Frequency response is everything!?...

[Tim Link]

That is a long time ! Hopeless for video sync ( I am going to run apple tv sound through convolution filters on another system)
What program is that.. I imagine all FIR are similar

Yes, it's a long time! Before I trimmed the silence off the ends of the impulse response it was about 1 whole second! The lip sync is noticeable on movies at 0.1 seconds but I've watched some stuff with it on - AppleTV stuff come to think of it. The latest Foundation episodes - featuring a robot that gets half her head chopped off right in the middle of having sex with a cloned human. No big deal for her. Just a few minutes in the repair shop.
I'm using Audio Hijack to run the convolution filter. They just recently added an FIR filter module, so that allowed me for the first time to try convolution on my system. My main reason for using Audio Hijack is to up-mix 2 channel stereo for my 3 speaker imaging array that I'm gaga about. The FIR filter block is a recent added bonus.

 

[pjwd]

Yes, it's a long time! Before I trimmed the silence off the ends of the impulse response it was about 1 whole second! The lip sync is noticeable on movies at 0.1 seconds but I've watched some stuff with it on - AppleTV stuff come to think of it. The latest Foundation episodes - featuring a robot that gets half her head chopped off right in the middle of having sex with a cloned human. No big deal for her. Just a few minutes in the repair shop.
I'm using Audio Hijack to run the convolution filter. They just recently added an FIR filter module, so that allowed me for the first time to try convolution on my system. My main reason for using Audio Hijack is to up-mix 2 channel stereo for my 3 speaker imaging array that I'm gaga about. The FIR filter block is a recent added bonus.

Thanks Tim .. software akways takes time to conquer!
How did you get apple tv audio through to audio hijack filters.. I am having trouble getting it through jriver where I intend to have the convolution filters

 

[Tim Link]

Thanks Tim .. software akways takes time to conquer!
How did you get apple tv audio through to audio hijack filters.. I am having trouble getting it through jriver where I intend to have the convolution filters

Audio Hijack can take system audio as input, as well as external devices. I bought a toslink to USB dongle to put on my Mac Mini. That feeds everything from the TV's optical output into the Mini. This only works for stereo in my setup, so no multi-channel sound capabilities. Audio Hijack only works in stereo anyway. This setup is kind of interesting because I can mix inputs together. I do this so I can listen to NPR while playing video games. I have the video game volume turned down so it's just audible. This allows me to justify extended game playing sessions because, hey, I'm staying informed, or indoctrinated, depending on who you ask.

BTW, by I managed to get the lag time on the FIR filter down to 14 ms! I limited it to 20Hz to 150Hz duty, and trimmed all the silence off both ends of the impulse file. That really goes a long way!

 

[Tim Link]

This has been my subjective experience with horns - much better dynamics than any other type of driver. However, I have never seen measurements that can confirm or quantify this. Are you aware of any?

This is my subjective experience too. I've tried to quantify it with my own measurements but have nothing to show for it. Any quick impulse I could get a horn to do, a direct radiator could do just as quickly - up to whatever the playback volume limit is of the speaker. A friend of mine who got me in to horns said you've got to turn them up to really hear their advantage. They can play louder, and that no doubt has been shown in measurements. Even very good direct radiators will start to compress sooner and experience distortion and non-linear response from the speaker elements heating up. But that's louder than I normally want to listen, and I perceive the better dynamics of horns at much lower volumes, where other types of speakers are still working well within their limits.

My current belief is that it's mostly about how the speaker is interacting with the room. This has been confirmed to me in some degree by hearing the same speaker and amp combination in different rooms or different placements in the same room sound considerably different to me in terms of perceived dynamics.

 

[Tim Link]

More direct sound, will be perceived as more dynamic.

I suspect this is the biggest part of it. You can turn up the direct sound to get to the same overall sound volume in the room if more of the sound is direct. Of course with a horn being so much more efficient you probably don't actually have to turn the volume knob.

An interesting thing to try if you have some gobos to work with is to build a little box around your listening position, effectively shielding yourself from a lot of early room reflections. When I tried it with some 2' x 4' gobos right behind me and to the left and right, I was amazed at how loud I could turn the sound up without it seeming too loud or out of control. I think I even put a roof on it so I had a gobo right above my head. I heard a super fast, punchy, hard hitting sound. It worked great on some rock/jazz fusion stuff I listened too. I think I turned it all the way up to max. It still sounded good, but I feared for my hearing. With orchestral it was a bit dry sounding in my little cocoon. Very clear though.

 

[Tim Link]

Only in the bass and only if the top end completely bypasses any singnal processor - ie you need a bi-amping system with DSP only appled in the bass amp.

DSP built into a full-range amp may well flatten the frequency curve (particularly if the owner can't or won't do this in less invasive ways), but it will reduce the sparkle and goosebump factor that un-proceesed top end can and should provide.

I've been playing around with trying to EQ only the bass using FIR filters and comparing that to letting the filter correct the entire bandwidth. What I've heard so far is a mixed situation. The bass really sounds best when the FIR filters have access to more of the bandwidth. Unfortunately, what works best for the bass seems to add some coloration to the highs, which I can easily notice by listening to ambient recordings of things like waterfalls or rolling surf in the background. That kind of broad spectrum, white noise like stuff is very revealing of coloration, and I can definitely hear it sounding less open and natural. It sounds a bit "roomy", which is supposedly what the EQ should be getting rid of. I suspect that what I'm hearing is too much room correction that's not quite working, so I'm effectively hearing an inverted version of my room's sound coming back through the speakers. It's like hearing a recording of your speakers playing in the room fed back through your speakers. It's not as noticeable with a lot of music, and overall produces a very clear and musically involving effect despite the coloration. If I limit the FIR filters to just the bass, and then time and level align the highs I get a more natural sounding high but it's less impactful, less clear, and not as well integrated with the bass.

This gives me something to work on. Maybe with some effort I can get the best of both worlds.

As far as completely bypassing the signal processor, turns out I really don't have a way to do that with my current setup. I'm using signal processing for the crossover so I'm fully committed! I'm not sure my setup is good enough, or my ears good enough to hear the fundamental distortion or noise that the processor may be adding to frequencies it's not attempting to correct.

 

[pjwd]

Audio Hijack can take system audio as input, as well as external devices. I bought a toslink to USB dongle to put on my Mac Mini. That feeds everything from the TV's optical output into the Mini. This only works for stereo in my setup, so no multi-channel sound capabilities. Audio Hijack only works in stereo anyway. This setup is kind of interesting because I can mix inputs together. I do this so I can listen to NPR while playing video games. I have the video game volume turned down so it's just audible. This allows me to justify extended game playing sessions because, hey, I'm staying informed, or indoctrinated, depending on who you ask.

BTW, by I managed to get the lag time on the FIR filter down to 14 ms! I limited it to 20Hz to 150Hz duty, and trimmed all the silence off both ends of the impulse file. That really goes a long way!

No one is judging
Thanks for that info.. it seems jriver is not so flexible..

 

[schlager]

Last night I was doing multiple experiments, also trying to implement an FIR crossover into the impulse response for the bass. Not sure what happened there but I definitely did not do it right. It was a total mess of weird reverb.

From what you describe it sounds like you got severe pre-ringing in the bass. FIR filter is a strong tool and if not careful, the cure can be worse that the disease. It all comes down to time-windows, aka how long the correction works with a particular frequency. In general we can use relatively large time-windows under 100 hz, up to a second in some systems. In the DRC world we use cycles (frequency dependent window) , so 10-15 cycles at 20 hz is not a problem (500-750 ms). As we move up higher in frequency I found it best to gradually reduce the cycles lengths and above 5 Khz I use 1-3 cycles, that would be considered only speaker correction, as we can´t do anything with the phase anyway, because of the multi reflection pattern. The only thing one should correct for above 1000 hz in the phase, is the excess phase generated form the crossover. What software are you using to generate your filter?

 

[Tim Link]

From what you describe it sounds like you got severe pre-ringing in the bass. FIR filter is a strong tool and if not careful, the cure can be worse that the disease. It all comes down to time-windows, aka how long the correction works with a particular frequency. In general we can use relatively large time-windows under 100 hz, up to a second in some systems. In the DRC world we use cycles (frequency dependent window) , so 10-15 cycles at 20 hz is not a problem (500-750 ms). As we move up higher in frequency I found it best to gradually reduce the cycles lengths and above 5 Khz I use 1-3 cycles, that would be considered only speaker correction, as we can´t do anything with the phase anyway, because of the multi reflection pattern. The only thing one should correct for above 1000 hz in the phase, is the excess phase generated form the crossover. What software are you using to generate your filter?

Thanks for the response. I'm using REW to generate the file. It has frequency dependent windowing but I don't see a way to change the number of cycles as the frequency goes up. Last night I succeeded in getting a good sounding result by limiting the FIR to 200 Hz, and then applying low Q parametric filters to the higher frequencies to bring their smoothed average close to the target curve. The bass is very even and clear and integrates nicely with the treble. I don't hear any hint of ringing now. So it's pre-ringing in the bass that causes the treble to sound weird?

Come to think of it, I suppose I could create different impulses with different time window periods for various frequency ranges and then just multiply these all together to create a full range impulse.

I don't know a whole lot about how this FIR method works, and what it's reasonable to try to do with it. My folded bass horns have a flaw that produces a very deep notch in their output at around 85 Hz. Because the bass horns are so loud and efficient I have to attenuate them to be more in line with my mids and highs I"m currently using. I can do that attenuation by turning down their volume on the amplifier, or I can let the FIR filter EQ them down to the target curve. If I let the EQ do that, it can almost completely erase the notch, and it seems to minimize the group delay all the way down to 20Hz. I'm not doing that right now, allowing the narrow notch to live, and allowing group delay to exist as the frequency approaches the lower cutoff. Do you have any opinion on letting the FIR filter get that aggressive? I know I'm losing about 10dB of digital headroom by EQing it down so hard, so the noise floor will come up, but I don't think I can hear that in the bass horns. But conversely, I don't think I can usually hear that notch at 85Hz either, nor do I think I can hear the 20ms group delay at 20Hz.

@pjwd I've gotten the latency on the FIR filter down to 21 ms! This is good enough for me to watch movies and play games. I had one filter that was clocking in at 14 ms. I'm not sure what all determines the latency but I know the number of samples in the impulse file is a big deal, so trimming off silence on the ends seems to make a difference.

 

[Tim Link]

I must say I really like the sound of my system equalized at the listening position to the Harman target curve. If your system and room aren't balanced to that target, and you haven't heard it, I'd recommend giving it a try. Going back to the OP's question: is frequency response everything? It's a lot, especially in the bass. Using an FIR filter on the bass to get a very smooth response while bringing down modal resonances is a really big deal to my ears. Bass traps can work brilliantly together with dsp and some kind of array of woofers to get excellent bass in a lot of rooms where it might have seemed impossible. The higher frequencies need to have a whole room response that's flat and downward tilted when the measurement is smoothed. It's going to have lots of little peaks and dips that should definitely not be messed with. However, if the broad, 1 octave smoothed levels are a bit too high or low in places it can be very effective to gently take them back down to the curve with simple low Q parametric equalization. I personally feel that a digital equalizer properly used will end up being a win every time if it's used correctly, even if it's a lower grade component inserted into a very high end system. The only time it wouldn't be helpful is if the smoothed response at the listening position is already dead on the listener's preference curve, which would be the ideal case, and extremely unlikely I'd think to occur without some measuring and adjusting. A good speaker/amp/source combo with speakers positioned in a properly treated room could probably get you there in the higher frequencies, albeit with a lot of measuring and experimenting. It'd be a tall order to really get the bass smoothed out completely without at least using a little equalization and perhaps a distributed array of woofers.

 

[pjwd]

Thanks for the response. I'm using REW to generate the file. It has frequency dependent windowing but I don't see a way to change the number of cycles as the frequency goes up. Last night I succeeded in getting a good sounding result by limiting the FIR to 200 Hz, and then applying low Q parametric filters to the higher frequencies to bring their smoothed average close to the target curve. The bass is very even and clear and integrates nicely with the treble. I don't hear any hint of ringing now. So it's pre-ringing in the bass that causes the treble to sound weird?

Come to think of it, I suppose I could create different impulses with different time window periods for various frequency ranges and then just multiply these all together to create a full range impulse.

I don't know a whole lot about how this FIR method works, and what it's reasonable to try to do with it. My folded bass horns have a flaw that produces a very deep notch in their output at around 85 Hz. Because the bass horns are so loud and efficient I have to attenuate them to be more in line with my mids and highs I"m currently using. I can do that attenuation by turning down their volume on the amplifier, or I can let the FIR filter EQ them down to the target curve. If I let the EQ do that, it can almost completely erase the notch, and it seems to minimize the group delay all the way down to 20Hz. I'm not doing that right now, allowing the narrow notch to live, and allowing group delay to exist as the frequency approaches the lower cutoff. Do you have any opinion on letting the FIR filter get that aggressive? I know I'm losing about 10dB of digital headroom by EQing it down so hard, so the noise floor will come up, but I don't think I can hear that in the bass horns. But conversely, I don't think I can usually hear that notch at 85Hz either, nor do I think I can hear the 20ms group delay at 20Hz.

@pjwd I've gotten the latency on the FIR filter down to 21 ms! This is good enough for me to watch movies and play games. I had one filter that was clocking in at 14 ms. I'm not sure what all determines the latency but I know the number of samples in the impulse file is a big deal, so trimming off silence on the ends seems to make a difference.

Good info.. thanks
How narrow is that notch .. is it obvious..I have an accumulation of room reflections that cause a narrow notch at 130hz .. I can correct it so it measures not to badly but I cant really hear the improvement ... its about Q3
Totally agree on the harman curve .. Linkwitz has a very plausible explanation for the down tilt at hf and goosed bass sounds great .. esp at low listening levels

 

[racerxnet]

You can try the hang loose convolver. The number of taps you use effect the lip sync delay for A/V usage. Lower them a bit and see how that works once you find a filter you like. Mitch Barnett developed the convolver with auto rate switching. For FIR filters you need to use RePhase and REW.

https://www.youtube.com/channel/UCcSs5v_xbH-uI4KLnIUUfuA/playlists

 

[schlager]

I'm using REW to generate the file. It has frequency dependent windowing but I don't see a way to change the number of cycles as the frequency goes up.

Yes, in REW you can only set one time window for all frequencies (cycles or octaves). The best DRC programs let you play with different windows for both frequency amplitude and phase (time domain) and also different cycles in the bass, midrange and tweeter. So they are powerful tools to integrate your specific speaker, in your specific room and also accounting for speaker placement and listening position.

So it's pre-ringing in the bass that causes the treble to sound weird?

Pre-ringing is only happening in the bass, so the artifacts that you get in the tweeter is most likely caused by too aggressive correction in the time domain. In the tweeter the correction should only be speaker correction and not including the rooms reflections. To do that you need to run low cycles 1-3 and only phase correct the crossover. But you're probably better left off, without time correction in the tweeter.

 

[schlager]

Come to think of it, I suppose I could create different impulses with different time window periods for various frequency ranges and then just multiply these all together to create a full range impulse.

That would be quite tricky, but go for it and let us know how it works

Do you have any opinion on letting the FIR filter get that aggressive? I know I'm losing about 10dB of digital headroom by EQing it down so hard, so the noise floor will come up, but I don't think I can hear that in the bass horns. But conversely, I don't think I can usually hear that notch at 85Hz either, nor do I think I can hear the 20ms group delay at 20Hz.

That is mainly frequency correction and as long as you have enough headroom, you should be fine. And no, 20 ms at 20 hz is no problem, we are not so sensitive for group delay below 100 hz, we are more sensitive to FR peaks and ringing, but taking the peaks down also reduce the ringing, so that is what is most important.

 

[schlager]

I can recommend DRC Designer, it's free and run on most convolver engines. DRC Designer is build from the same cloth as Audiolense and Acourate and is very powerful. I tried to do FIR filter in REW and Rephase, but DRC Designer beats it every day and it is a much faster process to generate different filters to compare. Recommended.

Digital Room Correction Designer Help

 

[Tim Link]

Good info.. thanks
How narrow is that notch .. is it obvious..I have an accumulation of room reflections that cause a narrow notch at 130hz .. I can correct it so it measures not to badly but I cant really hear the improvement ... its about Q3
Totally agree on the harman curve .. Linkwitz has a very plausible explanation for the down tilt at hf and goosed bass sounds great .. esp at low listening levels

I'd guess it's a 1/6 octave. I modified the speakers to improve it, which made it much narrower but didn't kill it all together. It's not generally obvious although if something I'm really familiar with has a prominent note close enough I might notice it being a little weak. 99 percent of the time it's not at all obvious.

 

[Tim Link]

That would be quite tricky, but go for it and let us know how it works

That is mainly frequency correction and as long as you have enough headroom, you should be fine. And no, 20 ms at 20 hz is no problem, we are not so sensitive for group delay below 100 hz, we are more sensitive to FR peaks and ringing, but taking the peaks down also reduce the ringing, so that is what is most important.

It didn't work too well. By the end of last night I was exhausted and disappointed.

 

[Tim Link]

I can recommend DRC Designer, it's free and run on most convolver engines. DRC Designer is build from the same cloth as Audiolense and Acourate and is very powerful. I tried to do FIR filter in REW and Rephase, but DRC Designer beats it every day and it is a much faster process to generate different filters to compare. Recommended.

Digital Room Correction Designer Help

This looks interesting. I'll have to get some hardware to make this work.

 

[Atmasphere]

It'd be a tall order to really get the bass smoothed out completely without at least using a little equalization and perhaps a distributed array of woofers.

Room correction can't sort out the bass if there is a standing wave in the room and if your room is regular in dimension, then there likely is.

Room correction will simply dump power into the standing wave, and being a standing wave, the power will be cancelled. The only way to really fix that is by using a distributed bass array. That breaks up standing waves very effectively.

If the bass isn't right, the rest of the spectrum will sound bright even if its flat. This is because the ear brain system has its own tone control- too much bass and the highs will sound muffled on the same account. If you get the bass right, in most rooms the DSP isn't doing all that much in the highs so you might not even need it.

 

[Tim Link]

Room correction can't sort out the bass if there is a standing wave in the room and if your room is regular in dimension, then there likely is.

Room correction will simply dump power into the standing wave, and being a standing wave, the power will be cancelled. The only way to really fix that is by using a distributed bass array. That breaks up standing waves very effectively.

If the bass isn't right, the rest of the spectrum will sound bright even if its flat. This is because the ear brain system has its own tone control- too much bass and the highs will sound muffled on the same account. If you get the bass right, in most rooms the DSP isn't doing all that much in the highs so you might not even need it.

I agree. A distributed bass array is a great solution, especially if room correction is intelligently used in conjunction with it, along with room treatments. Smooth, tight, and extended bass in small rooms can be achieved!