They look consistent between angles which is what a good waveguide would do. Some EQ on the design axis would likely make it look a lot better but there does seem to be reflections in there.
If you bring up the Impulse window you can see where the right window will gate out the reflections and adjust to just before the first one.
gate before the first reflection and have another look at the measurements.
audio blog: Gating loudspeaker measurements
audio blog: Gating loudspeaker measurements
My take is gating is mostly worthless.
Whatever time window you set to eliminate reflections, the one we commonly equate to be the lowest valid frequency of the measurement, also becomes the resolution of the measurement.
IOW, you gate to maybe 3 ms for close range work. We assume that gives good data to 333Hz. What we forget is that freq resolution throughout the measurement is 333Hz.
Bogus imo, makes for pretty graphs is about all.
Whatever time window you set to eliminate reflections, the one we commonly equate to be the lowest valid frequency of the measurement, also becomes the resolution of the measurement.
IOW, you gate to maybe 3 ms for close range work. We assume that gives good data to 333Hz. What we forget is that freq resolution throughout the measurement is 333Hz.
Bogus imo, makes for pretty graphs is about all.
This is incorrect because you dont want to eq a speaker using a measurement which includes a reflection. it simply will not sound good. its very easy for you to try.
It also doesn't tell you what the speaker is doing. it tells you what the overall system (room + speaker) is doing, a terrible approach for developing a horn
you could possibly avoid gating by doing a lot of spatially averaged measurements, but this will likely just be similar the result you would have gotten by just gating. if you want resolution, then go outside. and still gate the measurement.
like I said easy to try for yourself. i'm not saying its causal, or that its absolute.
I have personally tried averaged spatial ave measurements, single point room measurements, and gated measurements. in my experience, EQing bass is best with spatial ave. but high frequency EQ has only ever worked well for me if i'm doing gated spatially averaged measurements.
We are talking about the analysis of HF horns in this thread. Mark100 is just adding useless noise. actually wish the moderators would remove this type of thing. its actually completely off topic.
I have personally tried averaged spatial ave measurements, single point room measurements, and gated measurements. in my experience, EQing bass is best with spatial ave. but high frequency EQ has only ever worked well for me if i'm doing gated spatially averaged measurements.
We are talking about the analysis of HF horns in this thread. Mark100 is just adding useless noise. actually wish the moderators would remove this type of thing. its actually completely off topic.
Off topic but...
"If this measurement is made in an ordinary room the acoustic reflections from the room boundaries will usually dominate the measurement. Since what was
wanted is to measure the loudspeaker, such a measurement is not useful for anything other than characterizing the room".
"There are several methods for doing this, but they all depend on windowing the time signal to eliminate the reflections before they contaminate the measurement".
Cited in attached Audio Engineering Society Convention Paper 6218 by Eric Benjamin, Dolby Laboratories...
"If this measurement is made in an ordinary room the acoustic reflections from the room boundaries will usually dominate the measurement. Since what was
wanted is to measure the loudspeaker, such a measurement is not useful for anything other than characterizing the room".
"There are several methods for doing this, but they all depend on windowing the time signal to eliminate the reflections before they contaminate the measurement".
Cited in attached Audio Engineering Society Convention Paper 6218 by Eric Benjamin, Dolby Laboratories...
That doesn't mean it cannot represent the device accurately. It all depends on how large and how important portion of the actual impulse response is changed by gating. The impact of gating can be very small in case of CDs and waveguides (HF sources in general), i.e. the measurement is typically quite accurate for the whole passband of the device.
- You can gate out silence without any effect on the resolution of the measurement.
Last edited:
Let me state it more precisely, perhaps: It depends on how much are the real anechoic IR and the gated IR different. In case of HF sources, they are typically not very different (unless the device is terribly bad), i.e. the resolution of the measurement is good enough to accurately describe the DUT.
So, please do gate waveguide measurements, otherwise it's pretty useless.
So, please do gate waveguide measurements, otherwise it's pretty useless.
Last edited:
My bad, i forget that waveguides are being thought of predominantly as solely HF devices.
Every wavequide I've used has xover no higher than 650Hz.....
and most my horn/guide work goes to 500Hz, and lately even 300Hz.
That said, having compared literally thousands of outdoor measurements to 1m indoor measurements, 1000Hz or higher shouldn't require any gating to clean up (at 1m indoors).
This is with dual channel. I assume REW gives the same results, as every time I've compared measurement programs, they look pretty identical.
Every wavequide I've used has xover no higher than 650Hz.....
and most my horn/guide work goes to 500Hz, and lately even 300Hz.
That said, having compared literally thousands of outdoor measurements to 1m indoor measurements, 1000Hz or higher shouldn't require any gating to clean up (at 1m indoors).
This is with dual channel. I assume REW gives the same results, as every time I've compared measurement programs, they look pretty identical.
The problem with this discussion is that it is based on a fallacy that "What we forget is that freq resolution throughout the measurement is 333Hz." The resolution across the bandwidth is set by the sample rate and total time window and not by the gating window.
Marcel was correct when he said: "You can gate out silence without any effect on the resolution of the measurement."
Hence the entire discussion is based on a false premise. Everything above the gating frequency is completely accurate with a very high resolution. Not gating is completely wrong.
Marcel was correct when he said: "You can gate out silence without any effect on the resolution of the measurement."
Hence the entire discussion is based on a false premise. Everything above the gating frequency is completely accurate with a very high resolution. Not gating is completely wrong.
Then I ask your help, to please fix my understanding....
Given:
Frequency Resolution = Sample Rate / FFT Size = 1 / Time Window
or expressed as:
Time Window = FFT Size / Sample Rate = 1 / Frequency Resolution
What Sample Rate and FFT size can give a Time Window of 3ms, and provide better Frequency Resolution than 333Hz?
Once a 3ms gate (post impulse peak) is put in place, how do the above relationships not hold mathematically?
Thank you !
Think about it this way - if the impulse response of the DUT falls into noise after 3 ms, there can't be any loss of resolution if you simply trim everything after that. You can then simply zero-pad the data to whatever FFT length you want. Compression drivers in waveguides seldom have impulse responses lasting much longer than that (above-noise-level, that is).
Guys, I don't mean to be dense....
May I use REW 'IR Windows' to try to understand what you are saying?
When Right Window is set to 3ms, Frequency Resolution is 333.33Hz.
(And magnitude graphs smooth out to show the resolution reduction.)
I totally get gating works fine once above a frequency where the reduced resolution is no problem. (like for the waveguides in this thread).
But for understandings sake, I still do not get how resolution doesn't decrease with gating......
Do any of our measurement programs use this 'zero-pad technique' you mentioned, to maintain resolution after gating ?
May I use REW 'IR Windows' to try to understand what you are saying?
When Right Window is set to 3ms, Frequency Resolution is 333.33Hz.
(And magnitude graphs smooth out to show the resolution reduction.)
I totally get gating works fine once above a frequency where the reduced resolution is no problem. (like for the waveguides in this thread).
But for understandings sake, I still do not get how resolution doesn't decrease with gating......
Do any of our measurement programs use this 'zero-pad technique' you mentioned, to maintain resolution after gating ?
It does, but as long as you don't trim the actual IR significantly, this resolution is still high enough to describe the DUT accurately (and can be "increased" further by zero padding, that's another thing).
- How long gate would you need to measure a device with a perfect IR? A few samples should be plenty without loss of anything, because in that case the frequency response is also perfectly smooth - you don't need much resolution to describe that. The better the device, the shorter gate can be used (i.e. the shorter its impulse response).
Last edited: