Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Michael Lawrence]

I can tell you how it works in Smaart, which is an approach commonly implemented in other real-time analyzers as well. Typically your Delay Finder is just looking for the highest magnitude value peak in the IR (which is calculated as a realtime IFT of the TF). It finds the highest peak (absolute value, so if the largest peak is negative-going it'll still grab on to it) and then simply sorts out how much delay is needed to move that peak to t0. Optionally this can be done using ETC if desired, it might shift you a few samples either way.

https://imgur.com/ZslZDDr

Here that answer is "just less than 4 ms". (I am measuring relatively close to a lousy loudspeaker in a small reflective office.)
Once that delay is inserted:

https://imgur.com/TDD6ON1

As Ivan has pointed out we are now timed to the HF range in particular (flat-trending phase trace there) and the LF is arriving later (downward trend from left to right). The Live IR is biased towards high frequencies so it doesn't work very well for subs. That is for a few reasons but we can bandpass filter the IR if we want to lock onto a particular frequency range. For example, here is what we get with 125 Hz bandpass filter:

https://imgur.com/RSQdUmQ

After the filter is applied, it's the same game: find the peak in the IR. As you can see it's often not so clear cut and can be fooled by other arrivals from room boundaries and so forth, but our answer is now a full 15 ms later than it was before. If we insert that value into the measurement delay:
 
https://imgur.com/wvnfpmo

Now the phase trace is trending flat in that region (we are timed to that energy) and the higher frequencies are trending upward left to right (arriving earlier than that).

Thanks to the group delay of the LF there isn't one clear cut arrival time, as it's smeared out a bit - Ivan I really like your metaphor about the line on the road - , so the best we can do is to say "at what frequency." That's where the bandpassed IR trick comes in quite handy.

To clarify - we are not syncing to a particular point in the measurement signal's waveform - we are taking a source independent measurement. We are looking at the waveform of the impulse response of the system under test.

[Mark Wilkinson]

So use a auto delay finder, with just a sub (no full range cabinets) with low pass filter engaged.  See what it tells you.  Now see if that distance is what the distance from the sub to the mic is.  I have never seen it be anywhere near close.

If you disengage the low pass filter, it will be more accurate, because there is more high freq information available (ie shorter wavelengths).  Of course the actual phase trace will tell you (depending on how it tilts) how close the measured distance is from the actual distance.  Of course you also have to take into account any digital delays (either actually set or AD/DA convertors).

It is not necessarily the peak, but is not the "onset" of the waveform getting there.

Yep, I guess most all of us have seen how much an auto delay finder bounces around, when just measuring a low-passed sub.

I had a delay finder shootout comparing Smaart, Systune, and REW some time ago, where the test was to see how much each program varied through about a dozen 'finds'.
As expected, they all gave the same time, with essentially zero variance, when there was sufficient HF content.
But sub range of course got interesting....and plenty squirrely.

But back to onset vs peak...
I've found the better the transfer function measurement, the flatter the mag and phase of the measurement, 
that it becomes clear the programs are "trying" to make calculations that align onsets together.

"Trying"....because like you say, without enough HF content, the measurement programs don't have enough data to nail down timing.

An electrical transfer function of a full range, flat phase signal, will give a perfect looking Impulse Response, with let's say peak at t=0. 
(Like what we get running a transfer function, with  the exact same signal output on reference and measurement channels.)
 
Looking at that full-range Impulse with bandwidth limiting in place, shows the timings inherent for impulses of different octaves. 
Whatever  frequency range is chosen, the bandwidth limited impulse can be seen to have its onset at the t=0 timing of the full-range Impulse.
I've found REW is great for seeing this, it has more bench type analytics than Smaart.
I also find i get the same measurement results and timings using either program, so i do believe REW is just digging in a little deeper graphically to show how it works.

That timings are set by onsets is easy to see with such a clean full range impulse response, because the measurement program has all the data to make it all come together perfectly.

Cleary not the case with a sub.  But I think the point is, the program is trying to do for the sub the same thing it does with the clean full range impulse.
So the more the delay finder looks like it timed to peak rather than onset,  the crappier the measurement.

[Michael Lawrence]

Acquisition method is a big variable there.
Doing a sweep of a few seconds as in REW will net you a higher signal to noise ratio than a realtime analyzer's Live IR (which is made to be as fast and reactive as possible - 8K FFT with 4 averages by default in Smaart - which is enemy of averaging and longer acquisition times). If you flip Smaart over to IR mode and use either sweep or period-matched pseudo-random pink with a few averages you can trade in some acquisition time for really digging down deeper into the dynamic range. Here is 16K FFT, 0 averages, 32K FFT 8 averages and 256K FFT 8 averages all with period-matched pink.

https://imgur.com/yRqGUcU

We get the same answer in terms of frequency response which is why if you are just measuring to EQ / time align a PA, I am not going to use a slower acquisition method to increase SNR, I already have my answer. But if we are studying the room acoustics etc, the algorithms that calculate all those associated metrics are a lot happier and more accurate if we have a higher SNR in the measurement and it is worth the extra acquisition time.

Horses for courses.

[Mark Wilkinson]

Acquisition method is a big variable there.
Doing a sweep of a few seconds as in REW will net you a higher signal to noise ratio than a realtime analyzer's Live IR (which is made to be as fast and reactive as possible - 8K FFT with 4 averages by default in Smaart - which is enemy of averaging and longer acquisition times). If you flip Smaart over to IR mode and use either sweep or period-matched pseudo-random pink with a few averages you can trade in some acquisition time for really digging down deeper into the dynamic range. Here is 16K FFT, 0 averages, 32K FFT 8 averages and 256K FFT 8 averages all with period-matched pink.

https://imgur.com/yRqGUcU

We get the same answer in terms of frequency response which is why if you are just measuring to EQ / time align a PA, I am not going to use a slower acquisition method to increase SNR, I already have my answer. But if we are studying the room acoustics etc, the algorithms that calculate all those associated metrics are a lot happier and more accurate if we have a higher SNR in the measurement and it is worth the extra acquisition time.

Horses for courses.

Hi Michael, thx for reply.

Yep, horses for courses. And please don't think i was saying REW is a better tool for finding sub timing.

My 'course' is always the same, that of a speaker builder trying to obtain the best quasi-anechoic measurements as possible (without a chamber other the great outdoors haha)
Haven't worked on room acoustic's in years.

So for my two horses, Smaart and REW, on their outdoor course, I've found I get more consistent sub timing with Smaart. 
It appears REW needs more HF spectral content than Smaart does, to do its transfer function math, with the same repeatability Smaart has. 
I don't know if this is due to Smaart's multi time windows, or to the temporal averaging capability.
I just know Smaart works better for low frequency work and is my go-to measurement program. (and that pink noise doesn't piss off neighbors like log swept sine sirens do lol)

What I like about REW is it lets' me look under the hood of the measurements, which I believe are probably much alike between programs. It's various high rez graphical breakdowns help, particularly for impulse responses, and other time domain looks. I often export IR's made in Smaart to REW to look into them deeper (it's easier to choose IR FFT size and averaging in Smaart.)

Anyway, apologize to the thread for swerving into technicalities regarding transient response, and even bringing up Smaart vs REW, etc.

But so help me, when i cut through all the horse crap audiophool impressions, and putting too much attention on measurements that don't matter cause they are buried so deep under SNR, or impossible to fix multiple acoustic source issues...etc
Well, simple mag and phase keep proving (to me anyway) that they are the gut science underneath it all that i can actually do something about....and the key to transient response.
I like to think my speaker builds have moved beyond 'punch' and into the world of 'slam' 

[Michael Lawrence]

Mark - no worries at all, it is definitely an interesting conversation, if derailed a bit from the original topic. I don't ever view it as a competition - at the end of the day you want to use the best tool to get the results you need. Simple as that.

m

[Marcel de Graaf]

Hi All,

Wooo i catched this thread to late. This sort of threads keep me visiting this forum board.

There is an ongoing debate about the existence off punch, transiënt slam etc...... It keeps us interested because its recognizable when its there, but its still not defined with a measurement in a manner we can reprocedure it. Here`s my point of view.

When we notice a low frequency speaker has a lot off punch it`s mostly related to the high frequency and the distortion components of the speaker itself. Well it can sound musical to our ears, its not part off the original input signal. As the goal of a speaker system to reproduce the input source its techical wrong to have a source that has a signature at his own. All this signature is mostly attributed by the loudspeaker diaphragm.

A low frequency loudspeaker still can sound very musically (and much better to me) as it`s very fast and only producing low frequency content. Therefore the most important factor is to have a motor structure with a very low inductance and a rigid, non flexing cone. A example, the contrabass servodrive subwoofer use a motor that has the lowest inductance vs motor strenght, the system has very heavy riged diaphragms and still is the fastest, cleanest low frequency speaker i have ever heard. It`s still sound very fast even with steep low pass and high pass filters engaged. In terms of musical terms i would quote this as "slam" rather than punch.

A measurement number thats expressing the difference of the waveform (acoustic) vs the input source would be a more appropriate way to quantify this.

Gr. Marcel

     

[Ivan Beaver]

Hi All,

Wooo i catched this thread to late. This sort of threads keep me visiting this forum board.

There is an ongoing debate about the existence off punch, transiënt slam etc...... It keeps us interested because its recognizable when its there, but its still not defined with a measurement in a manner we can reprocedure it. Here`s my point of view.

When we notice a low frequency speaker has a lot off punch it`s mostly related to the high frequency and the distortion components of the speaker itself. Well it can sound musical to our ears, its not part off the original input signal. As the goal of a speaker system to reproduce the input source its techical wrong to have a source that has a signature at his own. All this signature is mostly attributed by the loudspeaker diaphragm.

A low frequency loudspeaker still can sound very musically (and much better to me) as it`s very fast and only producing low frequency content. Therefore the most important factor is to have a motor structure with a very low inductance and a rigid, non flexing cone. A example, the contrabass servodrive subwoofer use a motor that has the lowest inductance vs motor strenght, the system has very heavy riged diaphragms and still is the fastest, cleanest low frequency speaker i have ever heard. It`s still sound very fast even with steep low pass and high pass filters engaged. In terms of musical terms i would quote this as "slam" rather than punch.

A measurement number thats expressing the difference of the waveform (acoustic) vs the input source would be a more appropriate way to quantify this.

Gr. Marcel

   

A speaker that is a "faster" than another, is one that has a longer excursion with a given input voltage.

At any freq, the speaker will only move as fast as that freq.  At 60 hz (for example) all loudspeakers will move back and forth 60 times per second.  If they don't then something is REALLY REALLY wrong.

But one with a longer excursion per voltage will travel further, so it MUST go faster from one end of the excursion to the other.

Of course this extra excursion brings its own set of problems.

[duane massey]

A speaker that is a "faster" than another, is one that has a longer excursion with a given input voltage.

At any freq, the speaker will only move as fast as that freq.  At 60 hz (for example) all loudspeakers will move back and forth 60 times per second.  If they don't then something is REALLY REALLY wrong.

But one with a longer excursion per voltage will travel further, so it MUST go faster from one end of the excursion to the other.

Of course this extra excursion brings its own set of problems.

And this is kind of the question that I obviously didn't iterate well. Back in the 70's we built very, very large horns, straight horns with restricted throats. Compared to other designs the speaker excursions were very small to achieve what we perceived to be higher outputs, especially at lower freqs. The signal to them was hi-passed at 25hz and low-passed at 80hz. To my ears, there was no "overshoot" or "boominess", just a solid thud or thump. We were very low-tech, no measuring gear, so everything was relative to what we heard.
Most of the conversations here are well above my knowledge, but it has made fora very interesting dialog. Thanks!

[MikeHarris]

been hesitant to bring up what had been a glaring issue to me.
some variable rails amplifiers...Crown MA for example..have difficulty with the leading edge of bass wave..notably kick drum attack.
i have seen over power related driver failure from using level to recover that feel.
Linea..Powersoft...Lab dont exhibit this

[Marcel de Graaf]

been hesitant to bring up what had been a glaring issue to me.
some variable rails amplifiers...Crown MA for example..have difficulty with the leading edge of bass wave..notably kick drum attack.
i have seen over power related driver failure from using level to recover that feel.
Linea..Powersoft...Lab dont exhibit this

I have experienced the same with some other amplifiers. One time i had the very solid reputable amplifer EV P3000 to my low frequency part and switch over
to a crest pro 9200. I couldn`t believe how much faster (if that is the correct word to use) and impact that amplifier gave to the system.   

[ProSoundWeb Community]