Hello,
I just wanted to share my two cents about binaural recordings.
I have been experimenting with binaural over the past few years and to me replicas of both the ear as well as the ear canal are necessary to achieve more realistic results.
Now, while with some materials some people claim to achieve good result with microphone capsules located at the pinna with little to no equalization, with the ear canal the equalization is absolutely necessary.
This might seem like an unnecessary extra step at first, but sound localization is derived from the difference between left and right eardrums pressure waves.
A big part of that difference is due to the ear canal, and recording with a ear+canal replica ensures that that difference is encoded in the microphone signal.
If one could separate the part of HRTF function due to the pinna from the one of the canal, and the total effect were the simple addition of these two 'filter' responses, then I can see how the ear canal could be set aside.
Unfortunately, ear+canal is a complex acoustic system that cannot be broken down as a simple A+B type equation.
I made an orchestra recording some time ago with a pair of pinnae replicas and the capsules at the end of approximate ear canals made simply out of straight pieces of plastic tubing.
That recordings was recently mentioned in an audiophile article for its quality (I'm a little proud of this I might say, as it is not easy to satisfy those people
).
Here is the link to the article. Flying Blind: Another Visit to the Magnepan Skunkworks - The Audio Beat - www.TheAudioBeat.com
Encouraged by the results achievable with this technique, in my perennial quest for the perfect natural reproduced sound, I went as far as modeling a 3D printable ear+canal, the only one I am aware of that has an anatomically accurate ear canal.
It took me a ridiculous amount of time but I have to say that the results are much better than with the simple straight plastic tube.
The ear+canal models are available for sale on Shapeways.
I'm not sure if it's ok to post the link to them here, so if anybody is interested please message me privately.
I just wanted to share my two cents about binaural recordings.
I have been experimenting with binaural over the past few years and to me replicas of both the ear as well as the ear canal are necessary to achieve more realistic results.
Now, while with some materials some people claim to achieve good result with microphone capsules located at the pinna with little to no equalization, with the ear canal the equalization is absolutely necessary.
This might seem like an unnecessary extra step at first, but sound localization is derived from the difference between left and right eardrums pressure waves.
A big part of that difference is due to the ear canal, and recording with a ear+canal replica ensures that that difference is encoded in the microphone signal.
If one could separate the part of HRTF function due to the pinna from the one of the canal, and the total effect were the simple addition of these two 'filter' responses, then I can see how the ear canal could be set aside.
Unfortunately, ear+canal is a complex acoustic system that cannot be broken down as a simple A+B type equation.
I made an orchestra recording some time ago with a pair of pinnae replicas and the capsules at the end of approximate ear canals made simply out of straight pieces of plastic tubing.
That recordings was recently mentioned in an audiophile article for its quality (I'm a little proud of this I might say, as it is not easy to satisfy those people
).Here is the link to the article. Flying Blind: Another Visit to the Magnepan Skunkworks - The Audio Beat - www.TheAudioBeat.com
Encouraged by the results achievable with this technique, in my perennial quest for the perfect natural reproduced sound, I went as far as modeling a 3D printable ear+canal, the only one I am aware of that has an anatomically accurate ear canal.
It took me a ridiculous amount of time but I have to say that the results are much better than with the simple straight plastic tube.
The ear+canal models are available for sale on Shapeways.
I'm not sure if it's ok to post the link to them here, so if anybody is interested please message me privately.
I want to also add a consideration about S/N ratio.
The canal further hypes the frequencies that our ears are most sensitive to.
The equalization reverses theses peak effects though correspondent nulls, and in doing so it also lowers the microphone noise floor signal at those frequencies by the same amount.
It is not out of the ordinary to increase the S/N ratio by 15 dB by using the ear+canal replicas. This comes from the fact that the peaks of the replica response are easily in the order of 15-20 dB
As a side note, I use omnidirectional capsules for their more accurate frequency response and lack of proximity effect.
The ubiquitous Panasonic WM-61 does the job very nicely, even though I ended up using a different capsule as my final choice. This is because the above mentioned peaks can drive the WM-61 capsules above their SPL handling. This can happen with material that is not even that loud, like it has happened to me with a piano recording, for example.
The canal further hypes the frequencies that our ears are most sensitive to.
The equalization reverses theses peak effects though correspondent nulls, and in doing so it also lowers the microphone noise floor signal at those frequencies by the same amount.
It is not out of the ordinary to increase the S/N ratio by 15 dB by using the ear+canal replicas. This comes from the fact that the peaks of the replica response are easily in the order of 15-20 dB
As a side note, I use omnidirectional capsules for their more accurate frequency response and lack of proximity effect.
The ubiquitous Panasonic WM-61 does the job very nicely, even though I ended up using a different capsule as my final choice. This is because the above mentioned peaks can drive the WM-61 capsules above their SPL handling. This can happen with material that is not even that loud, like it has happened to me with a piano recording, for example.
@ Epicyclic
I'm glad you like the models.
Headphones would work just as good without specific equalization. In the case of binaural recordings there is no speaker's cross talk effect, which is why the effect is more realistic.
However, I don't see any issue with going the extra mile and equalizing the headphones to tailor them to your own canals.
It is a field that I haven't experience with though, and I don't have any prediction as to the cost/benefit or the potential signal degradation of this process though.
@Chris
Yes, the recording is played through speakers.
The canal is important because it is part of the ear+canal system.
I'd like to expand on the ear+canal system not being able to be described as ear response + canal response.
Let's take the pinna response, for example.
This is measured by putting a microphone capsule where the ear canal opening is.
This changes the boundary conditions (it force the wave speed to be 0), effectively changing the pressure of the sound wave in respect to the instance where the capsule is not there.
There is also the effect of the microphone body, which is likely and usually comparable to the ear dimensions, adding to the whole wave form change.
In other words, the measuring is affecting the measured entity.
Same thing for the canal. Measuring the canal by putting and earbud at the canal opening and a microphone at the place of the eardrum changes the canal from a open-close ends tube to a closed-closed ends one.
This changes the characteristics of the standing waves inside.
In general, both headphones and earbuds are pre-equalized to take into account their intended location in respect to the eardrum.
This is not to say that specific equalization for separate individuals can't make them sound even better (for those individuals separately), but as I was saying I don't have much knowledge of this process and the potential benefit/degradation it may add to the signal.
I'm glad you like the models.
Headphones would work just as good without specific equalization. In the case of binaural recordings there is no speaker's cross talk effect, which is why the effect is more realistic.
However, I don't see any issue with going the extra mile and equalizing the headphones to tailor them to your own canals.
It is a field that I haven't experience with though, and I don't have any prediction as to the cost/benefit or the potential signal degradation of this process though.
@Chris
Yes, the recording is played through speakers.
The canal is important because it is part of the ear+canal system.
I'd like to expand on the ear+canal system not being able to be described as ear response + canal response.
Let's take the pinna response, for example.
This is measured by putting a microphone capsule where the ear canal opening is.
This changes the boundary conditions (it force the wave speed to be 0), effectively changing the pressure of the sound wave in respect to the instance where the capsule is not there.
There is also the effect of the microphone body, which is likely and usually comparable to the ear dimensions, adding to the whole wave form change.
In other words, the measuring is affecting the measured entity.
Same thing for the canal. Measuring the canal by putting and earbud at the canal opening and a microphone at the place of the eardrum changes the canal from a open-close ends tube to a closed-closed ends one.
This changes the characteristics of the standing waves inside.
In general, both headphones and earbuds are pre-equalized to take into account their intended location in respect to the eardrum.
This is not to say that specific equalization for separate individuals can't make them sound even better (for those individuals separately), but as I was saying I don't have much knowledge of this process and the potential benefit/degradation it may add to the signal.
No problem, Chris.
This is my passion and I'm glad to share what I know with others.
Please get back and share your results. I'm curious to see if this takes you closer to where you want to be. With simple ear replicas + tubes one can definitely achieve very good results, but again, equalization is pretty much mandatory in this case.
This is my passion and I'm glad to share what I know with others.
Please get back and share your results. I'm curious to see if this takes you closer to where you want to be. With simple ear replicas + tubes one can definitely achieve very good results, but again, equalization is pretty much mandatory in this case.
I'm finding EQ pretty much mandatory in any case. Figuring out what to measure and EQ is difficult, though. Taking an average of many angles for each ear didn't work so well, though it's been a little while since I played around with this. With Christmas, NYE etc out of the way, I'm hoping to continue making progress with this.
Chris
I'm with you. Equalization is necessary in any case to me too. Just wanted to point out that some people do seem to be happy without it, with certain types of material and when there is no canal involved.
As far as the equalization, I use only one measurement from speakers at about +/- 45 deg.
The function of the equalization is to undo the second ear+canal pass that happens when one listens through speakers, so measuring the response based on the source being at the standard listening position makes sense to me.
The 45 deg instead of the more standard 30 deg comes from the fact that at that angle the sensitivity of the corresponding ear is at its maximum, while the sensitivity of the other is at its minimum.
This is to try to lower speakers' cross talk as much as possible.
Headphones and earbuds are also equalized to make music sound consistent in respect to when it is played through speakers at standard position, so the single measurement works with them as well, in my experience.
As far as the equalization, I use only one measurement from speakers at about +/- 45 deg.
The function of the equalization is to undo the second ear+canal pass that happens when one listens through speakers, so measuring the response based on the source being at the standard listening position makes sense to me.
The 45 deg instead of the more standard 30 deg comes from the fact that at that angle the sensitivity of the corresponding ear is at its maximum, while the sensitivity of the other is at its minimum.
This is to try to lower speakers' cross talk as much as possible.
Headphones and earbuds are also equalized to make music sound consistent in respect to when it is played through speakers at standard position, so the single measurement works with them as well, in my experience.
Thanks for your insight, I'll give that a try.
I come from a live sound setting, where everything gets (as a minimum) a highpass filter, and usually a whole lot more besides. The studio guys go another level again with EQ and processing, so I've never really understood the audiophile "no EQ in my system, ever!" approach, since there's already a spectacular amount of EQ done to the music.
For those following along, I'll use my little Bose speakers (small point source 80Hz-15kHz, rather useful), put the measurement mic up to derive a calibration file for the speakers, and then put the head up. EQ the head's response flat and off we go!
I suppose I could break out the ribbon tweeters and go all the way up past 20kHz, too. Hmm. Then put the head in front of a 12" sub and get a VLF measurement...
Things to do!
Chris
I come from a live sound setting, where everything gets (as a minimum) a highpass filter, and usually a whole lot more besides. The studio guys go another level again with EQ and processing, so I've never really understood the audiophile "no EQ in my system, ever!" approach, since there's already a spectacular amount of EQ done to the music.
For those following along, I'll use my little Bose speakers (small point source 80Hz-15kHz, rather useful), put the measurement mic up to derive a calibration file for the speakers, and then put the head up. EQ the head's response flat and off we go!
I suppose I could break out the ribbon tweeters and go all the way up past 20kHz, too. Hmm. Then put the head in front of a 12" sub and get a VLF measurement...
Things to do!
Chris
Single-driver systems have their place, it's just a power handling compromise. I've got a single cone per side that can do 50Hz-15kHz, though they tap out at 85dB in a medium sized living room. There's something to be said for the lack of crossovers (and phase anomalies) in the critical bandwidth, though I'd like them to go louder.
Edit - I've had a thought, and I'm going to take the ear canal a step further by making them slightly fluffy inside. Something like...
- apply glue to the inside
- stuff with cotton wool
- leave it to dry
- pull out the cotton wool core
- keep removing until the surface is just covered
I think that'll start to emulate the frequency-dependent behaviour that might occur in a human ear canal, where there are lots of hairs and occasionally earwax.
Chris
Edit - I've had a thought, and I'm going to take the ear canal a step further by making them slightly fluffy inside. Something like...
- apply glue to the inside
- stuff with cotton wool
- leave it to dry
- pull out the cotton wool core
- keep removing until the surface is just covered
I think that'll start to emulate the frequency-dependent behaviour that might occur in a human ear canal, where there are lots of hairs and occasionally earwax.
Chris
Last edited:
@ Chris
I mean the whole shape of the canal. Both the curve as well as the variation of the section
@ turk
The equalization I used comes form the process I described above (speakers at 45 deg and so on...).
I use match EQ in Logic pro to reverse the measured response and apply it directly to each channel.
After that, not much more is necessary to get a realistic sound.
If the material/room isn't too good to begin with though, one can play with EQ and compression (not too much though, in my opinion, before messing the sound stage irreparably)
I mean the whole shape of the canal. Both the curve as well as the variation of the section
@ turk
The equalization I used comes form the process I described above (speakers at 45 deg and so on...).
I use match EQ in Logic pro to reverse the measured response and apply it directly to each channel.
After that, not much more is necessary to get a realistic sound.
If the material/room isn't too good to begin with though, one can play with EQ and compression (not too much though, in my opinion, before messing the sound stage irreparably)
I've ordered some rubber tube and glue. I wonder if I could make odd-shaped holes in a series of bits of wood (like a series of tight-fitting washers) to alter the cross-section of the tube.
Interesting that you find the shape and curve to be important. I would have thought that, with dimensions of the order of millimetres, the sound pressure would propagate with little regard to whether or not the tube is curved along its length.
Similar to a ported subwoofer - it doesn't mind much if the port has a 90-degree elbow, so long as the length comes out correct.
Chris
Interesting that you find the shape and curve to be important. I would have thought that, with dimensions of the order of millimetres, the sound pressure would propagate with little regard to whether or not the tube is curved along its length.
Similar to a ported subwoofer - it doesn't mind much if the port has a 90-degree elbow, so long as the length comes out correct.
Chris
These are all experiments that can be done. I certainly would not encourage you not to try them. It looks like you have a few hours of fun planned out for in the next future
Keeping in mind that everyone's ear canal and pinna's geometry vary greatly from one another, the curve dimension is comparable to the length of the canal and it can be expected to play a non indifferent role in the final system response.
In my model I used a canal derived from several hundred canal impressions to average this great variety.
What is very important to me is a realistic replica of the transition between pinna and canal (canal opening). I always disliked how the plastic tube interfaces with the pinna, creating an edgy surface profile.
This surface irregularity is again comparable in dimensions to the canal, and that is why I decided to create my model.
However, I want to reiterate that it is possible to achieve great results with the plastic tube too.
Keeping in mind that everyone's ear canal and pinna's geometry vary greatly from one another, the curve dimension is comparable to the length of the canal and it can be expected to play a non indifferent role in the final system response.
In my model I used a canal derived from several hundred canal impressions to average this great variety.
What is very important to me is a realistic replica of the transition between pinna and canal (canal opening). I always disliked how the plastic tube interfaces with the pinna, creating an edgy surface profile.
This surface irregularity is again comparable in dimensions to the canal, and that is why I decided to create my model.
However, I want to reiterate that it is possible to achieve great results with the plastic tube too.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.