Hi John
I think the Behringer and several others do the same trick - CARA and some hack of CATT being the first convolution based SW I've used ( pretty old stuff) - now I use PC based Accourate for the purpose of speaker response shaping and nowadays there are other PC based solutions of course - but - I'm not (was not and will not) arguing about any *details* of UE being capable of or not (going 6 way currently - its not any serieous option for me anyway). What I was referring to was just the general possibility of linearizing frequency and phase at will.
If you wish, we can continue anytime where we got stuck - taking UE as the workhouse in exploring restrictions I claim with respect to correctability of CMP behaviour.
I will possibly throw in some additional wavelet analysis and Joachim – if got interested - may do some Cepstrum analysis too – so it might be juicy fun...
I'll have a look how to purchase UE standalone version and give it a try anyway (hope it works in my setup)...
Michael
PS
BTW, John your page linked is not correctly displayed on my Ubuntu/ Opera 11.01 (latest version) – old story ...
Certainly the theory for amplitude and phase linearization is not new. But the concept behind the UE is making this theory accessable to the average DIY speaker builder without needing any serious background. That is what it does that nothing else does, AFAIK.
I'm not interested in wavelet analysis because, as I have said, it is just another way of looking at the same thing and you know where I stand on CMP.
As for my pages, the problem is most likely that the correct fonti s not installed on your system. Most likely the pages that don't display use Arial MS Unicode. If you save the web page on your PC and then edit the source, search for "Arial MS Unicode" and substitute "Arial" andit should display correctly.
Well, thats all I was referring to.
Sent Bohdan an email regarding UE - will see how I can get it - and get it to work...
Michael
Last edited:
Ohh - could you possibly send me the preliminary version then ?
Release of Ultimate Equalizer V1.0 program - Techtalk Speaker Building, Audio, Video, and Electronics Customer Discussion Forum From Parts-Express.com
Michael
Thanks a lot for that - might be Soongsc finally can take it from you ?
---------
The "real" problem with EQing break up modes is that break up modes actually are CMP effects (build up of standing waves in different orientation within the diaphragm).
So even if it looks as if we can EQ it *perfectly* by convolution techniques - this is not really the case. Those limitations regarding EQing of CMP systems (any CMP system !) I've clearly outlined.
Michael
I'm in agreement with John on this topic. An off-topic debate here will just detract from the thread.Thanks a lot for that - might be Soongsc finally can take it from you ?
---------
The "real" problem with EQing break up modes is that break up modes actually are CMP effects (build up of standing waves in different orientation within the diaphragm).
So even if it looks as if we can EQ it *perfectly* by convolution techniques - this is not really the case. Those limitations regarding EQing of CMP systems (any CMP system !) I've clearly outlined.
Michael
Dave
Last edited:
This is exactly why digital EQing requires an experienced hand as Joachim had mentioned.
Not going to start an OT discussion but I will just post the following comment.
Let Ta(s) is the on axis response of a system and To(s) be some off axis response of the same system. Then if Eq(s) is the transfer function of the respons ethe Eq's Ta(s) to flat amplitude, linear phase, then the CSD or any other presentation of Eq(s) x Ta(s) will be that of a perfect system. If To(s) is not identical to Ta(s) then the relative degree of degredation of To(s) compared to Ta(s), by any measure will be the same as the degradation of Eq(s) xTo(s) relative to Eq(s) x Ta(s). This follows directly from the observation that To(s) / Ta(s) = [Eq(s) x To(s)] /[Eq(s) x Ta(s)].
Let Ta(s) is the on axis response of a system and To(s) be some off axis response of the same system. Then if Eq(s) is the transfer function of the respons ethe Eq's Ta(s) to flat amplitude, linear phase, then the CSD or any other presentation of Eq(s) x Ta(s) will be that of a perfect system. If To(s) is not identical to Ta(s) then the relative degree of degredation of To(s) compared to Ta(s), by any measure will be the same as the degradation of Eq(s) xTo(s) relative to Eq(s) x Ta(s). This follows directly from the observation that To(s) / Ta(s) = [Eq(s) x To(s)] /[Eq(s) x Ta(s)].
This doesn't square with my experience. I've tried on several occasions to use EQ to correct response anomalies created by physical/mechanical properties of systems under test and each time more distortion was generated by the EQ process with little or no benefit overall. In theory, it sounds great but in practice, it seems to resemble more of a "wack-a-mole" exercise.
It depends on the source of the problem and whether or not it is minimum-phase. If it's minimum-phase, as most drivers are, it can be EQ'd on most any particular axis. Off-axis it will change as John described. I've been able to take a 10" woofer with rather severe breakup and create a model that is minimum-phase and match measured phase up to 20KHz on-axis. The effort took quite a lot of time as the software (CALSOD) requires that a minimum-phase element (MPE in its parlance) be created for each non-linearity, but it worked. The phase match was very good. Were any of it not minimum-phase, it could not work. But it did.
Likewise, with the new UE from Bodzio Software (SoundEasy) the output is as good as the input. One could, I'm sure, EQ above 20K with success, but that becomes an issue of mic placement and calibration as well, so to get truly flat response at high frequencies with short wavelengths, the mic would likely need to be re-positioned well and with a mic calibrated much higher. I have one calibrated to 40K+, but I've not mounted it as my current one does what I've needed.
Look at my post #403 on page 41. Below the limit I set at 20K and above the low end limit due to measurement conditions, you have the evidence of just some if capabilities and how easy it is. Were I to use measurements without any smoothing and re-measure at the same position, the response would be even smoother than that is.
Dave
I think the argument for equalization to address resonances would be a bit stronger if your graphs included 2nd through 5th harmonic distortion plots. Also not sure what you mean by a "minimum phase" driver. All the same, I can recall on several occasions introducing some high q "adjustments" to active crossovers and watching 2nd and 3rd order harmonics skyrocket with just 4 or 5 db of change. Maybe the crossovers haven't been up to the task or I've been asking too much from them.
"run in circles" that's one of ways to breakup the breakup modes.wow ! you tend to accept CMP ??? - cant believe !!!
On the other hand - why else is an experienced hand needed ? - I guess you (and anybody else) still run in circles unless you can answer *this* question - from a technical point of view of course, mind you !
Michael
I still have no idea what CMP is.
I suppose you're going to wiki it?
CMP - Wikipedia, the free encyclopedia
I haven't made distortion measurements, but I would expect even order linear distortion to drop since non-flat FR is linear distortion. A flatter FR should have reduced linear distortion. I certainly don't hear any change that I can attribute to any increased distortion. Not knowing the specifics of your cases, I can't comment on them, either.
This is off-topic here, though, so let's not hijack the thread.
Dave
On the contrary, I think the practicality or lack thereof for what you're doing is very relevant to this thread. Frankly, I see holes in the theories you're promoting that are big enough to drive a Mack truck through. Have you ever seen a driver with flat frequency response that had gobs of 2nd order and (even order) distortion? I have. Have you ever seen a driver with uneven frequency response that had excellent even order distortion characteristics ? I have. The fact is, non linear distortion cannot be readily dealt with using techniques that rely on strict causality. You can't sit there and tell me a driver at 1V applied signal is going to need the same response "correction" that it will need when 10 V is applied. Energy storage is a complex phenomenon that can't be simply mapped with frequency response and a Hilbert transform. If what you are promoting were so effective at dealing with driver response anomalies, there would be no need to build highly linear, critically damped, wide bandwidth drivers. Any crap design would do and we could rely on EQ correction to make it "sound good". The physical manifestations of irregular driver response could easily be overcome by applying corrective amplitude and phase EQ. This is simply fairy tale stuff and my physical experience has made that evident - to me at least.
Btw, I did a Google search for "minimum phase driver" and your web page was the only thing that came up. Are you trying to create a new theory about drivers? The only sparse commentary other than yours on the subject that came back from Google indicated that drivers can exhibit minimum phase behavior "for the most part" under small signal conditions.
Discussion of Group Delay in Loudspeakers
Again, not trying to side track this thread. But for me, if the discussion is to have any value whatsoever, there needs to be some reckoning with reality and I'm not sensing that you've done that yet. Distortion plots would go a long way in that direction.
I'm happy to see that Joachim is fine with this. I have my own questions on several aspects, so I can't answer all of your questions myself.
Regarding new theory, no, I've reported on my empirical findings on this. If you do a search here you'll find threads in which minimum-phase in drivers is discussed. I invite you to do so.
With regard to reality, I report what I measure and how that relates to and matches models I've created in this case. It is small signal, yes. But since my focus was the FR response that is essentially the steady-state response, it has been related to what would fall under linear distortion.
You'll note that I have not once tried to imply that any of this EQ improves non-linear distortion. It relates strictly to linear distortion. The changes you mention about small to large signal change is certainly non-linear. The EQ will not correct that, this is not a claim I made. It looks like you have misinterpreted what I wrote.
With that said, the change due to signal level will remain, with one possible exception that I can think of. That is the change associated with the ability to use much steeper slopes in DSP than one might do passively or even actively (non-DSP). The abrupt cutoff that reduces output above Fc may limit excitation of a driver in an area that might otherwise more highly energize higher order distortion products. That could only be determined by a series of distortion measurements, of course.
I'm really not sure what you're disputing in what I said because it looks like you're thinking primarily non-linear distortion when I was addressing linear distortion.
Dave
The only reason I decided to post in this thread rather than simply read with amusement was soongsc's post about breakup modes and your reply which based on standard theory and practice was highly incorrect. Break up modes are widely considered non linear distortion essentially because the physical manifestations that cause them are not linear in any real sense. Guessing about lower frequency excitation of upper breakup modes is folly - simply not predictable in any practical sense from one driver to the next since there are way to many contributing variables, some that are exposed as lower order energy storage and some that aren't. In any case, I appreciate your candor in acknowledging the limited scope of usefulness for this exercise in the small signal regime. Given that, I will bow out and not pursue the issue of break up modes in this thread's context any further. Thanks.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.