markus76 said:
That issue has been of considerable interest for a number of years and the answer is not clear because "it depends ...". Perhaps the better references would include: Elizabeth Wenzel et al (in JASA) and Simon Carlile et al (in JASA). There is also a good deal of anecdotal work but that is more often hallway conversation rather than published work
WithTarragon, then tell me on what does it depend? Should excluding the ear canal help in extracting one of those "generally working" HRTFs? They didn't work in the AKG BAP-1000 and obviousely they don't work in one of the newest incarnation, the Beyerdynamic Headzone. There are HRTFs that work for a couple of people (nobody knows for how many) but there are HRTFs that don't work. As long as this is not quantified we're all talking about hot air.
Best, Markus
Best, Markus
His dummy head is the hard surface type, I don't agree that this type of dummy head is appropriate. Further more, the mic should not be in the ear rather it should be at the mouth of the ear because if you record in the ear, and then listen using speakers, you are getting the double ear effects.markus76 said:
markus76 said:
It heavily depends on "expectation".
Where have I heard that before?
Soongsc - the surface impedance issue has been studied and for most of the surface its a minimal effect. In and around the ear canal there is some effect of the surface characteristics, but its a secondary effect not a primary one. And many of the head systems do use a softer material for the pinae and canal.
Indeed HRTF's are something highly individual. Still some distinct "typical" HRTF categories seem to be possible, there is an university paper which discussed that:
EXTRACTING AND MODELING APPROXIMATED PINNA-RELATED TRANSFER FUNCTIONS FROM HRTF DATA
I have tried various HRTF-based tricks and gimmicks, including some simple algorithms of my own (mostly based on the rather old KEMAR HRTF data), and they usually work fine -- alas, all of them rely on the standard stereo setup and are not compatible to a 2-->3-Channel rematrixing known as "trinaural", which is my favorite reproduction (or shall I say, projection) format for two-channel stereo source material.
- Klaus (and ahem, we are getting way OT here)
EXTRACTING AND MODELING APPROXIMATED PINNA-RELATED TRANSFER FUNCTIONS FROM HRTF DATA
I have tried various HRTF-based tricks and gimmicks, including some simple algorithms of my own (mostly based on the rather old KEMAR HRTF data), and they usually work fine -- alas, all of them rely on the standard stereo setup and are not compatible to a 2-->3-Channel rematrixing known as "trinaural", which is my favorite reproduction (or shall I say, projection) format for two-channel stereo source material.
- Klaus (and ahem, we are getting way OT here)
markus76 said:
Markus, It looks like you don't want to read the suggested papers to start with.
Well, your question needs to posed more precisely. First off, I think you are making too big of a deal out which headphone is being used (that is probably not a high priority issue). Second, you need to define what you mean by "works well". Third, you really don't expect me to summarize a large body of research to a general audience in 25 words or less do you? This why I use the phrase, "it depends..."
Let me steer you in the right direction. The problem with non-individualized HRTFs (assuming that they are correctly measured in the first place and correctly implented DSP-wise) is that there are front-back confusions and elevation confusions. The "degree of sidedness" is typically not a problem. This problem is further confounded, since the front-back confusion may not occur everytime, but rather it may be infrequent. This makes it difficult to quickly determine.
The second thing to remember is that in simulations that either
1) include motion (the source following a plausible trajectory) or
2) include plausible simulations of room acoustics (reverberation)or
3) allow head tracking (thus updating the HRTFs based on relative head position) then,
the issue of non-individualized HRTFs is less of a problem.
Please note that when you read this literature you must keep in mind how the authors are presenting the "error" data. In some cases the "front-back" reversals are being "corrected" since the study may have been focused on the precision of the "degree of sidedness" judgements.
Let's go on with something the humble reader really expects from this thread:
You need to drill a hole in one side of the shorter inner braces. It's for the screw that connects the brace to the back board. Try to drill the hole as perpendicular as possible. Then mount the brace to the back board and apply some wood glue to the open end of the brace:
Lay the enclosures on the floor front side facing down and put the back board in place by screwing in a couple of srews so everything is centered. Don't move the enclosure to ensure that the wood glue stays in place:
When the glue is hard enough glue in the horizontal brace:
Best, Markus
An externally hosted image should be here but it was not working when we last tested it.
You need to drill a hole in one side of the shorter inner braces. It's for the screw that connects the brace to the back board. Try to drill the hole as perpendicular as possible. Then mount the brace to the back board and apply some wood glue to the open end of the brace:
An externally hosted image should be here but it was not working when we last tested it.
Lay the enclosures on the floor front side facing down and put the back board in place by screwing in a couple of srews so everything is centered. Don't move the enclosure to ensure that the wood glue stays in place:
An externally hosted image should be here but it was not working when we last tested it.
When the glue is hard enough glue in the horizontal brace:
An externally hosted image should be here but it was not working when we last tested it.
Best, Markus
Although I am interested in helping out folks trying to understand HRTFs (and there is already too much speculation and not enough scholarship occurring), I am more interested in hearing how Markus is coming along on his build. I think the kit, the degree of difficulty in building, and the results are of a greater interest to folks reading this particular thread.
Markus
Clever idea on gluing the front-back brace. Thats not the way that I do it, but it probably works better. I just glue then in, and when dry measure the location and mark it on the back panel. That works too but does everythng in one step instead of two.
I did read the post about wood glue being the stronger glue and I don't doubt it. But I like the better damping of the polyurethane glues as they don't tend to cure very hard like wood glue.
Clever idea on gluing the front-back brace. Thats not the way that I do it, but it probably works better. I just glue then in, and when dry measure the location and mark it on the back panel. That works too but does everythng in one step instead of two.
I did read the post about wood glue being the stronger glue and I don't doubt it. But I like the better damping of the polyurethane glues as they don't tend to cure very hard like wood glue.
WithTarragon, one last thing: obviously you did not read post #654. I do own a Headzone and it does all stuff necessary (real time head tracking, room simulation, etc.) but IT DOES NOT WORK FOR ME. I don't experience an out-off-head localization. So individual HRTFs are a crucial part of such systems. Otherwise it won't work.
Best, Markus
Best, Markus
markus76 said:
In spite of the tone of your question, I will reply since it is an otherwise reasonable question.
The phenomenon (out-off-head localization) that you are mentioning is referred to (in the literature) as "externalization". It raises two points.
First, the awful truth is that externilization with current setups, even when you use your own HRTFs, is typically not that good. The degree of externalization can be rather modest. In your case it sounds like it is negligible.
Second, externalization can not be objectively measured since it is a subjective percept. There have been some indirect measures (inferences really) in contrasting HRTF presentation to "lateralization" (conventional manipulations of "simple" combinations of interaural delay and level).
This is in contrast to some of the other measures of localization precision (using HRTFs) which can be objectively measured.
I bring up this last point, since some will swear that they are getting "great externalization". Usually, accurate room reverberation (not simulation) and motion are required for "better externalization" (my subjective experience). But what one person may think is "great" another may think is "modest"
WithTarragon, I don't understand what you're trying to say. The topic of binaural recordings came up because soongsc proposed it as a system superior to stereophony (or multichannel). But a headphone based VR system has to provide externalization in each and every case. There's no such a system in the mass market today so all discussion about replacing stereophony with it is academic.
Best, Markus
Best, Markus
markus76 said:
You were asking about HRTFs, externalization, etc and mentioning some of the research on HRTFs etc. I am a scientist and this is one of my topic areas. I thought you were asking about understanding the research findings.
Perhaps we are talking at cross-purposes.
In either case, I am interested in how you are coming along with your own project (at one time it was the topic of this thread).
Please keep us up to date.
gedlee said:
Hi Earl:
IME most good woodglues don't glue as hard as you're implying. Do you have a specific glue you're thinking of? Titebond 2 is my favorite (better than 3), and dries to something very similar to hardened gum- thick and strong, but can bend under force.
The nice thing about woodglue is also a nice thing about shellac- they both soak deep into the wood. I've taken to shellacking the inside of cabinets to seal them.
And I hope you don't get this post until your vacation's over, stop checking in, markus will be here when you get back
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.