Trying to isolate nulls in system frequency response

Greg Dunn

Member
Thread Starter
Joined
Oct 4, 2017
Messages
107
I have a room issue which is not serious but I'm struggling to get a handle on the cause. The situation is that I have a pair of Magneplanar Tympani speakers, which each provide a sound source approximately 4 feet wide by 6 feet tall, so there is no single location for each speaker which can be identified as the "source". Each of the 3 panels on each side covers a different portion of the frequency range, some overlapping, others not (the system is tri-amplified). The room is small enough that I also don't have much of a choice about where they are placed. I've moved them around as much as I can, but they are where they are and it's not going to change much. They are crossed over to a Dayton sub at 63 Hz.

The problem is that I have a few large nulls in the response between 150-250 Hz, likely due to room size and LP location. Most of the response irregularities yield readily to EQ and panel position, leaving only these. I have attempted to simulate the room in REW, but the problem is that the speakers are too wide to simulate, and even the portion which covers the frequency range in question for each speaker is about 3 feet by 6 feet. So the room simulation of the speakers doesn't match the actual room very well. It seems that the nulls somewhat relate to the distance between the perceived midpoints of the low/mid panels for each speaker, but again, this distance isn't going to change so I can't alter it. Moving the measurement mic causes the nulls to appear at different frequencies to some extent; however, I've done the moving mic measurement average, and it actually tends to accentuate the major null at 140-160 Hz over a traverse, rather than smooth it out as you might hope. Moving the LP forward or to the rear doesn't seem to help with the major null; moving to either side puts me far enough off axis that it starts to impair the stereo image. So it does seem to be a real phenomenon and not an artifact of measurement.

Up till now I've been ignoring it because it subjectively seems to be minor and doesn't impair the enjoyment of music. But I'd like to try to address it if I can. I have another null near the upper crossover frequency (1300 Hz) which seems to be an interaction between the individual panels, so I wouldn't be surprised if this is another similar effect. But without being able to simulate the room/speaker situation, I'm only making slow progress.

I've attached the mdat file of a nominal measurement (which should include the room sim), a MMM RTA capture, and a picture of the speakers (not quite the current positioning) to give a sense of my environment. You can see that the RTA of both speakers together exhibits the 150 Hz null, while it's only hinted at in the swept measurements. The speakers are definitely in phase electrically, which is why I'm suspecting some kind of inter-panel cancellation effect, reason unknown.
 

Attachments

  • T-1C home!.jpg
    T-1C home!.jpg
    73.3 KB · Views: 263
  • RTA 050120.jpg
    RTA 050120.jpg
    104.3 KB · Views: 266
  • L-R-Sub Time Delay test 050120.mdat
    5.9 MB · Views: 47
Last edited:

DanDan

Moderator
Staff member
Joined
Mar 10, 2018
Messages
743
The measurement file is not downloading successfully here. You will need to identify the type for each null. Modal ones don't change frequency when you move the mic. Non resonant SBIR nulls change frequency will location. I suspect a floor bounce for at least one of your dips. Try moving the mic towards a spot on the floor half way between you and speakers. If the frequency rises it suggests you are moving towards that reflection.
I guess with those beasts a ceiling bounce is possible also. Have you treatment anywhere? Cloud? Side Reflection? Back Wall?
 

Greg Dunn

Member
Thread Starter
Joined
Oct 4, 2017
Messages
107
I re-uploaded the mdat file; it looks OK at this end, though I did create it with b49 of REW, just FYI.

I'll do some more measurements today. I have heavy carpet on the floor, the ceiling is spackled plaster, and I do have plenty of absorbing material at strategic spots on the walls. The rear wall is actually diagonally bisected by a stairwell with some stuff stacked along it, and the upper part of that wall is textured with some hangings. It's not a reflective room by any standards. :)
 

DanDan

Moderator
Staff member
Joined
Mar 10, 2018
Messages
743
I don't think heavy carpet would have much or any effect on a 150Hz floor bounce. The walls are brick and perhaps the floor is concrete? The combination as you describe it would lead to a very bass heavy room tone. But more importantly perhaps, the LF reflections would be strong, causing robust nulls. Edit, the.mdat downloaded....... so perhaps ignore what I just wrote......LOL The Room Tone Topt looks absolutely fine on the L and R. Something odd with the sub measurement though, 7 Seconds of Decay at 500Hz seems unlikely... I tried the A+B trace arithmetic and your lack of 'warmth' became fairly noticeable. Have you tried different listening positions?
 
Last edited:

Greg Dunn

Member
Thread Starter
Joined
Oct 4, 2017
Messages
107
Walls and ceiling are all drywall (not plaster - that was a dumb typo), but the floor is poured concrete. Actually, the bass was a little thin on most recordings until I adjusted the lower octaves a bit upward. The sub does not see anything much above 63 Hz (24 dB/octave crossover); the only reason I measured it full range was from comments in the REW forum that I needed to measure all drivers full range to get valid impulse results. Anything above probably 120 Hz is mostly from the main panels.

I hope to get some measurements tonight when the house is a little quieter. Different listening positions do not seem to affect anything audible except the stability of the stereo image, but after doing the new measurement positions I'll see if a dramatic shift in distance from the speakers has any effect on the bass.
 

DanDan

Moderator
Staff member
Joined
Mar 10, 2018
Messages
743
Good. I would continue as you are, adjust the Bass but also the HF roll off. Your response is already spectacularly good in many respects, Some simple tonal talioring with cut Eq should make you very happy. As ever, I personally rely on and recommend Dirac Live. I have been assured by trusted friends that Accurate is also pretty amazing, maybe even better, but my rig is Mac based so I will stick with my personal experienced recommendation. I have a full blown Mastering scenario and have tried Eqs of that standard, but DL just does it better for some unknown to me reasons. They are pretty much friends of mine, but please do not consider my endorsement as advertising. I am just sharing my experiences.
 
Last edited:

Greg Dunn

Member
Thread Starter
Joined
Oct 4, 2017
Messages
107
Edit: I too am a Mac user, so Acourate is out. I have heard great things about Dirac but it seems very expensive just to automate something I can do with my own gear - and potentially learn more about my system / room at the same time.

I found time to run the RTAs when the house was quiet; they are captured at the LP ± several feet, all on the same F-R axis. Looks like getting closer (-2 feet) really accentuates some of the dips, while moving away (+2 feet) helps a bit. The ~150 Hz dip is not changed a lot though it is smoother at some positions.

The 2k dip stays pretty much the same too, and it is definitely an intra-panel interaction as moving to the sides (far enough to disturb imaging) fills it in a bit. It doesn't respond to EQ either; I tried overlapping the panel XO frequencies and all it did was make the imaging worse. I'm thinking it might be wiser to knock the 300-500 Hz region and the bass down rather than try to fix this one.

For a typical MMM RTA measurement centered around the LP, I get a much flatter response except for the 150 Hz region, but perhaps I need to concentrate more on the on-axis response since I don't move around much during serious listening?
 

Attachments

  • RTA - varying distances.mdat
    1.7 MB · Views: 40
Last edited:

linearphase

Member
Joined
Jan 17, 2019
Messages
59
Wow that is a big speaker in a small room.
The two dips noted are likely due to floor and wall bounce.
As noted absorbing energy below 300 Hz requires lots of thick material or limp mass absorbers. To validate this theory you could place a large chunk of fiberglass or pillows stacked two feet deep in a pile about half way between the speakers and the mic.
The tight fit does not allow you to move the speakers away from the walls. I don’t think there is much you can do here except to try to boost about HALF of the dip out.
 
  • Post hidden due to user being banned.

Greg Dunn

Member
Thread Starter
Joined
Oct 4, 2017
Messages
107
I tried putting a very large and thick cushion on the floor midway between the speakers and the LP; it helped the 270 Hz dip noticeably, but did nothing for the 130-140 Hz one. Moving it around and raising it off the floor made essentially no additional difference. I couldn't hear any changes at all in sound character, which isn't too surprising.

I have a list of the panel resonances from the speakers as originally manufactured, but not after the factory refurb. They're probably pretty close. The LF panels had resonances around 45-47 Hz; the mid panels had resonances around 53 Hz, 65 Hz and 71 Hz. These are all treated by the panel tie-downs added during the test and validation phase. You can barely see peaks in the bass due to these resonances in the swept FR test, though the lowest ones are below the crossover frequency and essentially buried.

I think the 160 Hz peak is probably the result of too much boost in that range on the EQ. I'm limited by having only a 31-band GEQ to use, though I'm not sure a PEQ would be that much help in dealing with what are essentially room issues. Still investigating options in that regard...
 

DanDan

Moderator
Staff member
Joined
Mar 10, 2018
Messages
743
PEQ is very much better at addressing room issues. Modes are typically only a few Hz wide. While one can sort of create a Target Curve using Parametric and Shelving Eqs I much prefer the flexibility of a freely shapable curve. A bendy Eq if you will.
The old Bruel&Kjaer curve is nicely, well, curved. I have tried all sorts of variations using Mastering Eqs etc. etc. Ultimately the curved bits do sound different to straight lines between waypoints. DiracLive uses waypoints but you can add enough to get bent! There is a Harrison Mastering Eq which is totally bendy and which works well also
 

Greg Dunn

Member
Thread Starter
Joined
Oct 4, 2017
Messages
107
I haven't finished tweaking filters yet, because I'm still working out the proper frequency and Q values for each anomaly. But I think it's on the right track now. The 140 Hz, 270 Hz and 1.5 kHz dips are greatly reduced now and once I get the bottom octaves level matched with the rest, it should be pretty nice.

This is probably old news to some, but it was necessary to use a fairly high Q filter to attack the deep nulls without smearing the width of them instead. It's not high enough to cause ringing, just a lot narrower than a 1/3 octave EQ will give you. Once I get all the levels to my liking, I'm going to run a sweep to see if there are any issues with the RT60 and waterfall measurements. Subjectively, though, the transient response seems fine; at least, no worse than it was.

Also, if using a digital filter in your system software, keep an eye on the signal output. Adding some boost in the low end can readily cause digital clipping with some sources, especially if you're already normalizing source signals!
 

damic

Registered
Joined
Jun 17, 2017
Messages
3
You have big deeps at 57Hz, 140Hz, even one at 190Hz, could be noticed. A have similar setup, of course, smaller Apogee Duetta Sigs, but with four subs. System is active. IME, remedy is in additional subs.

Agree with DanDan, PEQ is much better than GEQ. Some kind of target curve will give better bass.
 

linearphase

Member
Joined
Jan 17, 2019
Messages
59
I tried putting a very large and thick cushion on the floor midway between the speakers and the LP; it helped the 270 Hz dip noticeably, but did nothing for the 130-140 Hz one. Moving it around and raising it off the floor made essentially no additional difference. I couldn't hear any changes at all in sound character, which isn't too surprising.

I have a list of the panel resonances from the speakers as originally manufactured, but not after the factory refurb. They're probably pretty close. The LF panels had resonances around 45-47 Hz; the mid panels had resonances around 53 Hz, 65 Hz and 71 Hz. These are all treated by the panel tie-downs added during the test and validation phase. You can barely see peaks in the bass due to these resonances in the swept FR test, though the lowest ones are below the crossover frequency and essentially buried.

I think the 160 Hz peak is probably the result of too much boost in that range on the EQ. I'm limited by having only a 31-band GEQ to use, though I'm not sure a PEQ would be that much help in dealing with what are essentially room issues. Still investigating options in that regard...

I would not worry about the peaks at this time. They are very minor. The 130-140 dip is the next thing you should investigate. Generally the audibility of deviations are proportional to the the “area” under the Curve. That is the width times the height. That dip may be due to a ceiling or wall reflection. Does it change in frequency as you move the mic? The fact you can not hear a difference likely means there was not enough energy in the program material in the frequency range. Pink noise is a good tool to use ONCE YOUR EARS BECOME CALIBRATED.
It is annoying to listen to though.
 

linearphase

Member
Joined
Jan 17, 2019
Messages
59
Observations/
Dealing with the dips with EQ does not work very well usually.
The floor absorber sounds like it works on the higher dip. It was probably just not large or thick enough to attack the lower frequency one.
Given that the dips are multiples of one another it is quite possible they are both caused by the same reflection spot that cause delays at the frequencies that are 1/4 and 1/2 wavelength. These are about 4.2 and 2.1 feet of added path length. It is easy find these paths just google Calibrate Acoustic String. Very easy to make and effective. I use this tool almost everyday. in this case you are simply looking for reflecting points that will yield the indicated distance differences.
I also note a response rise from 500-1.3Khz of s couple of dB/ I would EQ this out as it will cause a general muddiness or thickness in the sound
of voices.

Given that an absorber big enough to be effective at these low frequencies is NOT likely to find comfortable home in your living room it may not be th best solution. But at least you know where the problem is.
 

ddude003

AV Addict
Joined
Aug 13, 2017
Messages
1,403
Location
Somewhere Northeast of Kansas City Missouri
More  
Preamp, Processor or Receiver
PrimaLuna Dialogue Premium TubePre (2 channel+sub)
Main Amp
McIntosh MC152 SS Amp (2 channel)
Additional Amp
Yamaha RX-A850 Pro (the other 5 channels lol)
Computer Audio
MacBook Pro, Custom i7 7700k De-lid 2xAsus1080ti GFX Audirvana Studio, Hang Loose Convolver, Pulsar Massive & 8200, LiquidSonics, SoX
DAC
Chord Electronics Ltd. Qutest
Universal / Blu-ray / CD Player
Sony UBP-X700 /M Ultra HD 4K HDR & PS5
Front Speakers
Martin Logan ElectroMotion ESL
Center Channel Speaker
Martin Logan Motion C2
Surround Speakers
Martin Logan Motion 4
Surround Back Speakers
Martin Logan Motion 4 (yes, another set of these)
Subwoofers
Martin Logan Dynamo 700
Other Speakers or Equipment
Cifte 12AU7 NOS & Genalex Gold Lion Tubes in Pre
Video Display Device
Samsung The Premiere LSP7T UST Laser Projector
Screen
Elite Screens Aeon CLR3 0.8 Gain 103-inch
Remote Control
PrimaLuna, Lumin iApp, Samsung & Yamaha
Streaming Equipment
Netgear Nighthawk S8000 Streaming Switch, Lumin U1 Mini Streamer Transport
Streaming Subscriptions
QoBuz Studio Premier, Amazon Prime & Netflix
Other Equipment
ThrowRug, SaddleBlankets, WideBand & Bass Traps...
I wonder how a convolutional approach to equalization might work in this case... I have been messing around with Roon and its convolution system and using Audirvana with Liquidsonics convolution system with some interesting results with my system in my room...
 

linearphase

Member
Joined
Jan 17, 2019
Messages
59
I wonder how a convolutional approach to equalization might work in this case... I have been messing around with Roon and its convolution system and using Audirvana with Liquidsonics convolution system with some interesting results with my system in my room...
I have not used it but have helped some who have.
It works well. The biggest issue with EQ involves know what not to touch. This is even more important than EQING everything.
 

linearphase

Member
Joined
Jan 17, 2019
Messages
59
I have not used it but have helped some who have.
It works well. The biggest issue with EQ involves know what not to touch. This is even more important than EQING everything.
With simple convolution of the impulse response I have no control over what does not get EQd.
 

ddude003

AV Addict
Joined
Aug 13, 2017
Messages
1,403
Location
Somewhere Northeast of Kansas City Missouri
More  
Preamp, Processor or Receiver
PrimaLuna Dialogue Premium TubePre (2 channel+sub)
Main Amp
McIntosh MC152 SS Amp (2 channel)
Additional Amp
Yamaha RX-A850 Pro (the other 5 channels lol)
Computer Audio
MacBook Pro, Custom i7 7700k De-lid 2xAsus1080ti GFX Audirvana Studio, Hang Loose Convolver, Pulsar Massive & 8200, LiquidSonics, SoX
DAC
Chord Electronics Ltd. Qutest
Universal / Blu-ray / CD Player
Sony UBP-X700 /M Ultra HD 4K HDR & PS5
Front Speakers
Martin Logan ElectroMotion ESL
Center Channel Speaker
Martin Logan Motion C2
Surround Speakers
Martin Logan Motion 4
Surround Back Speakers
Martin Logan Motion 4 (yes, another set of these)
Subwoofers
Martin Logan Dynamo 700
Other Speakers or Equipment
Cifte 12AU7 NOS & Genalex Gold Lion Tubes in Pre
Video Display Device
Samsung The Premiere LSP7T UST Laser Projector
Screen
Elite Screens Aeon CLR3 0.8 Gain 103-inch
Remote Control
PrimaLuna, Lumin iApp, Samsung & Yamaha
Streaming Equipment
Netgear Nighthawk S8000 Streaming Switch, Lumin U1 Mini Streamer Transport
Streaming Subscriptions
QoBuz Studio Premier, Amazon Prime & Netflix
Other Equipment
ThrowRug, SaddleBlankets, WideBand & Bass Traps...
You have an interesting problem with those panels pointed crossing the other parts... Highs going one way with mid and mid lows going another... Could be glorious and then again could be chaos... Might be a fun science experiment varying how your panels are positioned and trying a few different EQ approaches... I am not very informed about how the panels are laid out or positioned... Are those highs in the very center crossing or beaming out? Might be fun running a ray trace simulation to see where all the sound is going...
 

linearphase

Member
Joined
Jan 17, 2019
Messages
59
He said those panel will have very odd polar response. They are also very hard to measure correctly. That having been said they have excellent transient resonse.
 

DanDan

Moderator
Staff member
Joined
Mar 10, 2018
Messages
743
Narrow boosts seems a dangerous experiment. I would recommend going as far as you can with Subtractive Eq.
Do remember that narrow nulls are not necessarily audible. Also they may not be present at both ears, in which case the brain will sum and average.
 

Greg Dunn

Member
Thread Starter
Joined
Oct 4, 2017
Messages
107
OK, I've had some time to experiment with the system and report back.

Thanks for the "acoustic string" link - I'm going to make one and put it to use later this week.

Magnepan recommends trying multiple positions for the panels to optimize the imaging. Using the attached diagram as the starting point, I did the painstaking job of moving them around to make the presentation as listenable as possible. What I'm using is not so much different, except in minor details of angle. The speakers are constructed with the tweeter on one end panel, the mids on the other end, and the low/mid in the middle to decrease the frequency at which LF front-rear cancellation occurs. The mid and low/mid are laid out slightly differently, making a sort of acoustic crossover occur due to panel resonances and routing of the wire "voice coil". The resonances are damped via the tie-downs mentioned earlier. Putting the tweeters on the "inside" is preferred, but the speakers need to be several feet apart in order for this to give sufficient stereo separation. With my setup, slightly modified from the diagram, imaging is excellent and reasonably stable with head position. It's the best I can achieve with the limitations of my environment, I think.

I already have introduced a delay (calculated by relative distance as well as measured via REW) into the sub to give good impulse alignment; I think this is vital to integrating the two different types of drivers and presenting a homogeneous sound quality in the bass region. With that taken care of, the transient response of things like kick drums and plucked bass seems more realistic. I actually tried several delay values while listening, and the one I picked as "sounding" better lined up surprisingly well with the measurements. These speakers do have excellent transient response, audibly as well as in the measurements - it's the main reason I bought them originally. Fixing the sub time delay and flattening the response lets them really shine IMHO.

The first cut at improved EQ has been done; I didn't go overboard with the gain, but by adding narrow boosts in the 130-140 and 270 range, I managed to reduce the steep dips noticeably. The same with the 1.4 kHz trough, that didn't yield to broader EQ filters. I chose to adjust the mid amp (70-1000 Hz) slightly down and edit the PEQs rather than add more filters in that range; that also helped minimize the amount of boost needed for the lower frequency dips and improved the transition to the sub (which could then be treated with cuts rather than boosts for the most part).

What I have now is definitely better than it was to my ears; even the RT60 and decay waterfall plots look a bit nicer. I do plan some listening tests where I reduce the gain of the LF boost filters to see if they are really audible or just psychological, because I know that the ear/brain will average these out. My initial purpose was to see if they could be ameliorated, not to eliminate them blindly and assume it was an improvement. I won't let the improved measurement dictate my ultimate choice.
 

Attachments

  • 1-C positioning.jpg
    1-C positioning.jpg
    241.6 KB · Views: 31

linearphase

Member
Joined
Jan 17, 2019
Messages
59
Depends on the cause and depth but yes in general. These are also less audible unless on a musical note or harmonic.
OK, I've had some time to experiment with the system and report back.

Thanks for the "acoustic string" link - I'm going to make one and put it to use later this week.

Magnepan recommends trying multiple positions for the panels to optimize the imaging. Using the attached diagram as the starting point, I did the painstaking job of moving them around to make the presentation as listenable as possible. What I'm using is not so much different, except in minor details of angle. The speakers are constructed with the tweeter on one end panel, the mids on the other end, and the low/mid in the middle to decrease the frequency at which LF front-rear cancellation occurs. The mid and low/mid are laid out slightly differently, making a sort of acoustic crossover occur due to panel resonances and routing of the wire "voice coil". The resonances are damped via the tie-downs mentioned earlier. Putting the tweeters on the "inside" is preferred, but the speakers need to be several feet apart in order for this to give sufficient stereo separation. With my setup, slightly modified from the diagram, imaging is excellent and reasonably stable with head position. It's the best I can achieve with the limitations of my environment, I think.

I already have introduced a delay (calculated by relative distance as well as measured via REW) into the sub to give good impulse alignment; I think this is vital to integrating the two different types of drivers and presenting a homogeneous sound quality in the bass region. With that taken care of, the transient response of things like kick drums and plucked bass seems more realistic. I actually tried several delay values while listening, and the one I picked as "sounding" better lined up surprisingly well with the measurements. These speakers do have excellent transient response, audibly as well as in the measurements - it's the main reason I bought them originally. Fixing the sub time delay and flattening the response lets them really shine IMHO.

The first cut at improved EQ has been done; I didn't go overboard with the gain, but by adding narrow boosts in the 130-140 and 270 range, I managed to reduce the steep dips noticeably. The same with the 1.4 kHz trough, that didn't yield to broader EQ filters. I chose to adjust the mid amp (70-1000 Hz) slightly down and edit the PEQs rather than add more filters in that range; that also helped minimize the amount of boost needed for the lower frequency dips and improved the transition to the sub (which could then be treated with cuts rather than boosts for the most part).

What I have now is definitely better than it was to my ears; even the RT60 and decay waterfall plots look a bit nicer. I do plan some listening tests where I reduce the gain of the LF boost filters to see if they are really audible or just psychological, because I know that the ear/brain will average these out. My initial purpose was to see if they could be ameliorated, not to eliminate them blindly and assume it was an improvement. I won't let the improved measurement dictate my ultimate choice.
Good work Greg, sounds like you down most everything about as good as can be done. I have found that small boosts, a couple of dB can help dips as long as they are not more that 6 dB or so. I never try to correct them completely. I do this only when room treatment is not practical.
That having been said I have had only one guy that was willing and able to do everything suggested. He spent at least 6 months with adjustments and treating the room. The final results are extremely good both from a measurement and a listening standpoint.
 

DanDan

Moderator
Staff member
Joined
Mar 10, 2018
Messages
743
Curious, are your speakers, if that is the right term running full range or does the entire signal cross over to the subs?
 
Top Bottom