WDR Control Room 7 and integrated subwoofer RIRs #4
Comments
|
Hi Capoei, I attempted to improve the magnitude and phase response of the subwoofer response using the original BRIRs as a reference and have updated the dataset with the results. Please let me know if there is an improvement to the low frequency response. If you are interested in analysing or improving the subwoofer RIR, here is a link to the new and previous versions of the IR in WAV format. I have also included raw BRIRs (without EQ) for control room 7 for reference. |
|
Hi. just giving some feedback. Thanks a lot. I have been trying things out with the no eq files the whole day. leave this open, please. |
|
Hi. check out my EQing. not so sure about the result. I need some fresh ears. tell me what you think https://drive.google.com/open?id=1C1ySvQV5SpEjBAB6O04k8ebqFHTKJjFC EDIT: balance is still off. +30 needs about 3dB reduction |
|
unfortunatly even the unequed IRs are full of problems in the low end. not sure if their room was poorly treated or why. |
|
Do you think If adding a 30 and a 45 degree IR I get 37.5 degrees? I think 30 is a little too steep (yea, I know it's the text book stereo triangle), and 45 is too wide |
|
hearing the equing you did you might like these more. they have a B&K house curve; going linear down above 100Hz and end -6dB at 20.000 |
I will add more azimuths to the dataset so that you can select one closer to your preference. Adding the two IRs together would not be ideal.
It could be due to a number of reasons including the acoustics of the room, the ability of the speakers to produce low frequencies, low signal to noise ratios, or even the measurement system. |
|
great to see the more azimuths. |
|
The no-EQ versions of room 4 were exported from the original miro data structures after resampling from 48kHz to 44.1kHz. The same BRIRs should already be available in WAV format in the original dataset . If you are having trouble with equalising the BRIRs, any convolution filters you create with DRC should be convolved with each channel of the BRIR or alternatively with the input/output channels if you are using Equalizer APO. As for the azimuths, more azimuths are available in the dataset I linked. Thanks again for the feedback. |
|
I had a look at the original dataset before and got very confused. |
|
Thats the correct download. |
|
I worked weeks on "room correcting" and frequency extending the WDR CR7 IRs. And yes, it is much "dryer" now. I think it sounds very good (the clipping ocured at ACC encoding stage. wont redo it for this simple demonstration): https://www.youtube.com/watch?v=3PrRXjyarpg |
|
I created correction filters against the stereo mic (Mk2) IRs and then aplied them to the binaural IRs. |
|
Hi caopei and Shanon, I am not an expert but from my experience, there is a "general" problem in binaural audio reproduction : results are poor compared to what could be expected from a real experience. Generally, this poor listening experience (very high timbral coloration or poor localization accuracy issues) is usually justified 1) by the fact that the HRTF filters are not individualized, or 2) by the fact that there is no headtracking that help the brain to reduce front/back localization confusion etc. The first argument is debattable and I can link some scientific papers to support this skepticism:
In my opinion, there is something to investigate somewhere else: maybe in the headphone compensation and/or in the measurement process (are the bone/cartilage sound conduction pathways really negligible ?). I tried to make an averaged HRTF over the entire ARI database, then I tried to equalize it with Apple Earpods + Ahastyle Covers. Here is the config I actually use : you just need to unzip the compressed folder and load the config-jean.txt file. The result is still not perfect but it seems better to me then the result obtained with the usual method (measurement with blocked meatus + headphone compensation filter). Also, averaged HRTF I am using have a nice transition between adjacent azimuts and hence is nicely interpolable. I am investigating another method for synthesizing an HRTF using a Koss Porta Pro headphones, which are nicely transparant (i.e. when you listen to an external sound source, and place the headphones on your ears, you can't perceive any timbral or localization variation). I hope to get some results in the next weeks for the Koss Porta Pro within frequency range 100-2500hz (maybe up to 5000-10000hz but less sure about that). |
Sounds impressive and very clear. |
I think about it lol. going to reply @jeanibarz, too |
|
Hi Capoei, Which headphones do you use and give good results for you ? Headphones have completely different responses between each other, I would like to test the videos with the same conditions as you to give you some feedback. I am currently trying to find cheap headphones/earphones to share analysis with other people too : I thought about the apple earpods first (30€) but they have too poor low frequency extension. I tried also the Koss Porta Pro but different units (samples) seems to be not consistent (some have good bass, others not...). Not sure however about this (a friend of mine bought some Koss Porta Pro and told me they had good low frequency response, and I find the opposite so I guess we have different models). If any of you have any suggestions about headphones/earphones that would be:
it would be really appreciated ! |
|
responding your older post: about the headphone issue: |
|
I also found out the head phone entusiast are so used to the in-head sound that they wont hear the stage at all. |
here you are: @jeanibarz |
|
I consider them beta-version by the way. I am working on my fisical room atm. but I will get back to them |
|
Hi, Ok so @capoei I try to make a brief recap: to test your settings in the same condition as you, I need to buy the Superlux HD-681 (red) + the ToneBoosters Morphit VST plugin, and that's it ? I thought I achieved good results too in the past, but the placebo effect, results expectancy, and memory biased my evaluations. For example, when I tried some settings in another room as I used to, the spatialization was (suddenly) lost ! I was thinking it was due to the hardware, I checked everything, and it seems that it was just my brain.... Now I try more to validate my results with other people. Did you checked the Harman listening test software ? I would be interested to know which difficulty level you are able to achieve, just for the sake to be able to compare my ability to analyze sound with other people which may be much more experimented or gifted than I. Thank you for sharing your impulse responses, will try probably next months if you can confirm the hardware/software configuration needed to reproduce your listening conditions. |
|
To discriminate the reverberation, I recommand this passage: https://youtu.be/KkZM6Cbi-L8?t=78 To discriminate for spectral coloration, I recommand sounds like pink noise but pink noise you have been calibrated for in real life, such as claps that are very well recorded here : https://www.youtube.com/watch?v=hnXD6FRZtn0&list=PLlcZV4J86GBb30QXR1UVp-cEwJJfR88Yg&index=56&t=0s It is nearly impossible for me to discriminate small reverberation nor spectral coloration in the type of music like the one you used for your demonstration because it is electronic and my brain cannot know how it "should sound". |
|
@jeanibarz |
|
2 files I like to use for stereo image and stereo clarity: but what most shows you what is going on is hearing well produced music that you know well for decades. I like to use Thriller album. Billie Jean shows LF ringin well, especialy in the bridge. Baby be Mine has a nice stereo field |
|
To use your files correctly, can I just substitute left and right files for yours in a 2.0_Config_XX_XXX.txt config file from BRIR_Convolution ? (I use EqualizerAPO) Like this : I have the impression that the soundstage is a lot different with your files than with the original Control_Room_7 (original seems more "in front of me" while yours is more "left, into and right" of my head without being "in my head" either). |
|
Hi,
first question: sure
about the sound stage: I used 40 degrees angles (iirc) instead of the standard 30 degrees. compare to the 40 degrees originals
…On Jan 7 2020, at 8:13 am, Mpic ***@***.***> wrote:
To use your files correctly, can I just substitute left and right files for yours in a 2.0_Config_XX_XXX.txt config file from BRIR_Convolution ? (I use EqualizerAPO) Like this :
I have the impression that the soundstage is a lot different with your files than with the original Control_Room_7 (original seems more "in front of me" while yours is more "left, into and right" of my head without being "in my head" either).
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub ***@***.***/0?redirect=https%3A%2F%2Fgithub.com%2FShanonPearce%2FASH-IR-Dataset%2Fissues%2F4%3Femail_source%3Dnotifications%26email_token%3DAFLJROGZN5VBOKGOMD2ECTDQ4RPV7A5CNFSM4IMDGDYKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEIIRDAQ%23issuecomment-571543938&recipient=cmVwbHkrQUZMSlJPQ1BBTk1HWkcyU1BFNTdXQVY0RUdORjdFVkJOSEhCWk1OREE0QHJlcGx5LmdpdGh1Yi5jb20%3D), or unsubscribe ***@***.***/1?redirect=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAFLJROHBFITFKOVH2VDLEG3Q4RPV7ANCNFSM4IMDGDYA&recipient=cmVwbHkrQUZMSlJPQ1BBTk1HWkcyU1BFNTdXQVY0RUdORjdFVkJOSEhCWk1OREE0QHJlcGx5LmdpdGh1Yi5jb20%3D).
|
|
Thanks a lot !
|
|
what you think? |
I use headphones, so I use a correction IR just after the room IR. I still have to make subjective measures to know if my headphones correction is correct... That being said, in comparison I like your version more for its precision in the bass spectrum, but I still have the impression of less depth than the original Control Room 7 even at 40°, ruining a bit the virtual room immersion effect for me. I don't know where that could come from, maybe the lessening of reverb ?
Oh, so you don't use headphones ? Does virtual room IR work without a binaural device ? |
|
I loved you new IRs. you got all of the bad bass decay out (how?). |
|
The recent changes I have made to the BRIRs have been aimed at minimising excess group delays in the low frequencies. The improvements are partially due to the generalised low frequency BRIRs that have been integrated into the BRIRs. The low frequency BRIRs were created from sets of subwoofer room impulse responses and BRIRs which were combined through synchronous averaging methods. The resulting low frequency BRIRs have smoother frequency responses and reduced group delays compared to the original room response. For each room, an appropriate crossover frequency and delay was selected such that the poorer response of the original room was replaced by the low frequency BRIR without introducing any artefacts around the crossover frequency, which ranges from 100Hz to 200Hz depending on the room. For frequencies above that, spectrum analysis and listening tests (tones and sine sweeps) were performed to find any remaining frequencies with excess group delays and phase distortions, which were equalised out. The result should have less distracting bass decay as you noted. |
|
If anyone is interested in removing the room target curve from the BRIRs or applying your own preferred room target curve, I have uploaded a filter containing the room target curve that has been applied to the BRIRs in my dataset as well as an inverse filter that can be used to remove the response. https://drive.google.com/drive/folders/13h8nwxS_hQJoPa18YxuiVbEmg869Ze2r?usp=sharing |
|
well, in the link of my modifications I allready aplied my curve. Do you think applying you inverse and then correcting will make a diference? however, talking about targets: imho the sum of the IRs should be flat, because most headphones are allready targeted to Harman curve. and the Harman curve allready accounts for ear resonances. I know that your idea is to use the difuse field correction for headphones with the intention to kind of remove Harman from the phones, but I think it makes things too complicated. the most advanced headphone correction database can make virtualy every headphone a perfect Harman phone https://github.com/jaakkopasanen/AutoEq. about group delay: check out this topic. I made your Room 28 (very nice virtual reflective room btw) flat (in sum). One version only targeted the frequency response, the other version is (excess) group delay / step response corrected: https://www.audiosciencereview.com/forum/index.php?threads/example-binaural-files-of-a-normal-room-response-and-b-time-aligned-exess-group-delay-corrected.14489/#post-447147 |
|
I would recommend applying the inverse filter if you are equalising your headphones to the Harman curve, as both the room target curve that was applied to the BRIRs and the Harman target curve result in a similar bass boost. This is the only issue of compatibility that I can see. I checked out your modifications of room 28 and it sounds like you achieved better results in the 'y' version of the BRIRs by targeting the group delay / step response. |
|
Hello there, I'm sorry if this is a little bit "out of subject", but I wanted to share some learned things. Since last time, I've bought planar magnetic headphones: the Hifiman HE400i for only 180€ during a big discount. It seems that planar headphones have much less distortion in low frequencies, compare to any dynamic drivers I had tried in the past. I have even bought another planar magnetic headphones, the Gold Planar GL600, to confirm that. Binaural reproduction works also much better with these headphones. So, I've decided to try the model of headphones proposed @capoei in an old message, the Superlux HD-681 (red). I listened to them for 10 secondes before being convinced that the bass suffer from a lot of distorsion, and the headphones sounds awfully bad to my hears. Please find a picture of the distorsion measured with 2€ microphones (KEEG1538WB-100LB), with the blocked ear meatus measurement method, of the Hifiman HE400i and the Superlux HD-681 that I just acquired. Distorsion of the Hifiman HE400i (2014): Distorsion of the Superlux HD-681: The distortion for Hifiman HE400i is probably overestimated by a 50hz noise and probably other things, but the distorsion of the Superlux HD-681 is clearly not coming from the measurement chain, but from the headphones themselves. Also, the distorsion was measured at a "moderate" sound pressure level, but the Hifiman HE400i can handle a really high enough SPL compared to most of the headphones I've heard, without noticeable distorsion. For example, this is my current settings with a 7.1 configuration with individualized HRIR and subjective adjustments: Today I think that dynamic drivers are not able to reproduce the entirety of the audible spectrum without distortion (most of the time the lower part, but also the upper part), and that only planar magnetic (and probably electrastic too) headphones can do that correctly. Some planar headphones can be found for a relatively cheap price, for example the Gold Planar GL600. However, I think a model of about 300-500€ is a better investment. I personally hesitate to invest in an even more expensive one, one day maybe.... Have you ever tried planar technology ? If not, I really really recommand you do, once you've heard them, you will probably never come back to dynamic drivers ! |
WDR Control Room 7 has a terrible low frequency response. it's very muddy with loud extended reverbs. very, very diferent from what you would expect from a professional control room. I suppose that is due to the integrated subwoofer RIRs. Is there any way to get this subwoofer filter so that I can undo it's efect on the BRIR?
not sure if this aplies as an issue. It's just that the result is not a control room response.
I ran a channel of a 30 degrees filter through REW to show how bad the low end is:
https://drive.google.com/open?id=1cFESB_NmGR9EV0cHYUCh13z8WleyPZNu
might be an issue on other rooms, too. I'm focused on this because I love the (rest of the) overall response of it
The text was updated successfully, but these errors were encountered: