Jump to content

The Wandering Topic Audio/Video Discussion Thread


Ricci

Recommended Posts

Been talking with a few guys about a thread for whatever audio/video related discussion happens to be the subject of the day. Bass traps, how to design a FLH, the state of 4K video products, amplifier/speaker matching. Whatever.

Here it is.

 

I'll start off with some points I wanted to make related to discussion in another thread that is getting way off topic. Each one of these subjects could be its own topic.

 

1.) Winisd or other modeling and simulations. Bench racing with questionable small signal parameters and highly incomplete or inaccurate sims. We've all seen it. It's all over the place at many forums. Has been for years. People base buying decisions on it because they don't know any better. Has jack to do with reality usually. Totally agree. As Bosso said driver in hand / in cab with measurement equipment is required to know what you really have. I said my piece on this years ago, a couple of times. I've had people tell me the measurements must be wrong because their simulation doesn't match it. LOL. This is a BIG part of how I ended up doing the testing to begin with. I started doing measurements and realized the simulations I and others were doing were fantasy land. Started with multiple old car audio drivers I had and trying to DIY a few successful builds using them, which mostly failed utterly to perform anything like the sims. That was beginning of the rabbit hole for me, after realizing that the TSP's and a quick sim do not predict what you will get with a manufacturers driver. Basic TSP's are inadequate to simulate modern, large coil, long throw drivers in most cases and most any driver type at high output levels. Power handling and xmax specs are all over the map. This is why I started buying and testing so many different drivers. Can't trust the specs or sims. Got to verify it myself. Might as well let everybody else know while I'm at it and help the other guys like me out.

 

 

 

2.) Klippel: It is not infallible. I have talked with a few different people in the industry who have said that the setup for a measurement can be screwed up quite easily. It can give data that is not accurate due to setup, operator or equipment errors. Now I'm not saying that has happened in any specific cases because I don't know of any for sure. I am told it is possible though and not that hard to do. Like anything else don't just assume.

 

 

 

3.)There has been a few remarks about not knowing whether the driver was pushed to the excursion limits during the max long term or burst tests here. EVERY driver has been. There is no question about this. The burst testing is only stopped when there is a clear indication of driver distress (usually mechanical noise) in the deep bass or the amp runs out. Below 30Hz the amp never runs out. The one case was the XXX overhung coil. The amp clipped right about the time the driver was done anyway. This was because the suspension had tightened dramatically and the motor had lost a ton of force with the coil that far from center in an undersized box. Same thing with the long term testing. That test is almost always stopped by driver excursion and bad noises indicating mechanical limitations are imminent. Severe thermal compression in the upper bass of 3dB or more is another good sign to stop but this almost always occurs at a point where the driver is out of excursion anyway. The CEA-2010 data may not be the maximum output of the DUT. That info is also available for nearly all systems though under the static charts section. The numbers in red are the maximum level attainable by the device limited by driver excursion. In the case of the XXX you could mount it IB and it will never put out more than about 111dB at 20Hz same as in the 4 cube cab. Power required will go way down as would distortion but that is the displacement limit. It is not going to put out any more maximum SPL in the deep bass. The 2-120Hz sweep used for the long term test covers the entire bass FR. While you may see slightly more output from a system with the CEA-2010 or MAX bursts broad band there will usually be one band or another that is very close. Note that the bursts are 1/3rd octave spaced. The sine sweep covers every frequency in the bandwidth and it will find the worst case frequency for induced driver excursion in the sealed cab. It may be a very narrow small area, but if the driver threatens to blow apart or smash itself to death there, that becomes the voltage limit for the long term testing. If the driver coil doesn't melt first of course. Some drivers like the XXX this occurs up near the impedance peak at 30Hz due to the high qtc and resultant excursion profile. Others such as the 19 with a very low qtc in the same cab are limited by excursion at 10Hz or lower rather than 20-30hz.

 

 

 

4.) Long term versus demanding music program material: DIYSG I have seen you and another poster make a few posts here and there about how the 2-120Hz sine sweep long term test signal here is not that demanding and how some demanding music material could be worse. I disagree. It is not as demanding as say picking the absolute worst frequency for heating the coil and dropping a square wave there, but it is much worse than almost any music content I know of other than a few fringe bass tracks or something that just happens to be sine wave like into the impedance/excursion minimums. As explained above this will hit the imp/excursion minimums every time and drop significant power. If you look at typical impedance curves and tunings of bass systems and consider the high and low pass filters involved, etc...Most music is above 25Hz and will have a high crest factor comparatively. The tracks that do have sustained sine like bass tones are almost always in the 25-50Hz octave. (I recognize outlier tracks such as Lone Survivor. I'm talking 99% of the rest of it.) Impedance is typically high over some or part of that range. The sweep signal only last 23sec roughly then it is over. Yes you may be able to kill a driver with the same voltage run for 20min solid with a music track on repeat but if the sweep continued on repeat the driver would die much quicker. That's why it is considered to be an accelerated thermal test as well as judging things like compression, excursion, vibration and mechanical noises. Comparing it on paper is one thing but running that sweep through a system pushed at the edge of its capabilities should remove any doubt about how demanding it is

Link to comment
Share on other sites

  • Replies 52
  • Created
  • Last Reply

Been talking with a few guys about a thread for whatever audio/video related discussion happens to be the subject of the day. Bass traps, how to design a FLH, the state of 4K video products, amplifier/speaker matching. Whatever.

Here it is.

 

I'll start off with some points I wanted to make related to discussion in another thread that is getting way off topic. Each one of these subjects could be its own topic.

 

1.) Winisd or other modeling and simulations. Bench racing with questionable small signal parameters and highly incomplete or inaccurate sims. We've all seen it. It's all over the place at many forums. Has been for years. People base buying decisions on it because they don't know any better. Has jack to do with reality usually. Totally agree. As Bosso said driver in hand / in cab with measurement equipment is required to know what you really have. I said my piece on this years ago, a couple of times. I've had people tell me the measurements must be wrong because their simulation doesn't match it. LOL. This is a BIG part of how I ended up doing the testing to begin with. I started doing measurements and realized the simulations I and others were doing were fantasy land. Started with multiple old car audio drivers I had and trying to DIY a few successful builds using them, which mostly failed utterly to perform anything like the sims. That was beginning of the rabbit hole for me, after realizing that the TSP's and a quick sim do not predict what you will get with a manufacturers driver. Basic TSP's are inadequate to simulate modern, large coil, long throw drivers in most cases and most any driver type at high output levels. Power handling and xmax specs are all over the map. This is why I started buying and testing so many different drivers. Can't trust the specs or sims. Got to verify it myself. Might as well let everybody else know while I'm at it and help the other guys like me out.

 
While I agree with all this I've actually done quite a bit of work to improve my sims and they are much more accurate than your average WinISD sim posted on a forum.  I sent you the paper I wrote on large coil simulation accuracy and adjustment, not sure if you ever read it or not.  David McBean even ended up incorporating the generic version of the tweak into Hornresp so it's as simple as clicking a button.  I worked extensively with David and LTD02 to make the study, the paper, and the tweak as accurate as possible.  You got credit in the paper too, since all the measurements but one came directly from you.
 
Based on your measurements I was able to manufacture a simple tweak that produces remarkably accurate sims for this type of driver.  Of course you still need measured specs for best results (but that's nothing new) and the method (the generic version especially) isn't perfectly accurate but it's a huge improvement over a regular sim.  Based on about 50 of your measurements of sealed, ported and horns with a variety of drivers I get a remarkably close correlation between sim and measurement.  It works, it works for all drivers and it works for all enclosure types.  Not perfect but very notably better than a regular sim.
 
Take that a step further and add a bit of Re to my sims and you get pretty close to your large signal (high power) test measurements.  Again, not perfect, but a lot better than ignoring power compression.
 
So the problem isn't entirely the sims, it's a problem with basic sims and a problem with people that don't understand what a simulator is actually showing them.  If you do some tweaks based on educated knowledge of how these specific type of drivers perform and how all drivers perform at high power you can get pretty accurate assessments of real world conditions with a simulator.  I spend a LOT of time with simulators (mostly Hornresp but also Akabak, tl.app, MJK's worksheets, Bagby's suite of worksheets), and I've made significant strides simulation in accuracy.
 
This is why I crave accurate data and specs.  The sims are only as good as the garbage you feed them.
 

 

 

2.) Klippel: It is not infallible. I have talked with a few different people in the industry who have said that the setup for a measurement can be screwed up quite easily. It can give data that is not accurate due to setup, operator or equipment errors. Now I'm not saying that has happened in any specific cases because I don't know of any for sure. I am told it is possible though and not that hard to do. Like anything else don't just assume.
 
I'm a little bit disappointed that this is your only comment on the whole Klippel thing, but so be it.  It was never claimed that Klippel was infallible, the issue was some manufacturers were using the klippel name inappropriately.  In the CSS case it said Klippel right on the spec sheet but a 3rd party klippel test showed 2.5 times less klippel verified xmax than the published spec.  This is the real heart of the issue, not whether or not the klippel (or it's operators) can make mistakes.  In the CSS case they had proof that the driver was not klippel verified for the xmax they published but when questioned came up with a variety of excuses and attacks that have nothing to do with how klippel defined verification works.

 

 

 

 

 

3.)There has been a few remarks about not knowing whether the driver was pushed to the excursion limits during the max long term or burst tests here. EVERY driver has been. There is no question about this. The burst testing is only stopped when there is a clear indication of driver distress (usually mechanical noise) in the deep bass or the amp runs out. Below 30Hz the amp never runs out. The one case was the XXX overhung coil. The amp clipped right about the time the driver was done anyway. This was because the suspension had tightened dramatically and the motor had lost a ton of force with the coil that far from center in an undersized box. Same thing with the long term testing. That test is almost always stopped by driver excursion and bad noises indicating mechanical limitations are imminent. Severe thermal compression in the upper bass of 3dB or more is another good sign to stop but this almost always occurs at a point where the driver is out of excursion anyway. The CEA-2010 data may not be the maximum output of the DUT. That info is also available for nearly all systems though under the static charts section. The numbers in red are the maximum level attainable by the device limited by driver excursion. In the case of the XXX you could mount it IB and it will never put out more than about 111dB at 20Hz same as in the 4 cube cab. Power required will go way down as would distortion but that is the displacement limit. It is not going to put out any more maximum SPL in the deep bass. The 2-120Hz sweep used for the long term test covers the entire bass FR. While you may see slightly more output from a system with the CEA-2010 or MAX bursts broad band there will usually be one band or another that is very close. Note that the bursts are 1/3rd octave spaced. The sine sweep covers every frequency in the bandwidth and it will find the worst case frequency for induced driver excursion in the sealed cab. It may be a very narrow small area, but if the driver threatens to blow apart or smash itself to death there, that becomes the voltage limit for the long term testing. If the driver coil doesn't melt first of course. Some drivers like the XXX this occurs up near the impedance peak at 30Hz due to the high qtc and resultant excursion profile. Others such as the 19 with a very low qtc in the same cab are limited by excursion at 10Hz or lower rather than 20-30hz.

 

Based on your text descriptions of excursion noises in almost all of your write ups I was pretty sure the drivers were pushed as far as they could mechanically go.  Honestly, props to you for pushing things to the limits when it could be a costly proposition if the driver fails.

 

 

 

4.) Long term versus demanding music program material: DIYSG I have seen you and another poster make a few posts here and there about how the 2-120Hz sine sweep long term test signal here is not that demanding and how some demanding music material could be worse. I disagree. It is not as demanding as say picking the absolute worst frequency for heating the coil and dropping a square wave there, but it is much worse than almost any music content I know of other than a few fringe bass tracks or something that just happens to be sine wave like into the impedance/excursion minimums. As explained above this will hit the imp/excursion minimums every time and drop significant power. If you look at typical impedance curves and tunings of bass systems and consider the high and low pass filters involved, etc...Most music is above 25Hz and will have a high crest factor comparatively. The tracks that do have sustained sine like bass tones are almost always in the 25-50Hz octave. (I recognize outlier tracks such as Lone Survivor. I'm talking 99% of the rest of it.) Impedance is typically high over some or part of that range. The sweep signal only last 23sec roughly then it is over. Yes you may be able to kill a driver with the same voltage run for 20min solid with a music track on repeat but if the sweep continued on repeat the driver would die much quicker. That's why it is considered to be an accelerated thermal test as well as judging things like compression, excursion, vibration and mechanical noises. Comparing it on paper is one thing but running that sweep through a system pushed at the edge of its capabilities should remove any doubt about how demanding it is

 

I have to agree with all this, it is true.  But I do have to point out that those "fringe" tracks do exist and they are becoming increasingly more popular.  Both in the car audio bass track genre and in the electronic music genre.  And sometimes these contain much lower frequencies than 25 hz.  Not often, but they are becoming more popular as more and more people obtain systems that can reproduce them and actively seek out this type of media.

 

While it is true that most of these will have bass lines in the 25 - 40 hz range where the impedance usually means they will not draw that much power in regular sealed or low(ish) tuned ported boxes, people are also sometimes using this type of driver in horns.  And in horns the impedance is all over the map and you could very well have your largest power draw right in the 25 - 40 hz range.  Bass lines in this range of frequencies might not be a big issue for normal sealed and ported boxes but it can be an issue for other enclosure types.

 

Perhaps I could have made these two points a lot more clear, because even the high intensity tracks won't hurt drivers in "normal" sealed or ported enclosures as long as the bass lines stay in the 25 - 40 hz or region where they are protected by the impedance causing low power consumption.

 

(This isn't that important, but remember that almost everything I sim is horns.  Like maybe 100:1 horn sims vs everything else.  Horns are my thing.)

 
Link to comment
Share on other sites

649-02423434er.jpg

 

 

Klippl'd my dog last week...

 

200127425-001-male-office-worker-pulling

 

You want ANOTHER infraction?!

 

See folks what we have here is typically called an habitual line-stepper or an HLS for short. Don't be this guy. They usually end up pregnant and on drugs.

Link to comment
Share on other sites

I'm a little bit disappointed that this is your only comment on the whole Klippel thing, but so be it.  It was never claimed that Klippel was infallible, the issue was some manufacturers were using the klippel name inappropriately.  In the CSS case it said Klippel right on the spec sheet but a 3rd party klippel test showed 2.5 times less klippel verified xmax than the published spec.  This is the real heart of the issue, not whether or not the klippel (or it's operators) can make mistakes.  In the CSS case they had proof that the driver was not klippel verified for the xmax they published but when questioned came up with a variety of excuses and attacks that have nothing to do with how klippel defined verification works.

 

 

Based on your text descriptions of excursion noises in almost all of your write ups I was pretty sure the drivers were pushed as far as they could mechanically go.  Honestly, props to you for pushing things to the limits when it could be a costly proposition if the driver fails.

 

 

What else is there to say about Klippel ratings? I don't know the full story on the CSS or DW case. Have to read up on that. I agree that it is not correct to present your xmax spec as the minimum of all 4 major specs if that is not true. No argument there. On the other hand I don't really think that was the intent with SI. The way I look at it SI publicly posted at least a BL curve and I believe the KMS suspension curve? (been a minute since I looked.) Most companies don't do any of that especially smaller niche companies. I give props for even going there. It would have been a lot easier to just list an xmax like a lot of others with no explanation at all. I bet he wishes he could take that back now. posting that type of data seems to open them up for more criticism not less. Most drivers look like shit on a full Klippel workup save for those from really big companies with big R&D. Even those don't do well sometimes. Some of the Morel's I've seen in VC were um...Anyway if it were me and I had a driver company yeah I'd handle things a bit differently and every driver would have the whole Klippel work up posted along with more engineering data than called for. And I'd like to think the drivers would be spectacular.  I bet it'd fail epically. Some driver at half the price and with none of the verification would have better paper specs and that'd be it. Low sales and a couple of hiccups in delivery or QA and it'd be over. That's why I realized I don't want to start either a driver or subwoofer venture.

 

The driver excursion testing bit was really just to clarify that again. It seems like it may not have been widely known that every driver gets "tapped" out during testing. At least so far.

 

About simulations...Yes that's exactly why. I have been manually matching curves in sim land for many years with various little tricks. Most of which you know. I haven't seriously used winisd for a very long time. It is good for a record book of driver specs though. I posted a few times on the subject but never really put a concerted effort into raising the level of awareness. Guess I thought others would put 2 and 2 together from looking at their sims versus the real deal. Some do some don't. There are guys who were well aware of it before me and doing similar things. Most are industry people who do it for a living. I could live without that type of accuracy for sealed or IB systems and even vented to some extent but most of my designs end up being more complex higher order designs like horns, bandpasses,etc. I just find them much more fun and challenging. In those cases it has to be close or the whole thing could be a big turd at the end of it. The irritating part is that the number of drivers I feel confident enough in that the sim will be accurate, is a small pool. Most I've measured myself.

Link to comment
Share on other sites

What else is there to say about Klippel ratings?

I don't know.  You did mention a few times that you were going to say something, I just thought you would say a bit more than you did.  No harm no foul.

 

I really don't want to rehash the specifics again, but if you put the word Klippel on a spec sheet or in a review, people are going to naturally assume it's been put through the whole test like Patrick at Red Rock or Vance Dickason does, and the published spec is verified by klippel standards.  There's literally no way to interpret the klippel data of the CSS driver by any stretch of klippel definitions to indicate that it has the published xmax spec, which specifically has the word klippel on the spec sheet beside the word xmax.

 

Also in a thread at avs I posted several links where Nick specifically said some of his drivers were Klippel verified and/or mentioned Klippel without saying he only verified Bl with the klippel and the drivers were not actually klippel verified.  A few people were pretty surprised, as they were specifically told the drivers were klippel verified and/or were not told the rating was based on Bl alone.

Link to comment
Share on other sites

I really don't want to rehash the specifics again, but if you put the word Klippel on a spec sheet or in a review, people are going to naturally assume it's been put through the whole test like Patrick at Red Rock or Vance Dickason does, and the published spec is verified by klippel standards.  There's literally no way to interpret the klippel data of the CSS driver by any stretch of klippel definitions to indicate that it has the published xmax spec, which specifically has the word klippel on the spec sheet beside the word xmax.

 

Also in a thread at avs I posted several links where Nick specifically said some of his drivers were Klippel verified and/or mentioned Klippel without saying he only verified Bl with the klippel and the drivers were not actually klippel verified.  A few people were pretty surprised, as they were specifically told the drivers were klippel verified and/or were not told the rating was based on Bl alone.

 

Not sure I agree with this. That when people see that they immediately assume that the xmax and power handling are verified that way, I mean. That system can do everything from quality control, thermal testing, rub and buzz checks, xmax determination, Dc offset, small and large signal parameters, you name it. When I see the K word in the advertising or on a website it usually seems to imply that the system was used to help improve and develop the drivers or do quality control. I don't assume it has a direct connection to the xmax spec or any other specs unless it specifically says it was verified that way with a clarification note. Now if the xmax has / the K word right next to it saying hey look at me then yes I would assume that.

 

Like so from the RF website. Alpine's used to say similar under their Klippel link. Seems they took it down. B&C seems to imply the same. I don't know maybe I'm wrong and it does mean the specs are pulled from Klippel but the power handling at-least is stated to be under CEA-2031-A guidelines for RF.

 

post-5-0-70657700-1459973442_thumb.jpg

Link to comment
Share on other sites

 

 

The irritating part is that the number of drivers I feel confident enough in that the sim will be accurate, is a small pool. Most I've measured myself.

 

If the published small signal specs are garbage there's no hope.

 

But ideally you have measured specs to work with.  In that case, if you apply my simple tweak you will get pretty accurate results.

 

If you want even more accurate results, all you have to do is do what I did to come up with the generic tweak.  Use measured t/s parameters.  Measure frequency response in a sealed box of known size and dimensions.  Sim the box.  Adjust the simulated Bl until the sim matches the measurement.  Then use the new value of Bl for any subsequent sims and the sims will be remarkably accurate.  It works.  I proved it in the paper I wrote.  Simulated sensitivity will be a couple db low but everything else is pretty accurate.

 

I know you have your own methods, but that's mine in a nutshell.

Link to comment
Share on other sites

Not sure I agree with this. That when people see that they immediately assume that the xmax and power handling are verified that way, I mean. That system can do everything from quality control, thermal testing, rub and buzz checks, xmax determination, Dc offset, small and large signal parameters, you name it. When I see the K word in the advertising or on a website it usually seems to imply that the system was used to help improve and develop the drivers or do quality control. I don't assume it has a direct connection to the xmax spec or any other specs unless it specifically says it was verified that way with a clarification note. Now if the xmax has / the K word right next to it saying hey look at me then yes I would assume that.

 

Like so from the RF website. Alpine's used to say similar under their Klippel link. Seems they took it down. B&C seems to imply the same. I don't know maybe I'm wrong and it does mean the specs are pulled from Klippel but the power handling at-least is stated to be under CEA-2031-A guidelines for RF.

 

attachicon.gifRF K.jpg

 

That thing says klippel verified in big letters.  If their drivers are not fully tested and rated on the lowest performing of the 4 major klippel defined categories I'd say they are not being honest at all.  That implies full klippel verification pretty strongly through each step of the design, manufacture and testing process.  If that's not what's happening I would say absolutely they are abusing the klippel defined intent of the verification system.  Klippel probably won't complain as they are in business to sell VERY expensive systems so it would somehow fall to the consumer to figure out if these products are actually verified or whether they are just using these klippel graphics as misleading advertising.

 

I honestly don't know what the intent of that image is from the manufacturer's perspective but it seems pretty clear that they are saying the drivers are verified by all klippel standards.  Otherwise the word "verified" should be removed and it should be made clear that the klippel system was used and the drivers are not actually verified by klippel standards.  If they are using the word "verified" to mean they verifiably own the machine and know how to use it and that's it - I would say that's pretty close to an outright lie by means of being intentionally obtuse and unclear.

Link to comment
Share on other sites

3.)There has been a few remarks about not knowing whether the driver was pushed to the excursion limits during the max long term or burst tests here. EVERY driver has been. There is no question about this.

 

I'm not sure if you are referring to my remark in the SI DS4-18 thread, but just to clear things up, I don't doubt for any instant that your testing pushes the drivers to their excursion limits.  My comment concerned the visibility of Le non-linearities, which may not be revealed very well in your sine-sweep tests.  The reason is that your tests don't measure high frequency distortion *while the driver is simultaneously being pushed to full excursion*.  By the time the sine sweep hits the high frequencies where Le modulation is likely to show up, the driver just isn't being pushed out to its limits any more.  Only with an impractically large amp would it be possible to do this with sine sweeps.

 

I know in the past you've done multi-tone tests, and IIRC, you abandoned them because (1) they didn't necessarily provide more info than the sine-sweep HD tests; and (2) they were very time consuming.  I believe #1 is only mostly true, and I believe my case above represents a clear exception.  At the same time, I totally get #2 above and can't really complain because your testing still yields way more data on many drivers than is available anywhere else.

 

On that note, do you think it would be feasible to do more extensive distortion testing if better software were available to automate this work?  I imagine there are some delays in the measurement process (like waiting for hot voice coils to cool) that can't be avoided.  The beauty of the Klippel system is that it measures the various excursion-related phenomena that contribute to distortion directly and independently.  The resulting data should in principle be a lot more reliable for predicting distortion performance of the woofer in a variety of different enclosure types and with a variety of different content (including real-world content even).  The obvious downside is that the apparatus is just too expensive.  While I don't expect GP measurements could ever give data with the same granularity as Klippel, I think it would be interesting to try to close the gap using better test signals and analytical methods.  I think there's a fair amount of unexplored territory here.

 

All this is ignoring effects on non-linearity and distortion.  This is an area where I think there's an even bigger knowledge gap.  People routinely brag about how much power they applied to a voice coil without killing it, or whatever, but there is a big variable here and that is time.  Some woofers might use high thermal mass voice coils with excellent short-term power handling, but if they have poor heat transfer from the coil, they might easily lose out in a comparison with a driver using a much small voice coil and better heat transfer.  These things are very important because applications vary a lot.  Some people want lots of headroom for occasional transients while others want to beat on a driver continuously with as many watts as they can get away with all night long.  Complicating things further, the relative motion of the coil may have a significant effect on how well heat is transferred away from it, and because of this sine waves might actually be a lot easier on a coil than pink noise at the same average SPL, for which there may be much less cone motion to force air over it and cool it.

 

Anyway, I have more thoughts, especially with regard to simulations, which is an area I've had some experience with completely outside of drivers and audio overall, but they'll have to wait for when I have more time.

Link to comment
Share on other sites

Man, LOTS to catch up on. Spent the whole lunch hour and then some trying to catch up here...

 

Habitual line stepper? Oh, where can I go with this.... Pity I suck at Internet and can't make pictures happen. Maybe if I Klippeled them first??

 

I'm sure I am gonna miss a bunch that has been mentioned since the first post. 

 

Simulations? Guilty. I have done FAR too many. Long, lonely nights in front of the computer... Do I trust them? Yeah, but only as a starting point. They're NOT the same as a measurement. WinISD? Might as well throw darts...can be fine when properly used and configured, but most don't know how or bother. I developed spreadsheets for the basics (sealed, ported, and bandpass) long ago. I've purchased commercial software along the way. Much of what I do is in HornResp, the remainder is in UniBox. Still, the sims all have limits. It's not always possible to put the reality of a cabinet design into the simulation. Measurements of the real thing FTMFW. Predictions are not, and never will be data.

 

Do I match my results to my sim by tweaking the sim? Occasionally, usually when I am trying to figure out what the hell has gone wrong and trying to figure out why what I measured isn't matching what I modeled. It happens more than I'd like to admit. I do computer modeling as a part of my real job, the approaches I have learned there have been carried over to how I do audio sims. Garbage in, garbage out is absolutely valid. A sim can be NO BETTER than and NO MORE ACCURATE than the inputs. Like at my day job, I trust the math, and I measure the input values. You have to trust the smart kids and start with the things you know. Altering things that I can readily measure to make the math match the reality flies in the face of what I was taught to do. If I have to do that to "match" my results, it sort of calls into question the "Trust the math" assumption I made coming out of the gate. It also brings up the point that the system may be more complex in reality than can be simulated with the tools and data that I have at hand. I'm leaning towards the latter as a more accurate explanation.

 

Klippel? Neat tool, but it is only as good as the operator. Is it the be-all/end-all test? Nope. It is a great tool for evaluating a moving-coil loudspeaker driver. I'm a bit of a data junkie, and I prefer more data vs less. If I had to pick between quantity and quality? Quality. I will take less (but accurate and representative) data over more garbage data any time. The problem is knowing which is which, because manufacturers won't always tell you. IIRC, there was discussion about parallel vs series voice coil wiring altering the Klippel results on some drivers, curve shapes were significantly different. It would be awesome if all manufacturers delivered the same high-quality data that accurately reflected the production model of the driver in question. They don't. Odds are they never will. 

 

So what? If you need to know, buy the driver and measure it yourself. T/S parameters? Easy. Motor range and linearity? Harder, but can be done. While not a Klippel, it is entirely possible to measure effective excursion with ARTA and a $75 laser head. I'll be doing this with some drivers later this year, though I expect I'll need a larger range laser for one. My plan is to measure SPL, distortion, and displacement, then figure out when displacement means an increase in distortion. ARTA has a script for this. I have other ideas about how I can test drivers too, things like motor and suspension symmetry. Just need the free time (and reason to spend it farting around with audio). The rest of life has sort of been getting in the way of the hobbies.

 

Testing? A fundamental goal of ANY test I do is to find the limits in play. In many cases for me, it was the amp. No longer the case. Understanding the data and the limits in play are paramount though. Every driver will take a kilowatt, the question is: "For how long?". Ricci's stuff? Not amp limited. High power testing of high-dollar drivers is a freaking tightrope walk. Machine-gunning a coil into the backplate on a $40 woofer is funny. On a $1400 woofer, I'd probably puke. I try not to hurt things, I watch for compression. If more watts are going in and they're not all making more dBs, they're going somewhere as heat, it's not like they're playing hide and seek. Once at this point, adding even more watts will just let the magic smoke out.

 

Sine wave sweeps? I've killed more drivers with sine waves than music and movies combined. High power sine waves at an impedance minima will let the smoke out faster than I can turn things down. The damage is typically done before I even hear the problem. Once I'm running high-power sweeps, I try to minimize damage by sacrificing some resolution and shortening the length of the applied signal. That said, there is no substitute for high power testing that I am aware of. If you want to know how something behaves at high power levels and define the operational limits, build it and test it. Only way to know is to go there.

Link to comment
Share on other sites

That thing says klippel verified in big letters.  If their drivers are not fully tested and rated on the lowest performing of the 4 major klippel defined categories I'd say they are not being honest at all.  

 

That's what I mean though. you are strongly correlating the term verified with the xmax spec. I'm saying EVERY spec and parameter on the sheet would be under the umbrella or not along with, all other sorts of QC checks and other assorted things. Do you have a link to where it is defined in the Klippel literature that if their logo  or "verified" appears on a MFG's site it implies that every spec from the driver/s is supplied via Klippel testing and protocol? I've not seen where if the MFG uses Klippel for driver development that they must adhere to the strict Klippel xmax definition in marketing. I believe you may be applying that term in a strict sense that is not necessarily the case.

 

Back to the simulation subject. My point was the small signal TSP's many times are off or are not detailed enough to suss out what the sim should look like or whether it is accurate. For example a random long excursion car audio driver. Perhaps they give a 1kHz LE figure. Perhaps nothing in that regard. Can you be certain that the results will be on target? The hump is in the right place if there, the FR shape is generally the same, the impedance curve matches etc...I cannot or at least not with enough confidence to design a complex cab around. You almost have to have 3rd party hands on measurements.

Link to comment
Share on other sites

Not sure if this thread is for quick side-topics but I'll take my chances of getting Beast-like infractions. :). Anyway, here's something interesting on Audio Fluidization regarding landslides, which basically means the deep bass from rocks crashing create vibrations, which in turn reduces the friction resulting in longer landslides. This had been predicted in the late 70s but computers weren't powerful enough to simulate it until recently.

 

http://www.engadget.com/2016/04/06/now-we-know-why-landslides-can-flow-like-rivers/

Link to comment
Share on other sites

To both Lilmike and Ricci - my large coil adjustment method works.

 

I developed it by curve fit matching MANY sims to measurements with a very simple adjustment.  And then the cherry on top was when Ricci posted the measurements of several long excursion drivers in Othorn and my method worked remarkably well - the adjusted sims had a much higher degree of accurate correlation to the measurements than normal sims and that's when I knew for sure it really worked.

 

This is not a random tweak that works on some drivers in some cases, it works on every driver (of this type) in every enclosure.  It's not perfect but it's far and away better and more accurate than a regular sim.

 

Absolutely I trust it enough to design a multi thousand dollar sub horn.  I can't justify spending the money right now but if I could it would already have been built - with 4 Sundown 18s.  And it would measure reasonably close to the sim, just like all the 50 sims I compared to data-bass measurements, all the sealed ported and HORN sims.

 

It works.  Simple as that.  Not perfect, not perfectly accurate, but it turns a regular sim nonsense crap shoot into a pretty sure bet.

 

If a tool (simulator) is not equipped to handle a job (simulation of large excursion drivers) but you are able to manufacture a tweak that improves the accuracy 100 percent of the time compared to real life real world measurements, this is empirical evidence, which is scientifically sound.

 

My sims are accurate, that's all I can really say.  I've simulated almost everything measured on data-bass (the diy stuff anyway) I've simulated all kinds of horns, I've even simulated the 12 Dancing Shivas IB and accurately correlated my sims to the upstairs AND downstairs measurements of that IB because I was told simulating an array of woofers in IB couldn't be done accurately.  If I have a high quality measurement I can sim it and it will be at least very reasonably accurate.  What other proof is needed?

 

If there's any doubt read my paper on the subject, which includes all 50 comparisons of the sims and measurements (including horn sims).  Then tell me it doesn't work.  And that's just the generic tweak in the paper, you can get even a bit better if you do a curve fit (same way I came up with the generic tweak).

 

The premise is that these large coil drivers present as having a weaker motor than the small signal parameters indicate.  If you simply adjust Bl your sims suddenly start to match the measurements (although a couple db low in sensitivity).  Always.  100 percent of the time regardless of enclosure type.  it just works.  You won't get a perfect overlaid match but it will be dramatically more accurate than a regular sim.  Read the paper.  Try the method.

 

You really do need accurate 3rd party measurements of at least the t/s parameters if you want to have any faith in anything, but that's always been the case since the beginning of simulations.

Link to comment
Share on other sites

That's what I mean though. you are strongly correlating the term verified with the xmax spec. I'm saying EVERY spec and parameter on the sheet would be under the umbrella or not along with, all other sorts of QC checks and other assorted things. Do you have a link to where it is defined in the Klippel literature that if their logo  or "verified" appears on a MFG's site it implies that every spec from the driver/s is supplied via Klippel testing and protocol? I've not seen where if the MFG uses Klippel for driver development that they must adhere to the strict Klippel xmax definition in marketing. I believe you may be applying that term in a strict sense that is not necessarily the case.

 

If they are advertising "verified", yeah, I think the whole process should be under the klippel verification umbrella.

 

I'm sure klippel doesn't care though, they sell the multi thousand dollar machine and with it the rights to use their graphics and "verified" status.

 

There's a matter of ethics here.  If I purchase a klippel machine and cobble together some junk that doesn't pass any klippel standards in my garage and throw that logo on it, is that ethical?  if you advertise klippel verified it should mean something.  In this specific case of Alpine and Rockford I don't have any information on what it actually means but I know what it should mean from an ethical standpoint.

Link to comment
Share on other sites

Sorry not to be technical here.  But Chapelle is the funniest TV show I have ever seen.  No boundaries.  The bonus interviews with Charlie and Rick James make Chappelle a 'must buy'.  you can see that the truth lies somewhere in between Charlie's and Rick's recollections....those must have been some wild-ass times.

 

Only time I have laughed as hard is with some ATHF episodes and 'The gang goes camping' episode of It's Always Sunny... 

 

JSS

Link to comment
Share on other sites

Sorry not to be technical here.  But Chapelle is the funniest TV show I have ever seen.  No boundaries.  The bonus interviews with Charlie and Rick James make Chappelle a 'must buy'.  you can see that the truth lies somewhere in between Charlie's and Rick's recollections....those must have been some wild-ass times.

 

Only time I have laughed as hard is with some ATHF episodes and 'The gang goes camping' episode of It's Always Sunny... 

 

JSS

 

ATHF?

 

This season's first episode of Sunny had be crying with the Chardee MacDennis and main-lining white wine.... Those dudes man.

Link to comment
Share on other sites

Not sure if this thread is for quick side-topics but I'll take my chances of getting Beast-like infractions. :). Anyway, here's something interesting on Audio Fluidization regarding landslides, which basically means the deep bass from rocks crashing create vibrations, which in turn reduces the friction resulting in longer landslides. This had been predicted in the late 70s but computers weren't powerful enough to simulate it until recently.

 

http://www.engadget.com/2016/04/06/now-we-know-why-landslides-can-flow-like-rivers/

 

Very cool. That's exactly what this thread is for.

There's a pm group of about 6 of us that's probably been going on for a couple of years and the topics are all over the place but there has been some really good technical discussion as well that got some of us thinking that it was a shame it wasn't public. Figured it might be time for a thread for whatever.

Link to comment
Share on other sites

To both Lilmike and Ricci - my large coil adjustment method works.

 

 

I don't doubt the method. I checked it and read the paper when you sent it to me. It works quite well. Not 100% dead nuts but very close and worlds better than not using it. It's a good thing somebody took the time to verify and easy button it. I give you props there.

 

I still hesitate to trust it for complicated high order alignments, with unknown drivers, where you are using your best assessment and judgment from the info given. That's just me though. I wonder have you simulated a driver like the old MTX 9515 or the LMS 5400 and used it? I have not. Those are outlier drivers that despite having huge coil mass behave close to expected without the "large coil" tweak. Of course those types of drivers are few and far between.

Link to comment
Share on other sites

I'm not sure if you are referring to my remark in the SI DS4-18 thread, but just to clear things up, I don't doubt for any instant that your testing pushes the drivers to their excursion limits.  My comment concerned the visibility of Le non-linearities, which may not be revealed very well in your sine-sweep tests.  The reason is that your tests don't measure high frequency distortion *while the driver is simultaneously being pushed to full excursion*.  By the time the sine sweep hits the high frequencies where Le modulation is likely to show up, the driver just isn't being pushed out to its limits any more.  Only with an impractically large amp would it be possible to do this with sine sweeps.

 

Wasn't referring to your remark. I had meant to respond to it but forgot.

 

I agree that this is something that would be very intriguing to look at. It's been discussed a few times years ago. I even tried a few things at home but couldn't work out a feasible, quick way to get it done with what I had to work with. I don't think the spectral contamination the way Illka or myself tried to do it was a good way for a few reasons. Firstly generating the spaced individual tones as part of one signal took some work with NCH tone generator. It'd be best to work it into a wav file sample that could be triggered. That wouldn't be too hard. Most of the RTA or spectrum analyzer settings I looked at required quite a bit of time to settle in and capture the content with high resolution. The signal was murder on the systems. At least it sounded like they were dying a quick death. Even at levels well below maximum the spaced non harmonically related tones sounded awful. I couldn't really tell too much between different systems using it when looking at the recorded spectrum. In short difficult to setup, took time, caused a ton of heat in the voice coil, didn't clearly show much difference visually and made me fear damage to the drivers worse than any other signal I've ever used.

 

A lot better option would be to use a sine tone to produce large driver excursion and then run pink noise or perhaps a sweep through and capture it that way. I'd lean towards PN on this I think. Logistically though there are a lot of questions. How do you standardize this? Each system has a different native FR such that some frequencies are emphasized more than others. Do they get EQ'd flat? This alone becomes a ton of work. Let's assume no. What excursion level is the driver pushed to? That is difficult to standardize. Do we settle on a normalized voltage versus impedance? What frequency of sine wave is used to generate the driver excursion when impedance varies widely as does the resonance of the system and excursion profiles? 40v into one driver might cause 20mm with a 15Hz sine while another might be at only 12mm. I'd guess that you would want multiple measurements with the driver excursion pushed to various levels. I'd like to see how things change at higher excursion versus lower as would everyone I assume. What level is the PN or sweep signal for the response shape relative to the sine signal producing driver excursion? Equal? Do they both get scaled by the same amount for each measurement?

 

This is the kind of stuff I start thinking about with new measurement types. It has to be relatively quick, easily setup (relative to the gamut of other stuff already being done), intuitive to interpret the results and standardized enough that the results from various drivers can be compared meaningfully. If it takes a dedicated reconfigure of the measurement gear or an hour and a half to complete it for one system I'm just not going to be able to include it. I'm all for doing something like this but would need some help figuring out the best way to tackle it and standardize it.

Link to comment
Share on other sites

I don't doubt the method. I checked it and read the paper when you sent it to me. It works quite well. Not 100% dead nuts but very close and worlds better than not using it. It's a good thing somebody took the time to verify and easy button it. I give you props there.

 

I still hesitate to trust it for complicated high order alignments, with unknown drivers, where you are using your best assessment and judgment from the info given. That's just me though. I wonder have you simulated a driver like the old MTX 9515 or the LMS 5400 and used it? I have not. Those are outlier drivers that despite having huge coil mass behave close to expected without the "large coil" tweak. Of course those types of drivers are few and far between.

 

Both the 9515 and the LMS were measured by you and were included in the study.  Both of them simulated closer to your measurements with the adjustment than without.

 

In addition, the LMS was measured by you in both the sealed box AND the Othorn IIRC, and in both cases the simulations were closer to the measurements with the adjustment than without.

 

Sure, sims with those drivers aren't incredibly inaccurate even without the adjustment but the adjustment does make things closer to reality.

 

Just for the record, normalized inductance (Le/Re) for a few examples (t/s measured by you):

 

Dayton UM 18 = 0,36

Dayton HO 18 = 0.4

MTX 9515 = 0.4

LMS 18 = 0.94

UXL 18 = 1.63

FI SSD 18 = 3.07

 

In general the ones with the higher normalized inductance are the ones that deviate the most from the normal (unadjusted) sims.  Although it's not totally linear there is a pattern, this is the charted data for normalized inductance (Le/Re) vs Bl multiplier (the amount I had to adjust Bl in the sim to make the sim match the measurement).

 

35co94g.png

 

Looking back at the paper, the sample size was 22 drivers (every driver of this type measured by data-bass), some of which were measured in multiple enclosure, so there were 30 measurements.  (Not 50 like I said before, not sure why I thought that.

 

A larger sample size would help to make the generic adjustment method better, because as you can see the dots on the chart don't follow the line perfectly but you can see a very distinct trend.  And the trend line is mathematical equivalent of the generic adjustment formula.

 

Even better results are had if you don't use the generic adjustment but use the curve fit method for the particular driver (the method that was used to derive the generic formula).  This requires measured t/s parameters and a sealed box measurement though.  But if you have those you just adjust Bl until the sim matches the measurement and that's the Bl value you should use for subsequent sims with that driver.

 

So yeah, the LMS and the 9515 and all of the rest of the 22 driver samples all simulated closer to the measurements with the adjustment than without, in sealed, ported and horn enclosures.  And yeah, I trust the method enough to design horns.  I'd probably use the curve fit method  with measured t/s and a measured sealed box FR (not the generic formula) for a big expensive horn, but the point is that I trust it.  I'm not saying the method is perfect but it's close enough to design with and much closer than not using it at all.

 

Without measured t/s parameters at the very least, forget it.  If you don't have at least that you probably won't get a good result from this type of large excursion driver no matter what you do.

Link to comment
Share on other sites

Archived

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


×
×
  • Create New...