Agreed (also pointed out by Resolve above). The stuff above 5K should be dialed in by ear, but it’s nice that we have a tool like this to help get us started.
I have my LCD-X 2021 dialed in nicely now, but I’ll play with the AutoEQ and manual functionality of this later this week for fun. I never pass on the opportunity to nerd out on audio stuff!
I tried reverse engineering Mitch’s filter, and this is what I got:
Here is my modified oratory EQ:
Yeah, you’ll get a feel for it by trying to follow the shape with a a handful of filters.
His is essentially tens of thousands of micro adjustments? Is that the idea?
So where all these EQ settings by various people take a measurement graph with a wack of smoothing, and adjust the general shape, the convolution filter doesn’t? Adjusts each mini dip/peak of a unsmoothed graph? There is unit to unit variation yeah, but he did say the several different models he measured are all pretty close.
So if a hump between 2 and 3k for example, needs PEQ down, it’ll get one or two filters pushing it down. In reality, that hump without smoothing has its own peaks and troughs, maybe 1-2db or whatever, so pushing them up or down, as part of correcting the smoothed graph, can have a additive effect? The smoothing masks mabye a sharp -3db trough, and then the PEQ corrects it down further, exascerbating it. Based on that, I’d imagine if you applied his convolution filter to get the response pretty much flat, then applied PEQ to your desired tonality you’d be cheering.
I can’t get the level of refinement and balance I hear in the convolution filter, with any set of PEQ filters. I can get the general tonality easy enough. Even if its in my head, I’m down with that.
Very interesting read from Mitch’s work on convolution and DSP filtering. Thanks for the link and awareness of greatly advanced digital signal conditioning. Learning a lot about this topic that I never even knew existed.
Did you get the filters or were you just trying to emulate a peq similar graph?
I am assuming you haven’t or you wouldn’t have bothered. I just tested the top values that you reversed and there is a noticeable difference in top end. The best I can describe is that fine edge just before being shrill in certain sounds is just about completely eliminated but it doesn’t sound veiled or cut in dynamics. It is still full of clarity and tone. Hard to explain but all I can say is there is some wicked sorcery at work here.
Yeah 100%.
I think its down to how smoothed the measured graphs are before these EQ settings are then applied to correct. End up with an overall smooth graph that looks good, but realistically you’re boosting sharp little peaks and reducing sharp little troughs already.
I was fairly content with a modified version of Resolves EQ before taking the punt on the filter. No going back now :S
@Resolve Can you share your measured graph (up in this thread somewhere) vs the same one with no or much much less smoothing applied?
Yes. I fully understand now. 12 band peq ain’t got nothing on ten thousand band adjustments. Kinda like oversampling the graph.
Of the LCD-5? This might not be the best example for this since it has quite uniform diaphragm movement and you don’t get the kind of ‘wibbles’ that show up on other planars. I also only ever smooth to 1/12th anyway.
I don’t think that would be the case, since our EQ APO doesn’t do that kind of smoothing. Also, the differences in our ear shapes make it pointless to adjust every nook and cranny in an FR plot.
What I mean is the filters you derive in EAPO are based off looking at a smoothed graph.
You see a dip between say X and Y, and you use a filter to boost that back up. Actual fact is that dip on the graph you’re looking at it a bunch of dips and peaks that form, once smoothed, an overall dip.
Mechanically I wish companies would stop using REAN Mini XLR. They are too hit and miss. The spring loaded clip part uses a small sponge to work. I think that sponge shifts a bit sometimes and the clip then doesn’t depress properly, leading to some connectors being really hard to pull out.
I was going to replace mine with Furutech, but I think i’ll just make a whole new cable. Got some real nice copper and silver plated copper litz here, can’t go wrong. Thicker and softer than stock cable, less janky too.
I believe you might still be missing my point.
Anyway,
Using EQ, regardless if you obtained it through a purchase or a professional reviewer, is a highly diminishing returns endeavor. Eventually you get to a place where the mixing of the track is much more influential than the filter the end user enables.
@EminentOne, did you make sure to disable the Mitch filter when you enabled the peace add-on?
I simply created a mock-up based off of the image of the Mitch filter as seen in the EQ APO. I’m sure I could get a closer approximation if the graph was less than 10db measures 
Yes. I use the Matrix Pipeline in HQP for the filters and peq. Both lines for L & R mono .wav files were deleted along with my bass shelf peq before I restarted and entered your reverse peq values and restarting again.
Don’t get me wrong, the tonality is there but there is something else going on affecting the sound.
I have found the need for EQ on these to be overblown. I’m perfectly satisfied with the stock tuning and end up going back to it after trying various EQs. Don’t just assume you won’t like these in stock form. They’re very neutral and inoffensive from my TT2 and a bit more exciting (a bit more bottom and top end) from the Questyle CMA Fifteen. Personally, don’t really need EQ on either one.
Yeah. The little graph at the bottom is merely showing a general shape of what the .wav file is doing.
It has some voodoo going on that you couldn’t replicate with 10-20 bands of PEQ.
Well its all preference. There is no need at all.
For my ears, the stock tuning has way too much in the 3-4k region. Triggers my ears within a few seconds/minutes. But I’ve personally always felt most tunings people like, including Harman, has too much in that region anyway.
Probably why I really like the Final D8000 and ZMF Verite. Naturally have some dips there.
I’m gonna need a bit more proof explaining how an open source program is creating voodoo. There is no reason to believe smoothing is going on in APO’s graphical representation of the filter, APO likely shows exactly what the math tells it to.
As with all graphs, it can be cut into a nearly infinite amount of subdivisions. But that doesn’t mean it can’t have a best fit plot represented by 10 or so bands, matching the original high res graph to an indiscernible degree.
Besides, the overall attitude is very defeatist. Any problem can be simplified and solved.
Ha. Go back and read what I said dude.
The graphs of the headphone FR measurement are smoothed. Nothing to do with EAPO.
You, Oratory, Resolve, Crinable, Bob The Builder, whoever, look at these smoothed graphs, and start to derive PEQ profiles. They add filters to push dips and peaks to better fit their target curve.
What I was saying, is the the visually observed dips or peaks, on these graphs, with smoothing, are then corrected with PEQ. BUT, these dips and peaks, without smoothing, actually have their own smaller/sharper dips and peaks. Just when smoothed, visually, they form a more overall shape.
You’ve taken me saying “smoothing” and somehow figured I’m implying that is occuring within EAPO.
I can get the tonality of the filter approximately the same with say 10 bands of PEQ. That’s easy. Simplified and solved bud! BUT. It doesn’t sound as good as the filter to me, or to many that have also tried it.
I don’t need you to agree. Just don’t tell me your 10 band PEQ that looks about the same as Mitch’s convolution filter, as graphically depicted in the bottom of the configuration editor view, is indiscernible to my ears. The filter sounds better than any PEQ setting to me. I’d happily test a set of PEQ settings against it, but it doesn’t mean anything because you’ll just call it biased.
I am not sure about this but I read that the FR is just a part of what is in the convolutiion. There is phase correction data that isn’t represented in the FR graph.
Still missing what I’m saying, but that’s ok. I never mentioned FR once except to explain how our ears are too different from each others it to make sense to filter out the high frequency combing effects. Every graph reference was to the filter graph in EQ APO that I had shown in the bottom of my screenshots. That bottom graph is Mitch’s filter graph, in case you were confused about that.
Also, I never said the quick and dirty EQ I made was equal to the convolution filter. How could it be, since the image of the filter graph I looked at was so zoomed out and lacking frequency labels and dB label detail? I’m sure it would be a trivial process to completely match the filter with an APO EQ profile if I could see a more zoomed in plot of the filter APO graph, that I have shown in my screen shot I posted.
I’m not sure why you would think I would call you biased, is that something you tell people out of habit or something? I’m just trying to figure out the best sounding filter settings for everyone to try and experiment with. If you think I think it’s a money thing, I assure you it is not. I could slap down $200 and buy it myself, money is no issue, but that would be stealing intellectual property and would be very dishonest of me to give away after reverse engineering a simpler filter profile.
Anyway, seems communications have broken down so I’ll just go back into hiding.
@EminentOne Headphones are minimum phase devices, there is no phase correction. Everything in the filter is essentially high res EQ. The Hang loose convolver is just a smart switching EQ filter program that auto adjust the volume of each EQ profile, so you can get an honest look at which EQ profile actually sounds best. Otherwise, by switching EQ profiles in something like EQ APO / PEACE, you might get a misrepresentation as to which EQ actually sounds best because one EQ profile will have more “real” energy across all of it’s spectrum compared to another.
I thought the general rule of PEQ was to use as few filters as possible with wide-band Q? If Mitch’s “filter” uses hundreds or thousands of filters, doesn’t that go against the commonly touted conventional PEQ wisdom ? I’m not knocking his filter, of course - just trying to understand why an exception should be made for his method.
