If you are looking at different delay curves, at say 80 Hz and 500 Hz, what is the threshold of hearing group delay? 😕
Thanks
Thanks
The more fundamental question I should have asked first is: comparing vented & sealed alignments
- Unibox compares step response curves and shows major differences, while
- WinISD plots group delay and the differences look subtle.
What differences in sonics do each they translate to, and which is more useful?
Thanks
- Unibox compares step response curves and shows major differences, while
- WinISD plots group delay and the differences look subtle.
What differences in sonics do each they translate to, and which is more useful?
Thanks
Some reading on group delay of different alignments and its audibility. Sadly, there is not actually much info on the audibility as not much research has been done.
http://www.trueaudio.com/post_010.htm
http://www.linkwitzlab.com/frontiers.htm#F
http://www.trueaudio.com/basslst2.htm
http://www.trueaudio.com/post_010.htm
http://www.linkwitzlab.com/frontiers.htm#F
http://www.trueaudio.com/basslst2.htm
The step response is difficult to analyse. The overshoot is important, as well as any ringing. Even "critically damped" sealed alignments have some step overshoot, and teh ringing of the vented alignments obscures the actual amount of overshoot. Basically, if the ringing on the step response curve is at a frequency below 30Hz, there is essentially no way you will hear it under normal circumstances as long as the response is relatively flat.
The group delay is just as difficult to analyse. You can't really compare curves with systems having different resonant frequencies. A speaker with a 150ms group delay at 50Hz would sound much worse than one with the same group delay at 20 Hz. The group delay of room decay resonances is on the order of seconds rather than milliseconds, so experiments using loudspeakers are much less sensitive than those using headphones. IIRC, some suggest that around 20ms is the threshold of audibility at 100Hz using headphones. This is two cycles. If you keep the group delay below 2 cycle periods you should be more than fine. This is 100ms at 20Hz.
IMO, it is much more important to design a system with a relatively flat response that increases smoothly with frequency (no peaking) and integrates well with the room. Cutoff should be around the area where the room starts to exhibit room gain, so smaller rooms should have higher F3's, unless you want a "Car-fi" sound.
The group delay is just as difficult to analyse. You can't really compare curves with systems having different resonant frequencies. A speaker with a 150ms group delay at 50Hz would sound much worse than one with the same group delay at 20 Hz. The group delay of room decay resonances is on the order of seconds rather than milliseconds, so experiments using loudspeakers are much less sensitive than those using headphones. IIRC, some suggest that around 20ms is the threshold of audibility at 100Hz using headphones. This is two cycles. If you keep the group delay below 2 cycle periods you should be more than fine. This is 100ms at 20Hz.
IMO, it is much more important to design a system with a relatively flat response that increases smoothly with frequency (no peaking) and integrates well with the room. Cutoff should be around the area where the room starts to exhibit room gain, so smaller rooms should have higher F3's, unless you want a "Car-fi" sound.
Tenson
I re-read those articles, those thresholds for group delay are useful; though I suspect that the step response is better measure.
Ron E
You’re right, relatively flat response is probably most important; especially for systems intended to cover the bottom end, I see your point.
In this case it’s a “midbass” only over 75 – 500 Hz (where I was going to do a midbass horn, but for now it’s too hard). As you probably know 75 – 500 Hz is also where more of the music energy is, including most percussion fundamentals.
I am hoping that the right choice of (active) crossover slopes to the sub can provide the flat response.
While Unibox calcs step response, it’s not clear at what frequency value, or over what frequency range are they based on. My guess is at the system resonant frequency, and by the time you move say an octave either way, they would not apply, and there would no difference.
I re-read those articles, those thresholds for group delay are useful; though I suspect that the step response is better measure.
Ron E
You’re right, relatively flat response is probably most important; especially for systems intended to cover the bottom end, I see your point.
In this case it’s a “midbass” only over 75 – 500 Hz (where I was going to do a midbass horn, but for now it’s too hard). As you probably know 75 – 500 Hz is also where more of the music energy is, including most percussion fundamentals.
I am hoping that the right choice of (active) crossover slopes to the sub can provide the flat response.
While Unibox calcs step response, it’s not clear at what frequency value, or over what frequency range are they based on. My guess is at the system resonant frequency, and by the time you move say an octave either way, they would not apply, and there would no difference.
Step response is the response to a step input - no signal, then instantly raised to some value. All frequencies are there, but the low frequency response is all that is looked at in the simulations in Unibox because it assumes the driver is flat to infinite frequency. The response of a real driver with crossover looks like the bandpass step responses, which are MUCH harder to interpret.
To be blunt, if you didn't even know what step response was, I would shy away from using either it or group delay as a comparative and stick to frequency response.
Making a vented midbass is nonsense for the most part anyway. What makes you want to do it that way?
To be blunt, if you didn't even know what step response was, I would shy away from using either it or group delay as a comparative and stick to frequency response.
Making a vented midbass is nonsense for the most part anyway. What makes you want to do it that way?
I knew what step response is “at a point”, but not over what frequencies it applied. Why does the response of a real driver with crossover look as bad (terrible?) as the bandpass step response?
Do you think that the relative step response values in Unibox (vented/ sealed/ etc) have some validity?
“if you didn't even know what step response was, I would shy away from using either it or group delay . .”
If I didn't fully understand it, I would rather try to learn it.
“Making a vented midbass is nonsense for the most part anyway”
IIRC Dickason suggests vented midbass is ok, if its resonance is at least an octave away from the crossover frequency.
With the driver in question (JBL 2035), even just above about 70 Hz, unless the box is very big, vented is flatter than sealed.
Also if step response over the passband is either not relevant or comparable with vented, ie pursuing a low Q is of little or no benefit: then the box need only be say about 70 litres instead of about 190 litres.
In summary if there’s a benefit to had going big and sealed, for low Q and better step response, I will.
If a step response benefit is unlikely (whether or not I understand it 100%); and as the Hz ‘distance’ between crossover frequency and Fb won’t be known until the overall system the sub is done, I’d probably go sealed.
Do you think that the relative step response values in Unibox (vented/ sealed/ etc) have some validity?
“if you didn't even know what step response was, I would shy away from using either it or group delay . .”
If I didn't fully understand it, I would rather try to learn it.
“Making a vented midbass is nonsense for the most part anyway”
IIRC Dickason suggests vented midbass is ok, if its resonance is at least an octave away from the crossover frequency.
With the driver in question (JBL 2035), even just above about 70 Hz, unless the box is very big, vented is flatter than sealed.
Also if step response over the passband is either not relevant or comparable with vented, ie pursuing a low Q is of little or no benefit: then the box need only be say about 70 litres instead of about 190 litres.
In summary if there’s a benefit to had going big and sealed, for low Q and better step response, I will.
If a step response benefit is unlikely (whether or not I understand it 100%); and as the Hz ‘distance’ between crossover frequency and Fb won’t be known until the overall system the sub is done, I’d probably go sealed.
Well ..... I think you just gave the answer yourself ! A bandpass will not have a step response that is different form a bandpass' step response ! Simple as that !
Regards
Charles
Charles
I was asking Ron to say to elabaorate on his statement "The response of a real driver with crossover looks like the bandpass step responses, which are MUCH harder to interpret."
(I assumed) he was referring to either the vented & sealed alignments being discussed so far.
ie he seems to be saying that real drivers + crossover screw the step response. I can imagine this applying to cruddy drivers with bad crossovers, but find it hard applying it to good drivers in sealed boxes.
But if it's true, Unibox is calculating (around the Fsys) that vented is not as good, and bandpass is all over the shop.
Cheers
I was asking Ron to say to elabaorate on his statement "The response of a real driver with crossover looks like the bandpass step responses, which are MUCH harder to interpret."
(I assumed) he was referring to either the vented & sealed alignments being discussed so far.
ie he seems to be saying that real drivers + crossover screw the step response. I can imagine this applying to cruddy drivers with bad crossovers, but find it hard applying it to good drivers in sealed boxes.
But if it's true, Unibox is calculating (around the Fsys) that vented is not as good, and bandpass is all over the shop.
Cheers
There is a strong association between the step response and the frequency response of a device. Independant of the the latter being electrical, mechanical, optical ,......... or whatever combination of these.
As soon as you filter in the time-domain you will have a deviation from the ideal step response. Period.
In other words a prerequisite for an ideal step-response is a linear amplitude response from zero to infinite frequncies. Unfortunately this is not feasible and - even worse - it won't be sufficient either ! A system has to be minimum-phase also - which is very difficult to achieve as soon as the signal-path is split like it is done in a conventional multiway speaker.
There is however something that could be called the optimal step response for a speaker and it looks like the one of a wide fourth-order bandbass. But even this one is very rare.
Regards
Charles
As soon as you filter in the time-domain you will have a deviation from the ideal step response. Period.
In other words a prerequisite for an ideal step-response is a linear amplitude response from zero to infinite frequncies. Unfortunately this is not feasible and - even worse - it won't be sufficient either ! A system has to be minimum-phase also - which is very difficult to achieve as soon as the signal-path is split like it is done in a conventional multiway speaker.
There is however something that could be called the optimal step response for a speaker and it looks like the one of a wide fourth-order bandbass. But even this one is very rare.
Regards
Charles
Tenson said:
And on that note I don't think the perfect impulse response is that important in the grand scheme of things 😀
IMO of course!
Well thats just it, if we knew more about the audibility of group delay in a small acoustic environment with loudspeakers we could make a more informed decision.
I would say get the other things right first then think about group delay and step response etc.. Anyway, if you have a DEQX or computer Xover then you can do some pretty neat things with phase coherency.
I would say get the other things right first then think about group delay and step response etc.. Anyway, if you have a DEQX or computer Xover then you can do some pretty neat things with phase coherency.
The step response shown in the Unibox graph is normalized acceleration step response - or the pressure developed by the step signal. This makes sense in a way because it is what you hear.
The more intuitive step response is displacement, this should follow the input as closely as possible, but there is an associated rise time and overshoot, ringing, etc. It is a bit easier to understand because the graph and the input are more similar, and will be for any loudspeaker design that can accept a DC signal - so no series capacitor alignments - and for that matter the real step response will still be quite different because very few amps are DC coupled....
For the speaker you are crossing over at ~70Hz, I would prefer to go sealed and raise the crossover frequency to suit. The step or transient response (or group delay) of a speaker crossed over above its low cutoff and then lowpassed (bandpassed in the flat range) is pretty much a nonissue, IMO. The poles of the crossover will dominate.
Careful attention should be paid to the integration of the crossover, and this wil definitely be a learning experience for you if you are just learning about these dynamic response concepts. While you can get a handwaving qualitative understanding of the terms, truly being able to confidently interpret the graphs will take experience calculating dynamic responses for many systems.
Do you know LaPlace transform techniques, or circuit analysis?
The more intuitive step response is displacement, this should follow the input as closely as possible, but there is an associated rise time and overshoot, ringing, etc. It is a bit easier to understand because the graph and the input are more similar, and will be for any loudspeaker design that can accept a DC signal - so no series capacitor alignments - and for that matter the real step response will still be quite different because very few amps are DC coupled....
For the speaker you are crossing over at ~70Hz, I would prefer to go sealed and raise the crossover frequency to suit. The step or transient response (or group delay) of a speaker crossed over above its low cutoff and then lowpassed (bandpassed in the flat range) is pretty much a nonissue, IMO. The poles of the crossover will dominate.
Careful attention should be paid to the integration of the crossover, and this wil definitely be a learning experience for you if you are just learning about these dynamic response concepts. While you can get a handwaving qualitative understanding of the terms, truly being able to confidently interpret the graphs will take experience calculating dynamic responses for many systems.
Do you know LaPlace transform techniques, or circuit analysis?
“I don't think the perfect impulse response is that important in the grand scheme of things”
My idea is/ was, if more than one alignment can give a flat response over a passband, the next thing to focus on IMO is impulse response. With similar response over a passband, the Unibox plotted impulse responses for vented and sealed vary wildly.
I’m not expecting or chasing *perfect impulse response. Just other things being equal (as far as possible, ie allowing for a big box if it helps) the least imperfect response . .
“if you have a DEQX or computer Xover then you can do some pretty neat things with phase coherency”
Presumably SoundEasy (and others) could do this?
Ron
Thanks for your advice (no I have no engineering background and LaPlace transform techniques or circuit analysis are foreign to me).
Cheers
My idea is/ was, if more than one alignment can give a flat response over a passband, the next thing to focus on IMO is impulse response. With similar response over a passband, the Unibox plotted impulse responses for vented and sealed vary wildly.
I’m not expecting or chasing *perfect impulse response. Just other things being equal (as far as possible, ie allowing for a big box if it helps) the least imperfect response . .
“if you have a DEQX or computer Xover then you can do some pretty neat things with phase coherency”
Presumably SoundEasy (and others) could do this?
Ron
Thanks for your advice (no I have no engineering background and LaPlace transform techniques or circuit analysis are foreign to me).
Cheers
The Manger does indeed have a step response close to that of a fourth-order bandpass. But this is only valid when not used together with a woofer and a crossover.
Regards
Charles
- Status
- Not open for further replies.