AVS Forum banner

More on INDUCTANCE

6K views 117 replies 25 participants last post by  head_unit 
#1 · (Edited)
I've taken some measurements of one of my Infinity 1260ws and didn't really want to throw it in the HzHorn thread, even as gracious as BTH is. If you want the backstory, read posts 224 through 330 in that thread: HzHorn The TLDR would be: diy speaker guy realized that higher-inductance drivers, when measured, don't match the performance predicted in their models. The real issue is that most modeling apps like WinISD or even HornResp don't use a detailed inductance model. dsg found a way to get more accurate results by adjusting Re and Power, and LTD02 countered by suggesting we adjust Bl, Qes/Qts, and Le. Tweaking the TS params in this way gets modeled results much closer to the measured ones.

With that said, I've got measurements of an Infinity 1260w to share.

Here's my FR and Impedance sweeps and the calculated TS parameters. FR sweeps are close-mic'd at about 1/2" with a UMIK-1. Measured Qtc of the driver in the empty, sealed 38.6L box is .977.





Now, here's the WinISD models. All are in a 38.6L box, Qa=100, Ql=50.
The green line is Infinity's published TS parameters. Qtc of the driver in the enclosure was 0.7.
The blue line is based on my measured TS parameters. Qtc of 0.892.
The purple line is my measured parameters with BL adjusted to 14.5, Le to 6.5 mH, and letting WinISD recalculate Qes at .686 and Qts at .619. I manually entered all the mechanical parameters, i.e., Sd, Vas, Fs, Qms, Cms, Mms, so they wouldn't change. Modeled Qtc becomes 1.092.



And here's the purple line overlaid on the measured FR.



And for reference, a link to Infinity's published TS parameters: http://eu.infinitysystems.com/tl_fi.../Reference/Reference/1260w/REF1260W_PI_EN.pdf

Seems like this jives wih what dsg and LTD have been saying. I basically fiddled with BL and Le in WinISD until the peak of the curve as well as the points at 10 or 20 Hz, as well as 100 or 200 Hz, matched the curve of the measured FR. The higher end of the impedance curve (higher freq than what's posted above) and the Qtc aren't exactly on the money, but are much closer than they were.

When I get a chance I'll be measuring my HT18 D4s.

Comments? LTD or anyone else, did you have impedance-compensated TS params for any of the common drivers that you wanted to post?
 

Attachments

See less See more
8
#2 · (Edited)
Just a few comments.

1. I'd rather see a 1m (or 2m or 10m) measurement done outside well away from boundaries than a close mic measurement for this. Close mic is ok but it's not really the same.
2. I'm pretty lazy and pretty busy at this time of year, I still haven't tried LTD's method of adjusting Bl and Le. I've only ever done Re and Eg (power) to gain back the losses from the added Re. I still have to go through LTD's method and then go through ALL the measurements again to correlate to the different measured designs - all the high inductance sealed designs at data-bass, the single ported design at data-bass, Submaximus, and Ricci's high inductance Othorn measurements with UXL and ZV3 (which were also close mic). Finding the right parameters with sealed box measurements and correlating (matching) sims to measurements of the more complex enclosures like horns is key. If you can do that you know you have a really useful, accurate and powerful tweak to work with.

Just typing all that seems like a lot of work, not sure when I'm going to actually get any of that done. I'm really interested to see how much more accurate LTD's method is. Adjusting Bl instead of Re seems to make sense, not sure how important it is to adjust Le as well, and I'm interested to see how all this correlates to high quality measurements, especially the Le adjustment part. I already know the Re and/or Bl tweak alone gets pretty close to measured results.

If anyone else has HIGH QUALITY accurate and repeatable measurements of ANY high inductance driver in ANY enclosure (especially more complex enclosures like ported, tapped horn, front loaded horn) it would be nice to have more to compare to. Enclosure details are necessary, more detail is better than less, exact dimensions would be nice (not just interior volume) as well as measurement details (power input, mic distance, etc) and measured t/s parameters. It's also nice to see measured impedance of the cab, the inductance hump(s) should correlate pretty well to the impedance spikes.

Nice to see this concept catching on, there's way to many garbage sims presented of high inductance drivers that are assumed to be accurate.
 
#4 ·
I would model it like this with Unibox:
As previously discussed that's only going to work with enclosure types that Unibox can model, which isn't many. The inductance correction method I came up with works with any simulator, any driver, any enclosure.

And where did you get the Le2 and Re2 values from? I don't see anything in his measured parameters that says 4mh or 5 ohms. Have you verified that using Unibox for this is accurate and repeatable over a variety of different enclosures with the same driver?
 
#9 ·
Well, REW generates parameters that are quite accurate. It even spits out a very high fidelity inductance model and one that's not quite as good (but still very good). Unfortunately, there is no enthusiast level software that can utilize either of the models. In the case of my Unibox "tweak", I'm effectively creating a lower fidelity inductance model (still much better than ignoring inductance).
 
#6 ·
I am willing to mic my sealed ftw-21s ,but am not going to move them out of the basement to do it. So close mic is the best? 1" or what? And how much power? Does inductance increase with more power (seams like it would)?

Interesting to see how e.q. could be used to try and counteract the effects of inductance (if needed or desired). Though probably deminishing returns applies and each driver may react differently.

Can someone post links to inductance background and research? Maybe it has been shared, but I can't find it...
 
#8 ·
The measurement distance depends on what the actual effective radiating radius of the driver is, so measure across the cone from the half way point of the surround. It will likely be around 18"/2 9". The measurement distance should be 0.10 times this number, so likely .9". This will provide an accurate measurement from 10hz up to 4311/ the radiating radius so likely 479 hz.

It is completely possible to get accurate quasi anechoic in room measurements from 10 hz up if you know what you're doing.
 
#15 ·
And this flurry of speculation, opinion, and going off half cocked is exactly why I was happy to hide in the Hzhorn thread with a small audience to discuss this.

I hand you guys an inductance correction method on a silver platter and get nothing but complaints. This has been going on for months now every time I bring this up. Finally I got LTD on board, and that's most likely only because Ricci backed me up on this.

Come on guys, do a bit of reading before you start complaining. This really works, and it's endorsed by Ricci and LTD02.
 
#18 ·
For those that don't want to read all the suggested reading just read this post.

http://www.avsforum.com/forum/155-d...asses-one-sub-126db-16hz-23.html#post31202121

This is what I posted back in October on diyaudio, and reposted in Popalock's build thread in January. This info has been out there for months. LTD02 recently helped improve my method a bit and it looks like he's getting even more accurate results than I was. I have yet to try his method, it's just a bit different but early results suggest it's a bit better. For those details you have to read the suggested reading mentioned in post 1.
 
#33 ·
For those that don't want to read all the suggested reading just read this post.

http://www.avsforum.com/forum/155-d...asses-one-sub-126db-16hz-23.html#post31202121

This is what I posted back in October on diyaudio, and reposted in Popalock's build thread in January. This info has been out there for months. LTD02 recently helped improve my method a bit and it looks like he's getting even more accurate results than I was. I have yet to try his method, it's just a bit different but early results suggest it's a bit better. For those details you have to read the suggested reading mentioned in post 1.
dsg's method of adjusting re was really the solution--all this comes from that. at some point, it struck me that by increasing re, motor strength, which is bl^2/re, would be reduced. that was my personal "ah ha" moment because it tied together the apparent underdamped response that high inductance drivers present when measured relative to non-adjusted models.

i simply tried to adjust the numerator and it worked. since the numerator term is squared, adjusting it down by the square root of 2 provides the same "motor detuning" as doubling re.

the advantages of adjusting bl are two fold. first, if the physical parameters of the driver are entered as hard values in winisd (i.e. enter compliance, cms, sd, etc. and let the program calculate qes and qts), then the actual change in qts (damping) can be observed. the second advantage is the sensitivity automatically adjusts because, of course, sensitivity is a derived calculation from bl (in part).
 
#23 ·
Most simulation software uses a simple inductance model, and does not adequately model the effect of inductance on many drivers. However, it has been observed that high driver inductance effectively weakens the motor of the driver. While we can't exactly plug a complex inductance model into WinISD or Hornresp, we can modify other TS parameters of the driver to reflect the weaker motor.

Nobody is modifying measured data. The measured data is the impedance and FR curves I posted from REW above. The goal is to make the models match the measurements.

Yes, doing this driver-by-driver like the OP is not solving the general case by itself. But what I'm hoping we can do is gather more data and figure out the general case. Look back at all the work done by Thiele, Small, and others, it at some point became a question of measuring drivers, boxes, and then using that data to derive the equations and relationships which we use on a daily basis.

Everything that we do in software someone had to discover the hard way. It's not like inventing a flux capacitor, where all you have to do is hit your head on a toilet. We have to test, measure, and analyze over and over again.
 
#29 ·
Yes, it's been a known issue for as long as high inductance drivers have been produced.

That's why I was so excited to have a consistent and (fairly) accurate tweak to quantify this behavior. I'm a horn guy, I can't go building massive horns with high inductance drivers and just completely ignore the fact that a hugely expensive project is going to have completely unpredictable behavior.

My next project is looking to be a 125 cubic foot front loaded horn and I can't source or afford TC drivers and pro drivers don't have the linear excursion capabilities I want, so I'm looking right at ultra high inductance drivers, and I absolutely need an accurate sim before laying out $3000. (4x high excursion 18 inch drivers and at least a dozen sheets of baltic birch.)

I'm not saying you were not aware of the issue 2.5 years ago, I'm just saying that if all you have done is tweak the dials so a single sim matched a single measurement, that's not a proven consistent and accurate method that will work with all drivers in all enclosures. It might not even work with that specific driver in a different enclosure. You don't know until you prove it will work.
 
#30 ·
I thought the HT18s were not high inductance? On the border line but not in the same league as these other drivers. Data-bass measurements don't appear to show this behavior with the HT18. Ricci's overview even states the driver could be used as a woofer with good response up to 400hz.

Copied this from databass (if that's not allowed sorry!)

SI18d2:

"Measurements of this system are pretty good with no major issues. The shorting rings in the motor system and a modest voice coil mass compared to a lot of other drivers combine to keep inductance related impedance and response shape effects under control. The response shape as a result is very smooth and will not require much shaping with linear response up well past 400Hz and down only 3dB at 165Hz. Indeed it appears that this driver could be used as the woofer in a three way crossed over as high as 350-400Hz."
 
#31 · (Edited)
I thought the HT18s were not high inductance?
If it's a high inductance driver (Le:Re ratio 1:1 or higher) and it's measured in a sealed box and the level at 200 hz is not higher than at all the lower frequencies (in other words if you see a hump caused by high q near the impedance peak and response droops off on both sides of that hump), then you are seeing high inductance effects. (It could also be a high q AND high inductance driver, but then it would just be a generally pretty crappy driver.)

The HT18 isn't nearly as bad as some of the others like the HST or the ZV3 or the RE XXX, but it is a high inductance driver and as such it's sims will not match it's measurements.

Take a look at any of the pro drivers on the site, they all have a rising response in sealed boxes, 200 hz is way louder than any of the lower frequencies. If you are seeing a bump in response anywhere below 200 hz it's either a crazy high q driver in a box that's way too small or it's inductance causing it. (Inductance causes the same effect as high q, that's why tweaking Bl or Re work to consistently and accurately simulate the high inductance drivers properly; tweaking those parameters effectively tweaks q.)
 
#32 ·
@rhodesj


just saw your thread. very nice work! i wish there was a way to hit the like button a few more times!


maybe there is some sort of rule of thumb that can be worked out for how much to de-tune bl based on the relative measured le/re, but at this point it is at a minimum worth keeping in mind.
 
#34 · (Edited)
In UNIBOX you can put in additional values for parallel Le and Re. Those values are never, to my knowledge, published, but i have seen, a few years ago, a comparison between UNIBOX sim and an anctual box posted by this guy Illka that was once famous on the forums. They lined up pretty much perfect. If there ware a way to measure Le2 and Re2 before sim in UNIBOX I think that would be a way forward as opposed to fiddling with TS to make the sim fit the real box.

I believe it was here. Unfortunately the graphs are missing..

http://www.hometheatershack.com/for...-design-software-produces-best-results-2.html
 
#37 ·
In UNIBOX you can put in additional values for parallel Le and Re. Those values are never, to my knowledge, published, but i have seen, a few years ago, a comparison between UNIBOX sim and an anctual box posted by this guy Illka that was once famous on the forums. They lined up pretty much perfect. If there ware a way to measure Le2 and Re2 before sim in UNIBOX I think that would be a way forward as opposed to fiddling with TS to make the sim fit the real box.
I believe TC Sound's reported Lp and Rp are the same two parameters used by Unibox. I haven't seen any other manufacturer publish the 3 element model used by Unibox for inductance though. Sundown publishes a Lp, Rp, and Cp which is not the same model. That's the conventional motional impedance model (without the frequency-dependent damping).

Maybe we can ask John M (author of REW) to support more inductance models, specifically the Leach electrical inductance model used by Unibox.
 
#63 · (Edited)
I dont think its either or.

In the proposed method you actually have to build the box and measure it, and then fiddle with TS to make the sim match the real box. Thats kind of backwards. Has it been shown that the new TS is also valid in an other box? For instance Sealed > reflex? Small sealed > large sealed? Small sealed > Horn? If that can be shown, then you could have a small(ish) measuring box and use that to generate better TS for use in an other more complex build. If not, I completely fail to see the point?

The other method is to measure Re2 and Le2 before building anything, sim the data and then build a box that matches the sim. That have been show to work, at least for a sealed alignment and presumably other boxes as well.
 
#36 ·
Sounds more like the specs are right and the modeling software is wrong. You don't fix the software by changing the scientifically measured specification to match the model. You fix the software so it more effectively matches both the scientific measurements and real world results. Call me crazy but it all seems ass backwards.
 
#42 ·
I'm very intrigued by this inductance fiasco. However I think I agree with what Mfusick is saying. I keep waiting for someone to say, "If you adjust re to re/square Rt 2 then your model will more accurately match the measurement for high inductance drivers."

It may be too early for that but, telling me to model a driver then build it and measure it so I can adjust the specs to remodel and then build another box seems counter productive unless we can get some coorelations across multiple drivers. Maybe that's the goal. IDK, but this thought just makes me doubt modeling software's accuracy and to some extent relevance.

I'm glad to see this gaining traction. Even in Stereodude's thread from a couple years ago, it seems there is a resistance to this phenomenon, which I don't understand.
 
#48 ·
I haven't seen that so much recently. At least not with the horns. I think people want to figure out how to use high inductance drivers with horns. Save some cash.

I have two LMS. They work nice. Hard to get though, and expensive. I'd totally love to be able to use alternative drivers with as good results. Problem is modeling never jives with the reality- and that is a problem.

I'm all for building and testing enclosures too. I'm weird like that. I like building them. I don't mind burning something and starting over either. Well,... perhaps a little ;)

But you get the point. There is a decent community of folks that would like to see this problem solved. My viewpoint though is the modeling needs to account for the reality and effects better, not the other way around. Reality is reality. Measurements are measurements. Can't change those. So what is left to change? Change the model. If the model doesn't match reality then the model is wrong- not the other way around. That's not going to change. EVER.

But I get what some of the guys are saying- and the idea of working withing whatever is available to you. Some mental compensation and awareness of the effect is a good starting point until the modeling can be made more accurate. I don't see this thread getting into a flame war though.
 
#53 ·
The procedure I used for ALL the high inductance drivers was to double Re and add power to compensate for the losses. This will get pretty close but it's not 100 percent accurate.

You could probably ask LTD02 for the Bl+Le tweak to get a bit more accurate results. I haven't tried his method yet, I don't have time to go over it and correlate all the sims to measurements.
 
#55 ·
Probably, but you would have to create a whole simulator and then add the tweak.

It would probably be easier to create a whole simulator that accepted complex inductance parameters instead. Then there's no messing around, assuming the complex inductance parameters give a really accurate result. I don't know if that is the case or not, I don't have a simulator that will accept them to test.

The tweak is easy though, a lot easier than writing software for a brand new simulator.
 
#57 ·
I found some more information for those interested in the more complex models for voice coil inductance. REW can extract parameters for the so-called "FDD model", as described here in its help files. Unfortunately, the two AES papers referenced in the help are not free. So I looked around and found two free papers about the FDD models, one describing the model and the other describing how the model data are fit. The author is Thorborg, one of the authors of the AES papers referenced in the REW help. The free ones are here:

www.bnam2012.com/papers/Thorborg_31.pdf

www.bnam2012.com/papers/Thorborg_32.pdf

I haven't read them completely yet, but they do seem to give very clear explanations and graphs, suitable for hardcore DIY folks, showing the inadequacies of the standard model and how their new model addresses them.

SPICE could be used with Laplace controlled sources for the semi-inductor, but I dunno how to use SPICE with horns though. Leach has an article here about how to use SPICE for loudspeaker analysis, and I have some examples on my site here using the freeware LTspice, showing a simplified vented-box model that lumps all the box losses into QL as Small does. That particular simulation neglects inductance but could be modified using the Laplace sources as described above, assuming LTspice doesn't choke on the Laplace expressions.
 
#61 ·
I think models for subwoofers are overrated. Rooms change things so much that who cares? I only looks at power vs excursion these days, to make sure a driver isn't likely to self-destruct in a given box powered by a given amp. (And, in the case of any resonant system - vented, bandpass, etc. - out-of-passband resonances. But i don't generally do resonant systems, because multiples gets efficiency reasonable with simple closed boxes.)

But I've also found that drivers with high normalized inductance (Le/Re) just don't sound good. Even after you equalize out the upper bass hump. Discovered that years and years ago with Janowitz's old TC Sounds-made HE15. (Massive throw due to huge overhung coil, no inductance control.) Maybe with some frequency-dependent delay or something they can be made to work. For movies, not a big deal perhaps. As long as it gives that boom, it's fine.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top