Thank you

Will have a go at this when my dsp and amps have been installed.

 

Hanatsu,
have you try comparing the result between band passed excitation signal vs adjusting by looking at the knee insted of the peak? That's where the impulse start to peak... Point where it left 0v mark...

 

Eh not sure I follow. Compare what?

Sent from my Samsung Galaxy 3 via Tapatalk.

 

Excellent write-up. Is it just the numbers that are different or were the measurements totally different? Did they both give you the same delay calculations?

 

Measurements were completely different. The method is accurate though. Tried a speaker at 10cm distance, then moved it 0.5cm away and both programs picked it up easily.

Sent from my Samsung Galaxy 3 via Tapatalk

 

Looks good. Will have to give it another read and a try.

 

Nice! Thank you!

 

Thanks for this, Hanatsu! This is very timely for me...

One problem I had was with the "use loopback for timing reference" switch. Since I use a USB mic, it was not clear how I can set a loopback and use it for a timing reference. My output is the PC soundcard but my input can either be the mic or the input of my soundcard. It seems that what I would need to make this work would be to have the program (REW, aRTA, HolmImpulse, etc) take one input channel (e.g. Left) from the mic and the other input channel (e.g. Right, in this example) from the soundcard. Then I could loop back the right channel and use that for the timing reference. But that doesn't seem possible ...

Since I ran into the above constraint, I've tried an alternative approach as described in Advanced Crossover Design. The approach is not as direct as using the impulse measurement but it seemed to be the only reliable approach available given that I was limited to a usb mic. The basic idea is to measure identical sweeps using each driver separately and also one with both operating concurrently (with no delays). From the individual responses (both amplitude and phase needed) the sheet computes what the summed response would look like for any given delay between the 2 individual responses. All that is left is to choose the delay that makes the computed trace look identical to the one we measured with both drivers firing.

This works, but is more convoluted/involved and if possible I'd like to be able to do/check this using REW, aRTA, etc.

Which leads me to the question : Is there a way to have an appropriate timing reference that REW/aRTA can use for those of us that have usb mics?

 

I'm not 100% certain, but I don't think you can use USB mics. HOLMimpulse have a function which can be used for time aligning without loopback, dunno exactly how reliable it is though.

 

From what Erin was saying, a mobile pre or transit would be needed. I picked up a pre but haven't tried it yet.

 

"mobile pre" or "transit"? Can you elaborate on these? Also, I could have sworn NPdang had done impulse measurements with a USB mic. It would really suck if my EMM6 mic can't be used for setting TA via impulse.

 

Hanatsu-

It seems from your diagram that if I had a laptop with one "mic in" jack and one "headphone out" jack, that I would need 2 dual jack splitters to give me a total of 4 jacks, correct? if so, aren't "mic in" jacks on laptops usually mono signal inputs?

 

Yes you would need 4 jacks in total. Mic input on my laptop is 2ch as far as I know. Be sure to set the input as 'Line-in' and not 'microphone'.

Sent from my Samsung Galaxy 3 via Tapatalk.

 

So long as the latency doesn't change between measurements, then latency doesn't matter, since we're concerned about the RELATIVE delay between speakers, not the ABSOLUTE delay of any speaker. This means that the loopback measurement isn't necessary. It eliminates the latency from the displayed measurement.

Making a loopback won't eliminate the error caused by additional latency caused by running other programs, etc.

There's one HUGE issue that no one is talking about here and it's really important. Picking the top of the peak is NOT the way to set delay, unless every speaker is a high-frequency speaker. The slope of the peak IS high frequency content. When we measure a tweeter, we can pick the peak because the peak is sharp and it's close enough to the picking the point at which the impulse starts. This is NOT the case with low frequency signals, where the slope is more gradual. The tip of the peak for a tweeter is NOT the origin of the sound and neither is top of the hump in the measurement of a low-passed signal. The beginning of the rise is the origin. If you're using the %FS in REW, for example, back up (leftward) on each measurement line to a point that corresponds to something like 5% and compare the measurements there.

 

So long as the latency doesn't change between measurements, then latency doesn't matter, since we're concerned about the RELATIVE delay between speakers, not the ABSOLUTE delay of any speaker.

Good point.

There's one HUGE issue that no one is talking about here and it's really important. Picking the top of the peak is NOT the way to set delay, unless every speaker is a high-frequency speaker. The slope of the peak IS high frequency content. When we measure a tweeter, we can pick the peak because the peak is sharp and it's close enough to the picking the point at which the impulse starts. This is NOT the case with low frequency signals, where the slope is more gradual. The tip of the peak for a tweeter is NOT the origin of the sound and neither is top of the hump in the measurement of a low-passed signal. The beginning of the rise is the origin. If you're using the %FS in REW, for example, back up (leftward) on each measurement line to a point that corresponds to something like 5% and compare the measurements there.

Great point. My main speakers are basically "full range" (more precisely, coax), so this was a non-issue for me. However this will be an issue when doing TA between active component speakers, or main speakers and subwoofers.

 

The part I do not understand is if you do this in pairs like align tweeters, then mids, then midbasses, etc. with respect to each pair only in isolation from the group. Is that assumption correct? Or do we at some point need to align a tweeter with a mid, and/or a mid with a midbass for a tighter overall time alignment?

 

All drivers should be T/A to each other. In which order you do it doesn't matter. I like to T/A each mid, each tweeter etc to eachother first then move both sides with an equal amount to another set of drivers until the align properly...

Tapaaatalk!!

 

Hanatsu, Andy

Still have some reading to do here but just wanted to say thanks!

I tried to figure it out on the rew where the time measurements would be but never got back to it and on arta I got some redicilous results, they did work but I must have multiplied something somewhere.

You guys sure saved me a lot of time!

By the way I noticed you did not enter any calibration files, I guess that for time response it does not really matter.

I remember that I created a calibration file that has the loopback response in it so that is what I am using as calibration file when tried to test for the impulse. Need to try it again sometime.

 

you don't need a loopback because all you need to find is relative delay. The delay through the soundcard is a constant. (A+X)-(B+X)=A-B

Oops. Edited. I already wrote this.


there's one other thing to consider because all of this focus on accuracy to the tiniest increment isn't necessary. You can only delay the signal by one sample. So...if your DSP samples at 48k (most do), then the adjustment increments are .28 inches. Once the destructive interference caused by the phase error is pushed out of the passband of the speakers you're aligning, then the error doesn't really matter in terms of altering frequency response. For imaging, getting the midbass and midrange right is important. Backing up from the peak by 12dB is absolutely sufficient to determine the approximate location of low passed speakers. For really high frequencies, we don't hear phase very well, so errors in tweeter alignment are audible as frequency response aberrations, which can be "fixed" well enough with EQ.

 

You need loopback if the software reset IR at 0ms for each measurement. HolmImpulse can lock the relative delay through the soundcard but I'm unsure if REW/aRTA can do it.

Tapaaatalk!!

 

This is a very good topic and a great way to measure TA down to a ms, worked like a charm, thanks for that.

However, i would like to emphasize on a point that EQ and general db output of a speaker can have a much greater influence on perceived sound than TA.

As Andy has mentioned, when it comes to high frequencies and some of the low range spectrum, we pick up sound location from whatever source is the loudest, so if you EQ your tweets or midbass incorrectly, you WILL end up with a wandering sound stage even if your TA is spot on.

If someone could provide more info on how to combat this issue by other means than avoiding using separate channel EQ'ing in these frequency ranges, i would be really grateful as i'm struggling with this atm.

 

If someone could provide more info on how to combat this issue by other means than avoiding using separate channel EQ'ing in these frequency ranges, i would be really grateful as i'm struggling with this atm.

The best way to tackle L/R balance is to have an eq per driver. At the very minimum you need an eq that is L/R independent.

 

How did you measure FR?

Verify that the EQ is centered by listening to these tracks. Lower frequencies like 200-400Hz should originate from the same spot as 4000-8000Hz for example.

https://www.dropbox.com/s/zoc2vgb4e44ys2a/Disc%201%20%28Tuning%29.rar

https://www.dropbox.com/s/mi6lx03nqpp7lxy/Disc%202%20%28Testing%20%26%20Verification%29.rar

 

How did you measure FR?

Verify that the EQ is centered by listening to these tracks. Lower frequencies like 200-400Hz should originate from the same spot as 4000-8000Hz for example.

https://www.dropbox.com/s/zoc2vgb4e4...8Tuning).rar

https://www.dropbox.com/s/mi6lx03nqp...ication).rar

I use a laptop with external sound card and a good mic running spectraRTA software.
I use the 7 drums track to check for centre image

https://www.dropbox.com/s/e0qql0pulk0cqpw/7 drums.flac

Anyhow, after EQing seprate sides of my system that had a perfect centre image, it has shifted both in tweeter range and in midbass range, midrange was unaffected and i was still able to get a good centre image on midrange drivers despite heavy EQing.

Is my EQing technique flawed ? From what i can only assume is the reflection problem, midbass area requires pretty drastic EQ changes in different areas on different drivers, i wont go into detail as this topic is not the place for it, but I have read plenty of threads about RTA tuning to make sure i do everything correctly, but my L&R speaker response differences are pretty drastic without any EQing even though they are level matched and have a very similar response when measured on-axis.

 

Well my question was not really if I need a loopback - I've mentioned the loopback as a part of my calibration file.

The question was if really there is a need when measuring impulse response for a calibration file at all - but I guess the answer is no as all we are measuring is relative time so it does not matter if the MIC and PC are calibrated correctly.

Thanks!

Eddie

 

Don't understand that part. Which calibration file? For the soundcard? How can the loopback 'be part of the calibration file'?

Differences of 10dB is common between L / R midbass drivers even if they measure the same in nearfield. It's the modal region of the car, where the speakers are mounted in relation to the listening space affects the response.

So how do you measure exactly? Are you in the car? Averaging? Sweeps or noise? One driver at a time or?

Tapaaatalk!!

 

There is a way to create a calibration file that takes in acount the loopback and the MIC calibration file - I did it a long time ago and I have both files to choose from.

The measurments I am doing is when the MIC is inside the car placed where my head is usally at when I'm driving.

As for the Time delay - of course I measure one speaker at a time and then compare the overlay.

Thanks

Eddie

 

Noi, the mic is also a constant. If your mic rolled off at 1k, then it might not be appropriate for measuring impulses from tweeters, but it is probably not the case.

 

Subbed, will try this later on today.

 

fcarpio, how did your TA session go?
Which route did you use?

 

I used TA software for quick calculation of time delay from Acoustic Power Lab. Amazing 3D pictures! TA soft can be downloaded free (demo). Takes very short time to optimize time delay for the whole system. Measured only with PC, external sound card and mic with 48V. Took less than halv an hour to adjust passive front and sub.Twitters and midbasses (in the passive setup) were automatically TA aligned with the help of another sofware from the same manufacturer. It is important to EQ each speaker before time alignment. My tweeters were rather different in sound before EQ correction, therfore it was not possible to place scene just an a center. They should play equal for optimal scene.

 

tell us more about the TA software.