DIY Digital Room Correction with Linux | Hacker News Reader

DIY Digital Room Correction with Linux

th0ma5w.github.io

91 pointspcr91030324 comments

It's fun stuff to mess with, but one difficulty is that you can't really affect time-domain issues by modifying output frequencies.

That is, there are likely both issues induced by the phase response of the speakers themselves as well as issues introduced by reflections in the room which will cause uneven frequency responses in these measurements. These issues are inherent in speaker systems and acoustic spaces.

If you have a really resonant frequency in a room, notching that frequency can help, but then you're compromising that signal; a more typical solution is to address reflectivity in the space.

And I hate to be a snob about mics, but yee, I do not like that specific mic-- of the many dozens of mics I have used it's memorably bad. And you don't need an expensive mic to do these measurements; there are a lot of ~$60 omni-directional measurement mics that work fine, as their low/mid frequency response is good enough for these tasks.

So all in all: hooray for folks experimenting... once you start playing with frequency modification, start investigating phase response and modal reflections in rooms, as they are super interesting.

Like, if you want to hear something really neat, put on a recording of a 120hz sine in a very reflective room, and you can walk around and hear the nulls and additions. And then you can find different frequencies and start to come to terms with the complexity there. Quite a fun exercise.

> And I hate to be a snob about mics

I wonder if you just need a mic with a calibration file?

this one is less than $25:

https://smile.amazon.com/dp/B00ADR2B84

and you can use the serial number to download a specific calibration file

I have a Dayton Audio EMM-6 that I own and had in mind. It's not Earthworks, but it's good nuff and I like an XLR out for my purposes.

In the end a mic is just a converter of air movement to electricity fluctuations. The hard part in measurement mics is that they don't show any characteristic that deviates from the expectation both in the time and frequency domain.

Calibration files can help you in the frequency domain (provided that mic is stable over it's lifetime), but the time domain is a different beast as in: if we play a square click how well can the mic reproduce the square and/or what kind of artifacts are added by it.

In the end the question is, how important calibration is for your purpose. If you are a hobbyist, I would even say, you don't really need it more than once. If you mix/master productions that have high budgets maybe spending a little bit more on making sure everything works makes more sense.

If you are the hobbyist, maybe renting is a good option?

Yeah, for a one-off look at the room, then all this stuff is all overkill, IMO.

However, it's still fun to geek around with acoustics.

I mix stuff that other people listen to, and I operate sound systems in a variety of locations. My interest in this is mostly professional, though I think that it's easy to get a long way in that business without a lot of specific exploration, so really it boils down to being a fun and kind of nerdy hobby.

To your point about mics, Rational Acoustics, maker of the popular Smaart analysis program, have advised users that high-end mics might not be the most important element in a measuring system.

This is because most of the useful elements of acoustic treatment happen at lower frequencies, where the cheapo omni capsules work just fine, and these elements are not generally creating lower-frequency artifacts.

Can't you determine the time domain dynamically?

> It's fun stuff to mess with, but one difficulty is that you can't really affect time-domain issues by modifying output frequencies.

PEQ can take you a surprising distance. Many perceivable issues can be substantially reduced by attenuating signal at problematic resonant frequencies. At no point ever (IMO) should PEQ be used to boost the level of any frequency to make it more audible.

FIR filters are where you can fix time-domain issues. The only problem is that, depending on the amount of filtering required, you may add quite a bit of latency to the signal. IIR filters (e.g. for your crossovers and such) are typically much lower latency approach. IIRC FIR filtering will also allow for you to correct for phase issues.

At the end of the day, the room and its treatments are the most important part of the equation. The number of LFE radiators and their positions are probably #2. Everything else you can easily fix in software.

I feel like that is why the article writer said it got them 80% of the way.

Pareto's principle indicates that the biggest 20% of work will provide 80% of the expected results, whereas to get the last 20% of expected results will require 80% of the work.

Doing a basic room EQ with the equipment you have on hand would mean that the spot you EQ'ed from would have a listening experience calibrated to the quality of the microphone at that location (but with a margin for error since they were not using any heuristic other than "make line as flat as possible in a single pass")

That's ~20% of the work in making a pristine audio environment but with close to 80% of the end results. Room correction, reverb baffling, bass traps, better speakers, fancier receivers with complex auto balancing algorithms all could definitely improve the end result but will take far more time and effort (and money) to achieve any noticeable improvement.

It cannot be understated how much better even modest/mediocre speakers can sound when their in-room response is corrected via DSP.

(This is essentially why a lot of consumer electronics sound surprisingly good these days: onboard DSP is cheap and easy to implement. What a blessing!)

Conversely, even "high end" speakers can sound bad if not dialed in correctly, especially if your room is rectangular and there are a lot of reflections.

I used to run an outboard FIR filter for my subwoofer with weights calculated (in part) using REW. I can't recall the actual DSP hardware model, but it was a pretty amazing effect when you toggled it on/off.

Today, I just have a miniDSP that does basic crossover duty. I haven't bothered to do any parametric EQ or more advanced filtering in my new office yet. The passive acoustic treatments have done such wonders that I probably can't be now.

I don't really like running anything above 80Hz through digital filters that I have constructed myself. I've had some success in a few areas, but you can instantly tell something isn't quite right with certain content.

I too did this exact thing with hardware instead of doing it in Linux. I used an t.racks DSP 4x4 Mini to apply my REW-measured room correction between my sound card and amp, also use it to generate the split for my subwoofer.

How do you like the miniDSP? Thinking about getting one - room correction definitely seems like a job for hardware; i.e. easy enough for a DSP, more portable and hassle-free than setting up software on one computer.

It's pretty good. I have the 2x4HD and no complaints. The software is adobe air crap but it works well enough.

I am using it as a USB DAC driving a pair of stereo monitors and 2x subwoofers at the moment.

The biggest problem with home theater setups with surround sound, IME, is that no matter what you do most of the seats will get very uneven sound (typically, one or more surround speakers being much louder than the others). The only fix is to have a larger space so the effective "sweet spot" covers more of your seating (think: an large-aisle-width space around a 3x2 seat configuration) but at that point you're looking at sacrificing a mid-sized living room worth of square footage for those 6 total viewing seats (and even more, if you scale up from there).

[EDIT] In case it's not clear, the core problem is that for some seats, without a large buffer between the seating area and the speakers, the nearest surround speaker will be like 5-10% as far away as the farthest one. No amount of room-correction can help much for most of the seats in such an arrangement. All you can do is use a larger space so you can put the speakers farther away without changing the size of the seating area (so, add empty buffer space around the seating area) so the difference in relative distance between the farthest and nearest surround speakers is smaller.

I'm in the middle of a dedicated theater room project. We're doing 5 seats behind 4 seats, which is probably more than we'll likely use, but it does have the benefit of expected use will have people farther away from the side speakers.

Depending on the seating costs and budget and other things, I think it makes sense to go ahead and put seats around the sides... in case you end up having extra guests, they can sit in comfort, even if the audio isn't ideal.

On a small budget, you can do a lot with just two decent speakers rather than speakers built into an ultra thin flat panel. Thrift stores often have nice speakers that may not be visually pleasing but are likely to be aurally pleasing.

> On a small budget, you can do a lot with just two decent speakers rather than speakers built into an ultra thin flat panel. Thrift stores often have nice speakers that may not be visually pleasing but are likely to be aurally pleasing.

This is very true, almost any separate speaker is going to be better than TV speakers, especially in the age of flat panels.

And one thing software room-compensation can do is let you build a surround system with mismatched speakers, without having to lose a day manually tweaking levels. No need to buy a full 7.1 or 7.2 set at once. Thrift decent pairs of speakers and orphaned subwoofers and let the software figure it out.

Good to see the detailed write-up! Many others are just a brief guide to make you buy their service or product.

> That seems like a reasonable thing to do, but there is a lot of pseudoscience in the audio world that will get you to buy platinum tipped styluses for depressing buttons on your remote.

I think there are a LOT of these. I wish I knew when "reasonable" or "plausible" pan out.

Room EQ Wizard is your friend.

> often had to have the volume at a bare minimum of 50% in order to hear speech correctly. Now I could get away with 15%.

Sounds like you don't need a mic or anything other than the EQ to make movies watchable. Just reduce every frequency range other than that of human voices. No more having your ears blasted by loud music / explosions because, earlier, you had to turn the volume up to maximum in order to barely hear the dialog.

I recently bought a "t.racks DSP 4x4 Mini" which is an inexpensive hardware DSP box capable of the same. I just have it sitting between my soundcard and some active speakers.

Gives me peace of mind knowing my EQ curve always is on even if I'm using software that's outputting direct to my interface.

I've been interested in this topic recently, I hate my AVR, but there isn't a better alternative. I've been hoping instead of an AVR I could pipe HDMI into my PC and run DIRAC there, but I've yet to find a capture card that can capture DTS/Atmos etc, so it hasn't been fruitful

Check out minidsp stuff.

Does anyone know how this compares to something like Sonarworks, which does a similar process but for a series of measurements across your room with a specific microphone they provide?

Reply on news.ycombinator.com