Measuring Low Frequencies and Driving Both Speakers

[technocolour v2]

Some of you have read my previous posts asking for help dealing with modal behaviour and what not. For the uninitiated, here's a recap: Since I started treating my room, I only had one speaker. Thus, all my measurements were done with only one speaker (the left speaker). Below is the FR and waterfall of my results as I tweaked my room, culminating with me trying to tame a 75hz null:

(the yellow band is +/- 5dB, centred around 70 dB)




I figured that this was as close as I could get without having both speakers as I didn't want to go too far down the rabbit hole, focusing on just one of two speakers.

Eventually, my replacement pair arrived so I broke them in and started to do measurements. When I measured the left speaker again, I was shocked at the result. Clearly something had changed but I wasn't too sure what...




I wasn't too put out as I was pretty sure it was a positional thing. I think I had moved the speaker stand and tried to put it back in the same place but I honestly can't remember. Unperturbed, I measured the right speaker (the one I hadn't measured yet):




Not bad. They look pretty similar to the original left speaker measurements at the top of the thread. So then I drove both speakers at the same time, pointing the mic dead ahead:




Driving both speaker delivered the best result yet, +/- 6.5 dB in the bottom end. I reckon I can get it down to +/- 5 dB as I haven't tweaked the new speaker and treatment placement(s).

So my question is: does driving both speakers when measuring modal behaviour yield a truer picture as opposed to driving and measuring each speaker independently? The independent speaker measurements vary greatly and have some hideous nulls, but driving both speakers delivered some pretty impressive results as the sweet spot is pretty big, I tested +/- 1 feet from the best position from my old measurements and the variation is only a few dB. I swept a sine from 300hz to 40hz when I was sat at my listening position and it pretty smooth with only a slight perceived variation across the spectrum.

[Jens Eklund]

Since bass at these low frequencies are usually in mono (not paned), the most likely scenario is that both speakers will be driven when listening to, or mixing music material. I would measure them separately and combined. Just summing/averaging the two measurements will probably give a false picture of what’s going on in the real world when the two sources combine with room modes, each other and everything else.

[technocolour v2]

Thanks for the input, Jens. So where do I stand when there's such a discrepancy between independent measurements and driving both speakers when measuring? Initially, when I did a sine sweep there was a blatant hole circa 75hz that also showed up on the driving both speakers measurements. I tweaked my setup and the sweep sounded a lot smoother, something that was reflected when driving both speakers. Out of measuring independently or driving both, which would you give more weight to when they don't correlate with each other so well?

[Jens Eklund]

I would continue to treat the room until the goal (whatever it might be, +/- 5dB @ 1/24 octave smoothing perhaps and decent decay times) is reached for each measurement, independently and combined.

[technocolour v2]

Excellent, thanks for clearing that up for me.

[tINY]

Positioning 2 or 4 subwoofers properly in a smallish room can have huge effects.

You can actually control the amount that you excite the axial modes significantly.



http://www.harman.com/EN-US/OurCompa...s/multsubs.pdf

Good reading here.




-tINY

[DanDan]

I agree that it is best to measure individually AND both.
However, even LF response changes a lot with relatively small position changes. Try it.
Mid and certainly HF from one speaker do not make it to the opposing ear. The radical HF comb filtering seen when driving both speakers, central mic, simply does not happen in real life. The head blocks, particular a thick head.


I recommend moving the mic to the two ear positions when doing individual measurements. Left for Left and vice versa. When driving both speakers, I am not convinced about the centre sum thing at all. The nose doesn't hear. Why not continue to use actual ear positions?

FuzzMeasure offers a wide choice including Sum Average Convolve and Correlate. RMS used to be an optional extra.
I will ask Chris (FuzzMeasure) to take a look in here, I would love to know what all those combining choices are intended for.
For now I am hopeful that measurements at actual ear positions is an improvement on the Nose Mono.

DD

[Ethan Winer]

Originally Posted by Jens Eklund

Since bass at these low frequencies are usually in mono (not paned), the most likely scenario is that both speakers will be driven when listening to, or mixing music material. I would measure them separately and combined. Just summing/averaging the two measurements will probably give a false picture of what’s going on in the real world when the two sources combine with room modes, each other and everything else.

Exactly.

--Ethan

________________
The Acoustic Treatment Experts

[liscio]

Originally Posted by DanDan

FuzzMeasure offers a wide choice including Sum Average Convolve and Correlate. RMS used to be an optional extra.
I will ask Chris (FuzzMeasure) to take a look in here, I would love to know what all those combining choices are intended for.

There are both frequency-domain and time-domain combinations that are available in FuzzMeasure. Generally folks don't need to use them, but they can come in handy in cases like this.

For instance, a time-domain Sum will create a new measurement by summing two selected measurements. So, hypothetically, you'd simulate mixing two measurements with one another.

The same can be said for averaging, except it'd effectively normalize the result of the addition above. Depends on the type of mixing you're trying to simulate. In reality, two loudspeakers would effectively be a sum, and not an average, of one another.

Convolution and correlation are frequency-domain combinations that do exactly what they say. They correspond to multiplication and division in the frequency domain, respectively.

This is useful for very specific scenarios, and probably not room acoustics per-se. For instance, if you are designing a loudspeaker crossover, you can measure it in FuzzMeasure independently of the loudspeaker it's going into (taking care to measure each driver, driven independently, in separate measurements). Then, you can convolve the filter response with your measured driver responses to get a sense for how it will affect the final response of each driver.

You can also accomplish the same thing with EQs and measured results, so you can quietly test the modifications of an EQ in relation to a room's measured response.

The above combinations are fairly delicate, though, and require the operator to know what she's doing. Without fully understanding what you're doing, it'll be impossible to recognize when the output of these operations has been obtained correctly.

For starters, the addition of two time-delayed signals that are nearly identical would create comb filtering (seeing how that's the very definition of a comb filter), and you have to be ready to interpret the results accordingly.

Hope this helps!

[technocolour v2]

Thanks guys, I took what everyone's said on board and did some further experiments with speaker/mic/treatment placement. I paid particular attention to what Jens had to say and aimed to get good independent measurements as well as driving both speakers. The hard work paid off as I got both L and R within +/- 7.5 dB and LR within +/- 5 dB. I'll post up my results later on but I think I can get the independent measurements within a tighter range. Massive thanks to everyone who's chimed in with all the helpful advice over the months, I still have my work cut out for me but I feel a lot more confident about realising my goals

Edit: didn't see liscio's post, thanks for the info!

[DanDan]

Tech, that really is a very good result, good on you.
There was a time when Bass and Kick pretty much had to be central and mono and devoid of any strange phase. Vinyl disc cutting. However I believe those days are kinda gone and panning (and phase elements) are wide open these days.
Many of us pan LF these days. Certainly when there is a visual element. e.g. Drummer on Left in picture. Electric guitar goes down to 80Hz and is often widely panned. Keyboards frequently go to 40Hz and are often very widely panned. Furthermore I would often place the Bass or Kick in a small room ambience, an IR of Cello Studios live room for instance. This is typically not symmetrical, and even if it was, I feel free to pan the ambience for mix clarity. So wild level and phase discrepancies L vs R.

One of Ethan's tests, with thanks, has shown pretty strong variations at LF with mere inches of position change. I reckon the ear/brain does average what is heard by the two ears at LF. e.g. a strong but narrow null on one side only is not easily heard.
Ultimately I believe that the Nose position does not represent what either or both ears hear.

Chris, given the above, and the lack of comb filtering at HF already mentioned.
Which scenario do you think represents what the ears/brain actually hear? Put another way, which of your Sums or Averages gets closest to the two ear/brain system response?

All- How about this. Drive both speakers. Mic at ear location.
Place a (Jecklin) disc to shield the mic from the opposing speaker at HF/MF ish.
Now we have the combined realistic modal drive. We also have whatever comb filtering or other cancellations are actually occurring during music drive, at LF. But we have removed the erroneous pictures of HF combing normally caused by two drive. Lastly we have sampled two, actually used locations, rather than one which is never used.

Any takers?

DD

[liscio]

Originally Posted by DanDan

Chris, given the above, and the lack of comb filtering at HF already mentioned.
Which scenario do you think represents what the ears/brain actually hear? Put another way, which of your Sums or Averages gets closest to the two ear/brain system response?

Now you're way outside of my expertise.

I can't even begin to guess which of the two would be best. Frankly I think the most accurate test is going to involve a Y adapter to physically mix the signals in the air…

[DanDan]

Chris, you have created a monster! The Y thing is funny but I don't think the HF does combine due to Head blocking. I am sure there are Head Transfer Functions or such but let's not go there.
I will play around with this shortly and let you know what your monster is capable of. Intuition says to me-
Two speakers, two mics, ear spaced, with a Jecklin disc between. Or one mic at a time if one doesn't have two mics.
Do a Time Domain Combine of the two measurements.



DD

[Ethan Winer]

Originally Posted by DanDan

There was a time when Bass and Kick pretty much had to be central and mono and devoid of any strange phase. Vinyl disc cutting. However I believe those days are kinda gone

There's still a very good reason to pan heavy LF content to the center: That shares the load between both speakers, reducing distortion.

--Ethan

________________
The Acoustic Treatment Experts

[tINY]

With all the head shadowing and psychoacoustic phenomena going on above 1KHz, you are better off doing 1/3 or 1/6 octave averaging in that range...



-tINY

[liscio]

Originally Posted by DanDan

Chris, you have created a monster! The Y thing is funny but I don't think the HF does combine due to Head blocking. I am sure there are Head Transfer Functions or such but let's not go there.
I will play around with this shortly and let you know what your monster is capable of. Intuition says to me-
Two speakers, two mics, ear spaced, with a Jecklin disc between. Or one mic at a time if one doesn't have two mics.
Do a Time Domain Combine of the two measurements.
DD

Why stop there? Go get a dummy head! Dummy head recording - Wikipedia, the free encyclopedia

(Interestingly, the photo shows a measurement mic placed beside the head, versus the typical "silicone ears containing mic capsules" that I'm used to hearing about. The method depicted in the photo seems like something that wouldn't be too expensive to implement in practice, compared to the $10k units I've encountered…)

[technocolour v2]

Originally Posted by liscio

Why stop there? Go get a dummy head! Dummy head recording - Wikipedia, the free encyclopedia

(Interestingly, the photo shows a measurement mic placed beside the head, versus the typical "silicone ears containing mic capsules" that I'm used to hearing about. The method depicted in the photo seems like something that wouldn't be too expensive to implement in practice, compared to the $10k units I've encountered…)

That head looks freaky! I get strange enough looks with all the treatment in my room as it is

Interesting direction the thread has taken, I'll start a new thread tomorrow with more thorough results to avoid cluttering this one up with more pictures (minor update: things have improved even more, I will do a nice before and after of minimal treatment and full, tweaked treatment... the before and after makes for entertaining viewing )

[DanDan]

LOL. That is an odd version of the dummy head. Most of the pics I have seen have the capsules in fake ears, with simulated pinnae (the squid part) and canals. I have once or twice tried placing my tiny DPA4060 mic in my ear.
Couldn't hear a thing! .

I don't think the colourations caused by in-ear would be useful to us.
I suppose I mis-spoke when I said 'measure what the ears actually hear'
It should be more like, 'measure what is presented to each ear, at it's actual location'.
I suppose it's time I got on with this, i.e. construct a Jecklin disc and do some tests. Or maybe look up Grave Robbers.....

I think it may be quite difficult to get or make a fake head which is soft and absorbent (ish). I presume the Jecklin Disc is a reasonable attempt at the same thing but again without the actual ear canal.

Chris, I am not sure what the terms convolve and correlate mean in this context. Actually I don't really know what Sum and Average mean either?
I will RTFM!
My wish is for a summation of the two ear measurements. The result should represent all the energy presented at both ear positions.
I am fairly convinced that the Time Domain Combine is the most useful here. Do you reckon? I had an RMS plug-in in my older FM2. Does that still exists? Is it appropriate here?

Thanks for taking the time here. The thread probably has a very focussed audience, so worth it IMHO.

Best, DD

[HDJK]

Interesting thread. So why not just sit down in your actual listening position and place the mic next to your ear? That would give a real world measurement, no?

After a forced break I finally got my measurement mic and REW and I'll do my first measuring tomorrow.