Hi all,
For selecting a subwoofer driver (12" to 18") what TS parameters define better transient response (fast attack or accuracy) ??
As far as I know, acceleration factor (BL/MMS) should be high. What other TSP or What points in published graphs to be checked ??
(Designing the cabinet for good transients is a different story. This question is only for selecting driver)
Thanks in advance
Audfrknaveen
For selecting a subwoofer driver (12" to 18") what TS parameters define better transient response (fast attack or accuracy) ??
As far as I know, acceleration factor (BL/MMS) should be high. What other TSP or What points in published graphs to be checked ??
(Designing the cabinet for good transients is a different story. This question is only for selecting driver)
Thanks in advance
Audfrknaveen
Ideally we want a transient perfect sealed [0.5 Qtc] or critically damped vented to mimic one once the room is factored in, ergo Qes/Qts varies, ditto compliance [Vas], so historically folks normally picked a 'middle of the road' driver spec-wise and do whatever it took to make it work in room:
Fs at/below lowest desired BW, ~ 0.31 - 0.35 Qts' and highest Vas one could find.
Nowadays though, they pick ever lower Vas to limit cab size and use super high power drivers to offset the efficiency loss.
[Qts']: [Qts] + any added series resistance [Rs]: Calculate new Qts with Series Resistor
[Rs] = 0.5 ohm minimum for wiring, so may be higher if a super small gauge is used as a series resistor plus any added resistance from an XO/whatever.
edit: need lowest inductance and widest BW too.
Fs at/below lowest desired BW, ~ 0.31 - 0.35 Qts' and highest Vas one could find.
Nowadays though, they pick ever lower Vas to limit cab size and use super high power drivers to offset the efficiency loss.
[Qts']: [Qts] + any added series resistance [Rs]: Calculate new Qts with Series Resistor
[Rs] = 0.5 ohm minimum for wiring, so may be higher if a super small gauge is used as a series resistor plus any added resistance from an XO/whatever.
edit: need lowest inductance and widest BW too.
Last edited:
It's amusing (to me) that you ask about "fast transient attack" but you say you want to build a ported sub. This is because (for the most part) all ported subs have much worse transient response compared to sealed boxes due to ringing in the time domain. It's an inevitable consequence to the higher order of the ported design. The exception is when you design a ported box with a very drooping response around the box tuning frequency, and then you may as well just go sealed. Also, unless you overdesign the port there will be some port compression at high SPL as well as possibly some wind noise, and this can also change the frequency response as SPL increases.
For other who might read this post, you have to look at the loudspeaker system (not just the driver) to see what kind of transient response you will get. This can be done using some box modelers that will produce the time domain (impulse) response for the system. For example Unibox does this.
In the past, there were drivers that had a very small magnet and therefore very high Qts. These had horrible performance and suffered from a lot of ringing. This is no longer typical of most drivers, but the "myth" persists that high motor strength and low moving mass should be pursued over all other properties of a driver if you want "the best" performance.
Each driver has its pluses and minuses, in terms of response, distortion, energy storage, and so on. One should really think about all of these parameters in balance, and not prioritize one of the over too much.
For other who might read this post, you have to look at the loudspeaker system (not just the driver) to see what kind of transient response you will get. This can be done using some box modelers that will produce the time domain (impulse) response for the system. For example Unibox does this.
In the past, there were drivers that had a very small magnet and therefore very high Qts. These had horrible performance and suffered from a lot of ringing. This is no longer typical of most drivers, but the "myth" persists that high motor strength and low moving mass should be pursued over all other properties of a driver if you want "the best" performance.
Each driver has its pluses and minuses, in terms of response, distortion, energy storage, and so on. One should really think about all of these parameters in balance, and not prioritize one of the over too much.
Off course it's a matter of balances, and the box alignment plays a major role in that. But low MMS and high BL helps with that. I would also use a sealed box and first select drivers that fit that alignment and the wanted FS, and then look in the list of candidates fit for the sealed cabinet which one has a relative low mms and high BL.
But speakers are always a matter of balances of factors and compromises. It's not focussing on one factor that will make the speaker you want, it's finding the right balance and choose the right compromises to fit your needs.
But speakers are always a matter of balances of factors and compromises. It's not focussing on one factor that will make the speaker you want, it's finding the right balance and choose the right compromises to fit your needs.
Well, you can say that ported boxes are "less" than closed in terms of group delay, phase etc.
But when you introduce stuff like "Linkwitz Transform" or other forms of EQ/filters you introduce group delay/phase issues too, either that or excessive time delay which can be even worse for subwoofer integration.
The direct benefit of "controlled roll off" would be "controlled driver movement". I do not see it as "overdesign" to use the correct port, that's called "proper planning" in my book.
Yes, indeed.
It's all a balance, few people plan the system as a "whole" before building it.
If you plan and execute things well I do not see a huge difference between OB/closed/ported/more exotic solutions, there's always some compromise, it's all in the implementation.
I'm not sure what the "stuff" is that you mention above, but regarding the Linkwitz Transform what you said is not true. It does not introduce any kind of "delay/phase issues [...] excessive time delay" whatsoever. The LT simply transforms (thus the name) one second order response into another one by applying or removing power having a certain phase response. It can be used to improve the time domain response in cases where Q is too high, as long as the driver has the motor strength to do that. Apart from those of the second order response itself, I am not sure where your claimed phase/delay issues come into play.
Not necessarily. I was thinking of the case where one uses an alignment that has a more rolled off (above tuning) response. These systems tend to have better time domain response compared to maximally flat and/or any alignment that results in a small peak just around tuning such as when you use a very low Qts driver in too large a box. Sometimes the EBS type alignments have this hump/peak. This causes ringing, in fact the more "maximally flat" a ported alignment is the worse its transient response. It's the same for closed box systems, really, as you go from e.g. Q=0.5 (somewhat drooping) to Q=1 (has a small hump/peak before rolloff).
But the driver movement can often be just as bad with the drooping vented box as with the maximally flat - it depends on what is exactly going on since the box AND the driver are contributing to the total frequency response around the tuning frequency (I am not talking about frequencies far above that, where the port has little influence).
I couldn't agree more.
I respect you and think it's a fantastic contribution to us all that you choose to spend time here on this forum, and it's quite possible I will say something that is completely wrong in this post, as quite a bit of it is pure assumption (or perhaps more my limits in experience and knowledge). Feel more than free to correct me.
Re: LT
So you haven't run any simulation looking at what it does?
Break it down enough, and LT = EQ, EQ = SPL, Phase and GD variance.
Any filter using IIR (that I have seen) will alter: SPL, Phase and GD. It is basically a virtually "perfect" analog circuit, but still a virtual analog circuit.
Unless you use FIR, but as we are in a subwoofer section and use FIR, you need high tap count which give time delay which is an even bigger problem IMO. Especially so if you need more processing power than a 1st order filter at, say, 27hz. If you use the system for music playback ONLY and don't mind waiting from pressing a piano tangent to hearing the sound, or playing a game and you hear sound after the event has occured.
).
Nice to know that we can agree, even though we're splitting hairs.
Re: LT
So you haven't run any simulation looking at what it does?
Break it down enough, and LT = EQ, EQ = SPL, Phase and GD variance.
Any filter using IIR (that I have seen) will alter: SPL, Phase and GD. It is basically a virtually "perfect" analog circuit, but still a virtual analog circuit.
Unless you use FIR, but as we are in a subwoofer section and use FIR, you need high tap count which give time delay which is an even bigger problem IMO. Especially so if you need more processing power than a 1st order filter at, say, 27hz. If you use the system for music playback ONLY and don't mind waiting from pressing a piano tangent to hearing the sound, or playing a game and you hear sound after the event has occured.
Yes, but you brought up ported vs closed.
Not true. Ported will have lower driver excursion towards box FS (which will be lower but in a bigger box volume), sealed will have higher driver excursion and require more eq and therefore more xmax and power to reach the same SPL at the same frequency. AFAIK we are not talking huge membrane area in IB here, closed box vs vented with controlled roll off, we are also (to my knowledge) ignoring problems induced by pouring in power in an enclosed box with no ventilation (you brought up issues with port turbulence, so I can pick up the power+heat issues
).Nice to know that we can agree, even though we're splitting hairs.
There is no need or requirement to use IIR/FIR or DSP to implement the Linkwitz Transform. There are at least two forms of analog electronic circuits that can do it (maybe more). SL published details on one of these circuits HERE. These have zero processing delay - the output appears immediately. That means that there is zero contribution to additional delay, etc. apart from what the "new/transformed" second order function brings in and of itself. And it is the gain and phase changes produced by the LT filter that are doing the "transforming" of the response, nothing more.
Regarding DSP implementations of the LT:
Any DSP system will introduce some internal processing delay. This will be pure delay that results from samples taking some time to move through the DSP system, and is frequency independent. You can easily compensate for it. Apart from that, the IRR/FIR filter algorithm itself will add more delay, but again this is frequency independent. After that, the LT processing via DSP will just result in a new second order response. The gain and phase changes are exactly what is required by the LT itself, nothing more. So overall there is frequency independent delay, and then the LT EQ. That's it.
Perhaps you have tried to mix DSP with non-DSP within a system but did not compensate for the (frequency independent) processing delay involved?
Welcome to my world! That said, some car subs are fine for HIFI/HT apps, though normally only when some type of pipe/horn loading is used, so tend to be a bit more expensive.
Re: LT
Any filter using IIR (that I have seen) will alter: SPL, Phase and GD.
I think it's worth saying here that any filter will alter the SPL, phase, and GD. This is what filters do. Now while it is true that an IIR filter is a mathematically computed contrivance, and thus has limits to its resolution, an analogue filter is also susceptible to just the same things, albeit in a different form.
As far as I understand the sigmaDSP line of chips only introduces a single sample period of intrinsic delay from its input to output, which, for all intents and purposes, can be ignored. This is across all internal channels unless you want extra delay. So all channels are synchronous on the output. I don't know about you but I'm not concerned with a single sample of delay from input to output, especially if we're at 192kHz.
With regards to resolution it is true that IIR filters lose precision as frequency decreases and as they roll off. But this is what double precision computation at higher bit depths is for (64 bit in the case of SigmaDSPs latest chips), not to mention most modern software allows you to see exactly what the filter you've implemented is going to do too. So if it's lost precision in an area of importance you can retune the filter until it does what you want it to.
Now aside from these small limitations they are extremely predictable. You tell a filter to be at 434Hz and it will be smack bang on 434Hz, the signal to noise ratio of the system only being limited by that of the digital to analogue step in the signal chain. The rest of the filters roll off will follow text book until the noise floor.
What of analogue implementations? They are susceptible to instability with certain component values when used with certain opamps. Unless you pick your parts carefully you will introduce noise into the system and most of the time you will be forced to pick a balance between noise and distortion. Both of these cause the filter to lose precision and component tolerances do exactly the same thing, altering the filters desired transfer function away from what you intended.
FIR filters are completely unnecessary in DSP based crossovers except for one thing - absolute phase linearisation - if you're so inclined. Otherwise leave them alone.
Yes, but they also affect the same things. SPL, GD and phase.
I am saying that IIR, for all intentions and purposes = close to perfect analog circuit.
Therefore: Reasonable delay, but impacts all three of SPL, GD and phase.
While FIR on the other hand, require a fixed amount of delay varying with frequencies involved to provide the basis of processing power. Low frequency = more taps = more delay.
No, what I'm saying that if you have a subwoofer with a sealed box, and whether if you use analog filters or virtual analog filters (through dsp): When you adjust the response in a certain way, you WILL influence SPL, GD and phase. If you boost the low end and add something like a 2nd order Butterworth HP for protection the difference between various design choices are negligible, in some cases you may end up with even higher GD and phase issues than if you made a design that required less compensation.
In most use cases any of this does not matter unless:
If you plan the system and required compensation/adjustment from an earlier point, you can achieve a system that requires less adjustment and therefore may get lower SPL, GD and phase variations than if you do not plan for these things.
A design needs to be put into the context of which it will be used. Looking at the measurements of an un-adjusted closed box, then deciding you need to start adjusting and after fiddling about for a while just say "design choice A is superior to design choice B because it's simply better". You're not looking at the system as a whole, just a specific part of it before doing any adjustments.
Filters and EQ add GD, unless it's FIR, and FIR is better to be avoided in the subwoofer section IMO.
5th element, yes! Completely agree!
Last edited:
Now it is clear that selection of driver is not independent of other system design choices, contrary to OP.
Perhaps a good way to look at this matter is in terms of how high the sub is to play in the system. An 18-inch driver might not be good if expected to handle as high as, say, 100 Hz while a 15-inch unit might be better in that upper part of the band.
Which leads to the eternal questions about freq bands for crossing-over since that determines what kind of driver best handles the next band up ("woofer" or "mid-range").
Yesterday, I was experimenting with some manufactured smallish egg-shaped speakers on tall stands. They are really good right down to perhaps 50 Hz or so, and excellent at the top to 20 kHz, or so my mic tells me. So I tried XO for my very large subs at 60 Hz*. Why not**? The result seems quite good, although I am spoiled by being used to listening to ESL speakers.
B.
* for my usual speakers, large panel electrostatic, I use 130 Hz.
** the "not" is because you need DSP; but who would make any system today without DSP
Perhaps a good way to look at this matter is in terms of how high the sub is to play in the system. An 18-inch driver might not be good if expected to handle as high as, say, 100 Hz while a 15-inch unit might be better in that upper part of the band.
Which leads to the eternal questions about freq bands for crossing-over since that determines what kind of driver best handles the next band up ("woofer" or "mid-range").
Yesterday, I was experimenting with some manufactured smallish egg-shaped speakers on tall stands. They are really good right down to perhaps 50 Hz or so, and excellent at the top to 20 kHz, or so my mic tells me. So I tried XO for my very large subs at 60 Hz*. Why not**? The result seems quite good, although I am spoiled by being used to listening to ESL speakers.
B.
* for my usual speakers, large panel electrostatic, I use 130 Hz.
** the "not" is because you need DSP; but who would make any system today without DSP
Last edited:
You seem to bring up this terminology trinity quite often, with what (to me) seems to be some kind of insinuation that "SPL, GD and phase variations" are a bad thing. Strange. For example, the only difference between any two sealed box (second order) alignments is: SPL, and phase variation (by definition, group delay is simply another way to state the phase response). These are frequency domain characteristics, which are typically understood to represent steady state operation, e.g. continuous tones. So they are not really the same as transient response, which was the topic of this thread when it was started.
Also, if increasing group delay at low frequency is what you are concerned about, I wouldn't be. Again, group delay is a SYSTEM parameter, not one of the driver itself. So you need to look at what the driver in the box is doing and go from there. For example, you can look at the range of closed or vented box responses in terms of their group delay, decide which you would like to try and build, and then find a driver that can do it. But I would caution you to not go overboard in this direction, because there are tradeoffs in other system parameters like its frequency response, that may impact what you hear much more than group delay. Also, the ear is less sensitive to group delay variation at the extremes of the audio spectrum, and for bass this means under about 100Hz. Pretty much every subwoofer has increasing group delay like this. But because our hearing is not sensitive to it, you will not get some kind of audible delay to the sound unless you are doing something wrong. This is why I brought up the topic of DSP processing delay, since that can certainly create some audible issues if you do not match the delay between bands in the loudspeaker.
My guess is that you are aware that peaks in the frequency response, particularly one at the resonance frequency of the system, will result in ringing in the time domain transient response. For the near resonance region, this can be stated in terms of the system Q. This is why higher Q values (above 0.7) for a sealed box system increasingly have some ringing in the time domain, e.g. impulse response. Peaks elsewhere in the passband will also produce time domain ringing.
All of these things have very little to do with the Linkwitz Transform. As I mentioned above, the only difference between two sealed box alignments is their gain (SPL) and phase response. As SL shows on his web page (that I linked to earlier) the LT circuit simply generates the DIFFERENCE between the two sets of gain and phase as part of the input to the driver. The unfiltered response is changed into a new one because the filter response sums with the unfiltered system response via the electrical to acoustical conversion of the motor. The user can choose what the new gain and phase will be, as long as they constitute a second order response. The only areas where this breaks down are (1) TS parameters are not linear with cone excursion, so as excursion increases the apparent Qts and Fs of the driver change and (2) the motor only has so much excursion capability and as BL falls off at high excursion it often cannot do what the input signal is asking it to. But at small excursions you absolute can make your driver in a box act like whatever closed box response you would like it to.
Anyway, this is becoming quite a digression from the topic of transient response, so that is the last I will comment on about it.
Last edited:
Again, the pioneers researched it pretty good apparently since their 18" ~20 Hz Fs TA-4181 designed for ~100-300 Hz [originally?] is listed as 'wide range' and based on the specs of other 'wide range' 12", 15" drivers it's 10 kHz and assume it's -24 dB so figure ~5 kHz like my early 15" Altec 515B.
It's not a bad thing. I've never said it was. I've been saying repeatedly that you cannot argue the case to build one specific design only, and then ignore the consequences of the analog correction circuit, virtual or physical, when you evaluate the different designs.
Ofcourse, but it's also a very useful indication of potential for ringing.
So now you're telling me that it doesn't matter? That these things are not linked at all?
)
Exactly my point.
)
Ofcourse, and very fine tips for anyone starting out building. But I think perhaps I have stricter demands than what you seem to be implying here.
Indeed, and this is why I've been advocating hunting for and reducing resonant modes instead of "brute force" DSP correction. "Less is more" is my mantra.
Ok. But I fail to understand why a proper implementation of LT will automatically be clearly superior to a equally properly planned and built ported enclosure.
I've heard a lot of crappy speakers, but the type of design seem completely irrelevant to me. It's all in the implementation.
I am happy to "agree to disagree", but it seems to me you've not been able to experience a well implemented ported design and may perhaps be generalizing just a little bit too much.
Yes, closed boxes have some properties that MAY provide better transient attack, at least before introducing any compensation circuit to achieve more low end response.
Is it therefore automatically a superior solution? No. I've heard both horrible and fantastic sound systems with any kind of sub, closed, ported or whatever else.