TOPIC:

I'm new to FIR Design and tried to use the free software available but overall found it lacking. I've now bought FIR Designer as it seemed to be one of the most powerful and flexible while also being lower in cost than some of the others.

I'm now just trying to get to grips with the software and hoping someone with experience of the software or the developer can help?
I think I'm ok with the gain adjustments but I am struggling with the phase side.
I've taken a measurement 1m from one of the speakers. My logic is that at 1m I can mostly correct the speaker phase issues. In general I was going to mostly ignore phase issues created by the room (as they appear to be more of a mess).
In terms of my measurement, where do you think I should start?
Should I just try to correct phase in a certain part of the frequency range, perhaps just at the cross over?



On a different note I suspect that my filter end result will be some basic corrections on phase using a measurement close to each speaker plus a gain adjustment based on the average of a few listening positions around the room. REW can create average measurementa to fix which I can import but no software appears able to operate a phase shift on one set of data while performing a gain shift on a different set of data, this becomes a manual process. I'm guessing this means that I cannot use the auto element of the software for this?!
I'm thinking that the ideal would be to create a phase shift design and a gain shift design based on different sets of data and then to somehow combine these two results into a single filter. Does this make sense or is there another way to approach this?
Any advice people can give would be appreciated.

Also, as a test I auto adjusted the gain and phase over an entire measurement from a single listening position and uploaded that to me OpenDRC. The result was truly awful (music sounded like it was from inside a metal box). It was a very complex filter (perhaps too complex)but is the issue the limited number of taps on the OpenDRC (6144)? If so, for a decent quality sound there a limit to how many filter modifications the software can make with 6144 taps?.

Hi psearancke,

Firstly w.r.t. taps, the filter length or number of taps affects how low in frequency a FIR filter can work. The way to check is to look at the filter length in milliseconds. 6144 samples / 48000 Hz sampling rate = 128 milliseconds. For comparison, 1 cycle @ 20 Hz is 50 milliseconds. Since 6144 taps is over twice as long as 1 cycle at 20 Hz, a FIR filter of this length can achieve a reasonable amount of EQ down to 20 Hz and a little lower.

The "quality" of the result comes from the accuracy of the measurement and how the filter is designed from the measurement. The measurement you posted looks fairly hairy and a challenge to work with. I'd suggest measuring up close first - say starting as close as 200mm - and ensuring you can get a fairly clean measurement. This will also help to troubleshoot any problems in the measurement gear. From there, then try measuring at further distances. The MiniDSP site here has some pages on measurements
www.minidsp.com/applications/acoustic-me...speaker-measurements
www.minidsp.com/applications/acoustic-me...coustic-measurements

At distances representative of the listening position, it's good to average multiple measurements from multiple listening positions in the room. The averaging will even out much of the ripple (in both magnitude and phase) and give something better suited for creating room EQ. FIR Designer has an averaging tool - the last tab. Just open a directory containing a bunch of measurements - each as different files - and the averaging process will time align them all before calculating the average. Select one measurement - say the on axis measurement - as a Reference, and then tweak the time alignment and relative weighting of all the other measurements.

W.r.t. to working on phase and magnitude separately..... In any loudspeaker, and more generally any impulse response, phase and magnitude are inherently linked. There's no notion of measuring phase and magnitude separately. That said, once you have a measurement (averaged or otherwise) that is representative of the loudspeaker, FIR Designer can manipulate the phase and magnitude independently. That's the beauty of FIR filtering versus more traditional IIR's.

W.r.t. to auto correction ... generally it should be used sparingly and usually with a fair amount of smoothing - say 1/3rd to 1/6th octave but no finer. Very fine auto filters can lead to odd sounding results. Loudspeakers do vary over level and time. A FIR filter with very sharp magnitude and/or phase transitions and very fine, detailed, correction might look good on screen for the measurement being used, but the fine structure might be mismatched to the loudspeaker at other levels and as the loudspeaker changes its mechanical behaviour (ever so slightly) over time. Broader, smoother filtering is generally better. (I was somewhat reluctant to add the auto methods in FIR Designer, as they can easily be overused.)

Hope this helps,

Kind regards,
Michael

Thanks for the quick reply. I'll try to take some better measurements and use auto less.
One other thing just came to mind, I've just done some basic phase adjustments to the graph I posted earlier (32-2000 hz, 1/3 smoothing). Listening to music through this has significantly affected the base boom I was getting (even though I've not done any gain adjustments). If phase adjustments can significantly impact a base boom then should I make the listening position gain measurements with phase correction aleady in place?
Thanks

Hi,
I have taken some steps to improve my measurements (hopefully). I have now imported a new measurement and made some manual adjustments to phase. See image below.



Do you think this measurement looks useful or do I still have more work to do on this?
Also, do my manual adjustments look ok or am I approaching this incorrectly?

Thanks,
Peter

I've been playing some more and I think I've improved the phase measurement and I've put in some manual phase adjustments.
Anyone familiar with phase adjustments, does this look ok?



Thanks.
Peter

Hi Peter,

Those measurements look pretty good and your manual phase filters look fine. I'd suggest listening to this as is, before trying any auto phase. You could even try leaving out the AP2 at 50 Hz, and just start tracking the phase tracking from 100 or 200 Hz upward. From the magnitude spectrum, the room looks to have significant resonances below 100 Hz, and so phase corrections that low may not be as relevant, nor as audible.

Kind regards,
Michael

Thanks, I have now created phase adjustments and tested them out (no auto phase as of yet). There is no audible difference to my ear with music. However, when I run REW to get a test measurement, the sweep tone makes some strange electronic type noises through the range which disappear when I bypass the filter.
The filter contains 6 x 4th order and 2 x 2nd order adjustments so is not particularly complex (and auto phase is off). The REW measurement result shows the phase is now pretty close to zero throughout but it seems odd that the tone of the sweep has been so obviously screwed up.

Any idea what's gone wrong?

Peter

psearancke wrote:

Thanks, I have now created phase adjustments and tested them out (no auto phase as of yet). There is no audible difference to my ear with music.

That's fairly typical. Phase adjustments alone are close to inaudible. Some folks can hear high phase slope in high order crossovers - e.g. 6th order or greater in subwoofer crossovers - but the effect can be subtle. Some folks also say they can hear improvements in temporal behaviour and clarity with linearising phase to sharpen up the IR impulse. I don't doubt them, but IMO the effect is subtle compared to magnitude/EQ changes.

One of the main uses for phase adjustment, in FIR Designer, is matching different makes and models of loudspeakers in PA systems, where each make/model has a different phase response due to different driver, cabinet and crossover choices.

psearancke wrote:

when I run REW to get a test measurement, the sweep tone makes some strange electronic type noises through the range which disappear when I bypass the filter..... The REW measurement result shows the phase is now pretty close to zero throughout

Since the measured phase looks good, I'd say what you are hearing is correct and it's just the sound of the sweep changing phase as it moves across frequency. Maybe check too that the output of the DSP isn't clipping during the sweep.

Kind regards,
Michael

Update...
I was worried that the electronic noises were caused by some other issue so I emailed my project files to Michael at Audio Eclipse. He got back to me very quickly. He was spot on with his final comment above, the digital signal was clipping which was causing some unusual electronic type noises across the REW sweep. Up until yesterday I hadn't really understood how this was possible but clearly it is.
As a test, I made no modification to the filter but just turned down the gain on the digital inputs on the OpenDRC 2x2 Plugin. This immediately removed the strange electronic noises and the filter started to work as I had originally expected.
Ultimately I can either continue to use the gain control setting in the OpenDRC or I can make a new filter with a reduction in the gain on the export tab in FIR Designer.
Thanks Michael.