Nice! We're basically neighbors - I'm on the north shore (within 1 hr. drive I'm sure).
As for speaker/speaker + room correction, you can see I've done both looking at posts 401 and 402. My experience is that good results can be achieved either way but it can certainly be tricky when working with a single listening position measurement.
Great to know another DIYer is in my neighborhood!
I agree that single position measurement is tricky and misleading, but for low frequency, I find it is almost impossible to do reasonable correction based on multiple mic measurement in a room. I wonder how you could do it...
Well, I think it's all about keeping the correction to a minimum without losing effectiveness. A good place to start might be with REW since it's easy to play around with. Record sine sweeps from the sweet spot, and then use a 4-8 cycle frequency dependent windowing of the minimum phase portion of the signal as the basis for EQ that only reduces peaks above a given threshold. You may even want to reduce the overall (right) window length to 30-50ms since our hearing resolution is low in the bottom 2 octaves.
Anyway, If you're interested in giving it a shot in the future, I'll try to help with the specifics.
Anyway, If you're interested in giving it a shot in the future, I'll try to help with the specifics.
1201 (and anyone else interested),
I added a 2nd filter comparison folder containing the same files but A-weighted RMS level matched. I'm interested in whether or not this will affect results. Thanks.
I listened to it. they sound much more similar now but I still prefer the piano sound on the 4 cycle correction.
I guess my question would be what has you not boosting the peaks to get a more true frequency response? excursion? ringing in the time domain?
Ive boosted peaks a lot in the mid range to match the tweeter and there was no discernible impact to the sound.
Correcting the anechoic magnitude response of a speaker is a waste of time imo.
once you introduce the speaker into the room, the response changes completely and MUST be corrected for there to be a semblance of accuracy.
If you think it doesn't help then you are using the wrong tools.or using the right tools wrongly.
with all due respect
Thanks again. Just to be clear: do you feel that the 4 cycles filter sounds more like the reference track (not just subjectively better)?
The psychoacoustic filter is not boosting dips nearly as much because it is believed that dips are subjectively less important than peaks. I think that this might be something that can be taken advantage of when attempting to correct a farfield response, resulting in both reduced driver excursion, and reduced correction strength overall (better off-axis response). Reflections can be thought of in the time domain as superimposed copies of the direct sound which are both delayed in time, and reduced in amplitude, and in the frequency domain as comb filtering (series of dips). In my experience, with a set up where there is sufficient time delay of the reflections, this comb filtering is not detrimental simply because we don't perceive it as such. I feel that boosting the dips by more than a small amount, will at best marginally improve the perceived sound at the sweet spot (at the expense of increased driver excursion and degraded off-axis response), and in more extreme circumstances, fail to provide a satisfactory sound at all.
As for a truer *perceived* frequency response, that is supposed to have more to do with how we interpret (interpolate?) magnitude response (connecting peaks (envelope) vs. average level). If you ask me, the snare drum and vocals in particular in the psycho filter clip avoid a slight "hardness" in the mid freq region that is present with the 4 cycles filter.
Thanks for the feedback - good to know
. I'll probably need some time to get to that unfortunately, but it should be interesting.Hi, I will listen again, but IIRC the custom was closer in sound to the original. the original has a crunchy, instant, sound to the piano, that the psycho was missing. that to me was the only difference.
it makes some sense because piano has a lot of overtones and a dip at one frequency may make it sound less piano-ey.
I do agree that dips ARE subjectively less important than peaks, and are harder to hear but I also believe they are still quite important. it would be an easy thing to test. create a couple of realistic dips in rephase, convolve the filter with a song and see how perceptible the difference is. I wont be surprised if
1. you can clearly hear it
2. you don't prefer it.
I guess what I'm suspecting at this point is that the comb filter pattern (dips) change more with location than the magnitude response envelope does...
Anyway, thanks again, guys, for sharing. I'll try to keep you updated (you've been warned
keep it coming
if you haven't tried the moving mic method, it will be an eye opener for you.
it only shows peaks that are persistent across several points in space, and the same for dips.
imo that is the true response we hear and that may be the best measurement to fix
Let's do a virtual audition/comparison (using convolution) of two filters as "heard" from my listening position. The custom filter is an impulse response inversion with 4 cycles windowing, and the psycho filter is a magnitude response envelope inversion of an ERB windowed impulse response. Level matching between channels as well as between filters is based on peak level. Which track sounds more like the reference? Let me know what you think
I had a listen to the A weighted RMS matched tracks on headphones, what I got when I first heard the corrections was the same feeling I get when I move into a new house and listen to a system I am used to. How come it sounds wrong, what happened to it, I must have got something wired wrong.
Both sounded phasey on the vocals and with an emphasis on the higher frequencies but the custom sounded more phasey to me on first listen. Both sounded like there had been a fair bit of small room reverb added, which I suppose it had! After I had listened to them both 10 times or so I started to tune out the room sound that had been added and both started to sound more similar. I can still hear the added room reverb but it now bothers me less.
If I had to pick I would say the psychoacoustic filter sounds more like the original on the plucked/snapped guitar parts and hi hats as there is slightly less added high frequency emphasis.
I imagine because you are now used to the sound of your room that it is easier for you to notice the differences between the filters.
I guess what I'm suspecting at this point is that the comb filter pattern (dips) change more with location than the magnitude response envelope does...
Anyway, thanks again, guys, for sharing. I'll try to keep you updated (you've been warned
Which comb filter patterns? From a single driver? If there are no obvious early reflections and no gross diffraction happening from the cabinet walls/shape, what would be left to create a comb pattern? At least above ~300 Hz I wouldn't expect that to play a huge role.
You would be too if you were in that room. The recording (well virtual) emphasizes the room. If you're actually in it at the listening spot it wouldn't present itself this way or this obvious.
I can't say I'd describe the sound as phasey. The balance is off, yes... which puts the high frequency material in your face. But phasey sounds different to me. Can you describe what sounds phasey to you?
To me the custom filter sounds a tad more relaxed (when you listen past the tonal imbalance). A little less busy so to speak.
The reference is a bit cleaner and even more relaxed due to not having the high frequency emphasis.
Yes which was why I equated it to listening to a familiar system in a new room, after a while you tune it out and everything sounds normal again.
When I first listened to them it sounded like there was a random chorus effect on the vocals, the more I listened the more it went away as I got used to the sound. When the chorus effect is mistimed it is closer to a phasor effect which is what I would describe as phasey as if there is a delayed version of it coming and going.
I found the added room sound quite distracting and the psycho seemed to emphasise it less once I had tuned the worst of the room sound out. I find it very hard to listen through a track with what sounds like small room reverb added, the psycho filter was the least obnoxious to me in this regard so my opinion may be biased in that direction.
Thanks, fluid, for listening and making some really good points. It's really amazing that with repeated listening of the convolved tracks, we can start to tune out the room even though to a large extent it is "glued" into the recording (our processing abilities amaze me more all the time). And like you said, I obviously have a big head start here since it's my own room. I hear the psycho filter as being slightly more neutral in the midrange, but would I hear it that way if I was an outside participant? I'm intrigued by your experience, and again, thanks for sharing.
From the floor, sidewall, and ceiling reflections is what I was thinking. A 4 cycle FDW windows out a lot of it above 500Hz (not so much in the case of the ERB window...).
I have listened to your samples as well.
When you helped me apply the 4-cycle filter to my room and speakers a while back, I remembered I liked it a lot.
Now that I listened to both the 4-cycle and the newer one, I have to admit the 4-cycle sounded a bit too pushed up front, with transients accentuated to the point where they start losing their natural feel and feel like they were tinkered with.
So, I also agree that the newer filter sounded more natural and true to the original. Effects were subtle.
If I'd heard the original song one day, and with the new filter on another day, I'm pretty sure I wouldn't have picked any differences.
When you helped me apply the 4-cycle filter to my room and speakers a while back, I remembered I liked it a lot.
Now that I listened to both the 4-cycle and the newer one, I have to admit the 4-cycle sounded a bit too pushed up front, with transients accentuated to the point where they start losing their natural feel and feel like they were tinkered with.
So, I also agree that the newer filter sounded more natural and true to the original. Effects were subtle.
If I'd heard the original song one day, and with the new filter on another day, I'm pretty sure I wouldn't have picked any differences.
Thanks, perceval .
Yes, the difference is subtle so it may be more a matter of long term adjustment. However, as I previously mentioned, I think the difference is more obvious in reality than with the tracks.
BTW, I just caught up on your "FAT" project; nice work there! The new filter script calls on multiple configuration files (despite the simplicity of the actual process), so if you want to try it out, and need any assistance (I assume you're still using DRC designer?), I'll be happy to help.
. Yes, the difference is subtle so it may be more a matter of long term adjustment. However, as I previously mentioned, I think the difference is more obvious in reality than with the tracks.
BTW, I just caught up on your "FAT" project; nice work there! The new filter script calls on multiple configuration files (despite the simplicity of the actual process), so if you want to try it out, and need any assistance (I assume you're still using DRC designer?), I'll be happy to help.
Last edited:
Do you mean no difference between the reference and psycho filtered track? Are you listening on speakers or headphones?
gmad said:
Excuse the rant, its not personal, just my view after trying everything I could come up with over the course of (at least) a year with daily experiments (2015) and many more afterwards, though not as frequent.
The longer ERB filter would let in more of the room, agreed? Not something I would strive for.
It would still make more sense (to me) to use a shorter frequency dependent window. As wether you like it or not, a longer window would try to correct very specific room anomalies/reflections. A better choise would be to use a moving mic measurement or multiple measurements averaged. They only see things that are the same between those points or over that area.
Yes, our perception of tonal balance changes over a longer time frame. That will certainly result in a difference in perception between two different corrections.
If you want to get rid of those dips or the room addition you see in measurements? Treat the room. Look trough time using all tools and you will see when the phase changes. Set the gate longer and longer untill that phase plot deviates from the min phase. Using a longer window would correct that deviation.
If you don't want to clean up the room, use multiple measurements averaged or a moving microphone measurement.
A longer window with less dip correction from a single point measurement will not succeed in averaging out the room reflections but will still mess with phase a little later in time than a shorter window.
Of axis measurements will most probably not get better. It will sound different though.
But that could also be accomplished by using that shorter FDW and use an average or sweeping mic to adjust the tonal balance afterwards (room curve with broad stroke EQ, I used to do that when I didn't have all damping panels up yet, I've always used post EQ and still do). What would you win? More valid imaging queues. We wouldn't be adjusting level plús phase based on reflections. You'd be better off with gradual EQ.
Compare early waterfall graphs between both methods and tell me which one has the better speaker correction.
We as listeners are so easy influenced by tonal balance changes alone making this comparison sort of a crap shoot. A binaural recording might do better to hear what you are hearing. As we don't get the imaging and do hear the room. Room alone can totally change the perception too. Most people actually like a bit of room. Or even speaker anomalies, just look at the subjective listening threads for evidence of that
.Just try to only change tonal balance of the short window correction to match the newly preferred filter with simple PEQ. Believe me, very tiny differences do count, if the change is over a broad enough area. You'd need to decide on some filter resolution to compare them.
I'd love to see off axis comparisons, early waterfall graphs and IR/filtered IR for both.
Different alone isn't better by definition and its not convincing me in any way (just yet). Believe me I've tried any and all window sizes for low, mid and high frequencies separate from each other. They all sounded different. Just get the overall tonal balance right an everything will start to sound better. For that we need a longer view of the FR and broad stroke EQ, or a way better solution than that: room treatment.
End of rant. Must not start how getting rid of all early room stuff completely messes up the stereo principles or I'll be up all night typing. Two ears vs one mic does make a difference. Use it to your benefit I say
. More fun!
Last edited: