Typically a large rubber surround creates a frequency response dip that goes through most on and off axis angles, there is an obvious impedance issue but also a sharp spike in distortion at the same frequency. This can be seen in a lot of SB drivers. The SB17CRC has it just above 1500Hz.
https://hificompass.com/en/speakers/measurements/sbacoustics/sb-acoustics-sb17crc35-4
The resonance in your speaker does not behave like this in the measurements I have looked at.
The factory data sheet calls out the area of the driver at 82 square centimeters.
82 X 0.2cm = 16.4 cubic centimeters.
Yes the air suspension is more linear than the driver's mechanical suspension.
Keep in mind that at ~400Hz the cone is a "piston". The rubber surround is not. Take a look at the directivity Index plots.
A 12L enclosure has twice the volume and has space for twice the damping material.
Larger mid enclosures sound much less boxy to me.
Place a 1/4 inch calibrated microphone in a 6L and 12L mid enclosure and report back.
Thanks DT
For whatever reason, the amp-side R was dissapating about 4x as much heat as the two driver-side Rs in this particular layout. But you got me thinking about this a little more, and I have tried a combination of amp side and driver side R and it works better.
Thanks for this suggestion. I could not get the "series cap and coil, then shunt cap and coil" to work out. Cascading the tweeter after the mid high pass is interesting, and I was able to get a good result. I am not sure it is better, but it is nice to have this as an option.
A tanking cap across the lowpass coil (when used by itself) introduces a sharp high-Q notch that is very deep, in the range of -40 dB. However, because it is high Q, it does not have a broad range. This mid driver needs a more shallow Q notch, something that suppresses from 6k to 12k+. This is why I originally used a parallel notch in series with the driver.
You seem to be advising something here
and I am not sure what you are getting at by "LC across the driver". Could you describe this in more detail?
Again, thanks! Your suggestions are making me think about things in a different way, which is helpfull.
j.
The tank cap across the series coil can have a series resistor with it alone to shape the response as well. This essentially makes a parallel notch with the coil and cap. Essentially, connect leads in a triangle, and the in and out on the coil leads. Usually this is a broad-range filter, from coil cutoff to upper treble. This is because there is bleed through in the top octave that increases the larger the cap gets. This can also cause low impedance drops in the upper treble because the impedance comes back down in the upper treble. Normally I have seen and used nothing larger than 1.5uF here, smaller the better, and the series R keeps the impedance benign.
An LC or bottomless trap (or pit, because of how it works) across the driver of an L and C in series, or series notch, though sometimes an additional R is added, is more capable of breakup and energy storage suppression not unlike flushing the frequency range down the toilet.
In my experience, sometimes a parallel notch like you have on the midrange will not even touch a bad resonance, and it takes the shunt/trap/series notch to actually kill the issue.
The LC across the driver looks like a short in impedance, save for the coil resistance or added R. If you have a lowpass ahead of it, or a series resistor, then this short is compensated as the amp sees it, but still is very effective at its job. However, it should be noted that without the voltage divider out front of either type, that the LC shunt will not affect the freq response.
It should also be stated here that the Purifi document about reducing HD by using a parallel notch placed in series with the driver calls it a series notch. Most electrical textbooks will be at odds with this classification, calling the circuit as they are arranged; parts in parallel = parallel, and in a string = series; no matter a transducers relation to the circuit.
An LC or bottomless trap (or pit, because of how it works) across the driver of an L and C in series, or series notch, though sometimes an additional R is added, is more capable of breakup and energy storage suppression not unlike flushing the frequency range down the toilet.
In my experience, sometimes a parallel notch like you have on the midrange will not even touch a bad resonance, and it takes the shunt/trap/series notch to actually kill the issue.
The LC across the driver looks like a short in impedance, save for the coil resistance or added R. If you have a lowpass ahead of it, or a series resistor, then this short is compensated as the amp sees it, but still is very effective at its job. However, it should be noted that without the voltage divider out front of either type, that the LC shunt will not affect the freq response.
It should also be stated here that the Purifi document about reducing HD by using a parallel notch placed in series with the driver calls it a series notch. Most electrical textbooks will be at odds with this classification, calling the circuit as they are arranged; parts in parallel = parallel, and in a string = series; no matter a transducers relation to the circuit.
I have mainly used some book binders glue and mod podge. Also tried some flexible construction glue that seemed pretty good. I think there are some threads about it on the forum too.
Hello,
When I selected this driver based on the prototype testing, I considered the 2 dB wiggle in the response between 1.5k – 3k to be minor. Now looking at the response in the full cabinet, it is a 3 dB wiggle, and to me it seems no longer minor. It now seems like something I will have to design around.
Looking at the inCab Frequency Response there is a dip at ~1600Hz and a peek at about 2100Hz.
I am thinking about what I call Tsunami Effect, the trough of the wave comes on shore before the crest of the wave.
Seems to me that passive filter equalization may not do so well at attempting to remove a resonance.
In testing, so far the driver has been held constant but the prototype enclosure has changed to the new modified "finished" version.
I suggest just for A x B testing that the driver be changed out with the aluminum sb15nbac30-8 driver. It will drop in the enclosure cut out. Same everything except the cone and dust cover material. The A x B testing holds the enclosure constant while the driver variable is changed.
Thanks DT
https://www.madisoundspeakerstore.com/approx-5-woofers/sb15nbac30-8-5-black-aluminum-cone-mid-bass/
New measurement amplifier. IcePower 125ASX2.
The V-notches on the side panel are for ventilation, but after running it for a few hours, I decided I need more ventilation. So I added spacers on the top and bottom panel to create a 4 mm gap around the whole perimeter. Now there is plenty of air flow.
Well that was optimistic... more like 5 hours and 45 minutes. I am happy with the result though...
The V-notches on the side panel are for ventilation, but after running it for a few hours, I decided I need more ventilation. So I added spacers on the top and bottom panel to create a 4 mm gap around the whole perimeter. Now there is plenty of air flow.
Well that was optimistic... more like 5 hours and 45 minutes. I am happy with the result though...
Now that I have a new testing amp, I re-ran my far field gated measurements. The first set of scans (a month ago) were made in my basement workshop, and due to the low ceiling, the longest time window I can achieve is 3.3 ms. For this new set, I moved upstairs, and with the high ceiling and elevated speaker/mic, I can get a time window of 5.5 ms.
The old amp is an FX Audio 1002A, which is a low cost Class D amp from Parts express. I used this same amp in the LCCAM-10.3 semi active speaker, and in that thread I discussed its limitations. It has a load dependent frequency response error from 3k up.
https://www.diyaudio.com/community/threads/compact-low-cost-active-3-way-speaker.402812/post-7575669
In the LCCAM application where DSP filtering is used and the load impedance is fixed, it is easy to compensate for the amplifier response with a calibration. But this amp is not ideal for making measurements. The new amp is an ICEpower 125ASX2. The ICEpower ASX series have a flat frequency response above 20k regardless of load impedance. The distortion is lower as well… it is certainly lower than my Audix microphone.
In the following plots, the solid line is the current on-axis scan with the 5.5 ms time window and the new amp. The dashed line is the older on-axis scan with the 3.3 ms time window and the older amp. The difference in high frequency response is clear to see, even with the woofer. With the tweeter, the difference is substantial.
Woofer
Midrange
Tweeter
j.
The old amp is an FX Audio 1002A, which is a low cost Class D amp from Parts express. I used this same amp in the LCCAM-10.3 semi active speaker, and in that thread I discussed its limitations. It has a load dependent frequency response error from 3k up.
https://www.diyaudio.com/community/threads/compact-low-cost-active-3-way-speaker.402812/post-7575669
In the LCCAM application where DSP filtering is used and the load impedance is fixed, it is easy to compensate for the amplifier response with a calibration. But this amp is not ideal for making measurements. The new amp is an ICEpower 125ASX2. The ICEpower ASX series have a flat frequency response above 20k regardless of load impedance. The distortion is lower as well… it is certainly lower than my Audix microphone.
In the following plots, the solid line is the current on-axis scan with the 5.5 ms time window and the new amp. The dashed line is the older on-axis scan with the 3.3 ms time window and the older amp. The difference in high frequency response is clear to see, even with the woofer. With the tweeter, the difference is substantial.
Woofer
Midrange
Tweeter
j.
I need to clarify the plots in the previous post #390. I was in a hurry to get to lunch and I posted too quickly...
These plots are far field gated measurements merged with near field measurements. The near field part did not change, only the far field measurements were re-scanned. In the "old" dashed line plots, the merge frequency was about 700 Hz for woofer, mid, and tweeter. In the "new" solid line plots, the merge frequency was about 500 Hz for woofer and mid, and 700 Hz for the tweeter.
j.
These plots are far field gated measurements merged with near field measurements. The near field part did not change, only the far field measurements were re-scanned. In the "old" dashed line plots, the merge frequency was about 700 Hz for woofer, mid, and tweeter. In the "new" solid line plots, the merge frequency was about 500 Hz for woofer and mid, and 700 Hz for the tweeter.
j.
I read on ASR that it will soon be available in a ready-to-go version. In the thread, the Fosi Audio M3 is given as an example as a direct competitor.
https://www.audiosciencereview.com/...with-pffb-is-coming.54799/page-5#post-2052482
I am very confused about the difference above 5kHz.
I can only explain that with either a difference in calibration or difference in position (angle)
But the changes you're describing don't explain this behavior.
Or maybe that's what you mean with; "where DSP filtering is used and the load impedance is fixed" ?
In that case that's pretty extreme and bad, to have a 5-6dB difference around 15kHz
(Yes I do also design Class-D amplifiers)
Now it reads like that the difference above 5kHz was improved.
Which is quite misleading.
Besides that part, I find the rest far from substantial and actually very similar.
There is the obvious and predictable shift in diffraction related issues because of the change in distance as well as a tiny bit more frequency resolution.
Perhaps I should have been more clear. Referring to the plots in post #390, the solid line is a frequency response scan, on axis, distance of 1 m, made with the new ICEpower amp. The dashed line is a FR scan, on axis, distance of 1 m, made with the old FX-audio amp. This old amp uses the TDA7498E chipset.
In a previous project, I measured the frequency response of the FX-audio amp while connected to a midrange + tweeter + crossover filter. This is a different set of drivers than what we are discussing in this thread, but this data illustrates the issue with this amp.
In yet another earlier project I measured an even larger frequency response deviation when the amp was connected to a tweeter with no filter network. At that time I did not understand what was going on, and I thought I had a broken microphone.
In this case, with the SB26STWGC-4 tweeter and no filter elements, I am seeing the largest frequency response deviation I have yet measured with this amp. The FX-audio amp is -1 dB at 5k, -2.5 dB at 10k, -4 dB at 12.5k, as shown in post #390.
Here is another way of looking at the tweeter data shown in #390. I imported the on-axis impulse responses into ARTA, and used a consistent time window of 3.26 ms. The blue line used the FX-audio amp. The two yellow lines used the ICEpower amp, with two slightly different microphone positions. Note that the y-axis is expanded to 2 dB/division.
Why two mic positions? After I made the polar response scans of the mid and tweeter, I wanted to get a longer time window, so I elevated the speaker as high as practical, and made a second set of polar scans. After reviewing the data, I discovered both sets of data had approximately the same reflection-free time delta, so the two sets are equivalent. However, they do show a slightly different tweeter response, particularly in diffraction effects. I have noticed that when waveguide tweeters are measured on-axis, they are very sensitive to small positional changes of the microphone. I first discovered this with the Satori TW29BNWG, and it drove me crazy for about a week...
So bottom line, if the FX-audio amp is used above 5k, it needs a calibration curve.
j.
It needs a calibration curve for every new change in load impedance ... which makes it very annoying as a piece of test equipment...
Thanks that does make it more clear.
That is a rather poorly designed amplifier, even for a simple Class-D
Personally, I would recommend getting a proper Class-AB amplifier.
Ideally one of those flavored composite amplifiers.
In that case you know for sure that you won't bump into those kinda issues anymore.
Even a simple 2nd hand receiver or so won't have these kind of issues.
I personally wouldn't even recommend that IcePower amplifier.
I have worked with those in the past, and they also have all kinds of funky stuff going on.
You mean totally unsuitable!
Anyway, if the biggest error here is the amplifier, I still don't see such a big difference in measuring method?
For measuring speakers any proper LM3886 or TDA7293 amp will do I guess… the composite stage is beneficiary, but a bit of an overkill, I’d say.
Maybe, although some people here like to measure such low distortion.
The price difference between the two is one additional opamp + handful of passives.
Here is an interesting crossover design, using the latest measured driver responses. This layout incorporates the alternate notch filter that @wolf_teeth recommended. I show two versions, one which is nearly flat on axis, the other which is sloped down by about -2 dB. My experience is that the second one will be close to the final subjective voicing.
Now the -2dB sloped response:
Now the -2dB sloped response: