B&C 21-IPAL Driver and Measurements Discussion

[Ricci]

Ok so about the ticking noise...I have a plethora of 21's in house from 4 different mfgs. Its not just the Ipals. One of the Lavoce 21's exhibits it and the other does not. And one of the Eminence 21's exhibits it and the other doesn't. It isnt audible in the sealed cab unless I put my ear right up to it. I dont think this is related to the spider gluing. It is present even with less than a few mm excursion. I almost wonder if it is the shorting action of the inductance rings even though that seems really far fetched but it does sound like a small spark jumping a gap. Really odd that Ive seen the same thing from 3 different major mfg's from 2 different continents on drivers with similar motor designs. That many unglued spiders out of such a small sample of drivers from multiple big name mfgs just doesn't seem possible.

[Ukko Kari]

Can anyone link a video of the sound? My hypothesis is that the shorting ring or rings are loose, and moving in response to forces generated by the coil. I should check my recently purchased 21DS115 for extraneous noises before building a cabinet.

[radulescu_paul_mircea]

First of all, wow, you already have that Eminence driver to test? I want a set of 4 but I have to wait for 4 moths at least. 

Second, that's odd, I have drivers from 4 different manufacturers, but only those from B&C make that noise.

I'll post a video on YouTube to for you to hear it and relate with.

[Ricci]

6 hours ago, Ukko Kari said:

 My hypothesis is that the shorting ring or rings are loose, and moving in response to forces generated by the coil. 

Thats another option. Might be the correct one too. It is present even with nearly imperceptible excursion levels. It is almost non existent once loaded in a cab and doesnt seem to affect the performance otherwise. 

[Ricci]

3 hours ago, radulescu_paul_mircea said:

First of all, wow, you already have that Eminence driver to test? I want a set of 4 but I have to wait for 4 moths at least. 

Im a 50min drive from them and they are production samples. Full production is a ways off still. Results are very encouraging so far.

[SME]

7 hours ago, Ricci said:

Im a 50min drive from them and they are production samples. Full production is a ways off still. Results are very encouraging so far.

That's awesome!  That's a factory I think it'd be a lot of fun to visit.

[radulescu_paul_mircea]

15 hours ago, Ricci said:

Im a 50min drive from them and they are production samples. Full production is a ways off still. Results are very encouraging so far.

Yes, I've seen that. The overhang is 21 mm, pole piece is 19 mm so the coil length should be 61 mm, huge. Probably the best 21 pro driver for vented alignment. Also, the BL^2/re product is around 330 and and that double separated suspension should make a very good sound 

[radulescu_paul_mircea]

Noise not present

 

Noise present

[Ricci]

6 hours ago, radulescu_paul_mircea said:

Yes, I've seen that. The overhang is 21 mm, pole piece is 19 mm so the coil length should be 61 mm, huge. Probably the best 21 pro driver for vented alignment. Also, the BL^2/re product is around 330 and and that double separated suspension should make a very good sound 

Coil overhang is 14.63mm. Xmax including 1/3rd of the gap is 21mm. It does that easily. Spiders are mirrored and large enough to maintain linearity. Big shorting ring, etc...Very similar to 21Ipal gap and coil specs. 1mm longer gap and just 0.3mm longer coil wind.

For comparison the coil overhang on a RCF 551 is 10.5mm and a 1/3rd gap xmax spec of 16.5mm.  The  DS115 is 11mm overhang and 15.67mm including 1/3rd of the gap. 

I think you may have had a different issue with your drivers. That's not what mine sound like. What I've been seeing is a higher pitched "ticking". Could be the video recording quality though. I'll see if I can get something to upload. 

[radulescu_paul_mircea]

What you hear in the video is the noise I get that keeps getting louder with time. It certainly is a suspension and glueing problem.

 

Very similar on the coil and gap but not magnet and coil material. If the coil is copper it means it is thinner for the same Re if the length is the same. This might mean that the gap width could be smaller so even if the amount of magnet is not the same, it might have a similar magnetic field inside. But that's just speculation.

 

[Ricci]

16 hours ago, radulescu_paul_mircea said:

What you hear in the video is the noise I get that keeps getting louder with time. It certainly is a suspension and glueing problem.

 

Very similar on the coil and gap but not magnet and coil material. If the coil is copper it means it is thinner for the same Re if the length is the same. This might mean that the gap width could be smaller so even if the amount of magnet is not the same, it might have a similar magnetic field inside. But that's just speculation.

 

That's reasonable speculation about the motor / coil specs. There are some differences, but the gap and coil overhang geometry is very similar. Weight is almost exactly the same. I like the spider setup with a spacer ring on the Eminence better though.

Did you recone a 21Ipal yourself? I'd like to see pics of the guts if you have them.

Your video does sound like a suspension issue. I'm hoping to be able to put up a video showing what I'm seeing on the Lavoce driver and the others this Thursday night. I won't have time until at least then. BTW the Lavoce driver seems to be quite capable so far.

[Kyle]

Could be a leaking surround, you would be surprised what they sound like. A ticking noise is not out of the question here. I can't say anything for sure but I do know if there is not enough glue on the pro cloth style surround, it can permeate air. It can be fixed with more glue applied right on top with a brush.

This might be the right stuff. I just did a quick google search so don't take my word for it! You need a little painters brush to apply on top.

https://www.simplyspeakers.com/speaker-repair-adhesive-mi-1291.html

 

 

 

 

[Ricci]

On ‎4‎/‎6‎/‎2019 at 1:46 PM, Kyle said:

Could be a leaking surround, you would be surprised what they sound like. A ticking noise is not out of the question here. I can't say anything for sure but I do know if there is not enough glue on the pro cloth style surround, it can permeate air. It can be fixed with more glue applied right on top with a brush.

This might be the right stuff. I just did a quick google search so don't take my word for it! You need a little painters brush to apply on top.

https://www.simplyspeakers.com/speaker-repair-adhesive-mi-1291.html

 

 

 

 

In my case it is definitely something inside the motor under the spiders. Driver in free-air. It's not really audible once in the cab. Eminence is replacing the woofer that ticks and want to diagnose the issue. The Lavoce driver exhibiting the same thing I haven't contacted them about yet.

[Kyle]

Ya, gonna be hard to guess at what that might be. Report back if they tell you anything.

Good news about your replacement.

[Ricci]

Speculation is it might have been the not enough glue on the joint at the former.

[SPAZ]

Forgive me for bringing up an old topic but in regards to the 21IPAL, what is the actual possible excursion? It seems like some people go by the XVAR which is 15mm. So can this woofer only go 15mm before THD goes up significantly? 

[SME]

On 8/12/2020 at 2:24 PM, SPAZ said:

Forgive me for bringing up an old topic but in regards to the 21IPAL, what is the actual possible excursion? It seems like some people go by the XVAR which is 15mm. So can this woofer only go 15mm before THD goes up significantly? 

Yes---for some definition of "significantly".  If you look in B&C's docs, they may describe their methodology for XVAR.  Someone else's definition of "Xmax" may yield a different number.  The measurements on DataBass don't give a specific number but simply show how compression and distortion increase as the driver is pushed toward its mechanical limits.

Realize also that actual distortion depends on more than just excursion.  Certain aspects of the cabinet design can suppress or amplify the effects of distortion in the motor.

[Ricci]

SME answered well. 

It's complicated. With almost any driver the performance starts to degrade as it is pushed to higher excursion. This really cannot be answered with a single number like xmax that mfg's try to use. The 21Ipal will go to about 30mm one way if not a bit more if you REALLY hammer on it but it does not sound good by then. It gets cleaner at 20mm one way, and again much cleaner at 10mm one way than 20mm. 

[kipman725]

I have been reading these papers by Kipple:

http://www.klippel.de/fileadmin/klippel/Files/Know_How/Literature/Papers/Green Speaker Design Part 1.pdf

http://www.klippel.de/fileadmin/klippel/Files/Know_How/Literature/Papers/Green Speaker Design Part 2.pdf

In the second paper they improve the efficiency of an overhung driver by reducing the coil height to the top plate thickness and then compensate for the resulting non linear behavior.  The benefit of this in their example is a 39% in voice coil temperature for a given output compared to an overhung driver (~3dB output gain in the non excursion limited region for the same driver).   Considering the IPAL driver already is designed to be a transducer with dedicated amplifier and it has a huge coil overhang it would seem that with just shortening the coil a bit and a new version of the IPAL mod that implements the kipple DSP work they could increase the efficiency/output substantially (IPAL2?).  There would also seem to be benefits form changing reducing kms but I understand that this is required to have small clearances in the voice coil gap due to rocking modes.

[SME]

Interesting discussion.  I'm definitely a fan of the idea of designing transducers and electronics to work together.  However I'm skeptical that the described approach will work as well in practice as it does in theory.  Even in theory, I don't really agree with the argument given in "5.2 Large Signal Modeling" in the first paper that it is adequate to approximate instantaneous large signal parameters e.g. (Bl(x,i), K(x), Le(x,i)) with their averages with respect to stroke and current.  To me this is basically a hand wave.  Among other things, it ignores the likelihood that the peaks (from instrument attacks) are likely most relevant to perception of the sound.

The practical example given in the second paper isn't all that gratifying to me either.  Distortion performance is reported in terms of "averages" with little attention given to distinguishing higher order vs. lower order contributions or the potential perceptual consequences of higher distortion on the peaks.  Improvements to 2nd and 3rd order distortion are impressive on paper, but those are likely to be the easiest components to reduce using DSP.  Also, the suggested 16 dB crest factor seems way too big for "common audio signals", and if the content really is that dynamic, I'm not sure this whole study would have much relevance.  That 16 dB crest means average power is 1/40th of peak.  So if we're delivering 4000W at peak, average power will be a mere 100W.  A more realistic crest factor is 9 dB, and that means average power is more like 1/8th of peak, or 500W for my example---much more significant!

I'll have to dig into the references some time to learn more about the KCS system used here.  The first paper suggests that the required DSP can be done in a feed-forward configuration "in theory", but in practice, some kind of "adaptive" control system is "required to cope with unavoidable production variances, aging of the suspension, changing climate conditions and other external influences" as well as for coping with "undesired effects generated by a voice coil offset".  OK.  The first bucket of things (suspension, climate conditions, etc.) are "slow" changes for which compensation need not necessarily be adapted in real-time.  However unless I misunderstand, voice coil offset is an issue that likely requires reacting quickly.  Furthermore, if we actually care about distortion levels at the peaks of current and stroke, we probably need a very fast acting *feedback* control system.  Is KCS a feedback system?  I have no idea, but the lack of detailed discussion on this point in the papers makes it harder for me to believe that this approach is as easy to achieve as the paper suggests it is.

Also feedback control systems have a major limitation due to lag in the control loop.  This is a problem that affects high frequencies more than low frequencies.  So while such a system may do a great job of reducing 2nd and 3rd harmonic distortion, it may be too slow to address higher order distortion problems.  Likewise, if high frequency content occurs simultaneous to high excursion from low frequency content, those high frequencies may not be rendered cleanly even with the feedback control.

So anyway while I like this idea in principle, I want to see more real world data and a more granular analysis of the resulting distortion artifacts.

[kipman725]

I think these area good points I also had some skepticism about the statistical basis for saying that the the small signal values can be used.  High crest factor signal is interesting and only reporting 2nd and 3rd harmonic distortion, this suggests some gaming of the results.

I think the practical implementation uses voltage and current sensors to inform the model along with reference data obtained from the kipple system of the transducer.  There are a few demo videos floating around like this one where parameter variation over time is shown:

 

[SME]

Interesting.  I'll assume the system updates the model parameters at least as frequently as the graphics do.  I imagine it does a good job of correcting thermal-related changes, and the ability to correct for environmental effects on soft parts could be quite useful.  These are things that can be approximated to be linear for relatively short intervals of time.

Now then, I see something important that I missed.  It is mentioned that a system that relies on actual K(x), BL(i,x), Le(i,x), etc. data taken from a "Klippel analyzer" device can theoretically model and cancel out the anticipated distortion from a given input signal.  This capability is independent of the ability to adapt to changes of these parameters over longer periods of time.  It looks like the purpose of the discussion of "effective" parameter values (averaged with respect to stroke or some short time interval) is to argue that a small signal model with periodically adjusted parameters works OK for identifying and correcting for the longer-term changes described above.

So really, there are two different kinds of corrections being made here.  AFAICT, the paper is not clear on how these two corrections are applied together, but I would assume they are applied relatively independently such as by multiplying.  For example, Kms changes might be modeled as:

    Kms_now(x) = alpha_kms * Kms(x)

where Kms_now(x) is used in the distortion cancellation model, Kms(x) is the original Klippel-measured stiffness, and alpha_kms is a multiplicative factor that descries changes observed by the live impedance monitoring.

Something notably absent from consideration here are so-called complex inductance models, which I expect to be important for cancellation of distortion in large, heavy voice coil subs.  These papers would generally recommend that one keep Mms as low as possible to maximize efficiency and to compensate for sensitivity problems around F0 by using lower Re (and a suitable high-current-stable amplifier).  However, one reason to have a higher Mms in a deep bass sub (aside from the fact that bigger coils also tend to handle more power), is that it acts as a kind of acoustic high-shelf filter that reduces higher frequency distortion harmonics.  (Note that high Le does not have this effect because it acts before the main source(s) of distortion are introduced).  Higher Mms may also improve system behavior in multiple driver systems or in highly loaded scenarios.  IIRC, this was a potentially important factor in @Ricci's M.A.U.L. design.  So back to Le.  Complex Le is well known to be essential for modeling small signal behavior of large subs, so I think having this be part of a distortion cancellation model is important.

====

Anyway, this work is definitely intriguing.  Several years ago, I contemplated experimenting with a feed-forward approach to reduce distortion in my own subwoofers, but I haven't looked into it yet.  Instead, I've been focusing intently on linear aspects of response for many years now.  Linear response seems to be absolutely critical to sound quality, and even seemingly minute flaws in linear response can profoundly alter the sound.  I believe the reason for this is, most ironically, because essentially all real-world content has a lot of harmonic distortion, some of which is naturally occurring but much of which is intentionally synthesized, often very late in the production chain in order to enhance the sound quality.  This added "distortion" relates to the original signal in a very regular way, one which listeners seem to adapt to very readily (given that this is also a property of most natural sound sources).  However, linear response flaws can disrupt that regularity making it much harder for the listener's brain to correlate the spectral content.  This leads to what I would term "perceptual spectral leakage" in which some spectral content becomes disassociated from the rest of the sound.  For example, perceptual spectral leakage in the low frequencies can cause muddiness.  Perceptual spectral leakage in the high frequencies can cause harshness.  IMO, both of these problems are widespread under typical reproduced sound conditions.

With that said, non-linearity surely matters at some point.  On an indoor system with ample headroom as I'm using to develop my optimization technology, I don't worry much about non-linearity, but I reckon it will be a lot more important for systems being run to their limits, including in most live sound applications.  I think I'm less worried harmonic distortion, especially low order HD which is much less likely to sound offensive, and more worried about effective frequency response changes.  These papers describe an approach which appears promising for long time-scale changes to sensitivity and frequency response such as from thermal and environmental effects. The papers also describe distortion cancellation, albeit with unclear performance for higher order distortion products which may be more offensive.  However, the papers don't appear to discuss short-term frequency response shifts such as when a kick drum hits at 16 dB above the average level.  Hypothetically on some system, my optimization methods may yield superb results for everything but those kick drum hits, which would be quite tragic.  At this point, I have no idea how important these problems are because my system for R&D has ample headroom for all but the biggest (under 20 Hz) movie effects.  We'll see, but it won't surprise me at all if I find that my optimization makes it a lot easier to hear and compare the underlying non-linearities of sub systems.