Home Theater Forum and Systems banner

1 - 20 of 45 Posts

·
Registered
Joined
·
43 Posts
Discussion Starter #1
Hi, Recently miniDSP announced the miniDSP2x4HD. This contains FIR and IIR filters. Which means that it makes minimum phase and linear phase happen.
On the other hand the EQ generation facility of REW assumes minimum phase only. Does this idea correct? Or REW provides (or will provide) linear phase EQ too? like Dirac's mixed phase.
Yuichi Arai
 

·
Registered
Joined
·
43 Posts
Discussion Starter #8
Thanks Wayne start to look rePhase. Yuichi
One question about REW and rePhase relations.
Assumption is the EQ generation of speaker sound property.

In REW, minimum phase EQ is generated after compensation of time delay (sound travel time).
To generate FIR filter EQ (Linear phase) , measured data has to be exported from REW to rePhase. Exported REW data has time delay, therefore, phase graph is out of scale in rePhase screen. So, is there any way to generate minimum phase in rePhase or any other way to generate linear phase EQ on rePhase?
rePhase works good for very small delay(phase) contained in the data. Does this mean that it is not a good way to use this combination to generate for the speaker sound EQ?

Yuichi
 

·
Registered
Joined
·
43 Posts
Discussion Starter #10
I think so. Therefore, IIR filter is properly generated by REW. But to generate FIR filter for miniDSP2x4 HD, it is recommended to use rePhase. Where, measured data which is obtained by REW is exported to rePhase as original. Which means that phase information contains delay time. This means that the phase curve on rePhase screen changes so sharply. So, I can not adjust phase to linear on rePhase. Does this mean that this method (REW->rePhase) is not applicable speaker measurement and EQ generation because of large delay (sound travel from the speaker to the mic)?

Or simply saying, how to generate FIR (Linear phase) filter for speaker sound?

Yuichi
 

·
Premium Member
Joined
·
1,562 Posts
Yuichi,
There are lots of options regarding how to approach EQ. I am not aware of any consensus on the best scheme. There are good guidelines here in HTS for IIR EQ. FIR methods tend to be more complicated and have less coverage here. There are many other sources that can be easily found for more complete coverage.

The general overall scheme I have been using is one that fits my equipment and level of understanding.
> Use IIR minimum phase filters for SPL EQ. There are lots of different ways to approach the EQ effort. FIR linear phase filters are not normally recommended for basic SPL correction.
> Use FIR phase filters to remove the phase rotation of the direct sound. Just filter out or ignore the impact of the room when designing the filter.

I use REW to help establish the IIR EQ scheme. I think it is best to first find a house curve and EQ scheme that works well for you. It may take some significant effort to find settings you really like. I typically use averages in the listening area for the full range EQ effort, but have also had good results using single LP measurements. I am currently using EQ settings based on close mic measurements for the high frequency portion of the range and LP measurements for the range under 600 Hz. That has worked well also. First set the final speaker/SW/LP positions and room treatments and then spend the effort to find a house curve and EQ settings that work well for you. This is a challenging part of the effort, but necessary for good results.

The final REW measurement can then be appropriately filtered and input into rePhase to create an FIR filter to remove the phase rotation. We are then left with a linear phase system. This makes the measurement look more ideal in terms of impulse, group delay, and step responses. With my setup and my substandard hearing, I cannot hear any difference using the FIR phase correction filter, but many others people state that they can hear clear differences. Most all the studies I am have heard referenced apparently indicate, no or very little, difference is to be expected.

There other approaches that can work well. RePhase and other utilities can create an FIR filter that contains both the IIR EQ correction and the FIR phase correction all in the same FIR filter if that fits your setup better. I have used the DRC program for efforts into mixed FIR solutions. I didn't find anything that sounded or measured any better than my current setup. Some of the more automated commercial programs may be easier to use and provide better results than my more manual efforts.

Sorry for the these rambling comments.

There is no 'delay' that is a problem for input of measurements into rePhase There is however the need to filter or ignore the impact of the room. We want to create a filter to correct the direct sound phase. Many just correct for the calculated excess phase of the SPL response so that is an option. If you can provide an example file and a more detailed description of your concern, I'm sure I can help you clarify that particular issue.
 

·
Registered
Joined
·
43 Posts
Discussion Starter #12
Hello jtalden

Thank you for your detailed explanation for my question. I have understood most of your parts.

The only thing I could not understand is the methodology how to export the REW measurement data into rePhase.
Or, how to remove phase rotation caused by group delay (sound travel time to the mic) on the rePhase.

In the listening point, the mic is located approx. 2m far from the diaphragm. So, there are significant delay (sound travel time from speaker to the mic). My question is how to and where to remove this time delay. Otherwise, I can not adjust phase on the rePhase screen because of very quick phase rotation.

Yuichi
 

·
Premium Member
Joined
·
1,562 Posts
For FIR Phase correction we use rePhase to create one filter that we apply to both channels of a stereo setup. Since the same filter is used in both channels the filter delay will be the same in both channels. The small delay is equal in both channels and thus has no impact to music.


REW automatically removed the time-of-flight delay if we did not activate 'loopback timing' or 'acoustic timing' when the measurement was taken. If we did have that active, or have any concern, we can we can activate 'Estimate IR delay' in the REW 'Graph Controls' panel on the Impulse graph. REW will then shift the IR to near 0 ms. If needed, we can also manually shift it to the exact ideal location using the manual offset features in that same 'Graph Controls' panel. It is not unusual to need to fine tune the IR location such that the phase tail at 20 kHz is falling at smoothly to a 0° phase angle. This fine tuning can also be done instead within rePhase using a control on its interface. Either method works fine.


If you need more detail please provide an .mdat file of your measurement. I will use your data it to post charts showing the steps needed. If you prefer, I can use my data. It is better to work from yours because IR appearance and the adjustments needed will be different thus possibly causing more confusion. Either way is fine with me.
 

·
Premium Member
Joined
·
1,562 Posts
When the REW file is ready to export we just use the REW export function and choose the 'Measurement as text' option.
 

·
Registered
Joined
·
43 Posts
Discussion Starter #15
Hi jtalden

So, I can remove the IR delay first, then export it to rePhase.
I have reproduced what you said. Thank you so much.

Yuichi Arai
 

·
Premium Member
Joined
·
1,562 Posts
Yes, you can remove the delay in REW and fine tune it in REW or rePhase to fall smoothly to 0° at 20 kHz.

It helps to remove octave smoothing (set 1/48 octave) and then set a FDW of about 5 cycles in REW before you export it. That way the phase rotation of the direct sound will be obvious. Post a graph or better yet a file if you have any issues with the phase trace.
 

·
Registered
Joined
·
43 Posts
Discussion Starter #18
Hi jtalden

I did my experiment as you suggested. It works fine. And can be exported to rePhase. And I could generate the linear phase filters on the platform.

One thing I like ask you about the difference between the [Generate Minimum Phase] function and [Estimate IR Delay and correction] function.
I compared both way on the same several source data. The corrected phase properties by both way are almost identical, but there are some small differences especially in the higher band such as higher than 10kHz.
But Impulse response wave forms look different in between. The Minimum phase IR is much front aligned. Does this mean that the Minimum phase generation process corrects not only sound travel time delay but also other phase shifts created by diaphragms and others?

Yuichi
 

·
Premium Member
Joined
·
1,562 Posts
Hi jtalden

I did my experiment as you suggested. It works fine. And can be exported to rePhase. And I could generate the linear phase filters on the platform.

One thing I like ask you about the difference between the [Generate Minimum Phase] function and [Estimate IR Delay and correction] function.
I compared both way on the same several source data. The corrected phase properties by both way are almost identical, but there are some small differences especially in the higher band such as higher than 10kHz.
But Impulse response wave forms look different in between. The Minimum phase IR is much front aligned. Does this mean that the Minimum phase generation process corrects not only sound travel time delay but also other phase shifts created by diaphragms and others?

Yuichi
Yes (if I correctly follow your question).

The IR is the calculated transfer function of the entire system as measured at the mic position. It include the phase impact of the room and XO. The IR contains all the data needed for the various more friendly charts in REW.

The minimum phase IR is created using only the SPL data from the captured mic data. The phase impact of the room and XO is discarded. It is the transfer function that represents the smallest phase rotation possible that can explain the measured SPL response. The minimum phase IR looks cleaner because the complex captured phase info is dropped and replaced with a simpler phase response. So one is calculated based on the captured real data and other instead is calculated to find the simplest phase possible for that SPL response.

'Estimate IR delay' just shifts the measured IR to near 0ms. It removes the Time-of-Flight and any other delays in the measuring system. It does not impact the phase so the measured IR shape is not changed. The phase difference you found above 10 kHz is most likely due to a difference in the settings in REW. There will normally be little or no difference there for the measured IR Vs the minimum phase IR when the reference time is properly adjusted .

rePhase only will allow the import of a text file containing the SPL and Phase response in table form. REW will not export the minimum phase response in that form. Thus, we need to first be sure that the phase response is windowed properly so we are not confused by the impact of the room. The actual sound phase response is the one we want to use in rePhase, not the minimum phase, nor the phase chaos due to the room impact.

Some info on your speakers and setup and an example REW data file would help me understand the situation and thus answer future questions more effectively.
 

·
Registered
Joined
·
43 Posts
Discussion Starter #20
Hi jtalden

I understood quite well. Thanks

What I am doing is to measure the Headphone transfer function and then correct the SPL and phase by using rePhase. Then, import generated filers to miniDSP 2x4 HD through its plug-in. It goes well.

Attached is the original REW data (before time shift) for your ref. The headPhone type is AKG Q701.

Thank you very much, anyway.

Yuichi
 

Attachments

1 - 20 of 45 Posts
Top