesome]

Looks good at first glance.

Thanks for giving it a quick look over, HarmonicTHD!

So, if I can believe the harmonic distortion measurement, the KEF R3 (plus the rest of the chain) has a THD of 0.065% or -64dB at 75dB SPL. The worse (right) of the two channels of the Loxjie A30 in Amir's test has total noise and distortion (dominated by the distortion contribution, if I interpret the graph correctly, since the noise floor is at -120 to -130dB) at -80dB or 0.01%, a factor 6 better than the combined distortion I measure. So the amplifier adds about 1.5dB to the overall distortion figure. This seems like an acceptable pairing to me. A ten times better (lower distortion at the same power) amplifier with SINAD of -100dB would add 0.15dB, and one with -120dB 0.015dB to the -65dB figure of the speaker.

Edit: actually, I need to use the amplifier figure at the SPL I am listening at. To produce 75 dB SPL out of the R3, one needs about a third of a Watt, which puts the A30 distortion at about -68dB ( https://www.audiosciencereview.com/...o-8-ohm-measurements-toslink-input-png.93597/). So the amplifier contribution is actually much larger, and dominates the total THD: the KEF R3 only contributes -72dB of the -64dB THD. So at the (low?) levels I am listening at I could paradoxically(?) use a better amplifier. A Topping LA90D would be at -110dB at 0.3W, a PA5 at -90dB, an NAD M33 at -85dB, a Denon PMA-50 about -81dB, and a Topping PA3s about -79dB. One probably shouldn't go any worse than that. All else being equal, I should have picked the PA3s instead of the A30, but not for example the A07, which is also a bit worse (~-72dB) with the 32V supply (see https://www.audiosciencereview.com/...255-measurements-power-into-8-ohm-png.102794/).

 

[Gruesome]

Update: these measurements might have the same 5dB slope error that I later found in other measurements with this microphone. So far I haven't been able to figure out the reason. Maybe a wrong setting in REW?

 

[raindance]

How can you be verifying room EQ at 1m for the speakers? Is that your normal listening distance?

 

[kemmler3D]

Speaker measurement looks about right to me, although I am not sure that measuring headphones this way is worthwhile. Even with purpose-built headphone jigs, measurement is difficult and unreliable.

When it comes to making room measurements, you should measure where you intend to listen. If you have more than one seat that you listen in (say couch vs. single armchair) then you might want to use the moving mic method, aka MMM. If you have got this far, there's nothing too tricky about it, looks like you're more or less on the right track.

 

[Anthony LoFi]

Hello G,
Your R3 before and after looks typical of any set of measurements for loudspeakers.
I have the same outcomes when using REW with a mniDSP 2X4HD.

What might be beneficial is whether you are using a sub? The 50hz sag looks to be room nodes or a problem with sub crossover settings.
I believe the R3 would be relatively flat to 40hz?

I note you are using EQAPO. Im not familiar with this.

REW gets you to firstly close mic your individual speakers to apply a design correction if needed.
Then we retest the addition of a sub, which means crossover and sub design correction settings.
Then as a complete system, we remeasure for room effects and correction.

With the mini I can individually provide each L+R speaker with its own treament curcves, I can then apply to the incoming stereo signal and overall treatment curve as well as separate crossover curves.
Thats how Im working on mine at the moment. I have two systems and I have been "tinkering" for almost a year.
You may also want to look at further treatment using freeware "rePhase" software in conjunction with REW.

I agree with 3D above. Well done with the headphones but really, HD 600 are pretty good, why not do it manually to some test sweeps?
Trial and Error

 

[Gruesome]

Thanks for all the comments. This is quite helpful.
Raindance and Kemmler, the 1m and 3.5" distances were just to get some practice with REW. To really measure the speaker alone (without the room effects) I might have to go outside, or apply very tight timing windows in REW, I believe. (Is that correct?) My listening distance will be 7 feet in this room, and I will be measuring from there, once I have a rough idea of how to use REW. (I believe I have sorted out my microphone problem in the meantime.)

If I understand Anthony correctly, one would first measure each speaker individually, from close up (or from the listening position?), determine a first correction (not sure that the equalizer apo can do left and right channel separately?), and then do a final stereo correction on top of that?

Regarding the HD6xx, I just wanted to see whether I can get a reproducible microphone setup.

 

[OCA]

Is that correct?

No need to go outside for HF equalization. If you drop down right window size of the impulse response to around 7ms, the FR will show only direct sound given your 7ft LP distance to speaker. You will lose FR information below 143Hz but that wouldn't matter for HF EQ. To avoid any degradation in the filtered sound, I would also suggest keeping the auto target level of the original response intact, applying psychoacoustic or better VAR smoothing to the short windowed response and applying auto EQ with 0dB max boost.

from the listening position?

You need to measure each speaker separately at the LP and correct for the bass/mid frequencies because you wanna correct for the room effects at the LP for frequencies up to roughly 500Hz and you wanna avoid all room reflections and correct only the speaker (direct sound) beyond that. You only need outdoors or near field woofer/port measurement if you intend to correctly linearize port phase shifts which I guess is a bit too advanced for beginners.

Regarding the HD6xx

Did you try the EQ settings on this website?

AutoEq

Automatic headphone equalization

autoeq.app

 

[Gruesome]

Thanks for replying directly to my questions, OCA!

I have tried canned/predetermined headphone corrections. At some point I might come back to comparing my headphone measurements with those (I assume the corrections are based on measurements.)