Some comments from Floyd Toole about room curve targets, room EQ and more

[Robbo99999]

Thanks for sharing your experiments, Robbo99999 !



I depends. I do my room eq up to 600 Hz.

It would be interesting to know if the strong variations between 100 and 200 Hz in your frequency response are physically caused on the speaker side or on the listener side.
The room simulator of REW can help you find this.



For me, the main variable is the target curve, rather than the smoothing. The success of my room eq strongly depends on slight differences between targets.

Last time I looked at REW for it's room simulator (think it was REW), I concluded that the simple room layout in the simulation did not equate to my room. My room is rectangular, but then imagine a recess on the back wall where I have a dining table, so not an L-shaped room, but a little like that. I also have recessed windows in deep thick walls on the front wall, so they're like a couple of deep boxes (with angled sides ) recessed into the wall behind the speakers. I think I remember concluding that the simulator didn't let me put in all the variables for the room size/shape & speaker placement. Let me know if I'm wrong in how I remember this, and I'll give it another go.

For the variations I see between 100-200Hz are you suggesting that there's a defect in my speakers? I looked back over some historical room EQ's & measurements with different room layouts, as I changed my room layout & speaker position a little since I had these speakers, and the peaks & troughs are not always in the same position which leads me to think that it's not the speakers, but the room. I'll attach my current and a few historical ones here:

The around 30-50Hz peak is always there along with it's dip afterwards. The dip between 100-200Hz changes position with layout by the looks of it.

I know that this is uncommon, but I am in the opposite situation : the room eq makes the sound much better all around the room (and even in the next rooms).
My situation is simple, without any eq, my system makes the noise of an elephant's herd. With eq, the elephants are gone.

I measured a set of frequency responses in various positions in the room. All measurements are single point sweeps, and therefore look quite jagged, compared to a proper MMM or multipoint measurement.

The red dot is the sweet spot. The blue dots are extra measurement positions.
Text, from top to bottom: armrest, center, sofa, table left, table center, table right.

View attachment 92468

In each of the following pictures, there is one measurement with room eq OFF, and one with room eq ON. The one with room eq OFF is always the one with strong peaks in low frequencies.
The room EQ is exactly the same in all graphs.

Sweet spot :
View attachment 92463

25 cm to the right of the sweet spot :
View attachment 92462

80 cm to the left of the sweet spot, sitting in the next place:

View attachment 92464

150 cm behind, sitting at the table, seat 1:

View attachment 92467

150 cm behind, sitting at the table, seat 2:
View attachment 92465

260 cm behind, sitting at the table, seat 3:
View attachment 92466

Yes, given your graphs at all your listening positions then I can see how roomEQ would help all of them, because there's the common peak between 40-60Hz for instance.

 

[Pio2001]

I think I remember concluding that the simulator didn't let me put in all the variables for the room size/shape & speaker placement. Let me know if I'm wrong in how I remember this, and I'll give it another go.

The simulator only works with rectangular closed rooms.
My room is not completely rectangular, but the simulator finds 75% of my room modes (the amplitude is wrong, but the frequencies are right).

For the variations I see between 100-200Hz are you suggesting that there's a defect in my speakers?

That's not what I meant. I was wondering if they were caused by a wall behind the listening position (worst case for a room eq working in the whole room), a wall behind the speakers (one of the best cases for a room eq working in the whole room, together with floor-ceiling modes), or something else.

 

[Robbo99999]

Consolidating Two or More Sets of EQ Filters
Previously (at this post https://www.audiosciencereview.com/...ve-targets-room-eq-and-more.10950/post-562722) I combined an Anechoic EQ of my JBL 308p with RoomEQ below 300Hz. This was 12 filters for the Anechoic EQ and 4 filters for the RoomEQ, for a total of 16 filters altogether. That's fine to use with EqualiserAPO for instance because unlimited number of filters, but I wanted to try using this EQ for my miniDSP 2x4 which has a max of 12 filters available, so I decided to work out how to combine them whilst retaining the same accuracy. Following is an outline of how I achieved this.....there might be a simpler way of consolidating multiple sets of EQ filters into a fewer number of filters & please let me know if you are aware of a simpler process! Following is the process I devised:

1. Applied all 16 filters to be consolidated in REW to a flat line frequency response so I could see the combined effect of all the filters (ie the Total EQ Curve):
   
2. Did a screenshot of the curve above and saved as jpeg file, then I used VirtuixCAD to SPL Trace that created EQ curve to turn it into frequency curve data points which I exported as a text file. I then set that text file as a Target House Curve in REW (obviously the same shape as the curve in the screenshot above).
3. I then created a flat 0dB flat line frequency response in notepad and saved as text file which I imported into REW as a frequency response, and I then used a mixture of REW's automatic EQ and my own manual EQ filters in REW to EQ the flat line to the EQ Curve. You can see the process in this step in the following screenshot. The blue line is the Target Curve, and the red line is the flat 0dB line that I've EQ'd to the "Total EQ Curve" which is essentially the heart of the EQ filter consolidation process. You can see that there are some small deviations from perfect, but you're talking only about 0.2dB in it's worst case places. Yep, so you can see I was able to consolidate down the filters down from 16 filters in the top pic to 12 filters in the pic below, which enables me to use this EQ in my miniDSP. (Note: zoomed in Y-scale in the following pic, to allow for greater EQ accuracy, so that's why it looks a bit different to the curve in my pic above, but they are the same curve)
   

Is there a simpler way of consolidating multiple sets of different EQ filters into fewer filters? I actually enjoyed working out this process for myself & applying with good results, but I'm open to easier or faster ways of doing this!?

 

[Robbo99999]

The simulator only works with rectangular closed rooms.
My room is not completely rectangular, but the simulator finds 75% of my room modes (the amplitude is wrong, but the frequencies are right).



That's not what I meant. I was wondering if they were caused by a wall behind the listening position (worst case for a room eq working in the whole room), a wall behind the speakers (one of the best cases for a room eq working in the whole room, together with floor-ceiling modes), or something else.

I'll check out the simulator again, but I'm betting my room differs too much from the standard rectangle to be valid, but might be interesting to see if there are some peaks & troughs of correlation.

 

[ernestcarl]

I know that this is uncommon, but I am in the opposite situation : the room eq makes the sound much better all around the room (and even in the next rooms).
My situation is simple, without any eq, my system makes the noise of an elephant's herd. With eq, the elephants are gone.

In my open plan living space it's quite the opposite. There is one seat in particular in my dining area where the bass is so bad and boomy hardly anyone sits there; the other seat directly adjacent is still bad, but somewhat a little more tolerable. In the kitchen, there is hardly any bass -- it's all mids and treble. On the couch it's more balanced (general tone control EQ [to lower bass level] is primarily adjusted for this area), but depending on where you sit at either ends of its length bass differences between is quite apparent/notable.

 

[Pio2001]

I wanted to try using this EQ for my miniDSP 2x4 which has a max of 12 filters available

The max is 20 filters, although the 8 remaining ones are not easy to configure.
In the MiniDSP crossover module, you can choose advanced, and type free text. This text can be a set of 8 biquad filters as exported by REW.

You must enter 8 filters. It doesn't accept less. Therefore you need to fill the missing ones with null biquads (REW can export an example).
The filters must be numbered from 1 to 8. Therefore you will have to change the numbers exported by REW.
The last row must not have a comma at the end. All other rows must have a comma.

I don't know how MiniDSP deals with a positive correction followed by a negative one (internal clipping ?), therefore I always put negative correction in the first module, and positive corrections in the last.

Edit : the identity biquad is:
biquad1,
b0=1.0,
b1=0.0,
b2=0.0,
a1=0.0,
a2=0.0

 

[Bliman]

Hi all,
Is there a ideal curve that I can load in REW to see if my curve follows it or that I can eq it? And where do I find it?
And how do I load the curve?

 

[Robbo99999]

Hi all,
Is there a ideal curve that I can load in REW to see if my curve follows it or that I can eq it? And where do I find it?
And how do I load the curve?

Harman Curve is often the most widely accepted standard. I'll attach it for you here:

 

[thewas]

Keep in mind that there doesn't exist the one Harman curve but it always varied depending on loudspeakers and time when it was published, https://www.audiosciencereview.com/...esponses-in-csv-txt-format.16401/#post-532004

If you also read the comments from Toole in the beginning it doesn't make sense to EQ a loudspeaker to a prefixed target as that depends on the loudspeaker directivity, room reverberation and listening distance, what you actually want is above transition a flat direct sound (of course ideally with a smooth directivity loudspeaker) and below adjustable smooth bass to room, taste, recordings and listening level.

 

[Robbo99999]

Keep in mind that there doesn't exist the one Harman curve but it always varied depending on loudspeakers and time when it was published, https://www.audiosciencereview.com/...esponses-in-csv-txt-format.16401/#post-532004

If you also read the comments from Toole in the beginning it doesn't make sense to EQ a loudspeaker to a prefixed target as that depends on the loudspeaker directivity, room reverberation and listening distance, what you actually want is above transition a flat direct sound (of course ideally with a smooth directivity loudspeaker) and below adjustable smooth bass to room, taste, recordings and listening level.

Good points on the variation of the Harman Curve over the years.

Although I've found a perfect match to the Harman Curve to be more preferable for me. Previously I had briefly done an EQ based purely on a flat Anechoic Measured Listening Window for my recently Amir reviewed JBL 308p Mkii speakers, and it was too bright, might have been unit to unit variation, but I actually prefer the following (which is HF Trim Switch on back of speaker at -2dB position):
Anechoic Listening Window:

Corresponding in Room "Measurment":

 

[thewas]

Nothing wrong with personal preference (often when listened in nearfield a HF trim can be useful as flat direct sound can sound a bit too hot with some recording), although I wouldn't call your match to your target really perfect as it has wide deviations which will still be significant even with psychoacoustic smoothing.
Of course there will be always also cases were a predefined matches well, for example in my room and preference the Harman targets don't work as good, even with my Harman loudspeakers (JBL 305 MKI and JBL 4321SE in the past).

 

[Pio2001]

Previously I had briefly done an EQ based purely on a flat Anechoic Measured Listening Window for my recently Amir reviewed JBL 308p Mkii speakers, and it was too bright, might have been unit to unit variation, but I actually prefer the following (which is HF Trim Switch on back of speaker at -2dB position):

It seems normal for mid-field listening in a normal room.
The Neumann KH-120 manual says that in this configuration, the treble trim should be set to -1 (and 0 for near-field listening in a treated room).

I have just succeeded in equalizing the JBL 305P mk2, with a result as good as, or better, than the Neumann KH-120, and one of the key was also to include a treble trim to avoid the result sounding too bright.

 

[Robbo99999]

Nothing wrong with personal preference (often when listened in nearfield a HF trim can be useful as flat direct sound can sound a bit too hot with some recording), although I wouldn't call your match to your target really perfect as it has wide deviations which will still be significant even with psychoacoustic smoothing.
Of course there will be always also cases were a predefined matches well, for example in my room and preference the Harman targets don't work as good, even with my Harman loudspeakers (JBL 305 MKI and JBL 4321SE in the past).

Oh yeah, I'm not talking about the bass being a perfect exact match, I'm talking about the treble lining up perfectly with the Harman Curve with HF Trim Switch at -2dB position....and I think the bass lines up well with the Harman Curve too in general tonal terms, but obviously big room modes there and I used the harshest smoothing variable in REW ("Var Smoothing" ie no smoothing on bass.). At the moment I actually don't want to apply RoomEQ to this speaker because it sounds great as it is, and I know ironing down the peaks will reduce the bass impact. I have experimented with boosting the dips by applying a Low Shelf boost and then cutting the peaks but it reduces overall clarity (as well as making the speaker sound worse at other places in the room) even though the graph looks prettier!

It seems normal for mid-field listening in a normal room.
The Neumann KH-120 manual says that in this configuration, the treble trim should be set to -1 (and 0 for near-field listening in a treated room).

I have just succeeded in equalizing the JBL 305P mk2, with a result as good as, or better, than the Neumann KH-120, and one of the key was also to include a treble trim to avoid the result sounding too bright.

I reckon the recommendation for "0" for near field in a treated room must be mainly due to it being a treated room, because my understanding is that in nearfield the speaker will sound brighter than in mid & far field?

 

[Pio2001]

I reckon the recommendation for "0" for near field in a treated room must be mainly due to it being a treated room, because my understanding is that in nearfield the speaker will sound brighter than in mid & far field?

Listening near filed means that you listen to a majority of direct sound and a minority of diffused sound. Which is the same thing as listening in a treated room. It decreases the amount of diffused sound.

The preference for a tilted treble curve is strange. In all cases, the neutral sound should be closer to reality.
One possible explanation would be that a speaker would have a sound power curve (the amount of sound radiated in all directions of space) higher than an average musical instrument in high frequencies, therefore giving a brighter sound in an average room.

 

[Bliman]

Harman Curve is often the most widely accepted standard. I'll attach it for you here:

Thank you

 

[Bliman]

Keep in mind that there doesn't exist the one Harman curve but it always varied depending on loudspeakers and time when it was published, https://www.audiosciencereview.com/...esponses-in-csv-txt-format.16401/#post-532004

If you also read the comments from Toole in the beginning it doesn't make sense to EQ a loudspeaker to a prefixed target as that depends on the loudspeaker directivity, room reverberation and listening distance, what you actually want is above transition a flat direct sound (of course ideally with a smooth directivity loudspeaker) and below adjustable smooth bass to room, taste, recordings and listening level.

Where lies the transition, or how do you find it? And how do you push the imported frequency curve upward because the Harman curve is at 0 dB but my measurement is at around 85 dB?
Is it also normal that it only goes to 30 Hz?
So basically you are saying that I should just tune it to taste?

 

[thewas]

Where lies the transition, or how do you find it? And how do you push the imported frequency curve upward because the Harman curve is at 0 dB but my measurement is at around 85 dB?

At usual room sizes around approximately 300-800 Hz, you will also see it when you do measurements that above a frequency the responses become smoother and closer to what you expect from the loudspeaker PIR.

And how do you push the imported frequency curve upward because the Harman curve is at 0 dB but my measurement is at around 85 dB?

At REW you can set the target level at the EQ window as you like.

Is it also normal that it only goes to 30 Hz?

On smaller loudspeakers yes.

So basically you are saying that I should just tune it to taste

Taste is always a good check to see what you did makes sense.
Better also have a look at this video

 

[Bliman]

At usual room sizes around approximately 300-800 Hz, you will also see it when you do measurements that above a frequency the responses become smoother and closer to what you expect from the loudspeaker PIR.

At REW you can set the target level at the EQ window as you like.

On smaller loudspeakers yes.

Taste is always a good check to see what you did makes sense.
Better also have a look at this video

Thank you. It helped me and I shall certainly watch the video.
I don't get why the curve stops at 30 Hz. I mean the loaded Harman curve because my measurement goes to 10 Hz to where I measured.

 

[thewas]

Thank you. It helped me and I shall certainly watch the video.
I don't get why the curve stops at 30 Hz. I mean the loaded Harman curve because my measurement goes to 10 Hz to where I measured.

Ah, now I see what you mean, just add in the beginning one more line with the relative target level you want to have at 10 Hz like

10 5.0

or use another target like from this thread https://www.audiosciencereview.com/...ker-target-responses-in-csv-txt-format.16401/
Mind you correcting down to 10 Hz usually is not needed and if you add too much boost there to reach your desired target you usually get only disadvantages.

 

[Bliman]

Ah, now I see what you mean, just add in the beginning one more line with the relative target level you want to have at 10 Hz like

10 5.0

or use another target like from this thread https://www.audiosciencereview.com/...ker-target-responses-in-csv-txt-format.16401/
Mind you correcting down to 10 Hz usually is not needed and if you add too much boost there to reach your desired target you usually get only disadvantages.

Thank you for helping me.