1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice

To crossfeed or not to crossfeed? That is the question...

Discussion in 'Sound Science' started by jasonb, Oct 21, 2010.
78 79 80 81 82 83 84 85 86 87
89 90 91 92 93 94 95 96 97 98
  1. castleofargh Contributor
    that's a hard one. how to get accurate data for our body using tools that introduce changes we don't want? I don't know how far along the research is on those videos or just pictures of the head/ears and some software tries to build the correct acoustic model. last AES stuff I read on this are a few years old already, and the simplified models seemed to consistently get too bright for the listener. Super X-FI is the obvious actual product that comes to mind, as they use pictures of the face and ears. but I didn't try and don't know much about it(knowing that the Realiser A16 would one day come to me, I've stopped purchasing all the little gizmos related to "3D" audio). I'm afraid it's creating a multichannel "surround" sound no matter what we feed it, I hope I'm wrong and it at least gives a choice.

    I guess if you could measure the signal really in your ear canal and then at a very similar position do just a speaker measurement(without you). the variation in FR could be fairly accurate and you could decide to only use that variation for your simulation. to remove most reverb, if your room isn't too small and you set it up not to have desk reflections right away or something, I guess you could just shorten the impulse and remove the trail of reverb(some reverb plugins let you play with that, but IDK if that's something you find in typical convolvers? I've stuck to free stuff so again I'm not very knowledgeable on what exists. I kind of butchered my impulses "by hand" like a monster when I made my own sauce a few years ago.
    or you can go the crossfeed way and just use the FR variations(or whatever you're comfortable with as FR alteration) and the interaural delay you measure between your impulses. or just the delay that makes you feel like a left channel only is at about 30° when you turn it ON and OFF. that way you have effectively removed almost all components of the room.

    250ms should be typo. you typically would get something in the ballpark of 250µs (modulo head size and crap). if the app doesn't allow such fine tuning, you can consider just editing the .wav files by counting the number of samples or something like that. there is probably a way to export from REW without the impulse being centered on the peak, but I don't know how. I'm really a noob for everything beyond FR in that app.
  2. bigshot
    I bet a VR helmet with a good sized listening environment would help. I imagine some of the directionality in speaker systems come from visually seeing the sources of the sound. I find that classical music concerts filmed in a concert hall and projected on a big screen feel different than the same sound without the visuals.
  3. Davesrose
    There's a problem that everyone's anatomy is a bit different. In the medical community, there's never an absolute value....but a range of tolerance. Especially when it comes to perceptual sound: all of our outer ears, ear canals, and inner ears are different.
  4. ironmine
    rePhase software has a preset for manipulating the phase, which is called "Shuffler" (one preset for the left channel and the second one for the right channel):

    Is it of any use when crossfeeding the audio signal for headphone listening?

    What is this Phase Shuffler effect good for?
  5. ironmine
    Yes, it's a good idea (variation in frequency). I need to try it.

    Also, the Room EQ Wizard lets you adjust the width of the impulse. You can also choose to time-align it or not.

    Yes, you are right. I need to put the Voxengo in another mode (it's called "dimension"). It uses meters instead of ms:


    "Audio Delay This group of knobs specifies delay time in a selected dimensionality (milliseconds, meters or feet). Note that each knob affects a single decimal position of the whole delay time value.During calculation of delay time expressed in meters or feet, it is assumed that the speed of sound propagation equals 340.29 meters per second". (from the manual).

    It's funny that using simple and sometimes even free VST plugins one can assemble a crossfeed. Why then do people pay money and buy crossfeed VST plugins?
    Goodhertz Can Opener wants to charge $65 for its CanOpener.
    Last edited: Oct 23, 2019
  6. 71 dB
    It requires knowledge/insight to assemble a crossfeed. Paying $65 for a CanOpener you don't need to think yourself. Someone has though it out for you.
  7. 71 dB
    1. This kind of "shuffling" isn't part of default crossfeed philosophy, but it can of course shape headphone sound to a direction you prefer.

    2. For creating spatial effects such as creating pseudostereo from mono sound or "wide" spatiality that is mono-compatible.
    ironmine likes this.
  8. ironmine
    Chris from AirWindows updated his Monitoring plugin. The option "Cans C" in it has changed.
    I don't like how it sounds. Discordant! Even the previous "Cans C" was better.
    The new version sounds as if there are two layers of music and they live their separate lives creating a cacophony.
  9. ironmine
    I was wrong! I used yesterday a poorly mastered album for testing, it's the fault of the recording I used, the plugin now sounds excellent! I retested it again today on a higher quality audio material and using big headphones. It's great.
  10. 71 dB
    Yeah, it's good to try a crossfeed plugin with different kind of recordings.
  11. bigshot
    I rarely use the hall ambience DSPs on my AVR. They tend to mush up the sound. But with Toscanini recordings made in Studio 8H, it is a vast improvement. That recording venue was too small for a full orchestra, so the recordings sound boxy and dry. Adding an envelope of hall reverberation improves them tremendously. Not all recordings are perfect. It's nice to have a toolbox to fix the crummy ones.
  12. ironmine
    71 dB,

    Why do you advise using the minimum phase? What's the logic behind for not using the linear phase?

    What is the best way to do it? Should I boost high frequencies in the Mid channel and reduce them in the Side channel? Or how?
  13. 71 dB
    1. The delay is known when we know the minimum filter type and cut-off frequency. In this case 1st order butterworth filter creates a low frequency delays of 1/(2*pi*800) seconds = 0.000199 s = 199 µs and delay drops to about 140 µs at 800 Hz. Together with the trembe boost phase shift this creates a ITD of about 250 µs which is very suitable for mimicking a typical stereo speaker angle. Minimum phase means minimum possible delay. If minimum is optimal, then everything else is too much. Linear phase needs FIR-type filters with a certain amount of filter taps and the delay is constant on all frequencies (phase is linear). That's is a very good thing itself, but the cost is increased phase shift/delay (delay is half of the taps) and also increased computational needs. Latency is a problem with linear phase filters. At sample rate 44100 Hz the needed delay (200 µs) is about 9 samples, so a linear phase filter of 17 taps would give a proper delay.

    2. The boost should be in "direct sound". So, the original left channel signal goes treble boosted to the left ear and low pass filtered/attenuated to the right ear (and vice versa). That's roughly what happens in reality. Sound on left get a treble boost on the left ear, because the head reflects high frequencies boosting them (the acostic energy that doesn't leak to the right ears remains on the left boosting the sound pressure level a few decibels). So, after crossfeed:

    - Mid channel: bass energy has increased, treble energy has decreased
    - Side channel: bass energy has decreased, treble energy has increaced.

    In other words crossfeed is kind of re-distributing energies of Mid and Side channels.
    ironmine likes this.
  14. ironmine
    I already implemented this treble boost in the direct sound after you wrote: "In your diagram you are missing the treble boost block between input and output in the lower branch. Just put another fabfilter plugin there (highself filter, 800 Hz, Q=1)."

    But then I asked you how to improve it even more, and you said "but you may want to reduce the channel separation little more at high frequencies". Reducing the channel separation at HF means equalizing HF down in the side channel and/or equalizing them up in the mid channel, right? Or am I misunderstanding you?
  15. 71 dB
    I'm sorry!

    If you want to reduce channel separation for treble do this: Start everything by adding an attenuated and channel swapped version of the original signal to the signal. This makes the original signal just a little bit more mono and it doesn't do much harm at low frequencies as the crossfeed reduces separation much more. Try attenuation levels -20 to -15 dB (gain 0.10 - 0.18).
78 79 80 81 82 83 84 85 86 87
89 90 91 92 93 94 95 96 97 98

Share This Page