3ll3d00d

let me know if you hit any issues, don't think anyone else has used it yet I could easily add some options to make it easier to use for this case though, for example 

auto detecting files dropped into a certain directory

providing some config option for tweaking the fft length

providing a quicker way to see peak/average on one chart (rather than having to select peak then select average each time)

fwiw the vibration app I wrote does the same sort of analysis as speclab for the target curve function - http://www.avsforum.com/forum/155-diy-speakers-subs/2681865-rpi-based-diy-vibration-meter-3.html#post52214001- and the relevant bit of user guide is http://vibe.readthedocs.io/en/latest/uiguide.html#wav-upload

It is currently tuned to produce output with ~1Hz resolution though, it decimates to 1kHz sample rate and then performs an analysis based on a 1024 point fft. Bit easier to use than speclab as you just drag and drop wav's into it, no options exposed to configure it though.

I think that sort of thing is most commonly packaged as a VST

where do you get the idea from that a PC is limited with respect to what sort of filters it can apply?

is it possible to approach this from a different angle? i.e. assume that port compression exists and model that effect (on driver excursion and output)

this would be analogous to the way you can model the effect of power or excursion so you'd set a port velocity limit of, e.g., 10m/s and then see what happens next.

deeper is always better obviously it sounds like I'm being deliberately obtuse but I'm not, honest!

Basically my approach here is model some stuff, see what can be achieved, compare to what can physically fit in my room, compare to what I have already and see whether that is something that makes sense as part of the system, repeat until I find one or more interesting things to build. This means that, atm, I'm asking "what can be built using this approach with 10-12" drivers?" rather than "I want something that can do 110dB at 25Hz" (or whatever).

Perhaps, not sure really, depends what the models come out like. By that I mean the main position available is fairly near field.

I don't really have a plan atm tbh, just want to try modelling some different designs and see what happens.

Size is the main constraint, about 2' high x 20" deep and maybe 4' wide. Budget I don't mind.

do you have any suggestions of drivers that would work in a scaled down version? e.g. using 10s or 12s. Just curious to try modelling a smaller version before I decide what to build this year.

As for REW FDW producing a spiky response, isn't that because it's not applying additional smoothing?  I recall a discussion on HTS about this where Accourate applies some kind of "psychoacoustic smoothing", whatever that means.  It likely involves some kind of weighting approach that maybe emphasizes peaks more than nulls or something.  That's OK and may be useful to its filter calculation approach, but is it really psychoacoustically relevant?  I don't think so.  I think that only the frequency response of the first-arrival is really psycho-acoustically valid.  Once the time windows get longer than a few cycles, the frequency response is not the best tool for understanding what we hear, whether smoothing is used or not.  Instead, some kind of a time-frequency analysis is more appropriate, IMO, because the brain will generally hear the first arrival and later arrivals as distinct (albeit related) events in time.

 

acourate does do something that looks like a form of dip limiting along with some sort of perceptual weighting, no idea how it is implemented. The REW version is done via the application of a cubic mean. I don't think it is an unreasonable approach psychoacoustically because we are less sensitive to dips, particularly the extremely high q high(er) frequency dips that can result from an FDW.

IIRC the typical filters used to a model the response of the basilar membrane looks more like a 4-6 cycle long window so 2.2 cycles seems a bit short to me if first arrival perception is the target but there aren't fixed rules here, if it works then great. Certainly using a shorter window should result in more robust (less position sensitive) results. The time between first and later arrival needs to be extremely large (relative to reflection times in a typical room) for it to be heard as separate events though.

I don't think anyone debates multeq xt is basically a bit rubbish do they?

I use a frequency dependent cycle length, I can't say it is a critical feature though. Perhaps I have been doing this too long but I am quicker to cut to the quick these days. I would like to think this is a nice benefit of spending a good few years doing this so that I can get to a good outcome fairly rapidly. I imagine I am lazier too though

Anyway I would say the issue with rew fdw is that it tends to produce a V spiky response when you go past a few cycles which makes it harder to use.

An easy online alternative is to paste it into https://www.mathpapa.com/algebra-calculator.htm

i.e. Paste into the box and click evaluate, you can then enter values for gain and q

1/((((1/Q)^2-2)/((10^(d/40))+1/(10^(d/40))))+1)

I mean measure the response of your combined set of filters by running a sweep through jriver, compare against the expected correction.

The formula to convert Q to S is listed in http://www.musicdsp.org/files/Audio-EQ-Cookbook.txtand is

 

1/Q = sqrt((A + 1/A)*(1/S - 1) + 2)

 

where

 

A = 10^(dBgain/40)

 

If you solve for S you get

1/((((1/Q)^2-2)/(A+1/A))+1)

You could run a loopback measurement through those filters to see what the shape looks like.

sounds like these ideas might help 

http://www.diyaudio.com/forums/pc-based/298248-sensorless-dsp-cone-excursion-limiter.html 

http://www.diyaudio.com/forums/pc-based/289698-idea-linkwitz-transform-ladspa-plugin-lookahead-boost-control.html

If it's the only source then your life is much simpler as you no longer need an audio interface with inputs and you don't need a dsp engine that can sit there spinning on those inputs.

If you're using windows then EqualiserAPO is one option, jriver is another. The former sits in the windows audio system and does not use ASIO (or WASAPI exclusive), the latter is more flexible as it has both a loopback interface for rendering system sounds and is a player with a highly capable dsp engine as well. jriver is probably the simplest option as well as one of the most feature rich from a DSP point of view. 

If you prefer linux then jriver is still an option, though lacks the loopback interface, brutefir, for a convolution option, and some sort of alsa/ecasound combination. 

An audio interface can cost as much as you want to spend, from a few hundred up into the stratosphere . My preference is to use the same interface for measurement as for playback so I like devices that have at least 1 mic pre and at least a few inputs. To that end, I've had a focusrite saffire pro 24 and an rme ff800 in my system though I currently use a motu 1248. A motu ultralite mk4 is probably one of the most compact units you can get with sufficient output channels, has solid drivers and should be a nice, clean interface. On the other hand, if you go for something that is purely a DAC then a cheap option is something like the minidsp u-dac8. Ultimately the right choice for you depends on how much you believe in the impact of electronics on audible SQ, certainly lots of options out there anyway.

Remember you need a 2 in/6 out (?) audio interface as well. Is it intended to be the source or just the processor?

How would a decent NUC plus audio interface be cheaper than 2 minidsp's?

I hope it works out for you!  I'm not sure it will work for everyone.  It may just work well for me because of the exceptionally even coverage from my SEOS horn-based 2-way speakers.  My thinking was to try to capture the anechoic response and make that flat.  Along those lines, the FDW window length must be long enough to allow for crossover delay (which usually peaks at the actual XO, in terms of FDW cycle count), but not so long as to allow any close room reflections to interfere.

 

I don't really follow that, a 2.2 cycle FDW is much smoother than an anechoic response.

From a modelling of view, I believe soundeasy has an extended box model that tries to account for non linear effects. There are some details in http://www.bodziosoftware.com.au/Chapter_4_2.zip

One thing I haven't found much data on is the accuracy of a model with respect to the simmed velocity. Anyone seen such data? or measured it themselves?

came across this paper recently which goes into this area in some more detail -> http://koti.kapsi.fi/jahonen/Audio/Papers/AES_PortPaper.pdf

 

1. Applying these filters in JRiver: the LCR and back/surround filters are pretty straight forward, but should the "Sub" channel be used to apply the filters labeled for "LFE" in this thread? I ask because I'm not sure if the LFE filters mentioned here are meant for the LFE channel only, or actually meant for the sub channel (where I send all <80Hz frequencies from all speakers too)

 

the filters are applied to input channels (e.g. LFE) not output channels (e.g. subwoofer). The JRiver sub channel simply refers to channel 4 which means the LFE input channel if the PEQ block is before bass management and the subwoofer output channel if the PEQ block is after bass management. This means you need to place a PEQ block before bass management and apply these filters there.

this works well -> http://andyc.diy-audio-engineering.org/mso/html/

v nice piece of software

it's rather hard to read a graph with loads of lines, different scales and no legend  

are you sure about that point about usable data being found under the window limit? I've never heard anyone say such a thing. The frequency resolution is defined by the window length after all.