Acceptable group delay for sq, question for box builders and sq guys.

[RSDXzec 10+ year member]

The first box I built I tried to keep group delay to a real minimum, as low as I could. But info I just found on the internet suggests that as long as the delay isn't more than a period or two of the frequency then it shouldn't be a problem. By that definition then:

at 30hz, anything less than 33.33ms is acceptable

40hz - 25ms

50hz - 20ms

60hz - 16.666ms

and so on...

But is it one period? or should it be more around half? I'm really curious to know what rule most of you box builders have as far as minimising group delay.

I'm asking because I'm planning on making a new box for my sub, because it really drops off fairly sharply at 50hz, and I enjoy bass between 30-50hz so it's a problem for me. But a flatter response = bigger box = more group delay. So let me know how you guys go about this.

Cheers.

edit:

For the record the design I'm looking at (with a much flatter response) building has the following group delay:

50hz - 7ms

40hz - 12.5ms

35hz - 17ms

30hz - 23.5ms

25hz - 27.5ms (it peaks here)

20hz - 21.3ms

My previous design: (the one currently in my car)

50hz - 7ms

40hz - 9.6ms

35hz - 10.7ms

30hz - 12.5ms

25hz - 17.9ms

20hz - 32ms (it peaks @18.3hz @34.8ms)

I don't notice the delay with my current box so I'd like to know if I would notice it with the new box. This is still very early in the planning stages so I'm open to changes I'm just trying to get a flat response because I really notice the decrease in loudness at lower frequencies with my current box.

Click to read more...

 

[CAT MAN 5,000+ posts]

@cablguy184

 

[Moble Enclosurs 10+ year member]

Well, the group delay would not be the only issue in a smooth response, first thing. There are other factors that control that as well. I have a way in my calculations to figure for db gain and loss based on phase changes as well, and it is very acceptable. That is where the direct audible output will be effected as far as output or the mentioned drop below 50Hz. Now, with group delay, the smaller the better in all cases. The delay can be acoustical or electrical also, so not all of it can be forgiving or controlled without proper equipment, so an acceptable one would definitely be below 30m/s for all frequencies...would be my answer. And though by calculation it is greater than 30 below 30Hz, for musical output, we do not want to try to control the phase or transience below 30Hz with as much concentration due to the fact that some of it cannot be controlled without a very low tuning, and that you will or can lose output in the upper range which will have a reverse effect on the quality of the music in general. One can effectively create a tuning resonance of around 70Hz in a design and still get smooth output around the 30-35Hz range before drop off if the coorelation of each delay is controlled all with each other as optimal as possible. There is, by my own calculations, an equalization of about 3dB difference for each 1/8th change in the phase delay from a 360 degree plane, which in turn directly effects the group delay in an anechoic calculation. This means that for every 45degree phase change in the response, there is a nominal 3dB difference with a nominal 2Hz bandwidth increase starting around 5-8Hz BW@0degrees of each 1/8th change without even accounting for time delay yet.

So, when you want to control the group delay, you HAVE to consider the phase changes that occur because they can effect the output equally as important as the time delay, which in a lot of cases for sub audio use, come from the equipment for the most part based on signal transfer over time. Frequencies are based on timing effects and when those are further from zero in any calculated variable, everything gets worse.

Even excursion can effect the time delay of a signal, especially when distortion is included or increased.

The idea is to have every resistive factor, or impedance change/rise, and phase changes to all be as minimal as possible because these all have effects on the response that is calculated and will usually not show up in a calculated response as much as you will measure in a real life graph. The audible effect is much smaller than the calculated change, but if you are minimized by drivers and cone area, and increase cabin volume, the effects become much noticeable.

When you want to get an sq response, minimal delay is always the key, whether it is on a timing basis, a placement configuration, power vs distortion factors, and even within the enclosure where less resonances occur with less manipulation. I have a pair of 3" full range drivers on my computer right now that I designed boxes for, that do not include a port length greater than the thickness of the wood, so it appears as just an opening in the box, but acoustically, this is needed to have all of that mentioned control, with the fact that I was able to concentrate on a higher frequency range than subwoofer use, but still the concept is the same and in fact over a high power, more sensitive with higher passband ranges especially since I designed them for a living room and not a car. All of the peaks and nulls in a response are important in figuring for actual response accuracy and controlling them is not easy if you are looking for a flat response. But this is why there is an acceptable range, and that range again, I would say for our situation in LF response, is less than 30m/s ideally. If you get around 70m/s, this can be acceptable for a more sql output, and up to 120m/s for SPL, BUT you always want it as low as possible, because 120m/s is terrible lol. That is a noticeable delay //content.invisioncic.com/y282845/emoticons/biggrin.gif.d71a5d36fcbab170f2364c9f2e3946cb.gif.

 

[RSDXzec 10+ year member]

Well, the group delay would not be the only issue in a smooth response, first thing. There are other factors that control that as well. I have a way in my calculations to figure for db gain and loss based on phase changes as well, and it is very acceptable. That is where the direct audible output will be effected as far as output or the mentioned drop below 50Hz. Now, with group delay, the smaller the better in all cases. The delay can be acoustical or electrical also, so not all of it can be forgiving or controlled without proper equipment, so an acceptable one would definitely be below 30m/s for all frequencies...would be my answer. And though by calculation it is greater than 30 below 30Hz, for musical output, we do not want to try to control the phase or transience below 30Hz with as much concentration due to the fact that some of it cannot be controlled without a very low tuning, and that you will or can lose output in the upper range which will have a reverse effect on the quality of the music in general. One can effectively create a tuning resonance of around 70Hz in a design and still get smooth output around the 30-35Hz range before drop off if the coorelation of each delay is controlled all with each other as optimal as possible. There is, by my own calculations, an equalization of about 3dB difference for each 1/8th change in the phase delay from a 360 degree plane, which in turn directly effects the group delay in an anechoic calculation. This means that for every 45degree phase change in the response, there is a nominal 3dB difference with a nominal 2Hz bandwidth increase starting around 5-8Hz BW@0degrees of each 1/8th change without even accounting for time delay yet. So, when you want to control the group delay, you HAVE to consider the phase changes that occur because they can effect the output equally as important as the time delay, which in a lot of cases for sub audio use, come from the equipment for the most part based on signal transfer over time. Frequencies are based on timing effects and when those are further from zero in any calculated variable, everything gets worse.

Even excursion can effect the time delay of a signal, especially when distortion is included or increased.

The idea is to have every resistive factor, or impedance change/rise, and phase changes to all be as minimal as possible because these all have effects on the response that is calculated and will usually not show up in a calculated response as much as you will measure in a real life graph. The audible effect is much smaller than the calculated change, but if you are minimized by drivers and cone area, and increase cabin volume, the effects become much noticeable.

When you want to get an sq response, minimal delay is always the key, whether it is on a timing basis, a placement configuration, power vs distortion factors, and even within the enclosure where less resonances occur with less manipulation. I have a pair of 3" full range drivers on my computer right now that I designed boxes for, that do not include a port length greater than the thickness of the wood, so it appears as just an opening in the box, but acoustically, this is needed to have all of that mentioned control, with the fact that I was able to concentrate on a higher frequency range than subwoofer use, but still the concept is the same and in fact over a high power, more sensitive with higher passband ranges especially since I designed them for a living room and not a car. All of the peaks and nulls in a response are important in figuring for actual response accuracy and controlling them is not easy if you are looking for a flat response. But this is why there is an acceptable range, and that range again, I would say for our situation in LF response, is less than 30m/s ideally. If you get around 70m/s, this can be acceptable for a more sql output, and up to 120m/s for SPL, BUT you always want it as low as possible, because 120m/s is terrible lol. That is a noticeable delay //content.invisioncic.com/y282845/emoticons/biggrin.gif.d71a5d36fcbab170f2364c9f2e3946cb.gif.

I took some time to take all that info in lol. I'm unsure how to read and/or understand the phase graph properly so I'm not sure what to look for when I look at that graph. I have a very basic understanding of subwoofer response because up until now I've only seen things like this in RLC circuits so I'm still trying to make the link to audio.

If 30ms is still good for sq then I'm doing great, I can't imagine what a 120ms delay would be like.

Thanks for the solid explanation.