Home Theater Forum and Systems banner

1 - 20 of 24 Posts

·
Registered
Joined
·
66 Posts
Discussion Starter #1
I am working with the IR window and the Impulse graph, but I am not sure if my impulse graph should look like this. It is looking so messy.

Also, I can't find out what the left and right windows mean and is there perhaps some advise or info available on what these different type of "algorithms" are good for like turkey 0.25?

 

·
REW Author
Joined
·
6,711 Posts
I've made this a separate thread since it doesn't relate specifically to the beta version.

It is difficult to tell a lot from the graph image since it only shows part of the response, if you attach the actual measurement file (the .mdat file) it will be much easier to check whether it is OK.

Different window shapes offer different trade-offs between resolution and artefacts arising from the truncating effect of the window, for most purposes you are best leaving the window types at REW's default settings. There is plenty of information about window functions in wikipedia, amongst other places.
 

·
REW Author
Joined
·
6,711 Posts
Left and right are relative to the Impulse peak, left is before and right is after.

The measurement looks fine. There is some activity in the impulse response before the peak occurs, though not enough to be a concern. That can sometimes be caused by sample rate conversion in the computer (for example, REW sample rate set to 48k when the soundcard is running at 44.1k or vice versa) and sometimes by FIR equalisers (like Audyssey or Dirac).
 

·
Registered
Joined
·
66 Posts
Discussion Starter #5
How useable will my measurement be in building a FIR filter by using these USB audio devices. I am starting to believe that I have bought the wrong mic.
 

·
Registered
Joined
·
66 Posts
Discussion Starter #7
I want to create a room acoustic filter for use in JRiver.

The measurement is that of a bookshelf speaker, on axis and at tweeter height, at 80 cm distance.
 

·
Registered
Joined
·
84 Posts
You want to create a room correction filter at that close distance? Is that actually the listening distance? A measurement for room correction is taken at the listening position. A measurement as close as you've taken it is for speaker design/correction with a gated impulse.
 

·
Registered
Joined
·
84 Posts
Well that depends on who you ask. I don't do any rc over about 400hz but if I did I would take into consideration the anechoic response. That said your ir looks too chaotic to be useful for a quasi-anechoic measurement. You need to get the speaker on a stand as far away from all boundaries that you can, and have the mic halfway between the floor and ceiling (assuming a typical 8' ceiling). Then you gate the data before the first reflection.
 

·
Registered
Joined
·
66 Posts
Discussion Starter #11
That is what I did. I calculated the first room reflection, which is the floor and started gating my measurements. I cannot do multiple sweeps because this creates time alignment issues. I do have impulse activity before time = 0 which is not due to resampling or conversion, because I checked it.

My question is, if I can get reliable measurements by using USB devices. There seems to be to many factors involved that are beyond any control without a proper driver or software in place.
 

·
Registered
Joined
·
84 Posts
USB mics work fine as far as I've seen, though I've never used one. I use a usb interface with REW all the time with no issues. The IR in the mdat you posted looks fine to me other than the pre-ringing or whatever is going on before the first peak. Like John said, stick with the default window (Rectangular). The other windows basically "round off" the end the the gate in different ways. If you have your gate close to the first reflection then this reflection can be partially included with the other gate types.
 

·
Registered
Joined
·
84 Posts
I'll start the gate 1ms before the first peak. In your IR it looks like there may be a reflection at 2ms? What kind of speaker is this? With a 2-3ms gate the data below 1khz is going to be quite smoothed and rolled off.
 

·
Premium Member
Joined
·
1,562 Posts
I understand you want to create an FIR filter for EQ purposes.

To determine an EQ filter to improve the sound quality in our listening area we are normally only concerned with the combined (speaker/room) response in that area. In that case there is no significant benefit to near field speaker measurements. I haven't seen any of the automated commercial systems that utilize a near field measurement in the filter creation process. People often refer to "room response" or "Room EQ", but it is normally intended to mean the combined response given a particular setup in a room.

For those that are DIY regarding the speaker XO or other speaker design considerations then quasi anechoic speaker measurements are helpful. Some of us are just curious; nothing wrong with that.

Regarding some of your other specific questions/concerns:
> natehanson66 his providing good direction regarding gating for quasi anechoic measurement, but, like he, I am not sure how this helps in designing EQ filters.

> Your comment regarding "impulse activity before time = 0": You are correct this is not due to resampling. This is a result of the anti-aliasing filter chosen for the DAC(s) in your measurement system. This is no concern to the sound quality in any practical sense. The anti-aliasing filter type applied can result in this IR appearance. It is not something that can be heard and there are pros and cons related to the choices made. A DAC that shows this type of "pre-ringing" is not "better" than one that doesn't it is a design tradeoff. Some DACs provide an option for the type of anti-aliasing filter to apply. While some audiophiles indicate they can hear a difference, I doubt they can agree on which type is "better".

There are other causes of pre-ringing that can be a problem, so not all pre-ringing is harmless. In this case it is. The DAC that caused this is operating at 48K. I can tell that because the period of the ringing is ~20.83µs. The envelope shape of this pre-ringing is also similar to that as found in other anti-aliasing filters.

> Single sweeps have no disadvantage in the development of an EQ filter; no worries there. :)
 

·
REW Author
Joined
·
6,711 Posts
Like John said, stick with the default window (Rectangular).
The default window shape for the IR is Tukey 0.25, which is flat apart from the final 25% which is a Hann window. That reduces the artefacts from the discontinuity at the edge of a rectangular window.
 

·
Registered
Joined
·
66 Posts
Discussion Starter #19
I understand you want to create an FIR filter for EQ purposes.

To determine an EQ filter to improve the sound quality in our listening area we are normally only concerned with the combined (speaker/room) response in that area. In that case there is no significant benefit to near field speaker measurements. I haven't seen any of the automated commercial systems that utilize a near field measurement in the filter creation process. People often refer to "room response" or "Room EQ", but it is normally intended to mean the combined response given a particular setup in a room.

For those that are DIY regarding the speaker XO or other speaker design considerations then quasi anechoic speaker measurements are helpful. Some of us are just curious; nothing wrong with that.

Regarding some of your other specific questions/concerns:
> natehanson66 his providing good direction regarding gating for quasi anechoic measurement, but, like he, I am not sure how this helps in designing EQ filters.

> Your comment regarding "impulse activity before time = 0": You are correct this is not due to resampling. This is a result of the anti-aliasing filter chosen for the DAC(s) in your measurement system. This is no concern to the sound quality in any practical sense. The anti-aliasing filter type applied can result in this IR appearance. It is not something that can be heard and there are pros and cons related to the choices made. A DAC that shows this type of "pre-ringing" is not "better" than one that doesn't it is a design tradeoff. Some DACs provide an option for the type of anti-aliasing filter to apply. While some audiophiles indicate they can hear a difference, I doubt they can agree on which type is "better".

There are other causes of pre-ringing that can be a problem, so not all pre-ringing is harmless. In this case it is. The DAC that caused this is operating at 48K. I can tell that because the period of the ringing is ~20.83µs. The envelope shape of this pre-ringing is also similar to that as found in other anti-aliasing filters.

> Single sweeps have no disadvantage in the development of an EQ filter; no worries there. :)
Thank you for all the response guys, it is certainly helping me to understand better the relationship between things, and learning to trust my own measurements. I am certainly the curious kind of type, and messuring and under standing my speaker response is my way of learning, because I can compare this with a refference.

How does the impulse response relate to samples? 1/48000 = 20.83us

Do I need a work around for my in room measurements, 5 channels, since I have the Dayton calibration file which seems to be calibrated for on axis measurments. Could I take a gated on axis speaker measurment and off axis measurement and compare the differences?
 

·
Premium Member
Joined
·
1,562 Posts
It would have more correct and clear if I had indicated the expected period of the ringing to be at the Nyquist frequency (24kHz) or 41.7µs. Below is a zoom of your impulse so we can better see the pre ringing. Note that ringing is spaced uniformly and the time for 5 cycles is roughly measured as 207.8µs so; 1 / (207.8µs / 5cycles) = ~24,062Hz. This agrees closely with 24kHz Nyquist frequency for a 48kHz DAC.

I really don't know if this represents a normal feature of using a linear phase anti-aliasing filter, or possibly something else. I only know that the 2 DACs I tested with this characteristic measured very well for SPL, phase, and THD characteristics. Since you are not setup for a loopback measurement on the DAC you will not be able to measure those items in your case. It may not be ideal for a DAC but it is not a fatal problem for REW purposes.

On a loopback test you can see the ringing continue after the IR peak also. When measuring the room with a mic the post ringing is buried in the measured IR response and thus not noticeable.

I just wanted you know that it is a DAC issue not a speaker/room issue and also that it isn't a rare situation.

I don't understand the premise of your question about needing a workaround so I have no comment there.

preringing.jpg
 
1 - 20 of 24 Posts
Top