ProSoundWeb Community
Sound Reinforcement - Forums for Live Sound Professionals - Your Displayed Name Must Be Your Real Full Name To Post In The Live Sound Forums => Audio Measurement and Testing => Topic started by: Irwan Prasetyo on November 12, 2012, 03:07:59 PM
-
Hi all...
what is group delay??and what the different between group delay & phase delay??
Thanks for any advice
-
Hi all...
what is group delay??and what the different between group delay & phase delay??
Thanks for any advice
http://en.wikipedia.org/wiki/Group_delay_and_phase_delay
-
Well that certainly cleared that up. Thanks Dick.
-
Try this one instead, http://www.excelsior-audio.com/Publications/Phase_Response_&_Receive_Delay.pdf (http://www.excelsior-audio.com/Publications/Phase_Response_&_Receive_Delay.pdf).
After that you might find this one interesting http://www.excelsior-audio.com/Publications/Subwoofer_Alignment.pdf (http://www.excelsior-audio.com/Publications/Subwoofer_Alignment.pdf).
-
http://en.wikipedia.org/wiki/Group_delay_and_phase_delay
Thanks Dick
-
Try this one instead, http://www.excelsior-audio.com/Publications/Phase_Response_&_Receive_Delay.pdf (http://www.excelsior-audio.com/Publications/Phase_Response_&_Receive_Delay.pdf).
After that you might find this one interesting http://www.excelsior-audio.com/Publications/Subwoofer_Alignment.pdf (http://www.excelsior-audio.com/Publications/Subwoofer_Alignment.pdf).
Thanks Charlie….I'll read it first and if i have some question i'll post in here
-
Hi all...
what is group delay??and what the different between group delay & phase delay??
Thanks for any advice
Read Charlie's article. Limitations on receive delay fidelity are an important issue in measurement.
Mathematically group delay is the negative first derivative of the phase wrt frequency, as discussed in the Wikipedia article. Now that's not very useful if you're not mathematical.
So, practically, what does group delay tell you?
Group delay takes the phase, which is one way of book keeping relative arrival times of each frequency, and provides a different view of this information. Rather than having to discern how the phase angle at any given frequency corresponds to the arrival time, a group delay plot instead shows the delay time (in ms) vs. frequency directly. A group delay plot makes it easy to see whether a woofer or horn is leading/lagging. It also make it easy to see the extra delay time added by the phase response of the filters we apply to audio signals.
In summary, it is a different and more clear means of looking at the relative arrival time of each frequency that is emerging from a device/speaker that is being measured.
-
Group delay takes the phase, which is one way of book keeping relative arrival times of each frequency, and provides a different view of this information. Rather than having to discern what the phase angle at any given frequency is, a group delay plot instead shows the delay time (in ms) vs. frequency directly. A group delay plot makes it easy to see whether a woofer or horn is leading/lagging. It also make it easy to see the extra delay time added by the phase response of the filters we apply to audio signals.
In summary, it is a different and more clear means of looking at the relative arrival time of each frequency that is emerging from a device/speaker that is being measured.
Nice explanation.
Mac
-
From my experience (and perhaps I am incorrect), the most apparent audio demonstrations of excess group delay occur in regard to transient response, esp in the SLF range. High-order bandpass boxes have enhanced steady-state sensitivity, but at the cost of hearable group delay.
In simpler terms: Keyed & plucked steady-state bass tone envelopes are fine, but kick drum and the like suffer.
-
From my experience (and perhaps I am incorrect), the most apparent audio demonstrations of excess group delay occur in regard to transient response, esp in the SLF range. High-order bandpass boxes have enhanced steady-state sensitivity, but at the cost of hearable group delay.
In simpler terms: Keyed & plucked steady-state bass tone envelopes are fine, but kick drum and the like suffer.
Just thinking through what you are saying...not that I agree...That the coherant phase relationship of all the frequencies (from the lowest to the highest) in a bandpass system is poor, and a gd error (relative to something else) can be heard when a steep/short attack signal like a kick hit is propagated through the system (although a plucked string can have a steep attack also), and the gd/phase errors can be heard, those errors are between the original sound (beater) and the system sound (PA)?
Maybe if LF system phase alignment is out of wack, the short kick signal is more suseptable to phase cancelations, where a sustaned bass is not as it natuarlly rolls along it's natural envelope.
Phillip, can I have clarification, pleeeeeeeeeeze?
-
From my experience (and perhaps I am incorrect), the most apparent audio demonstrations of excess group delay occur in regard to transient response, esp in the SLF range. High-order bandpass boxes have enhanced steady-state sensitivity, but at the cost of hearable group delay.
In simpler terms: Keyed & plucked steady-state bass tone envelopes are fine, but kick drum and the like suffer.
There are multiple things in play here:
- Group delay is always higher in the low frequencies as a consequence of the fact that the world we live in is full of mechanical/physical high pass filters, and time travels in one direction. We're used to additional group delay in the LF from infancy, albeit not as dramatically as we induce through processing choice or box design.
- Very few of the signal we associate as music are in any way steady state from a signals and system perspective. This is to say they do not remain consistent over a length of many time constants.
- The rapidity of the rise and fall of a percussive source's envelope is directly related to the ultimate bandwidth of the reproducing device. This is fundamental reality of information theory. The rapid rise and fall of an envelope function involves high frequencies outside of the passband of subwoofer. Ignoring the additional content generated by the nonlinear operation of the subwoofer, the nature of the attack and decay lies almost exclusively with the mid-high enclosures.
- It is fairly straightforward to match the group delay of the mid-hi boxes to the subwoofer at the crossover point.
So, if the relative group delay between the subs and the tops is not the defining factor, and much of the attack and decay is defined by the mid-hi loudspeakers, then what is going on? The following points are my personal thoughts on the matter:
- As an industry we dramatically understate the wide range of undesirable effects from the ports of our loudspeakers as the output level increases. A good amount of the ugliness we hear from subwoofers, vented or bandpass, is this.
- All minimum phase highpass and lowpass filters re-shuffle the relative phase of the frequency components in the in coming signal, with the net effect of distributing the energy of the input signal over an longer period of time. This expresses itself as group delay.
- In physics it is common to use bandpass filters (HP and LP in series) to distribute the energy of an incoming signal (e.g. neutron emission from radiative decay) over time. They call this type of filter "pulse spreading." A pulse spreader makes the peak of the pulse less tall, which makes characterizing the incoming signal easier to record electronically because of the reduced peak input level.
- A subwoofer, with is high pass and low pass filter, has the same effect as the pulse spreader filter above. It takes the incoming signal and spreads its energy out in time, causing it to take longer to decay.
- A traditional vented box has a single low pass filter and two high pass filters. One of the high pass filters is the electromechanical response of the helmoholtz resonator, and the other is the protective high pass filter to remove the energy below the tuning frequency. Lets call the LP 4th order, and the net HP 8th order (2 4th order in series).
- A dual vented enclosure with HP and LP is more like an 8th order LP and 8th order HP. This is a steeper roll off rate than the standard vented box.
- The higher order response of the bandpass box causes more pulse spreading, which increases the length of the overall decay envelope of the incoming signal.
In conclusion, any two boxes that exhibit similar pulse spreading behavior will have sonically similar character. Whether this spreading comes at the hand of electrical filter sets, or the box's inherent electromechanical behavior, the aggregate sonic character will be similar.
-
Just thinking through what you are saying...not that I agree...That the coherant phase relationship of all the frequencies (from the lowest to the highest) in a bandpass system is poor, and a gd error (relative to something else) can be heard when a steep/short attack signal like a kick hit is propagated through the system (although a plucked string can have a steep attack also), and the gd/phase errors can be heard, those errors are between the original sound (beater) and the system sound (PA)?
The sad fact of the matter is that we as humans suck at hearing phase, especially if it changes slowly with frequency. We are pretty good at picking out magnitude artifacts that arise from phase issues, but not the actual phase itself. Now, I'm not saying that flat phase response is a bad goal. In the process of flatting the phase response, you also fix minimum phase amplitude errors, and sometimes improve how the boxes play together. So its a lofty goal, but at the end of the day it is only the last extra sheen on the overall sound. Getting a loudspeaker well behaved enough to finally consider linearizing its phase response is 90% of the battle towards great sound.
Maybe if LF system phase alignment is out of wack, the short kick signal is more suseptable to phase cancelations, where a sustaned bass is not as it natuarlly rolls along it's natural envelope.
It doesn't work this way. A steady state LF tone has a given phase relationship defined by the system, as do all the higher frequency components that enable the envelope to start and stop. The only thing that is changing between a steady state tone and a transient one is the rapidity with which we are exposed to each aspect of the fixed phase relationship.
-
A steady state LF tone has a given phase relationship defined by the system, as do all the higher frequency components that enable the envelope to start and stop. [emphasis mine]
Nicely stated...
This is off-topic, but I was having a conversation the other day with someone telling me about the square waves going to subs. There is a widespread myth-conception (out there that is unfortunately promoted by many who should know better) that a square wave at say 60 Hz is exactly that and makes it through to the subs after the lowpass filter. The gentleman I was talking to had been told that the square wave low frequencies in dance music had been causing his subs to blow by "holding them" in a fixed position until the waveform starts to return to zero. I was trying to explain that if nothing was electrically clipping, the sub was not seeing the square wave at all, but rather just the low freq components of it.
Anyway, I liked the eloquence of that above statement.
-
...
6. A dual vented enclosure with HP and LP is more like an 8th order LP and 8th order HP. This is a steeper roll off rate than the standard vented box...
Why do you suggest a dual vented BP design has a net result of 16th order BP, when in literature this design is described as 6th order BP ( for example X1, Rog Mogale, speakerplans.com ) consisting of 4th order HP and 2nd order LP.
Add 4th order protective HP and 2nd order LP electric filters and maybe out of band EQ and the design will end as a net 12th order BP.
Another example is the Eighteensound kit 18 ( dual 18" BP ), which is a 5th order BP design, similar to B2 from d&b or B10 from Kling&Freitag ( Access series )
In the net result there would be no difference to the direct radiating vented design regarding Group Delay.
But electric filters do not change in demanding applications :)
Uwe
-
Why do you suggest a dual vented BP design has a net result of 16th order BP, when in literature this design is described as 6th order BP ( for example X1, Rog Mogale, speakerplans.com ) consisting of 4th order HP and 2nd order LP.
The classic double vented bandpass enclosure alignments exhibit a sixth order acoustic roll off for both the high pass and low pass sections. Add a second order high pass and a second order low pass at either end of the system, and you've got an 8th order HP and an 8th order LP.
Of course there are plenty of other ways to string together acousticallly reactive elements and produce responses of a variety of roll off rates, but that is a rabbit trail off the key physical insight I was describing.
Please note that I'm not dismissing bandpass boxes, merely noting that the spreading of energy in the time domain is a consequence of the roll off rates of the bandpass filter the signal is passed through (acoustisc + electrical). This behavior is a physical consequence of any system. The spread of the signal in time, while a reality, is not the only factor in play when designing a loudspeaker to perform a given task. It can, however, help explain people's perceptions of certain devices in a reasonable manner.
-
The classic double vented bandpass enclosure alignments exhibit a sixth order acoustic roll off for both the high pass and low pass sections. Add a second order high pass and a second order low pass at either end of the system, and you've got an 8th order HP and an 8th order LP.
It's been a while since I looked at these, but I think a double vented bandpass enclosure is sixth order, total; fourth order HP and second order LP. Adding second order electrical HP & LP results in a sixth order HP and fourth order LP system response.
-
It's been a while since I looked at these, but I think a double vented bandpass enclosure is sixth order, total; fourth order HP and second order LP. Adding second order electrical HP & LP results in a sixth order HP and fourth order LP system response.
Charlie,
Ooof, I went and looked at the equivalent circuit model and you're right. That's what I get for talking about a class of cabinets i haven't visited in a long while :(
Thankfully everything else is correct except the intrinsic roll off orders of the bandpass cabinet.
-
Since the topic of group delay is being discussed,can anyone describe the procedure to delay cabinets or subs using the group delay function? Is it as simple as delaying until the delay in ms is the same or close at the crossover frequency?
Thanks.
-
Ignoring the additional content generated by the nonlinear operation of the subwoofer, the nature of the attack and decay lies almost exclusively with the mid-high enclosures.
What direction does the nonlinear artifacts push things? Vs more group delay or less? Also how do you distinguish between where it goes from linear to nonlinear, its a gradual process right?
In conclusion, any two boxes that exhibit similar pulse spreading behavior will have sonically similar character. Whether this spreading comes at the hand of electrical filter sets, or the box's inherent electromechanical behavior, the aggregate sonic character will be similar.
Applies for MF/HF too or?
/R
-
Since the topic of group delay is being discussed,can anyone describe the procedure to delay cabinets or subs using the group delay function? Is it as simple as delaying until the delay in ms is the same or close at the crossover frequency?
This might be of interest.http://www.excelsior-audio.com/Publications/Subwoofer_Alignment.pdf. (http://www.excelsior-audio.com/Publications/Subwoofer_Alignment.pdf.)
This video of a presentation I did at an AES Convention several years ago might also be informative http://www.youtube.com/watch?v=yoASUFGPWwg (http://www.youtube.com/watch?v=yoASUFGPWwg). It's difficult to see the slides but a PDF of them is here, http://www.excelsior-audio.com/Publications/AES129_RH_Charlie_Hughes_Subwoofer_Alignment_with_a_Full-Range_System.pdf (http://www.excelsior-audio.com/Publications/AES129_RH_Charlie_Hughes_Subwoofer_Alignment_with_a_Full-Range_System.pdf)
-
Since the topic of group delay is being discussed,can anyone describe the procedure to delay cabinets or subs using the group delay function? Is it as simple as delaying until the delay in ms is the same or close at the crossover frequency?
Thanks.
Yes it is possible, but group delay is very sensitive to noise. With very good measurements it is a simple matter to adjust delay so that the frequencies of interest are aligned to the extent possible.
I did demonstrations of this very procedure a couple of years ago at InfoComm using SysTune.
-
Since the topic of group delay is being discussed,can anyone describe the procedure to delay cabinets or subs using the group delay function? Is it as simple as delaying until the delay in ms is the same or close at the crossover frequency?
Thanks.
Alfredo,
A derivative tracks the rate of change of a signal. With noise on the signal, the derivative can oscillate wildly. If you have a "smooth," low noise phase trace, then doing the alignment with group delay is as simple as you imagine. The trick is getting a phase trace smooth enough that you can generate a well behaved derivative thereof.
-
What direction does the nonlinear artifacts push things? Vs more group delay or less? Also how do you distinguish between where it goes from linear to nonlinear, its a gradual process right?
It doesn't really push them. The nonlinear artifacts generate new frequency components from the transducer, and to a first order approximation the relative phase of each of those components follows the global, linear phase response output of the speaker. The big issue is not shifting of the relative arrival, but rather production of new frequency components.
Applies for MF/HF too or?
The pulse spreading phenomenon is universally applicable for minimum phase (or greater) bandpass filters, independent of the frequencies involved. It is most obvious at low frequencies, as this is where the group delay is the largest.
-
Thanks Phil,
It is not possible to change the Group delay of a speaker with out changing the Phase response, or ?
mvh
R
-
Thanks Phil,
It is not possible to change the Group delay of a speaker with out changing the Phase response, or ?
mvh
R
You can change the total group delay by moving father away :) as the slope of the phase response due to pure acoustic delay is a straight line, and therefore the derivative is therefore a constant.
Otherwise, since the group delay is defined mathematically from the phase, they are linked at the hip.
-
You can change the total group delay by moving father away :) as the slope of the phase response due to pure acoustic delay is a straight line, and therefore the derivative is therefore a constant.
Otherwise, since the group delay is defined mathematically from the phase, they are linked at the hip.
Thanks Phil,
Really appreciate you commenting, Its such a pleasure to be able to get a clear, complicated, but correct answer to all the "grey zones" i audio knowledge, especially for one with a non mathematical background.
Its actually really fun because i have just been "grilled" in the field this past season by two parties, a competing software manufacture and a speaker manufacture, about group delay and its "difference and comparison" from the phase trace, and after showing them this post, it turns out I was right from the start, and got an apology..
/R
-
Try this one instead, http://www.excelsior-audio.com/Publications/Phase_Response_&_Receive_Delay.pdf (http://www.excelsior-audio.com/Publications/Phase_Response_&_Receive_Delay.pdf).
After that you might find this one interesting http://www.excelsior-audio.com/Publications/Subwoofer_Alignment.pdf (http://www.excelsior-audio.com/Publications/Subwoofer_Alignment.pdf).
Hello, this post is around one year ago.But i hope it's not too late to "up" this topic.
After read Charlie article, i try many times to consider group delay when alignment beetwen fullrange & sub.
I attach some pict from one of my measurement.The distance beetwen mic to spkr is around 20m.
Pict 1 is response of fullrange & sub has aligned.And we can see the summation.
But after look inside into group delay curve, this system have high group delay (pict 2).
Pict 3 is response of system after fine tune, so the group delay getting lower than before.
Pict 4 is response of system which have low group delay & we can see the phase response is not align anymore.But the freq response is still show good summation.
Pict 5 is compare group delay beetwen before & after fine tune.
Pink curve = before & Blue curve = after
My question is :
1.Why the summation is not change after the phase response not align anymore?
2.which one that you choose beetwen this two comparative response??because frequency response doesn't change, only the phase & group delay is different.
Thanks & any advice will be appreciate.