Speaker Measurement,Room Measurement & Treatment

for generating ETCs,
you need to be utilizing the hardware loopback connection and such that the signal leaving the source (speaker) is T=0 , NOT the direct signal arrival at the receiver (mic) set to T=0.

there is also a checkbox to 'utilize hardware loopback' in REW Options/Analysis and also uncheck 'set direct signal to T=0' (or equivilant --- don't have REW in front of me at the moment).
Edited by localhost127 - 6/15/12 at 12:10pm

you can't just subtract the direct flight path from the indirect (to determine indirect total) without utilizing hardware loopback as it's not linear..

the ETCs you provided show the direct signal set to T=0. that is not what you want - you want T=0 set to when the signal leaves the source, and then the direct signal will be displayed at whatever time it takes for propagation to the receiver (mic) ... eg, straight vector distance from acoustic center of source (speaker) to receiver (mic) in feet * 1.126 = direct signal flight time in milliseconds.

eg, if your receiver is 15ft from your speaker, you should see the direct signal energy on the ETC at approx 17ms NOT 0ms.

once you have done this and thus have the total flight path of a particular signal (sparse, indirect specular reflection spike of energy as measured on the ETC) - then you can work backwards to identify the particular boundary or source of the indirect reflection. for the early-early (1-3ms) energy, look for sources of edge diffraction nearby the speaker (or the speaker cabinet itself?)

http://www.avsforum.com/t/1415741/speaker-measurement-room-measurement-treatment#post_22135051

"use loopback as timing reference"

if you care about maintaining accuracy with respect to intelligibility, localization, and imaging - then yes. you will want to utilize the ETC to identify and attenuate any destructive high-gain early arriving indirect signals (reflections) that arrive within the haas interval of which the ear-brain lacks the resolution to identify them as discrete (separate) signals and thus, fuses them with the direct signal into a single auditory event. richard heyser referred to this as "time smear distortion". attenuating these high-gain early arriving signals gives the ear-brain adequate time to process the direct signal only - an effectively anechoic time-period referred to as the InterSignalDelay (ISD) gap. no high-gain indirect signals (that the ear-brain keys on for localization, imaging, etc) are allowed to impede the listening position within this time period - the direct signal is all that is processed. you naturally have a larger ISD-gap as the room's dimensions are increased (boundaries are further away and thus, indirect boundary reflections take longer in time to impede the listening position). if you have personal tastes for a completely damped room, then this effective anechoic time period (within specular region) would be infinite and no later arriving specular energy would be reintroduced to the listening position. you would utilize the ETC to identify any boundaries incident of specular indirect energies and attenuate (eg, absorb). if you do not want a completely damped room, then you can reintroduce specular energy back to the listening position at a certain time period (corresponding to a specific indirect flight path in distance), effectively terminating (sharply delineating) the ISD-gap - preferably as diffused as possible (NOT as later arriving high-gain sparse reflections) for spaciousness. eg, an exponentially decaying laterally arriving (semi)diffuse-field. unfortunately, in a small acoustical space there lacks a statistically developed reverberant sound-field and instead we deal with focused specular reflections that reflect within the acoustical space like laser beams of sound - and you can see these sparse (focused) spikes of energy within your ETC. so this is why complex geometric or reflection phase grating diffusers (or scattering surfaces) are used to break up these sparse reflections into highly complex/mixed returns. utilizing the ETC here in such a scenario will detail this as well - you can identify a boundary incident of a high-gain sparse specular reflection, place a diffuser (eg, a reflection phase grating diffuser that also offers temporal dispersion) - and see the energy on the ETC be changed from a sparse spike, to many spikes lower in gain AND spread out in time.

the ETC displays how ALL of the specular energy impedes the listening position. from the direct signal, to the sparse, high-gain early reflections, to the later reflections, to the specular room decay until the last of the energy is damped and below the ambient noise floor. you can even utilize it to identify coupling issues - as you may come across a scenario where you see measured energy arriving before the direct signal reaches the receiver (mic) --- which may confuse you at first until you realize transmission speed is much faster via solid than air medium.

do you run audyssey or equiv?

is the test signal routed directly to the channel of interest? seems there is some processing going on somewhere in the chain.

would you mind detailing your USB DUALPRE physical cabling setup?

Place your microphone a known distance from the speaker. Assuming you have an accurate test kit, the first arrival of sound to the microphone will be the direct sound. Take that time. Subtract from that time, the time sound takes to travel the known distance to the microphone. That will be your total processor related delay. If your kit allows you to adjust the gate, you can make that adjustment in your kit. A more simple approach is to take the delta-T between the first arrival and the second, third, etc, arrival. That delta represents the difference in the path length between the first and subsequent arrivals.

Example: If your speaker/microhphone distance is 8' and the second arrival took path of 8'-6", stretch a string 8'6" long to the mic position and the speaker. Any surface that string can touch would represent your reflection point. What if you cannot find such a surface? Look for a port on the rear of the speaker. Consider the number of drivers in the speaker...if there are three of them, you can easily have 3 first arrival times. The face of the speaker is also a reflection point.

Note, that the deltas between subsequent arrival times are not linear differences as most interpret them. In other words, to use a simple example, if the difference is 1ms between arrivals, it means that the 'triangulated' path from source, to incident boundary to mic is 1.13 feet longer.

As local has observed, something is wonky in your configuration and you are not employing a 'direct' path through the AVR where the signal is only being amplified. Assuming things are connected and configured correctly ('use loopback as timing reference' is set and the 'set IR to t=0' is NOT engaged), additional latency due to 'something' in the AVR is in play.

Either this needs to be corrected such that the arrival time correlates to the actual distance from source to mic, or the AVR itself needs to be included in the hardware propagation delay compensation loopback, by taking the output of the AVR and looping that back to the input of he other channel in the mic pre.


Also, if you have a non signal aligned source (a good reason we typically utilize aligned signal sources with a unity acoustic origin), the driver offsets can be determined as well with an ETC, including any diffracted virtual sources such as the baffle. But note that a driver offset (delay) of 6inches would correspond to a time offset of ~.44ms. This is typically part of the initial grunge associated with many direct arrivals (which typically includes diffraction sources from speaker mountings which must be physically corrected, as well as driver offset that should ideally be corrected with active crossover delay adjustments) that must be corrected - as this initial grunge is very detrimental.

As should be becoming apparent, the ETC can be used to evaluate multiple 'scales' of behavior, from the gross levels of speaker to mic, to the finer degrees of driver-driver signal alignment, to even finer scales of internal driver reflection issues such as the behavior of sound within the throat of a horn. What we are proposing to do here simply scratches the surface of the capabilities of the ETC response.

You can do that with REW; but, understand processor delay is just what it is. If you know what it is, you can still determine where year early reflection points are simply by taking that delay into account. There's also no point in locating these surfaces to an silly level of precision (your panels will be larger than any non-linear error). That level of accuracy is useful in driver/cross over alignments...also, take into account that ol' angle of incidence = angle of reflection thingy. There are potentially several reflective surfaces which are not affecting listening areas (useful to consider AFTER the low hanging fruit has been picked).

We use digital input signals for calibration, so we'll have processor delay 100% of the time. In your case, the bigger issue is determining why "direct" doesn't appear to be "direct". If you cannot get everything except DA conversion out of the loop, all your calibration measurements are going to be "cooked".

One other thing you ought to do is measure the noise floor in your room. You don't really care about reflections/sound which is below the noise floor ... no sense chasing crud down in the weeds which is masked by the noise floor.

For a very simple reason! The issue is that this allows one to use the ETC to determine a much more precise total time of flight, which incorporates a very precise determination of the initial time offset - much more accurately than you can by imagining the acoustic origin and trying to manually measure it with a precision orders of magnitude greater than that provided by the ETC itself.

And as mentioned earlier, the deltas do not correspond to the differences most imagine them to be. They are triangulated (assuming only a first order reflection path), and this distance does NOT correspond to the direct distance. For a second order reflection there would be three legs which would deviate even further than the direct distance.

The ETC is orders of magnitude more accurate than your ability to estimate based upon not even knowing where the acoustic origin is of he source, let alone your ability to measure the offset to the same degree of precision.

If you are not willing to configure the setup correctly, this error is compounded the longer the path, as you are not simply measuring a direst path from acoustic origin of the source (which is not the center of 'a' driver) to mic, but of an indirect pathway for each subsequent arrival.

The fact is, in proper acoustical testing we normally do not even use the system amplification for exactly the reasons you are encountering, but instead use a dedicated 'test' amplifier such as a Crown D75A.

In a small space, you have several options that are not available in other applications. So while they may seem to be a solution to the issue, they are limited only to this special case. One is the 'string method', a concept originally used by Don Davis to illustrate the concept of a specular reflection in a small bounded space which is possible ONLY if the boundaries are close enough to reach. trying this in a larger venue, or for that matter in the throat of a horn , you will find your arms either not long enough, or far too large to fit. Likewise, another option more commonly used is the blacking method whereby one iteratively places a piece of absorption at angles around the mic such that it blocks a particular vector path from source to mic. When a particular pathway corresponding to an arrival 'spikes' path is interrupted, the spike is diminished (but not necessarily 'removed' due to the relationship of wavelength to absorptive obstacle size), and one may then iteratively walk the path back to the point of boundary incidence.

Much of this is eliminate entirely by the TEF PolarETC program whereby after measuring the response and selecting a threshold level, a plot is generated showing a 3space rendering of the various arrival times and gain, whereby for specular arrivals above the established gain threshold 3 space coordinates are generated allowing us to replace the mic with a laser pointer, and the mount manipulated such that the 3space X, Y & Z coordinates are dialed in and the laser points precisely to the point of boundary incidence without any further mucking about.

So, at this point, instead of debating the nature and use of the ETC, may I instead suggest that one might want to instead concentrate on simply configuring the setup and addressing the issue with the AVR in order that it can be properly utilized at this cursory level, as there are already a myriad options for less accurate and precise methods. Your purpose here is to establish the travel time of the test signals WITHOUT inclusion of additional hardware propagation delay.

And yes, noise floor is a basic factor that should be the first measurement made in any space and is properly a function of isolation, not of the measurement of specular signal levels.

But, in any event, we are not going to be concerned with any early arriving low level signal that falls below ~15 dB SPL of the direct signal. And if you cannot achieve at least 15 dB SPL S/N in the room, I would pack this up as the room is not useable for the intended purpose. As far a later arriving signals outside of the ISD, you will be concerned only with sparse high gain signals (relative to the gain of the surrounding smoothly decaying sound field) that can affect localization and coloration, and these will typically be diffused in order to proves a spatially and more temporally dense soundfield.
Edited by dragonfyr - 6/20/12 at 10:44am

Sparse high gain signal = narrow spike on an ETC

Quote:
Originally Posted by dragonfyr

For a very simple reason! The issue is that this allows one to use the ETC to determine a much more precise total time of flight, which incorporates a very precise determination of the initial time offset - much more accurately than you can by imagining the acoustic origin and trying to manually measure it with a precision orders of magnitude greater than that provided by the ETC itself.
And as mentioned earlier, the deltas do not correspond to the differences most imagine them to be. They are triangulated (assuming only a first order reflection path), and this distance does NOT correspond to the direct distance. For a second order reflection there would be three legs which would deviate even further than the direct distance.

Calimark;
shall I try another way...what does T=0 do to the measurement that isnt done when t=0 is not set?

Dragonfyr;
We have already explained this. You do not have the total time of flight (TOF) of the test stimulus from speaker source to measurement mic. Unless one is solely going to use the blocking method (a technique limited to use due to scale in a reasonably small room), one cannot determine the actual time of flight of the various indirect signals and hence determine their precise corresponding distances, and if you have ever tired extrapolating such information in a larger room, where such error can easily lead to the mispositioning of treatment, not to mention innumerable trips up a ladder, glacially fast scissors jack or auditorium catwalk strategically placed at the farthest point in the room from wherever you are working, you would not debate this.

Without an ACCURATE TOF determination, the string method and other methods are close guestimates at best, especially when you add the error of the actual location of the acoustic origin at which one end must be terminated.



Quote:
The ETC is orders of magnitude more accurate than your ability to estimate based upon not even knowing where the acoustic origin is of he source, let alone your ability to measure the offset to the same degree of precision.
If you are not willing to configure the setup correctly, this error is compounded the longer the path, as you are not simply measuring a direst path from acoustic origin of the source (which is not the center of 'a' driver) to mic, but of an indirect pathway for each subsequent arrival.

Calimark:
How does this apply when I read most folks are using 2'*4' panels? If one is applying such huge panels, how much resolution does one need when trying to find a point of reflection ?

Dragonfyr:
Since you apparently have plenty of time to debate why things aren't done differently instead of finding out why your configuration of the setup is not functioning properly, while you are at it, you might want to take a moment and investigate just how large a treatment must be to be effectively SEEN by a broadband signal that extends down to say 300 or 400 Hz.



Quote:
The fact is, in proper acoustical testing we normally do not even use the system amplification for exactly the reasons you are encountering, but instead use a dedicated 'test' amplifier such as a Crown D75A.

Calimark:
Hmm...I dont recall reading that in any of the threads here.

Dragonfyr:

Hmmm..then apparently the advise must not be valid. I will promptly toss mine! And out of curiosity, how many people here have been actively using said measurements for the past umpteen years? And a resourceful lad, if they had a bit of initiative and desire, might find one on EBay for about $175.
The larger issue is that those who do this routinely do not mess with trying to figure out every screwy AVR switching matrix on the planet, nor trying to figure out how to get the input to play through both the full range speaker and sub simultaneously. At least not until they have memorized how to program every VCR on the planet first. And this may confuse more, but for larger SR systems, they don't use AVRs. So you can just imagine some poor nut wandering around wondering what to do with having an AVR to use for a switching matrix when doing studio, club or SR test.



Quote:
So, at this point, instead of debating the nature and use of the ETC, may I instead suggest that one might want to instead concentrate on simply configuring the setup and addressing the issue with the AVR in order that it can be properly utilized at this cursory level, as there are already a myriad options for less accurate and precise methods. Your purpose here is to establish the actual travel time of the test signals WITHOUT inclusion of additional hardware propagation delay.

calimark:
I think I have outlined all the settings, I specifically used this older AVR thinking it has no bells and whistles and will be literally direct- but its not.
Do you have some troubleshooting tips to offer that doesnt cost $500 to buy a crown amp?

Dragonfyr:
Here's one: you can use some initiative and query EBay.
Short of that, we have already explained how to include the AVR in the hardware propagation delay loopback. One can only imagine what might be accomplished if we would simply do what has already been suggested instead of thinking of new issues to complain about, or debating basic testing procedures and suggesting new ways to perform testing despite not having the bare basics down. Especially if what you have said is correct, and everything has been configured correctly, either this situation does not in fact exist, or you DO have some form of processing included in the AVR that must be in some manner, effectively 'removed' from the test results.

And has already been mentioned, you can either correct the loopback to take the AVR into the loop, or you can remove the offending instrument from the signal chain, or one can in what seems to be the preferred route, complain to us and demand that we solve the problem with your configuration for you.



Quote:
And yes, noise floor is a basic factor that should be the first measurement made in any space and is properly a function of isolation, not of the measurement of specular signal levels.
But, in any event, we are not going to be concerned with any early arriving low level signal that falls below ~15 dB SPL of the direct signal. And if you cannot achieve at least 15 dB SPL S/N in the room, I would pack this up as the room is not useable for the intended purpose.

Calimark:
Done previously but again I need wrap my head around this dilemma first.



Quote:
As far a later arriving signals outside of the ISD, you will be concerned only with sparse high gain signals (relative to the gain of the surrounding smoothly decaying sound field) that can affect localization and coloration, and these will typically be diffused in order to proves a spatially and more temporally dense soundfield.

Calimark:
Can you define this with some granulatiry for the layman?
what is definition of sparse as used here? How 'relative to the gain' ?
Graphical example are good here if you have.

Dragonfyr:
Geesh, you haven't even gotten the rig successfully configured and you are chastising me for not explaining the entire measurement and analysis process "with granularity" as well as providing detailed examples?! I can guarantee that complaints will accomplish everything you have demanded...

And when you DO finally get around to generating measurements, PLEASE post the raw REW response file for the sweep, with each source speaker measured individually and clearly labeled if you want us to take the time to open and evaluate the various derived responses. And please don't apply ANY smoothing to the measurements! NONE! It is infinitely easier than trying to tell one how to appropriately window a static display.

Oh, and if the tongue in cheek humor in the reply is objectionable, next time consider the tone in which assistance was, ah, 'requested' (sic)...
Note, the lack of such a tone won't prevent the resultant wry wit, but then it won't come as a surprise either...


Oh, sorry for the #^$&% quotes, but I can't seem to figure out how to quote the original and the original reply both - assuming it can be done...
Edited by dragonfyr - 6/20/12 at 1:33pm

I know you did and that they do...
They also use REW for little more than waterfalls for modal analysis and setting PEQ filters below 80 Hz as well...


For the money its a very welcome product. But those on that forum are hardly aware of the full potential of the available functions, nor do they commonly use it to its full capabilities.

They also feature advisers for whom opening the data files is considered too complicated and where suggesting that the raw data files be posted so that anyone can open them using the very software distributed there and view the data from a variety of available perspectives literally causes all kinds of trouble - in addition to confusing the resident whoseits and resident LA redneck to no end.

It also serves a practical value in that we can quickly open and properly window the results to see exactly what we need without an endless round of trying to explain how if you only move this that way, and that this way, and move that a smidge and that a hair...only to finally obtain a useful windowing of the results several days later and instead cut to the chase in short order. When you are ready with such data we will be glad to show you the method to the madness in order to get you quickly up to speed.

That is precisely why we offered the modification to their advice.
Edited by dragonfyr - 6/20/12 at 4:58pm