disappointing lower midrange performance from woofers, why do people cross high?

[thadman]

I became aware of some testing Zaph did regarding the RSS315HF and RS270. I was surprised to say the least, the distortion looked pretty bad. The RSS315HFs 3rd harmonic levels reach .3% at 200hz (wasn't measured beyond this) and the RS270 maintained .3% 3rd harmonic levels from 200-500hz and ~.4% 2nd harmonic levels from 350hz-->500hz. It doesn't look like they (although possessing a much greater amount of displacement) can swing with the the midranges Zaph measured. Although appearing bad, I'm not sure what to make of it since Zaph, other than leaving us some context clues, doesnt appear to list the SPL he measured them at. Are they measured at the 96dB @ 1/2m that the 6.5s were measured at or something else?

Zaph mentions that the RS270 was reaching ~2/3 xmax around 100hz, so I plugged the figures into spl_max1 (330cm^2, 4.4mm xmax, 100hz) and got 106dB. If we account for baffle step, thats roughly 100dB. Is this about right?

I'm not sure if Zaph likes people borrowing the pictures from his blog (so to be courteous I saved them on my computer and hosted them myself).

RSS315HF


RS270


(FYI I plan to use these drivers in a dipole alignment)
I have been under the impression that a high displacement midbass/lower midrange driver crossing to a midrange around 300hz was the way to go...have I been misled or interpreting information incorrectly? Would a multitude of RS225s (say 4/panel) crossed around 100hz to a deep h-frame/w-frame be a better option than taking the subwoofers (specifically RSS390HF) up to the midrange crossover point on a flat baffle? The RS270 doesnt seem to measure all that impressively either, is it truly one of the best midbass drivers when crossed ~500hz, or is it bested by its smaller brothers the RS225/RS180 when used in pairs costing the same $$$?

[CeramicMan]

I looked it up on Parts Express and the cone has a loud resonant mode at around 1.6kHz. That'll be the problem. A small amount of distortion from non-linear springs and flux will be amplified at various divisors of 1.6kHz.

Maybe the input signal can be calibrated to cancel out that distortion?

[thadman]

Originally posted by CeramicMan
I looked it up on Parts Express and the cone has a loud resonant mode at around 1.6kHz. That'll be the problem. A small amount of distortion from non-linear springs and flux will be amplified at various divisors of 1.6kHz.

Maybe the input signal can be calibrated to cancel out that distortion?

?The RSS315HF was only measured up to 200hz, well below the F2,F3,F4, and F5 harmonics.

[CeramicMan]

?The RSS315HF was only measured up to 200hz, well below the F2,F3,F4, and F5 harmonics.

I don't know then. It could be anything that might be non-linear in some way – "yielding" in the rubber surround? Cross-over distortion from the spider?

It's a balancing act: "high excursion" parts are probably relatively more accurate at high SPL but less accurate at low SPL.

[thadman]

Originally posted by CeramicMan

Maybe the input signal can be calibrated to cancel out that distortion?

How would I go about doing that?

[CeramicMan]

Originally posted by thadman


How would I go about doing that?

Measurement microphone + microphone amp + software.

You'd need to filter all the music through a digital system, such as a PC. Sorry that I can't be more specific at the moment, but I haven't done that before either and it's on my to-do list of projects.

[DougL]

Measurement microphone + microphone amp + software.
You'd need to filter all the music through a digital system, such as a PC. Sorry that I can't be more specific at the moment, but I haven't done that before either and it's on my to-do list of projects.

This "procedure" is not theoritacally or practically sound.

The inverse transfer function would be non-trivial even if the drivers had no energy storage. Certainly more than a microphone would be required.

Think of it this way, once you strike a bell, you can't cancel the sound by striking it again.

Doug

[joe carrow]

Originally posted by thadman


How would I go about doing that?

One method that's been used in the past is to attach an accelerometer to the subwoofer cone, and use its signal for feedback to generate the correction signal, which is summed with the analog input. You'll see these called servo subs, motional feedback subs, and other things. This method is, in my opinion, fairly exotic and not as cost effective as just buying a high quality woofer and putting it into an appropriately sized box.

Here's a review of one, with a link to a commercial offering:

http://www.hometheaterhifi.com/volum...er-4-2004.html

Due to the speed and accuracy of the accelerometers and other components available, there is a fairly strict limit on how high this method is effective. The reviewer shows the performance of the sub up to 100 hz. Based on what I know of the performance of available accelerometers (It's been two or three years since I've researched them extensively), none have the range, accuracy, and sampling rate that would be required to go much higher than that. Laser and capacitive methods might be feasible to higher frequencies, but then you're talking about tens of thousands of dollars in lab grade instrumentation.

Another thing that, in theory, could work for DIYers is to get a series of fairly accurate distortion measurements of a driver at different playback volumes, and to generate a model of the complete transfer function of the woofer. You would be able to create an inverse of certain undesirable parts of this function and use it to convolute the incoming signal such that the final output more closely matches the desired output.

A purely predictive software method would only take a DSP to process the signal to the amp, and with the effective upper frequency limited by the accuracy of the simulation and the speed of the DSP.

Essentially, I wouldn't worry about it. As far as I know nobody's really doing that, and unless you program DSPs and/or motion control for work it's not something you could DIY too easily.

[CeramicMan]

Originally posted by DougL

This "procedure" is not theoritacally or practically sound.

The inverse transfer function would be non-trivial even if the drivers had no energy storage. Certainly more than a microphone would be required.

Think of it this way, once you strike a bell, you can't cancel the sound by striking it again.

Doug

I never said it would be easy!
The ringing of a bell can be cancelled out by striking it with a negative force at precisely the same position and time as the original strike. By applying the negative force to the same hammer, the bell never gets hit in the first place. So the analogy is that instead of using feedback with the associated delays and whatnot, a digital system anticipates things that haven't happened yet.

A purely predictive software method would only take a DSP to process the signal to the amp, and with the effective upper frequency limited by the accuracy of the simulation and the speed of the DSP.

The lower frequency limit is also likely to be a problem because of the memory requirements. Not all DSPs support external memory interfaces, and at sub-100Hz frequencies a FIR window could take up a lot of space!

Re: accelerometers, I think there's a different issue. They are differentiators, while most loudspeakers have a frequency dependent combination of integrating, proportional and differentiating properties. Ordinarily, an accelerometer can't be just thrown into the feedback system as it would produce positive feedback at some frequencies.

What I would try is to put an oversized subwoofer in an undersized box. That way it has a simple straight-line response that increases by 12dB/oct throughout its usable range. Or, I could try to linearize the output by using a negative output impedance amplifier, which compensates for the speaker impedance (edit: and box response). Then an accelerometer can be put into the system.