What does an "accurate" in-room response look like?

Sigberg Audio

Sigberg Audio

Audioholic
I saw some somewhat similar threads here but not with a lot of replies, so I decided to start a new one. The topic came up in the Sigberg Audio Manta developmpent thread I guess derived from my post #115 there, which derailed that thread into a discussion between @TLS Guy and me. I think this discussion is somewhat tangential to the topic of that thread, so I'm starting this thread in an attempt to continue the discussion here.

The core topic here I think is how an in-room response should look in order to get accurate sound reproduction across the audio band.

So I'll start off by quoting the final post from TLS Guy on the matter from the other thread, if you want to read how we ended up here I guess you need to go to the #115 post from the other thread as linked above and read a couple of pages forward. :)

So to cut to the chase. What is my opinion of Target Curves. I say quite unequivocally that they are nonsense. You don't need a curve, you need a flat straight line and the flatter the better. However there is some license on the last octave or so 20 Hz to 40 to 50 Hz, in the area of room gain. To get extremely "hair shirt" in this region can make reproduction excessively dry. So a small amount of leeway in this area is required. I think this is actually best judged by ear, on a room to room basis.
I absolutely do not agree with the hot sub and attenuation above sub range. That in my view has come about by pandering to speakers with inadequate power response on the two octaves or so above sub range. I will state this again for the record that this defect likely affects the majority of speakers, with the resulting lack of balance at power, limiting realistic reproduction. This became clear to me as a result of the invaluable experience of making live recordings.

So this has been a long winded way of saying we need more straight lines and less curves.

As evidenced from your shared graphs (see below), your own system has a 20dB drop from 20hz to 10khz (10dB from 100hz to 10khz)? So you need to elaborate on how that follows your idea of a straight line.

I'm very aware that I may be misunderstanding either what you mean by flat/straight and/or what this graph visualizes (actually I'm pretty confident it's one or the other). So I'm not saying you're contradicting yourself or that you are wrong, I'm asking for more information to understand your standpoint. :) When I understand that, I will be happy to share my own views on this interesting subject as well. :)

(@TLS Guy in-room graph as shared in the other thread, if I understand his references correctly)
1665383989698.png
 
TLS Guy

TLS Guy

Seriously, I have no life.
I saw some somewhat similar threads here but not with a lot of replies, so I decided to start a new one. The topic came up in the Sigberg Audio Manta developmpent thread I guess derived from my post #115 there, which derailed that thread into a discussion between @TLS Guy and me. I think this discussion is somewhat tangential to the topic of that thread, so I'm starting this thread in an attempt to continue the discussion here.

The core topic here I think is how an in-room response should look in order to get accurate sound reproduction across the audio band.

So I'll start off by quoting the final post from TLS Guy on the matter from the other thread, if you want to read how we ended up here I guess you need to go to the #115 post from the other thread as linked above and read a couple of pages forward. :)




As evidenced from your shared graphs (see below), your own system has a 20dB drop from 20hz to 10khz (10dB from 100hz to 10khz)? So you need to elaborate on how that follows your idea of a straight line.

I'm very aware that I may be misunderstanding either what you mean by flat/straight and/or what this graph visualizes (actually I'm pretty confident it's one or the other). So I'm not saying you're contradicting yourself or that you are wrong, I'm asking for more information to understand your standpoint. :) When I understand that, I will be happy to share my own views on this interesting subject as well. :)

(@TLS Guy in-room graph as shared in the other thread, if I understand his references correctly)
View attachment 58099
Well, that is a good question. However, that curve is taken at the MLP. So, the HF will taper at a distance which at the MLP in that room is about 16 ft.

Now the ear expects that. It will be the same taper for someone speaking in the room or playing an instrument. As I said, the ear expects that. Where you need the straight line is on axis of the tweeter at a distance of one to two meters. I use 2 meters as a rule. Then you have to let the room do the rest. It is my observation that you can not "force" the room by what your notion of what the room should do. I think somehow our brains are wired to know what the environment will do.

Even so, with speakers there is a fall off of HF as you move off axis, which will be dependent on tweeter design more than anything else. However as you move off axis the response must closely mirror the axis response except the off axis taper with increasing frequency. If not the our brains really cry foul.

So lets us look at the responses, the first on axis 2 meters distant from the tweeter. The taper above 15K is due to Omnimic, that is a well documented problem.



Now lets see one more time what the room does with it. There is room gain, AND the expected taper in the HF due to the room attenuation of the reflected portion of the HF spectrum. All rooms do this to a varying degree, and the ear expects this. If you try and modify it significantly the ear cries foul as it does not expect there not to be this room attenuation. I should state, there is no room Eq, using Audyssey, or anything like that.



That is exactly what you would expect a room to do at 12 to 16 ft. distances from the speaker. If you try an correct this, then it does not sound right at all.

The last two octaves are a different case though. Here we have room gain, and room long wave reflections with the potential for peaks and nulls. These are determined to a very large extent by the ratios of the room dimensions. If you have the luxury of designing and building your room, then you have the ability to optimize the room dimension ratios. This is an area where determining the best response here does benefit from room optimization, and critical listening here, without preconceived notions, achieves the best results. This is where your active design will give your customers a degree of optimization not generally available now.

These auto calibration programs, impose preconceived notions on the room, which likely have more chances of being suboptimal than not.

What, I and others have noted that from 1KHz to 10 KHz at 1 to 2 meters, the axis response needs to be as ruler straight as humanly possible, with the off axis responses falling off in a smooth fashion increasing their slope with frequency. That is the best a tweeters can currently offer.

The problem area is that range from 80 to I Khz or so, particularly from 80 to 500 Hz. This latter includes the Baffle Step compensation range of most speakers, and is a range where the involved drivers have to receive large amounts of power. This is where good motor design sorts the "wheat from the Chaff". It is a crucial band however. I pointed out previously that it contains the fundamental frequencies of a large number of instruments, especially the foundational instruments. This is not an area where you want to be lacking in power, or dip the response.

This is an area though, where room and speaker position do come into play. Here your active designs will have a huge advantage. I have found with my active feeds, with variable BSC, that the amount required is in fact highly room, and also speaker position in the room dependent.

So this does leave me with one piece of advice for you, to make the BSC variable, and also preferably have some way to allow the consumer to set it with instrumentation. Many members here already have that facility.
 
Sigberg Audio

Sigberg Audio

Audioholic
@TLS Guy Then we are largely in agreement. It was confusing that you were referring to straight lines when we (or at least I) were explicitly discussing and referring solely to in-room measurements at the listening position in the other thread.

I also agree that the 80-500hz area is cruicial, which is why we have dedicated drivers for this range (more precisely 90-600hz) in both SBS.1 and the upcoming Manta. And both speakers have, not dips, but a slight lift / tilt in this area, as evidenced by the anechoic response of the SBS speaker:

1665417434986.png


BSC is quite effectively adressed in the SBS speakers by its 2.5-way design (dual drivers in the 90-600hz range), and in the Manta by having a massive midbass driver that happens to have roughly 6dB higher sensitivity than the coax. Both are of course fine tuned in addition to that. If the consumer needs to alter this due to room issues, there are as mentioned in the other thread both boost/cut and shelf filters available individually in each speakers, so in practice the BSC can be adjusted. The final octaves (~20-100hz) are also easily adjustable by either EQ and/or adjusting the volume of the subwoofer.
 
TLS Guy

TLS Guy

Seriously, I have no life.
@TLS Guy Then we are largely in agreement. It was confusing that you were referring to straight lines when we (or at least I) were explicitly discussing and referring solely to in-room measurements at the listening position in the other thread.

I also agree that the 80-500hz area is cruicial, which is why we have dedicated drivers for this range (more precisely 90-600hz) in both SBS.1 and the upcoming Manta. And both speakers have, not dips, but a slight lift / tilt in this area, as evidenced by the anechoic response of the SBS speaker:

View attachment 58102

BSC is quite effectively adressed in the SBS speakers by its 2.5-way design (dual drivers in the 90-600hz range), and in the Manta by having a massive midbass driver that happens to have roughly 6dB higher sensitivity than the coax. Both are of course fine tuned in addition to that. If the consumer needs to alter this due to room issues, there are as mentioned in the other thread both boost/cut and shelf filters available individually in each speakers, so in practice the BSC can be adjusted. The final octaves (~20-100hz) are also easily adjustable by either EQ and/or adjusting the volume of the subwoofer.
Yes, I think we do agree pretty much. However I have a hunch that your room response is affected by a close boundary effect and not the total room response.
If it were me, I would be doing further investigation of that, but that is just me.
 
AcuDefTechGuy

AcuDefTechGuy

Audioholic Jedi
I think "accurate" means the in-room FR is within the +/-3dB window.

But some people may ask, will +/-2dB sound better than +/-3dB and will +/-1dB sound better than +/-2dB, which is another can of worms. :D
 
Sigberg Audio

Sigberg Audio

Audioholic
I think "accurate" means the in-room FR is within the +/-3dB window.

But some people may ask, will +/-2dB sound better than +/-3dB and will +/-1dB sound better than +/-2dB, which is another can of worms. :D
But +/-3dB according to which reference? A typical in-room response will be falling, so a slope rather than a straight line. So are we looking for +/-3dB deviation from that then perhaps?

Personally I would say using EQ to hunt for +/-2dB or +/-1dB is a slippery slope that quickly will become problematic and give the wrong sound, especially from 2-300hz and above. This is due to the fact that direct sound from the speakers will begin to dominate, and correcting for the measured response will give you an inaccurate direct sound.
 
Sigberg Audio

Sigberg Audio

Audioholic
Yes, I think we do agree pretty much. However I have a hunch that your room response is affected by a close boundary effect and not the total room response.
If it were me, I would be doing further investigation of that, but that is just me.
I agree. Where we (possibly) disagree is whether this is something to be addressed in the design of the speaker, or something specific to my room/listening position (and as such should be addressed by room correction).

That being said, the speaker is under development and the response between 100-600hz is one of the things that is not finalized, so more experimentation and testing with this range will be done. :)
 
AcuDefTechGuy

AcuDefTechGuy

Audioholic Jedi
But +/-3dB according to which reference? A typical in-room response will be falling, so a slope rather than a straight line. So are we looking for +/-3dB deviation from that then perhaps?
So that means in-room response of +/-4dB to +/-6dB? :D

Personally I would say using EQ to hunt for +/-2dB or +/-1dB is a slippery slope that quickly will become problematic and give the wrong sound, especially from 2-300hz and above. This is due to the fact that direct sound from the speakers will begin to dominate, and correcting for the measured response will give you an inaccurate direct sound.
Yeah, I agree. Chasing a flat-line in-room response is silly. :D
 
TLS Guy

TLS Guy

Seriously, I have no life.
I agree. Where we (possibly) disagree is whether this is something to be addressed in the design of the speaker, or something specific to my room/listening position (and as such should be addressed by room correction).

That being said, the speaker is under development and the response between 100-600hz is one of the things that is not finalized, so more experimentation and testing with this range will be done. :)
Yes, that is exactly the question I am raising. I just have, actually a pretty strong hunch that it is due to reflection from the back wall. The quickest way to test that is to see what happens is a couple of different rooms. I have understood, possibly incorrectly, that these speakers are designed to be used up against a wall. If that is true, then I think this does need addressing in the design if this occurs any time the speakers are up against a wall.

By the way, have you listened to that YouTube video I sent you? That is a true very high quality totally co-incident phase coherent recording. Those recordings are excellent for speaker evaluations. Apart from the fact, that I think that is a good and valid approach, I did it for speaker evaluation. As most recordings are far from phase coherent, it makes hard to evaluate the phase anomalies of speakers.
 
Sigberg Audio

Sigberg Audio

Audioholic
So that means in-room response of +/-4dB to +/-6dB? :D
I'm not sure defining these types of goals makes sense at all, at least not for the full spectrum. Below 100-150hz you can apply EQ pretty safely and make it basically as flat as you want. 150-500hz it gets dicey, personally I never correct more than 1-3dB either way, depending on the type of problem, and always listen to see if it has the desired effect. Above 500hz I wouldn't apply anything other than changing tonal balance (so tilting the response, not correcting deviations to the response).

All of this is assuming the speakers were accurate to begin with of course, and thus any deviations seen is then necessarily room related.
 
Last edited:
Sigberg Audio

Sigberg Audio

Audioholic
Yes, that is exactly the question I am raising. I just have, actually a pretty strong hunch that it is due to reflection from the back wall. The quickest way to test that is to see what happens is a couple of different rooms. I have understood, possibly incorrectly, that these speakers are designed to be used up against a wall. If that is true, then I think this does need addressing in the design if this occurs any time the speakers are up against a wall.
They are designed to be placed 10-20cm from the wall. Any closer than ~10cm is practically difficult due to power and XLR cables and being able to access the amplifier.

Deviations in response between 100-300hz is indeed often due to the wall behind the speakers or some kind of boundary, so that's a reasonable assumption. But as I've said repeatedly, the referenced measurements from the other thread were stereo in-room measurements done from the listening position, they really can't be used to figure out much without knowing anything about the room, the listening position or the speaker positions.

Here's the left and right speaker from the listening position. They're both at equal distance from the wall. The red graph (left) where we see most problems are located close to a corner. The green graph (right) is free standing (no walls directly on either side), and as you can see (this is still zero smothing) has minimal issues, and the 180hz dip is not present.

1665432464245.png


It's impossible to perfectly eliminate these types of boundary effects. What I can do (and will do) is measure this speaker anechoically (as we already did with the previous prototype), and control whether the cardioid effect in this frequency range is even. If it is, that's the best we can do with this approach. I expect that it will be. The main difference between this prototype and the previous is changes made to make the cardioid effect work even better in the lowest octaves (~50-200hz).

By the way, have you listened to that YouTube video I sent you? That is a true very high quality totally co-incident phase coherent recording. Those recordings are excellent for speaker evaluations. Apart from the fact, that I think that is a good and valid approach, I did it for speaker evaluation. As most recordings are far from phase coherent, it makes hard to evaluate the phase anomalies of speakers.
Not yet. Another way of evaluating the phase coherence of the speaker is to measure frequency response and phase anechoically, which we of course also do. Both SBS and Manta maintains linear phase through the crossovers between both midbass/midrange and midrange/tweeter.
 
Trell

Trell

Audioholic Spartan
They are designed to be placed 10-20cm from the wall. Any closer than ~10cm is practically difficult due to power and XLR cables and being able to access the amplifier.
The cables are inserted horizontally and not vertically?
 
TLS Guy

TLS Guy

Seriously, I have no life.
They are designed to be placed 10-20cm from the wall. Any closer than ~10cm is practically difficult due to power and XLR cables and being able to access the amplifier.

Deviations in response between 100-300hz is indeed often due to the wall behind the speakers or some kind of boundary, so that's a reasonable assumption. But as I've said repeatedly, the referenced measurements from the other thread were stereo in-room measurements done from the listening position, they really can't be used to figure out much without knowing anything about the room, the listening position or the speaker positions.

Here's the left and right speaker from the listening position. They're both at equal distance from the wall. The red graph (left) where we see most problems are located close to a corner. The green graph (right) is free standing (no walls directly on either side), and as you can see (this is still zero smothing) has minimal issues, and the 180hz dip is not present.

View attachment 58105

It's impossible to perfectly eliminate these types of boundary effects. What I can do (and will do) is measure this speaker anechoically (as we already did with the previous prototype), and control whether the cardioid effect in this frequency range is even. If it is, that's the best we can do with this approach. I expect that it will be. The main difference between this prototype and the previous is changes made to make the cardioid effect work even better in the lowest octaves (~50-200hz).



Not yet. Another way of evaluating the phase coherence of the speaker is to measure frequency response and phase anechoically, which we of course also do. Both SBS and Manta maintains linear phase through the crossovers between both midbass/midrange and midrange/tweeter.
Yes, those are the breaks. It is disappointing that the results are not better close to the wall. That red curve is pretty lumpy. It seems the bass driver is responsible. It is a pity the cardioid designing was not more effective in limiting this problem. The green trace is really very good and I doubt that narrow Q dip just above 130 Hz is audible. However I suspect that red trace is not optimal acoustically. So essentially these speakers actually need conventional placement. Adsorbing material behind the speaker might help.

This technique I used of the swept front baffle, is effective at dispersing these troublesome reflections.



I am indebted to the late John Wright of TDL who used this technique in his later designs. I also used his design of dual TLs tuned half an octave apart. He used this technique in his last design, shortly before his death. Few of those speakers were ever made, and their cost was in the stratosphere. I never turn any good idea down.

You have the advantage of digital crossovers and so can make a truly phase coherent speaker. In the above design the two 7" drivers are allowed to roll off acoustically, as they are robust drivers. The BSC is sent to the upper 10" driver and the SUB LFE signal sent to both 10" drivers. It is unusual, and many would say "oddball". They are however very effective speakers. Three Quad 909 power amps drive that pair of speakers.
 

Latest posts

newsletter

  • RBHsound.com
  • BlueJeansCable.com
  • SVS Sound Subwoofers
  • Experience the Martin Logan Montis
Top