Measuring Time Delay

[MZKM]

The thing about the delay of the subwoofer caused by DSP is that the main speakers are receiving signal from the same DSP.

I was referring to the internal DSP of the subwoofer. You don’t just hook up a speaker amp to a subwoofer driver.

 

[andreasmaaan]

The delay of the main speakers was less then 0.10 ms. I also found the delay of the sub confusingly long. It is not within site of the listening position but it is only a few feet back from the speakers at most. The thing about the delay of the subwoofer caused by DSP is that the main speakers are receiving signal from the same DSP. The signal for the main speakers is only being modified with a high pass filter and delay.

The subwoofer will have it's own group delay which will vary with frequency. Depending on the subwoofer, this group delay may exceed 30ms at low frequencies. There's a reasonably good explanation given here.

If the subwoofer has a high-pass filter incorporated into its design (for example to protect a vented woofer from over-excursion), this will increase the group delay even further.

However, I'm not sure I understand exactly what your software is doing, so I can't say whether or not this is an explanation in your specific case.

 

[Kvalsvoll]

I still don't understand what is causing such huge delay in subwoofer. Which miniDSP device are you using? Maybe it would be more clear if you post screenshot of your config screen.

Btw, are you using high pass filter on your mains to roll off bass to make life easier for them? If that is so I think that is a good idea.

Which sub are you using? Are you setting low pass filtering on the sub or within miniDSP software?

Those delay measurements are not useful for the intended purpose. They are wrong, because delay depends on frequency, and can not be expressed as a single number.

Typically, the delay is longer at very low freq - caused by the system roll-off. Moving up in freq the room starts to dominate, causing quite large variations. The overall delay also depends on the low-pass filter, lower cut-off means more delay, higher order filtering means more delay.

@mitchco has showed how this can be fixed, requires Audiolense, which many people may not have.

But you can still get the integration part right. Set crossover freq and slopes, set delay on main speakers, observe that mains and bass system actually integrates into one sound source. To get it right requires measuring, and being able to understand what the measurements mean.

 

[TimW]

Just to clarify, the subwoofer I am using was a kit from parts express. It came with the SPA250DSP plate amplifier. This amp has simple DSP capabilities and it is all that I am using in this system, no minidsp. My signal chain consists of TV>toslink>SMSL Sanskrit 10th>rca>SPA250DSP>rca>stereo amplifier. Here is the block diagram of the SPA250DSP:

I don't own Audiolense and I'm not aiming for perfection in this system but I do want to know how to do this correctly. I tried varying the delay while looking at an RTA of the system but there was no noticeable change in frequency response. I suppose I should be looking at other graphs to see whether the sub is integrating into the system properly.

 

[andreasmaaan]

Here is the block diagram of the SPA250DSP:

The inherent group delay of the sub + the group delay added by the subsonic filter is (I suspect) why you're seeing 30ms+ figures when you measure. Anyway, there's no need to try to do it this way.

I tried varying the delay while looking at an RTA of the system but there was no noticeable change in frequency response. I suppose I should be looking at other graphs to see whether the sub is integrating into the system properly.

The wavelengths are very long in that frequency range and the majority of the energy in the room is from reflections, which is why you probably aren't seeing much change in the frequency response while varying the delay, and also why you're not hearing much of a difference either. This is not a bad thing, it just means there's a wide margin for error.

I didn't catch earlier what software you were using, but is it possible to dial in whatever your desired crossover is, set the delay for both subs and mains to zero, and then use your measurement software to generate a step response similar to the one Mitchco posted in post #13?

 

[Krunok]

.. but is it possible to dial in whatever your desired crossover is, set the delay for both subs and mains to zero, and then use your measurement software to generate a step response similar to the one Mitchco posted in post #13?

I agree, that would be usefull to have.

@TimW , make a sweep measurement of both main speakers. Keep microphone at app 0.5 meters from the speakers otherwise reflections will affect step response too much.

 

[Krunok]

Those delay measurements are not useful for the intended purpose. They are wrong, because delay depends on frequency, and can not be expressed as a single number.

Typically, the delay is longer at very low freq - caused by the system roll-off. Moving up in freq the room starts to dominate, causing quite large variations. The overall delay also depends on the low-pass filter, lower cut-off means more delay, higher order filtering means more delay.

@mitchco has showed how this can be fixed, requires Audiolense, which many people may not have.

But you can still get the integration part right. Set crossover freq and slopes, set delay on main speakers, observe that mains and bass system actually integrates into one sound source. To get it right requires measuring, and being able to understand what the measurements mean.

If I understand correctly delay measured by REW is a difference between expected initial raise of IR and the one recorded by microphone, so, when measured with fixed timing reference, it simply reflects the time delay when sweep sound from various speakers reached the mike. In that context it is usefull when distance of various speakers varies significantly. It has nothing to do with delay variation caused by frequency - this one is a "physical" delay of the whole speaker caused by the time sound takes to travel from the speaker to the mike. That figure he can enter in his SW so it makes sense to measure it.

With @mitchco setup things are different as he has 2 subs and each sub is adjacent to the main speaker so they practically act as one speaker. If @mitcho decides to move his subs say 2 meters from main speakers he would also have to adjust the delay although phase of the subs would stay correct as the sound from subs will simply have longer distance to travel and this would cause the delay.

So, in my opinion the single number delay figure compensates for the fact that sub is far away from the main and you have to set it correctly. When you have done that than you should have a situation similar as when sub is below your main so you can start playing with phase of the sub to get frequency dependent timing right.

Btw, as @TimW 's concern is about configuring delay for mains I am assuming he has already set correct volume and low pass for sub and high pass for mains resulting in the linear frequency response.

 

[Kvalsvoll]

If I understand correctly delay measured by REW is a difference between expected initial raise of IR and the one recorded by microphone, so, when measured with fixed timing reference, it simply reflects the time delay when sweep sound from various speakers reached the mike. In that context it is usefull when distance of various speakers varies significantly. It has nothing to do with delay variation caused by frequency - this one is a "physical" delay of the whole speaker caused by the time sound takes to travel from the speaker to the mike. That figure he can enter in his SW so it makes sense to measure it.

With @mitchco setup things are different as he has 2 subs and each sub is adjacent to the main speaker so they practically act as one speaker. If @mitcho decides to move his subs say 2 meters from main speakers he would also have to adjust the delay although phase of the subs would stay correct as the sound from subs will simply have longer distance to travel and this would cause the delay.

So, in my opinion the single number delay figure compensates for the fact that sub is far away from the main and you have to set it correctly. When you have done that than you should have a situation similar as when sub is below your main so you can start playing with phase of the sub to get frequency dependent timing right.

Btw, as @TimW 's concern is about configuring delay for mains I am assuming he has already set correct volume and low pass for sub and high pass for mains resulting in the linear frequency response.

The delay number is not useful because it does not relate to the actual delay at the crossover. The easy way to do this is simply to measure the system including subwoofers, look at the measurement, and enter a delay that gets you close around the crossover. Then fine tune to get the summation correct- which in practical systems means as good as possible.

At 100hz one period is 10ms. If you are off by a multiple of 10ms, it will still sum correctly at the crossover, but it will be more or less wrong at freqs outside. If you do a textbook 4. order BW there will be a delay corresponding to the total phase shift of 360 deg, so you will not end up with flat phase and the lower freqs will be delayed. But in practical systems the roll-off is affected by speakers and subwoofer and room and placement.

 

[TimW]

I didn't catch earlier what software you were using, but is it possible to dial in whatever your desired crossover is, set the delay for both subs and mains to zero, and then use your measurement software to generate a step response similar to the one Mitchco posted in post #13?

I'm using REW for measurement and the Dayton Audio SPA250DSP software for changing DSP settings. I can only set delay for the mains, not for the sub. There is probably a way to generate a step response with REW but idk how or what to look for on the graph.

 

[RayDunzl]

There is probably a way to generate a step response with REW but idk how or what to look for on the graph.

Step Response is calculated from the measurement sweep.

 

[TimW]

Step Response is calculated from the measurement sweep.

So is this my step response?

It doesn't look anything like Mitcho's and I assume that's because I haven't set someting properly.

 

[TimW]

The easy way to do this is simply to measure the system including subwoofers, look at the measurement, and enter a delay that gets you close around the crossover. Then fine tune to get the summation correct- which in practical systems means as good as possible.

I messed with delay, phase, and crossover settings while looking at the RTA and this was the best summation I could get (right speaker + sub):

The dips from 100 to 300 Hz are as small as I could make them. This is with 12 db/octave high and low pass filters at 125 Hz, sub is 180 degrees out of phase, no subsonic filter, and there is a 3.4 ms delay on the mains. Interestingly I had tried to tune the delay by ear before ever measuring the system and I had ended up with 3.3 ms of delay.

 

[RayDunzl]

So is this my step response?

Step is observed over a shorter period, like, 10ms, with the % (not dBfs) display

 

[TimW]

Step is observed over a shorter period, like, 10ms, with the % (not dBfs) display

Ah I see. Well here is mine in all of its ugliness:

 

[RayDunzl]

Well here is mine in all of its ugliness

It's not ideal...

 

[Kvalsvoll]

I messed with delay, phase, and crossover settings while looking at the RTA and this was the best summation I could get (right speaker + sub):
View attachment 26774
The dips from 100 to 300 Hz are as small as I could make them. This is with 12 db/octave high and low pass filters at 125 Hz, sub is 180 degrees out of phase, no subsonic filter, and there is a 3.4 ms delay on the mains. Interestingly I had tried to tune the delay by ear before ever measuring the system and I had ended up with 3.3 ms of delay.

Good, now you can decide you are done, and put on some music and enjoy.

If you still want to explore and find out whether the delay you just set is correct, you can look at the group delay and step response. For bass you should have a time scale covering at least 200ms on the step response. But since it is not possible to add more delay than 10ms, you can choose to decide you are done, there is not much point in exploring something that can not be easily changed.

 

[Krunok]

The easy way to do this is simply to measure the system including subwoofers, look at the measurement, and enter a delay that gets you close around the crossover.

I agree. And getting the delay figure that will get timing to the sub/mains crossover point is exactly the purpose of that measurement with fixed timing reference signal described on the miniDSP site. Ability to have the same speaker playing timing reference signal has been built in REW exactly for that purpose.

Then fine tune to get the summation correct- which in practical systems means as good as possible.

Agree again. But that you have to do via phase correction on the sub signal and eventually on the main signal as you will most probably do a bass roll-off on them.

 

[Krunok]

I messed with delay, phase, and crossover settings while looking at the RTA and this was the best summation I could get (right speaker + sub):
View attachment 26774
The dips from 100 to 300 Hz are as small as I could make them. This is with 12 db/octave high and low pass filters at 125 Hz, sub is 180 degrees out of phase, no subsonic filter, and there is a 3.4 ms delay on the mains. Interestingly I had tried to tune the delay by ear before ever measuring the system and I had ended up with 3.3 ms of delay.

This looks pretty good and most probably sounds prety good as well. I wouldn't worry too much about the step response/phase response as those are far less important then getting the frequency response right.

 

[andreasmaaan]

Ah I see. Well here is mine in all of its ugliness:
View attachment 26776

Ok, where did you take that measurement in the room? Can you describe the room dimensions and where the speakers are in the room?