Home Theater Forum and Systems banner

1 - 20 of 33 Posts

·
Registered
Joined
·
162 Posts
Discussion Starter #1 (Edited)
My Sonotube build and measurements

I recently decided to upgrade my subwoofer and, because of the complete lack of measurement data for the popular drivers, I thought it might be useful to post some of the measurements I'll be taking along the way.

I was looking for a 15" driver and selected the SDX15 (mainly because I live in the same town as Creative Sound Solutions so it is easy to return if not suitable). Besides, Bob is a good guy. :) Whether I end up keeping this driver remains to be seen, I am interested in seeing how other drivers perform also.

Simulations give a notional idea of the driver performance but in practice they aren't as useful as most people believe due to key parameters being omitted or incorrect. So, upon getting my SDX15, the first thing I did was measure its T/S parameters. Below is the published parameters (yellow) compared to my measured parameters (red):

TS_parameters.JPG

Clearly my driver is stiffer than the published parameters with significantly more inductance.

Most manufacturers publish inductance value at a specified frequency (usually 1kHz). However driver inductance is a function of frequency (among other things) and not constant. The value at 1kHz, while common, is of little value for a subwoofer. Inductance has a strong effect on driver behaviour. Accurate results need a more complete model consisting of the Le/Re and Le2/Re2 pair. Praxis will calculate the Le2/Re2 values but they are rarely published. Unfortunately Praxis measures Le at 1Khz which further limits the modelling. Personally, I use ARTA/LIMP since it calculates a least squares curve fit to the three free variables of Le, Le2 and Re2. This usually gives the most accurate results.

The difference between using the fixed Le=1kHz value versus a curve fit Le can be seen in WinISD (which unfortunately ignores Le2 and Re2) as follows. (NOTE: The red color for my measured results wasn't very visible so I changed it to Cyan for this plot):
WinISD_Simulated_FreqResponse.JPG


Unlike WinISD, Unibox can model the Le2 and Re2 values. This usually gives the most accurate picture and the result can be seen with the blue frequency response curve below:

Unibox_Simulated_FreqResponse.JPG


This curve is starting to look quite different than the smooth, flat textbook simulation. And more like the actual results Ilkka found when he measured this driver (see post #2 in this thread: http://www.hometheatershack.com/forums/subwoofer-tests-archived/8158-diy-css-sdx15-sealed-100l.html ). NOTE: Ilkka's data doesn't keep dropping above 100Hz (and in fact starts to increase) due to cone breakup resonances around 500Hz (more on that later).

Anyway, that's more than enough for now. This may end up being the worst build thread ever...but if there is enough interest on the measurement side of things I'm happy to keep measuring and posting the results.
 

·
Registered
Joined
·
162 Posts
Discussion Starter #4
I haven't broken in the driver yet, the plan is to do a before and after test. It's got about an hour of free air excursion on it for this test.

If people do their own measurements they should realize that T/S parameters can vary depending on several factors. Even a broken-in driver will measure differently after sitting versus having just been used. For repeatable measurements one has to exercise the driver in a consistent way prior to measurement. Ambient temperature makes a difference too as does internal voice coil temperatures. The parameters vary with drive level and one needs to choose a power level to get the suspension moving yet not so strong as to cause temperature related issues. These tests were driven at 1.5v RMS and room temperature.

The driver needs to be firmly mounted and suspended in space so it's impedance isn't affected by nearby objects. The added mass can't vibrate or resonate with the cone, so the choice of weight makes a difference as does it's placement. And if one uses a computer with Praxis or ARTA then they have to be very careful about ground loops or return resistance throwing off the measurement. My test leads are 3meter 10 gauge wire.

The chose of stimulus has a strong effect on signal to noise ratio and therefore repeatability of results. The logarithmic swept sine that Praxis uses is quite good but unfortunately the demo version limits this to 500msec chirp. This is too short for serious use (I would use typically 10 to 20 times that length for significantly greater signal to noise ratio) but it isn't so bad for DIY usage. ARTA uses a stepped sine stimulus with excellent signal to noise albeit slow. I've never used an impedance tester like WT3 but unless it can output enough power to drive a big woofer like this and uses a good stimulus I would expect the results would be quite variable. Maybe someone who has used them can comment.

Anyway a lot to get right for good measurements. I sometimes wonder how carefully the manufacturers do it.
 

·
Elite Shackster
Joined
·
7,142 Posts
Anyway a lot to get right for good measurements. I sometimes wonder how carefully the manufacturers do it.
This may be of interest then. I recently took delivery of a pair of brand new AV-15X's. With each driver comes a sheet with a set of actual parameters on them. AE also say they test each driver prior to shipping, and do it enough to have run in the drivers a little for you. So the actual TSP's should be pretty bang on.
 

·
Registered
Joined
·
162 Posts
Discussion Starter #6
Yes, probably the difference between making your own drivers and having someone else make them. Those AV-15's are definitely interesting drivers. I'd love to see a close-mic measurement of one in a sealed box for both frequency and distortion. Or test one myself in direct comparision to the SDX's.
 

·
Registered
Joined
·
162 Posts
Discussion Starter #7 (Edited)
Here's complex impedance plots of the driver mounted in 155L Sonotube. Perturbations in the plots are always audible and need to be addressed. The first plot is unstuffed and the second is after filling with bulk polyester fibre.

First the unstuffed sonotube:
Impedance_UnstuffedSonotube.JPG

I've marked the key issues. Some are due to the driver and others the enclosure. The red lines indicate half pipe resonances at multiples of 170Hz due to the 39inch length of the Sonotube. These (especially the lowest one) are audible. Modes this strong will be clearly visible in the frequency plot. Filling with 100% stuffing should fix those.

The blue line indicates the area of cone breakup. This is a driver problem (all drivers will do it to varying degree) and the plot indicates fairly strong breakup modes. Once again, this will be clearly visible in any frequency plot. It would be nice if the mode was less strong and/or higher but it is satisfactory given this is a subwoofer. However I would not be tempted to use a high crossover point.


For comparision, here is the same enclosure stuffed with about 4.5 lbs of bulk polyester.
Impedance_StuffedSonotube.JPG

The stuffing has addressed the resonances nicely. Only the driver breakup modes remain and these can be addressed by the crossover (assuming at least a standard 24dB/oct slope).

It's interesting to note the system resonance has dropped from 31 Hz to 28.8Hz. Stuffing has increased the apparent enclosure size by about 25%.

At approximately 6 ohms typical impedance, this driver should not be a problem to drive. In fact two in parallel should not pose a problem for a 2 ohm capable amplifier.

The smoothness of the curve indicates an an overall well-behaved driver.
 

·
Elite Shackster
Joined
·
7,142 Posts
Yes, probably the difference between making your own drivers and having someone else make them. Those AV-15's are definitely interesting drivers. I'd love to see a close-mic measurement of one in a sealed box for both frequency and distortion. Or test one myself in direct comparision to the SDX's.
There is a member on here called Russ.will. He's just built one and usually jumps at the chance of some testing :T

FWIW, a recent convo with some fellow purchasers revealed the final TSP's of the AE's might not be as accurate as they used to be, although not that far out. After run in they reckon you get something in the order of an additional 10-15% performance out of them, but dont quote me on that one just yet :D.
 

·
Registered
Joined
·
373 Posts
Those are interesting measurements bjs. It is one thing to be told about internal resonance and how to deal with it, but seeing is believing. I expect that you could get away with less of a higher density material like OC 703.

I'm looking forward to a second set of T/S measurements to see if break in made any difference.
 

·
Registered
Joined
·
162 Posts
Discussion Starter #10
Here are nearfield frequency reponse measurements. The microphone is 3cm from the dust cap. The result is the same as an outdoor ground plane measurement except the absolute SPL is arbitrary. There is no smoothing of the response.

First is the unstuffed plot showing the effect of the sonotube resonances on the frequency response (indicated with red arrows).

Frequency_UnstuffedSonotube.JPG



Next is the frequency response in the stuffed sonotube. Residual effects of the 170Hz tube resonance is still visible. Some further experimenting with stuffing density and/or material may be useful. All higher order modes have been adequately damped and are no longer visible.

The remaining irregularities are due to the cone/surround. Resonances begin augmenting the response above 100Hz with major breakup beginning soon after. Clearly this is a subwoofer, not intended for the higher frequencies (like th AV15 is purported to be). However, until actually measuring other drivers I would not assume they are much (if any) better. Subwoofers, in general, are not designed to be used into the mid-bands.

Frequency_StuffedSonotube.JPG


Finally, below is the same stuffed sonotube plot but with a more conventional frequency range from 10 to 200Hz.
Frequency_StuffedSonotube_10_200Hz.JPG

This can now be compared to Ilkka's measurements with excellent agreement. As mentioned my plot has a slight residual 170Hz sonotube resonance (1dB). Also my enclosure is more than 50% larger than Ilkka's so my frequency response hump is lower and the low bass more extended. The ripple around 14Hz is not present in Ilkka's test and could easily be artifacts in my measurements. Given the frequency and amplitude (+/- 0.5dB) it is not enough of a concern for me to pursue.
 

·
Registered
Joined
·
2,431 Posts
Good stuff Bjs. There is not enough of this going on. I did some similar tests on my SDX's and never really posted much of them. I'll see if I can dig up my WFT3 data. I've since stopped using it as the ARTA method is much better IMO for big powerful subs. Anyway I eat up stuff like this so by all means continue to do things like this. Since you are in Canuckistan :D How about finding an IXL18 or 15 to play with for a few days?


My close mic'd SDX graph looked similar. There is some inductance roll off.
 

·
Registered
Joined
·
162 Posts
Discussion Starter #12 (Edited)
It'd be nice to test other ones for sure. With Illka's tests and mine the SDX15 is pretty well covered but these others aren't documented at all (AFAIK). People buy them based on marketing and faith...hoping for the best. The SDX15 at least is well documented. I'd like to measure up an IXL15...or Tempest-X2...or AV15 in comparison.
 

·
Registered
Joined
·
162 Posts
Discussion Starter #13
Time to look at power compression.

Illka's tests show very little compression until his final trace which hit Xmax. However they also show the peak in the frequency response moving around depending on drive level (from 36-44Hz). A peak in itself is not a big problem, it can be EQ'd out but not so when it is excursion dependant. For someone interested in accurate reproduction as I am, this poses a problem.

For those reasons, I was quite interested to see if the surround and spider updates in the current version affected this behaviour. Here are my results:

PowerCompression.JPG

These were taken with the mic 12cm from the dust cap. This has some room influence but the results are not significant. One has to be careful including too much of the room because room effects are not necessarily linear with power level. There is a trade-off to be made since measuring too close to the driver results in high SPLs and at some point the microphone's linearity becomes an issue. The effect varies with microphone. In any case, with my measuring microphone, 12cm provides good results. In fact 20cm produces identical results and works well too. However I use 12cm since it is my standard distance for distortion tests and means I only have two distances to worry about in the various tests (3cm and 12cm).

The maximum trace was within 1dB of amplifier clipping and, coincidently, reaching Xmax so there was no point in going further. Power compression is beginning at this point as there is almost one thousand watts being pumped into the driver. IMO, these are pretty good results and for those who like to live in the "compression zone", this driver can certainly take more although with diminishing returns.

Some drift in the frequency hump is visible. The top trace shows more relative change than do the lower traces. Overall it doesn't seem too bad.

IMHO the updated version of this driver, while similar to the old in terms of power compression, is noticeably improved in terms of stability of the frequency peak with drive level. Apparently the new surround and spider make a difference.
 

·
Registered
Joined
·
162 Posts
Discussion Starter #15
I might be missing the point of your question...but my setup is the standard one for LIMP. Specifically as shown in Figure 3.2 in the LIMP user manual. (http://www.fesb.hr/~mateljan/arta/download/LIMP-user-manual.pdf)

I use a 10 ohm resistor instead of their 27 ohm. Mainly because I had a bunch of 10 ohm, 1% resisitors. I used a number of them in series parallel arrangement to increase the wattage and accuracy. This value must be known accurately.

Probes are shown there also. Although I made mine with higher attenuation in order to measure the EP4000 at full power without blowing my soundcard. One doesn't need this for impedance measurement but useful for measuring the amp's behaviour (frequency/distortion etc).

The demo version is fully functional except you can't save files. Graphs can be copied to another file easily using standard Windows functions. You may not need the licensed version.
 

·
Registered
Joined
·
162 Posts
Discussion Starter #16 (Edited)
Distortion Tests

Distortion is an important test. Frequency response can be adjusted by the end user but non-linear distortion cannot. We're completely at the mercy of the driver manufacture, with only the ability to make it worse by our box designs but never better.

Distortion tests are tricky to do right. A lot of things can happen and most of them are bad. :eek: They are very sensitive to noise and, due to limitations in microphones, not so amenable to accurate indoor testing (due to high SPLs of nearfield measurements). So outdoor testing is the best scenario for most people provided they can find a quiet environment. Since the ear is insensitive to low frequencies, use a microphone and spectrum test to identify what noises exist in the environment. Wind especially is a problem and traffic and industrial HVAC often not far behind. A large open space is required as reflections from structures can impact readings.

Having said that, these tests are indoors (it's cold outside!). Indoor distortion tests are subject to noises from furnace, fridges, noisey computers, rattling furniture etc so these need to be managed. Indoor tests also suffer greatly from room modes and reflections, so near field measurements are essential. Microphone distortion increases with SPL and also increases with low frequency. So we walk a tightrope between measuring too close (with too high an SPL) or too far away (with too much room influence). Do a frequency response check at your chosen distance, if it doesn't match the near-field 3cm curve then your readings are compromised.

Not all software is equal in a noisy environment. ARTA in stepped sine mode (ie STEPS) and REW do quite well in this regard, most programs using some kind of swept sine do not.

Finally, not all microphones are the same. My Behringer ECM8000 has the highest distortion of any I own and is completely unsuitable for low frequency, near-field measurements. It is barely suitable for outdoor 2 meter measurements. My Nakamichi CM-300 (rated 3% at 140dB) is much better and my IBF Acoustic measurement mic better still. Frequency response errors will affect distortion readings so any microphone used must be calibrated for frequency response. All tests are with my IBF Acoustic microphone.

First I like to measure with REW in spectrum mode. This confirms that I'm achieving adequately low noise floor. I like to have at least 60dB. REW also provides a nice visual indication of the distortion products for the driver. Here is the result for 20Hz sine wave.

REW_Distortion_20Hz.JPG

This was measured with 32.3v RMS output (about 260 watts into 4 ohms), microphone at 12cm from dustcap, 161L sonotude stuffed with 6 lbs of bulk polyfill. Xdisp was measured at 15mm. SPL at 50Hz was about 116dB at the mic position.

The peak of the 7th harmonic is interesting. This coincides with a small driver resonance and the first sonotube resonance. Amplifying distortion products is an unwanted side-effect of box and driver resonances. These will not be removed by the crossover.

REW is great for this kind of visualization. However far too cumbersome for taking distortion numbers over a range of frequencies. For that I use STEPS (in ARTA suite). The results are as follows:

[ Distortion.JPG

It is interesting to see how the distortion is climbing rapidly soon below 30Hz. By 10Hz, distortion is almost 30%! Getting a clean 10Hz signal at any realistic level is not so easy!

Comparing to Ilkka's distortion plots, these show marked improvement. Part of this is probably due to the new surround and spider but the rest is due to the larger box (161L versus Ilkka's 100L box). Box size has a large impact on distortion and is a key reason why I decided to go with the 161L box (versus 100L).
 

·
Registered
Joined
·
162 Posts
Discussion Starter #17 (Edited)
Somone PM'd and asked why the distortion rises so quickly below 30Hz and why box size affects this.

Basically, this is due to the fall off in the fundamental below resonance (at a rate of 12dB/octave in a sealed system) while the excursion related distortion products remain at constant levels.

The previous post showed distortion of my subwoofer in the 161L box. The system has a resonance around 28 Hz. Therefore distortion begins to rise quickly starting soon below that point. I'll repeat the plot here:
Distortion.JPG


The same woofer in free air has a resonance around 20Hz. One would expect then that the quick rise in distortion would shift down to below 20Hz. Here is the plot for this.

IBF_Mic_Subwoofer_in_FreeAir.JPG


Exactly as expected, distortion rise now begins below 20Hz.

The rise in distortion can be predicted. In the initial case of a 30Hz resonance, one would expect the distortion to rise about 2.25x going from 30Hz to 20Hz (based on 12dB/octave falloff). With an initial value of about 3% around 30hz, then one would predict 20Hz to be around 6.5% simply due to this effect. Indeed that is what we see.

Similarly, in the second plot where the resonance is at 20Hz we would predict that 10Hz (one octave lower) would show about 4x the distortion. Given a base value of about 4% at 20Hz then we would predict a value around 16% at 10Hz due to this effect. Again, this is what happens.

One could address this problem by equalizing the response flat but I'm not convinced it will work due to the driver being compliance dominated below resonance and an assumption that most of the distortions are compliance related (this would be an interesting test). Also, smaller boxes have greater air spring distortion so the net result of using a smaller box and a "linkwitz transform" to flatten the response will almost certainly be worse than simply using a larger box.

Bottomline, when it comes to distortion, go big or go home! :)


EDIT: Ilkka's THD measurements (http://www.hometheatershack.com/forums/subwoofer-tests-archived/8158-diy-css-sdx15-sealed-100l.html) are another data point for comparison. With a 100L box this would have a higher resonance compared to a 161L box or free air test. The sudden distortion increase should begin higher at around 35Hz and indeed it does. Some people assumed this was a driver problem but it was a box "problem". If that version of the SDX15 had any problem (and I think it did) it was in the high level of base distortion evident elsewhere. Also the slightly steep increase in distortion at resonance was partly due to the frequency response peak and associated steeper than expected initial rolloff.
 

·
Registered
Joined
·
2,431 Posts
Bjs,

Any chance that you could do the exact same distortion measurements back to back with the ECM8000 and the IBF and present the data? I'd like to see how much worse the ECM8000 is. I know I need a better mic and I need some extra motivation to purchase one. :R Plus it'd just be interesting data.

Can you extrapolate the 1m or 2m spl from your ARTA distortion test reaching 30% at 10hz? Not sure if you used the same drive level as the REW test. How much displacement was the driver being subjected to at the lowest frequencies? Anywhere near 20-30mm one way(I suspect not)?
 

·
Registered
Joined
·
393 Posts
I've gotta take the time to set up so i can do this kind of measurement on my SDX 15. I'm in a larger box than you, 16 cubes (440l) with 20~25% fiberfill tuned for 15 hz. im curious as to where my distortion rates increase dramatically. it is nice to see that someone is testing it in a reasonably larger sized box. i guess i might have to be the one to step up to measure it in a Large box.
 

·
Registered
Joined
·
162 Posts
Discussion Starter #20
Ricci: I added a free air test of the ECM8000 and posted in the mic thread. I'm not sure how one finds a "good" mic for nearfield distortion testing...it isn't something that seems to be well spec'd especially for the low frequencies and high SPLs. Too bad Herb didn't test 10 and 20Hz with his good mic versus the ECM8000.

I think I used the same drive levels for the 30% test. Xdisp was about 15mm at 20Hz as I recall so probably easiest would be to simulate that in at 161L box and extract the SPL.

Binary: Go for it. Your box is way larger and ported. A completely different animal so it would be interesting to see distortion plots of the driver and the port. Unless you go outside you won't be able to get the combined numbers though...that's one of the problems with measuring ported boxes inroom. :(

PS. Why is it that the SDX15 is the one that always seems to be tested? Where are the measurements for a Tempest, AV15, IXL etc? Someone loan me one and I'll test it!
 
1 - 20 of 33 Posts
Top