Jump to content

The Bass System Setup, EQ/Correction Thread


maxmercy

Recommended Posts

  • Replies 76
  • Created
  • Last Reply
As I've said, I don't want to argue whether other methods are better, especially methods that include specifics that can't be revealed for whatever reason. I just showed my method, reasoning and results.

 

I would argue this is relatively uncommon set of requirements & would agree that EQ, via an offboard device that adds a rolloff, is clearly unnecessary for you. 

 

...uhg.  What is this AVS or something?  :lol:

 

Link to comment
Share on other sites

Let's see, small room, treated very well, throw 18's all over and big power. I have been performing the opposite of common wisdom, I EQ the top end of my speakers and don't EQ my subs. I let my room and multiple subs do the smoothing. The goal for me is flat response but I am so used to a boosted low end I always turn the gain up on my bass amp. I make every movie a 5 star if possible and run the 5 star movies not as hot. My biggest problem is shaky screen effect.

Link to comment
Share on other sites

I think SME made the point I was going to make (about how adding DSP can be a simpler and/or more practical and/or more cost effective solution then revisiting the speaker itself or the entire room setup). I also don't understand the resistance to DSP in the audio chain, a speaker is a mechanical device after all so why not use DSP to improve it?

 

In practice though the key issue, with respect to the applicability of DSP to the signal chain, seems to be that modern prepros are just so lamentably weak. The basic design of a prepro hasn't changed 10+ years while the amount of computing power in there has barely changed in that time too. This means add on boxes of varying quality OR accepting the limitations of a PC only approach OR spending a fortune on a trinnov. 

 

I thought this site had some interesting possibilities - http://www.servobass.com/Blog.html- though nothing seems to have come to market yet. The DIY DSP programmable servo controller (for a sub) looked especially promising but I think they pulled that.

 

Maybe I just don't embrace FIR DSP as much as others.  I think that a properly engineered speaker should minimize crossover deleterious effects, and not require FIR DSP just to have acceptable fidelity.  FIR DSP when applied to the new 'immersive sound' formats and the multitude of channels will not be inexpensive either.  

 

I agree pre/pros could have more power, but they are generally marketed to folks with less than insightful knowledge than those that post here, and manufacturers will not put any more processing power into a box than they absolutely have to as long as Joe-Bag-O'-Donuts is willing to pay.

 

Servo Control is something I consider a little 'boutique'.  Enough headroom built into a system reduces distortion just as well as servo control, sometimes at lesser cost.  If space were a significant constraint, servo control could be very useful, though.  

 

JSS

Link to comment
Share on other sites

There's more to it than roll off. There's an unnecessary analog =>digital =>analog conversion, the assumption that the DACs don't matter, increase in noise floor, increase in distortion, loss of headroom, the difference in input signal and...roll off. See the thread about clipping and the resultant huge increase in THD, for one example. Anytime you add an additional component, all of these things must be considered and usually are not.

all we need to do now is start arguing about DACs and we really will be back to avs  :lol:

Link to comment
Share on other sites

Maybe I just don't embrace FIR DSP as much as others.  I think that a properly engineered speaker should minimize crossover deleterious effects, and not require FIR DSP just to have acceptable fidelity.  FIR DSP when applied to the new 'immersive sound' formats and the multitude of channels will not be inexpensive either.  

 

it's certainly true that my current speakers, while still sounding nice, are not the best behaved once you measure them and hence may well benefit from correction more than some other speakers. I'm in the process of building/designing some replacements myself though which, if I do a good job, will be much better behaved. It will be interesting to see whether I still see a substantial benefit to PC based DSP once I have those in place.

Link to comment
Share on other sites

Just wanted to show you possible effects of different delay settings for the subwoofer system.

 

Unfortunately my example is kind of a failure, because due to how the subwoofer system is set up, there is not a very significant change in peak amplitude of the first initial impulse.

 

Crossover frequency is 120Hz, and the only difference is the delay of the mains - 7.30m in the first, and 3.80m on the other.

Both sum up to approximately equal frequency response, but the time domain behavior is different.

 

Step response:

post-181-0-91999700-1428453570_thumb.png

 

Recorded playback, excerpt from Avratz (Infected Mushrooms), 1. Original, 2. 7.30m, 3. 3.80m:

post-181-0-86830000-1428453602_thumb.png

 

 

You can see that the initial attack on the drum is slightly delayed and smeared in time on the 3.80m.

When running auto-setup like audyssey, this 3.80m is the best you can hope for.

 

This difference is audible, and not only at loud volumes.

The character of the bass is different between the two delay settings, and my impression is that the 7.30m setting is better, the attack is experienced as slightly more precise, more like a real drum, more powerful.

Link to comment
Share on other sites

I think SME made the point I was going to make (about how adding DSP can be a simpler and/or more practical and/or more cost effective solution then revisiting the speaker itself or the entire room setup). I also don't understand the resistance to DSP in the audio chain, a speaker is a mechanical device after all so why not use DSP to improve it?

 

In practice though the key issue, with respect to the applicability of DSP to the signal chain, seems to be that modern prepros are just so lamentably weak. The basic design of a prepro hasn't changed 10+ years while the amount of computing power in there has barely changed in that time too. This means add on boxes of varying quality OR accepting the limitations of a PC only approach OR spending a fortune on a trinnov. 

 

I thought this site had some interesting possibilities - http://www.servobass.com/Blog.html- though nothing seems to have come to market yet. The DIY DSP programmable servo controller (for a sub) looked especially promising but I think they pulled that.

 

This pretty much nails it.  I even find the MiniDSP equipment to be frustratingly limiting.  I'm a programmer who would not be the slightest bit intimidated to program the DSP directly if I needed to.  Ideally, a pre/processor would be an open platform like a PC or phone for which one could download custom 3rd party apps.

 

There's also the movie industry's paranoia about letting us see the bits flowing between the Blu-ray discs and our displays/DACs.  The HDMI specs are completely proprietary, only available officially (and legally) by membership in the group, which carries a steep fee.  On top of that, vendors are specifically prohibited from creating an HDMI device that makes it possible for anyone under any circumstances to copy content illegally.   Between the poor economics (not enough power users) and the legal restrictions, I believe bringing some sort of open platform pre/processor to the market would be extremely difficult.

 

All the same, if I get to the point where I feel the presence of too many ADC/DACs in my system is a major limitation to its audio quality, then I will likely be close to that point of diminishing returns where it's just stupid to keep spending more money.  Implementation bugs not-withstanding, that is.  While others here are discovering irreconcilable gain structure errors in the bass management implementations on some popular hardware, I believe I have discovered what is likely a firmware bug on my Denon 3313CI AVR that's introducing pre-ringing in my left channel response, even with Audyssey off.  I wouldn't be surprised if it affected several other Denon and Marantz models as well since they likely use the same software platform.  Anyone have a Denon or Marantz who wants to go looking for it?  A benefit to having custom FIR correction is: I can most likely cancel it out with a filter until D&M figures out their stuff's broken.

Link to comment
Share on other sites

When running auto-setup like audyssey, this 3.80m is the best you can hope for.

 

 

do you know why it has got the alignment so wrong? That adjustment is equivalent to 450 degrees at the xo freq isn't it? It is a much much bigger error than I have personally seen from an auto setup routine. Obviously rooms behave differently but it still seems like a big fail. Was it audyssey in that case?

 

It would be interesting to see the mdat of the underlying responses if you have it.

Link to comment
Share on other sites

If you're not sure what you're supposed to be seeing in that pic, in reference to your question, then there's not much sense in my continuing with the discussion.

 

Smoothing on a close mic measurement of a sealed sub is irrelevant.

 

The EQ "fix" grossly distorts the input signal, causing problems you apparently choose to ignore. It makes the FR flatter but so does the method Max and myself have mentioned.

 

Butchering the input signal with EQ adds no headroom. Headroom resides at the weakest point, which for any sub is <20 Hz. What happens to your response and headroom if the hard limiters squash the output into a completely different FR every time there is <20 Hz content?

 

Alter the native response of the sub, move the sub, move the seat, treat the room, change the cross point, adjust relative phase... voila!, no EQ needed. Or, skip those steps and just insert and apply ham-fisted EQ and pretend there is no price to pay... that it instead fixes everything. Hey, that's cool with me because what choice do I have in the matter? The majority use quick EQ, tout it's sonic improvement benefits and call it a day. But, I remain unconvinced by that sort of evidence.

Smoothing a close sub measurement is relevant, even if the two only differ by a few dB, particularly when we're using EQ to correct features on the order of a few dB in magnitude.  In any case, I still don't see the point about the input signal being altered.  We don't listen the input signal, and the room corrupts the signal substantially before it reaches the listening position.  As long as the rest of the signal chain is up to the task of reproducing that input signal output by the EQ, the EQ can potentially reduce the harm done by the room.  I'm not saying it can fix anything or everything, but it can do a lot more than you are giving it credit for. 

 

To the extent that your suggested method works, great!  That you achieved a +/- 4.5 dB response without needing EQ is remarkable and I believe it may be the exception more than the norm.  Every room and set of circumstances is different, and in many cases, placements and distances (assuming you are even setting those for the subs independently) aren't enough to get that kind of performance.

 

Then there's the question of whether +/-4.5 dB is really good enough.  It's a great result for a typical listening room, but I believe it may still leave a lot to be desired.  Even that difference may be relevant as far as the direct air-to-body tactile experience is concerned.  When I turn on a new system configuration, I'll typically listen to several familiar music albums and watch several familiar movie scenes before rendering a subjective judgment about it.  It usually takes at least a week of my spare time before I've decided whether I'm happy or whether I need to tweak the target curve here or there.  From what it sounds like, you only did a short audition of the EQed sound.

 

A few dB matters a lot more for bass than other frequencies.  The ear has both poor dynamic range and a low density of critical bands in the bass.  This means masking is also a bigger issue which translates directly into issues hearing a soundtrack consistent with the artist's intent.  On the opposite end of the spectrum are those who are thrilled to hook an integrated sub system to their AVR, turn the gain up to 75% of some such, and hear their windows rattle.  Maybe you are satisfied with your system the way it is.  There's nothing really wrong with that, but I personally like to try push the boundaries and see what's possible.

 

The drawbacks of EQ you mention are certainly relevant, but provided care is taken, most can either be worked around or are relatively unimportant.  ADC/DAC conversions with competent modern implementations do very little harm to the sound, and almost all of it occurs near the Nyquist frequency, i.e., the top of the audio range or above.  The low-end roll-off can be a real issue but if it's not too severe, then there is an effective EQ work-around.  And yes, here too you may run into trouble if your IIR implementation does not have enough numerical precision to work in the single digits Hz.  Yes, you may have to give up a bit of signal to noise ratio to accomplish these things.  At the same time, SNR on equipment is so low these days that there's usually plenty of room to work with.

 

As I see it, the only drawback to using EQ for bass response is the sophistication it demands of the practitioner.  Even auto-EQ is too much for a lot of people.  IIRC, the new Atmos-enabled Onkyo's where they dumped Audyssey advertise "New!  Only one measurement required."  So to be fair, I'm not saying everyone should go out and buy a MiniDSP or whatever and start dialing in settings.  There's a lot of stuff that one can mess up, and one needs to keep in mind the issues you mention, even if they can be adequately addressed under most circumstances.  I'm also aware that the readers of this forum are quite a bit more advanced than average.  I expect that readers won't just look at what I write here and assume that EQ is a panacea and then go screw up their response or worse still, their equipment.

Link to comment
Share on other sites

I just conducted a simple experiment to clarify my position on my setup method.

 

Here is a FR at the seats with an obvious problem. I used a simple delay setting adjustment to rectify the problem and have shown the signal and the result. I then injected an outboard digital PEQ and used 2 filters to rectify the problem and have shown the new signal and result, and finally, I've compared the resulting input signal and results measured at the seats to illustrate why I do not use post smoothing EQ.

 

So I take it the problem was that the sub was not in phase with whatever mains channel you measured it with?  You solved it by adjusting the sub delay.  Figuratively speaking, you've demonstrated why a hammer isn't the best tool for turning a screw.  Nowhere did I say that EQ is a replacement for delay settings, and in fact, I advocate using a separate delay on each sub if possible.

 

As it turns out, the whole reason I bought my MiniDSP 2x4 in the first place was to do separate sub delays.  Like you, I was skeptical that the EQ features would be of any practical use.  Now I'm here arguing that it can make a big positive difference.

 

Just because I can, I'll point out that FIR filters (as opposed to typical minimum phase IIR-based EQs) are very powerful.  Technically speaking, if you have enough taps available, you can implement an arbitrary delay as part of the FIR filter.  It's kind of a waste of computing resources to run such a filter, but the result is exactly the same as if you used any other digital delay feature.

Link to comment
Share on other sites

Just wanted to show you possible effects of different delay settings for the subwoofer system.

 

Unfortunately my example is kind of a failure, because due to how the subwoofer system is set up, there is not a very significant change in peak amplitude of the first initial impulse.

 

Crossover frequency is 120Hz, and the only difference is the delay of the mains - 7.30m in the first, and 3.80m on the other.

Both sum up to approximately equal frequency response, but the time domain behavior is different.

 

Step response:

attachicon.gifstep_730_380.png

 

Recorded playback, excerpt from Avratz (Infected Mushrooms), 1. Original, 2. 7.30m, 3. 3.80m:

attachicon.gifarz_1.png

 

 

You can see that the initial attack on the drum is slightly delayed and smeared in time on the 3.80m.

When running auto-setup like audyssey, this 3.80m is the best you can hope for.

 

This difference is audible, and not only at loud volumes.

The character of the bass is different between the two delay settings, and my impression is that the 7.30m setting is better, the attack is experienced as slightly more precise, more like a real drum, more powerful.

I'm not sure what you mean when you say "when running auto-setup like Audyssey, this 3.80m is the best you can hope for"), but I am very well aware of the effect that sub delay can have on bass performance.  I have also routinely witnessed Audyssey choosing a poorly performing distance setting.  I recognized this even before I had my measurement mic and could see just how bad it was.  Initially I optimized my delay by measuring test tones in the crossover region with an SPL meter and by listening to specific passages of music, usually kick drum.

 

These days, I take time-correct measurements at each of my listening positions using the trick I described in a previous post.  Then, I simulate the responses for each channel *and* each listening position using these measurements.  I also simulate different crossovers for different speakers.  I choose the distance and crossover that works best at all the seats with heavier weight given to the center/MLP.  Now that I also have the ability to EQ my mains channels, I use EQ to improve the crossover blend between each individual speaker and the sub.  I am hopeful I will soon be able to get even better results with the help mixed-phase filters.  I won't know till I give it a try.

Link to comment
Share on other sites

Servo Control is something I consider a little 'boutique'.  Enough headroom built into a system reduces distortion just as well as servo control, sometimes at lesser cost.  If space were a significant constraint, servo control could be very useful, though. 

I believe servo control may be very beneficial, to the extent that distortion at the lowest frequencies is audible.  (I already had a long discussion about that at these forums.)  It appears to me that at the lowest frequencies, distortion often lingers at a moderately high level even at reduced overall playback levels.  I haven't built a woofer, but I imagine one of the most difficult challenges is designing a suspension that is as linear as possible.  As I understand it, the "acoustic suspension" (basically, putting a woofer into a relatively small sealed cabinet) was successful because the air-spring provided much more linearity than was possible with suspension materials of the era.  Of course, an air-spring alone contributes distortion.

 

Another thing about a servo amp is that it could lead to a significant reduction in cost for woofer design.  To the extent that an otherwise sloppy woofer can be linearized via the control circuit, it may be possible to get more performance for the money.

Link to comment
Share on other sites

I'm not sure what you mean when you say "when running auto-setup like Audyssey, this 3.80m is the best you can hope for"),

The only reason I can think of for this is that modal ringing causes the peak finding algo in the auto EQ setup routine to misread where the actual impulse is and thus set an completely incorrect delay. Hard to tell without the underlying data though.
Link to comment
Share on other sites

The only reason I can think of for this is that modal ringing causes the peak finding algo in the auto EQ setup routine to misread where the actual impulse is and thus set an completely incorrect delay. Hard to tell without the underlying data though.

I really don't know.  I know that with regular speakers, in cases where I've had a strong back wall reflection that created a peak slightly higher than the initial arrival, Audyssey time-aligned to the later peak, even though this is psycho-acoustically undesirable.  The energy of a room mode actually builds within the impulse response over time, so I don't think that's it.  On the other hand, if two early reflections arrive at roughly the same time, the peak can definitely be shifted quite a bit relative to the initial arrival.

 

Curiously, the last time I ran Audyssey, it selected a sub distance that aligned my mains peaks with the very start of the impulse response in the bass.  For a change, this actually gave excellent performance overall and not just at the MLP.  When I set my sub distance, I definitely don't bother to try to find the peak or line it up with the mains peaks.  Instead, I look at the frequency response in the vicinity of the crossover and choose the distance that gives the best response, not just at the MLP but at the other positions too.  In many rooms, by the time you get down into the bass frequencies, the room effects are so strong effect that merely aligning the peaks often results in an inferior result.  Indeed, unlike the case with the mains where it is imperative to make sure that the first major arrivals occur at the same time for each speaker, what sub distance is "best" is not necessarily well defined.   Rather I treat the distance as just another knob to turn to get the best response at the crossover.  Typically, a compromise is necessary where one favors the response of one channel (like the center) over others.  This means one might prefer different sub distances for movies (center) versus 2 channel music (left and right).  Often the surround responses just suck.  Now that I have EQ for my mains, I'm able to get a much better crossover response for the non-center channels, and this is something Audyssey definitely can't do well because it optimizes each channel separately without considering the combined response.  Heck, last time I ran it, it rolled off my front right steeply at ~180 Hz.  With my use of EQ, I was able to easily obtain a much better result at MLP.

Link to comment
Share on other sites

Smoothing a close sub measurement is relevant, even if the two only differ by a few dB, particularly when we're using EQ to correct features on the order of a few dB in magnitude.  In any case, I still don't see the point about the input signal being altered.  We don't listen the input signal, and the room corrupts the signal substantially before it reaches the listening position.  As long as the rest of the signal chain is up to the task of reproducing that input signal output by the EQ, the EQ can potentially reduce the harm done by the room.  I'm not saying it can fix anything or everything, but it can do a lot more than you are giving it credit for. 

 

To the extent that your suggested method works, great!  That you achieved a +/- 4.5 dB response without needing EQ is remarkable and I believe it may be the exception more than the norm.  Every room and set of circumstances is different, and in many cases, placements and distances (assuming you are even setting those for the subs independently) aren't enough to get that kind of performance.

 

Then there's the question of whether +/-4.5 dB is really good enough.  It's a great result for a typical listening room, but I believe it may still leave a lot to be desired.  Even that difference may be relevant as far as the direct air-to-body tactile experience is concerned.  When I turn on a new system configuration, I'll typically listen to several familiar music albums and watch several familiar movie scenes before rendering a subjective judgment about it.  It usually takes at least a week of my spare time before I've decided whether I'm happy or whether I need to tweak the target curve here or there.  From what it sounds like, you only did a short audition of the EQed sound.

 

A few dB matters a lot more for bass than other frequencies.  The ear has both poor dynamic range and a low density of critical bands in the bass.  This means masking is also a bigger issue which translates directly into issues hearing a soundtrack consistent with the artist's intent.  On the opposite end of the spectrum are those who are thrilled to hook an integrated sub system to their AVR, turn the gain up to 75% of some such, and hear their windows rattle.  Maybe you are satisfied with your system the way it is.  There's nothing really wrong with that, but I personally like to try push the boundaries and see what's possible.

 

The drawbacks of EQ you mention are certainly relevant, but provided care is taken, most can either be worked around or are relatively unimportant.  ADC/DAC conversions with competent modern implementations do very little harm to the sound, and almost all of it occurs near the Nyquist frequency, i.e., the top of the audio range or above.  The low-end roll-off can be a real issue but if it's not too severe, then there is an effective EQ work-around.  And yes, here too you may run into trouble if your IIR implementation does not have enough numerical precision to work in the single digits Hz.  Yes, you may have to give up a bit of signal to noise ratio to accomplish these things.  At the same time, SNR on equipment is so low these days that there's usually plenty of room to work with.

 

As I see it, the only drawback to using EQ for bass response is the sophistication it demands of the practitioner.  Even auto-EQ is too much for a lot of people.  IIRC, the new Atmos-enabled Onkyo's where they dumped Audyssey advertise "New!  Only one measurement required."  So to be fair, I'm not saying everyone should go out and buy a MiniDSP or whatever and start dialing in settings.  There's a lot of stuff that one can mess up, and one needs to keep in mind the issues you mention, even if they can be adequately addressed under most circumstances.  I'm also aware that the readers of this forum are quite a bit more advanced than average.  I expect that readers won't just look at what I write here and assume that EQ is a panacea and then go screw up their response or worse still, their equipment.

 

 

Smoothing of a close mic measurement of a sealed sub is irrelevant, as seen in the smoothed/no-smoothing comparison I posted 5 years ago for a poster with the same errant comments to the contrary:

 

9ac72ed9c1e2b373e9d1d3aa27ccd845.png

 

The rest of your posts are a bit confusing. You say you prefer to push the envelope because a +/- 4dB response is not good enough, but you posted your response and it looks like +/-4dB to me:

 

5dfbf1532e927f772f4717441ce9089b.png

 

More importantly, lopping off nearly 3 octaves with a 4th order roll off hardly seems to be pushing any envelope regarding reproduction of transients, like the kick drum spectrograph overlain on the graph, or most of the movies we measure and discuss in these forums.

 

If your preference for ending up with a +/- 4dB response across 2 octaves of sub bandwidth is to just insert EQ filters and the result makes you happy, that's OK with me. My responses were triggered by your admonition that I rethink whether to use EQ, that gross distortion of the input signal somehow isn't distortion of the input signal and causes no audible effect and that that concept is nonsense.

 

My concept of how to achieve accurate low end reproduction in a room comes from 15 years of experimentation, discussion, measurements, average persons listening sessions and reading most every theory on the subject. I believe it remains a learning experience and so am open to ideas that get good results. So, post what ya got and it'll be appreciated as opposed to calling nonsense and such on methods you don't endorse.

Link to comment
Share on other sites

I've been tussling with an anomaly in my setup recently & I think I have been guided towards the root cause today by another acourate user. It is also directly on topic as it was my sub that was giving me grief.

 

My basic problem was that "something sounded off" so that I felt the integration between sub and mains was suboptimal. This hunch was reinforced by the fact that various different methods for determining the correct alignment of the sub to the mains were giving quite different answers. For example

 

- physical distance said there was either in alignment or 1-2 samples ahead

- a delayed XO method (create a crossover that delays the main by 2000 samples then align the main speaker impulse to the theoretical HF XO impulse & the sub impulse to the theoretical LF XO impulse, if the actual impulses are 2000 samples apart then the 2 drivers are in alignment), this suggested the sub was 60-70 samples (at 48kHz) ahead of the mains 

- the narrow bandwidth sweep method said the sub looked to be in opposing polarity to the mains (or ~240 samples ahead, take your pick)

 

Testing actual polarity with a 9V battery showed they were all wired correctly so it has been a bit of a mystery for a little while now.

 

The problem appears to have been resolved by investigating what is going on in my main, passive, speaker. The method was basically

 

- measure the main speaker full range ( A )

- measure the sw + mains full range ( B )

- convolve A with the lowpass of the XO (1)

- convolve A with the high pass of the XO (2)

- convolve C with the low pass of the XO (3)

 

This graph shows 1, 2 and 3 after alignment;

 

1 = blue

2 = teal

3 = brown

 

1 and 3 are now aligned but both are ~190 samples behind the tweeter. 

 

post-1440-0-78326800-1428522354_thumb.png

 

In step response terms, this misalignment is quite hard to pick up, or at least I didn't pick up on it. The arrow is at about the 190 samples point and shows that the step diverges from the aligned case due to the misalignment of the subwoofer.

 

post-1440-0-59288000-1428522474_thumb.png

 

Moral of the story is that manual alignment against a passive speaker might have some unforeseen pitfalls unless you know exactly how that speaker behaves.

 

The other moral is that the difference between the 2 final corrections is quite small, acourate has cleaned this up v nicely in either case. I think this is a nice demonstration of the power of FIR filtering as well as a possible argument in favour of JSS's point re "if you fix the root cause, do you need that filtering?".

Link to comment
Share on other sites

do you know why it has got the alignment so wrong? That adjustment is equivalent to 450 degrees at the xo freq isn't it? It is a much much bigger error than I have personally seen from an auto setup routine. Obviously rooms behave differently but it still seems like a big fail. Was it audyssey in that case?

 

It would be interesting to see the mdat of the underlying responses if you have it.

 

It is not necessarily an error.

The textbook-implementation of this is to see it as a 4. order crossover, and such a crossover does not have linear phase, and there will be group delay.

High up in frequency that may not matter too much, as long as both channels l/r are the same.

In this upper-bass/lower-mid region, which is crucial for the kick/transeint attack, not so.

 

There will be several delay settings that will sum up to approximately correct freq response, separated by one wavelength.

The one that sums exactly correct, is something like the 3.80m, provided the main speakers and the subwoofer system has responses that exactly matches 4. order rolloffs at the fc, and flat outside - we all know that is rarely the case.

Or even smaller, I just used this one as it sums up correct.

 

What you can do is choose to ignore that theoretical correct delay, and adjust for best possible impulse response.

Then you end up with a total system that has flat gd and flat phase.

 

Does it sound better?

Up to you to judge yourself.

Link to comment
Share on other sites

Smoothing of a close mic measurement of a sealed sub is irrelevant, as seen in the smoothed/no-smoothing comparison I posted 5 years ago for a poster with the same errant comments to the contrary:

Yes, but a single measurement does not convince me that this is always the case.  Perhaps if you said that you've compared the two measurements on a variety of different sub systems in a variety of different rooms over your two decades of experience, then I might be more convinced, but I'll likely still avoid using smoothing and retain as much resolution as I can, using smoothing only when it's necessary to make the data presentable.  My experience suggests 1/24 octave smoothing is a good compromise, but for bass I like to just leave things un-smoothed and run with higher resolution.  I generally agree with this article:  http://www.data-bass.com/data?page=content&id=81

 

Out of curiosity, is your measurement room response measurement done with REW?  And if so, what FFT size are you using?

 

The rest of your posts are a bit confusing. You say you prefer to push the envelope because a +/- 4dB response is not good enough, but you posted your response and it looks like +/-4dB to me:

 

Let me clarify then.  A +/-4.5 dB unsmoothed response is impressive for a residential sized space.  That's good.  It's so good, it's probably better than most of what's out there.  At the same time, it's possible for roughly adjacent sustained tones to differ in level by 9 dB, which may be subjectively perceived as a factor of 3 or more difference.  That's a big difference.  The more narrow the peaks and dips are, the more prominent ringing may be, which can harm bass clarity considerably.

 

Yeah, I'm running at +/- 4 dB right now too.  Before I added room treatments and a new wider sofa that curves forward, my sub response was a bit tighter.  On the plus side, my upper bass is much better than it was before, and my crossovers are also better with all my mains speakers than they were.  These improvements were made possible by the room treatments and by ditching Audyssey for my custom EQ.  The room treatments definitely did most of the work, especially with regard to performance outside the MLP and the reduced need to compromise overall.

 

In my sub response, the biggest aberrations are at around 57-60 Hz and 90 Hz.  The issue at 90 Hz is mostly one of headroom, and I expect to eliminate it with more bass trapping that I plan to install at some later date.  The issue at 57-60 Hz is a bit uglier.  I have a lot of headroom here because it's near my crossover, but the room is definitely fighting me here.  Bass trapping is not very effective this low, but I might be able to wrangle it using EQ involving all 4 subs.  The 57-60 Hz issue manifests in another way because all the extra energy I end up injecting there concentrates in another part of the room where it's more like a 15-20 dB issue.

 

More importantly, lopping off nearly 3 octaves with a 4th order roll off hardly seems to be pushing any envelope regarding reproduction of transients, like the kick drum spectrograph overlain on the graph, or most of the movies we measure and discuss in these forums.

 

If your preference for ending up with a +/- 4dB response across 2 octaves of sub bandwidth is to just insert EQ filters and the result makes you happy, that's OK with me. My responses were triggered by your admonition that I rethink whether to use EQ, that gross distortion of the input signal somehow isn't distortion of the input signal and causes no audible effect and that that concept is nonsense.

I have to work within the constraints of my current room, and getting those "last 3 octaves" will be difficult to realize in here because of the openness and overall size.  I'm more likely to save my time and money until I can construct a dedicated room.  As far as the kick drum transient is concerned, I can't make any kind of quantitative estimates without knowledge of the scale represented by the color gradations.  I looked around at some of your other posts hoping for hints, but I haven't worked it out yet.  it does look like the vast majority of the energy is above the bottom 3 octaves.  This is a relevant point.  The bottom end can add more weight to a transient, but most of what makes a transient a transient lies in the higher frequencies.  Now, a transient that's really bottom heavy is a different story.  I even wonder how audible that lower portion on that kick drum is because the stronger content above may mask the lower part completely, but perhaps you did hear a significant difference in your learning experience?  Was it really a big difference?  Or just a subtle added weight?

 

I'm also still a bit confused about your obsession with what the input signal looks like.  Do you actually listen to the input signal?  I don't think so.  Apart from headroom issues, why should it matter?

Link to comment
Share on other sites

The other moral is that the difference between the 2 final corrections is quite small, acourate has cleaned this up v nicely in either case. I think this is a nice demonstration of the power of FIR filtering as well as a possible argument in favour of JSS's point re "if you fix the root cause, do you need that filtering?".

I'm a bit confused as to what you are trying to demonstrate.  Are your measurements with your Accourate filters on or off?  Also, your description of the curves in your first plot are confusing to me.  Do your measurements use loopback measurement or some other mechanism for accurate absolute timing?  Are these LP and HP XO filters minimum phase, linear phase, or mixed phase?  If mixed phase, then I assume the XO filters are Accourate optimized FIR filters?

Link to comment
Share on other sites

I'm a bit confused as to what you are trying to demonstrate. Are your measurements with your Accourate filters on or off? Also, your description of the curves in your first plot are confusing to me. Do your measurements use loopback measurement or some other mechanism for accurate absolute timing? Are these LP and HP XO filters minimum phase, linear phase, or mixed phase? If mixed phase, then I assume the XO filters are Accourate optimized FIR filters?

It is just an example of how using impulse peak alignment can be misleading.

 

Precise timing is achieved by using the mic alignment tool which emits Dirac pulses down two channels at a known offset, measures the actual offset and tells you whether you have the mic perfectly centred or are some no of samples off. Precision is at the sample rate so I usually use 96kHz for this.

 

The XO filters are linear phase, 2nd order Neville Thiele to be precise.

 

The resulting measurement is post correction.

Link to comment
Share on other sites

It is just an example of how using impulse peak alignment can be misleading.

 

Precise timing is achieved by using the mic alignment tool which emits Dirac pulses down two channels at a known offset, measures the actual offset and tells you whether you have the mic perfectly centred or are some no of samples off. Precision is at the sample rate so I usually use 96kHz for this.

 

The XO filters are linear phase, 2nd order Neville Thiele to be precise.

 

The resulting measurement is post correction.

Thanks.  I think I understand a little better now.  Time aligning by aligning the impulse response peaks is definitely problematic.  An impulse response peaks contains much more high frequency than low frequency energy, so a peak is often more representative of the time of arrival of sound at the top end of the transducer's range.  While it's encouraging to see alignment between (1) and (3), I'm wondering why in (3) you convolved the LPF with a measurement of both the subs and mains bass-managed together.  I think to see how well the transducers are time aligned, I would want (3) to be a LPF applied to a measurement of just the subs.  I guess I can see issues with that not working if the mains speaker has its natural roll-off close to the crossover.  For these reasons, I advocating using frequency response around the crossover region to choose the best distance.  If by chance, your delay is off by an entire cycle, you will still likely see inferior performance at frequencies away from the crossover.

 

Of course, all this ignores the impact the room has on the response, which can and often do distort the temporal response much more than the speakers do.  To the extent that Accourate successfully cancels out these problems at a single measurement point, it should be possible to get a fairly smooth crossover at that measurement location by choosing the right distance.  What the room does without correction makes the distance choice a lot more ambiguous.

 

I do have to wonder here what's going on with

 

Most speakers actually have a slight time offset between the arrival of tweeter and woofer frequencies.  To avoid this, the speaker designer must either ensure that the radiators are placed at the exactly same distance from the listener, or use an active crossover solution that can apply separate digital delays to the transducers.  Of course in practice, the difference in distance may not be very much, especially as far as the wavelengths at the crossover between the mains and sub.  Even if the tweeter sticks out a foot relative to the woofer and arrives ~1 ms earlier, the impact on an 80 Hz crossover will be pretty minimal.

 

I do have to wonder here what's going on in your picture with (2) versus (1) and (3).  First of all, your impulse response plot does not show where (1) and (3) actually peak.  Both the LPF And HPF filters should delay the response of the mains similarly in the crossover region, and being that these are linear phase, they should actually delay the mains responses uniformly in *all* frequencies.  This suggests that your mains low frequency response is actually delayed by at least 20 ms relative to their high frequency response, assuming the low frequency impulse responses are at their peaks near the very end of the plot window.  That's a lot of delay!  A speaker will rarely do that on its own in an anechoic environment unless it's using something like an aggressively tuned bass reflex enclosure that adds a lot of excess group delay.  Even then, 20 ms is about 1.6 cycles at 80 Hz.  Certainly the crossover between the woofer and tweeter shouldn't be doing that.

 

What's more likely going on is that this delay is almost entirely induced by the room.  What confuses me is that you said these measurements were made with Accourate FIR filters engaged.  Isn't Accourate supposed to correct that type of problem?

Link to comment
Share on other sites

..

Most speakers actually have a slight time offset between the arrival of tweeter and woofer frequencies.  To avoid this, the speaker designer must either ensure that the radiators are placed at the exactly same distance from the listener, or use an active crossover solution that can apply separate digital delays to the transducers.  Of course in practice, the difference in distance may not be very much, especially as far as the wavelengths at the crossover between the mains and sub.  Even if the tweeter sticks out a foot relative to the woofer and arrives ~1 ms earlier, the impact on an 80 Hz crossover will be pretty minimal.

 

 

Actually, in modern speakers, the problem is often opposite - the hf driver is too far back due to the horn..

 

As for the time alignment;

It is quite possible to be one wavelength off, and the fr still sums up kind of good.

Finding the "right" value may be difficult since this frequency range often is severely masked by room reflections.

Using fr, step response and group delay, with/without smoothing, can be helpful.

Link to comment
Share on other sites

I do have to wonder here what's going on in your picture with (2) versus (1) and (3).  First of all, your impulse response plot does not show where (1) and (3) actually peak.  Both the LPF And HPF filters should delay the response of the mains similarly in the crossover region, and being that these are linear phase, they should actually delay the mains responses uniformly in *all* frequencies.  This suggests that your mains low frequency response is actually delayed by at least 20 ms relative to their high frequency response, assuming the low frequency impulse responses are at their peaks near the very end of the plot window.  That's a lot of delay!  A speaker will rarely do that on its own in an anechoic environment unless it's using something like an aggressively tuned bass reflex enclosure that adds a lot of excess group delay.  Even then, 20 ms is about 1.6 cycles at 80 Hz.  Certainly the crossover between the woofer and tweeter shouldn't be doing that.

 

It's certainly a conundrum, I don't think I have got to the bottom of it yet.

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.


×
×
  • Create New...