Post by elysion on Jan 16, 2015 3:07:24 GMT
I've had the idea long ago, but I was already absent from DIY audio, when I had the idea:
Before my career did go slightly towards IT, my profession was Polygraf (a prepress specialist). In the prepress, it is common to calibrate displays and output devices. Usually, colorimeters are used to calibrate displays, while spectrophotometers are used for measuring the output of output devices like printers, printing machines, proofers etc.
The idea behind calibration is simple: The characteristics of the devices are measured. Then you know what the device is capable of (more exactly how big its "gamut" is, which could be explained roughly as "colour range") and you can generate a correction curve (ICC profile). It's a lot of theory behind that and I'll focus on displays to explain the rest. If you have measured the characteristics of a display, you generate an ICC profile for it. The correction curve which is coming from this ICC profile is loaded in the LUT (lookup table) of the graphics card of the computer. The correction curve will compensate the flaws of the display (at least partially, depending on the capabilities of the display and the used output colour space, for example ISO coated V2). Calibration has two goals here: First goal is that the display is showing a particular colour exactly how it should be and the second goal is that the colour is not only displayed correct on the display but also on the output device. There's also a third goal: Colours should (ideally) look also identical on other displays and output devices (OK, a display is also an output device).
In other words: Calibrating can correct the flaws of a specific display (or output device) and it makes it possible to see the same colours on different displays (or output devices).
BTW: I have already at least as much colorimeters collected at home as I have headphones. I also own an older spectrophotometer. (My favourite colorimeter is still the rather old X-Rite DTP94B, which still gives excellent results.) I'm using the open-source colour-management system ArgyllCMS together with a GUI named dispcalGUI on Linux to do that. ArgyllCMS and dispcalGUI are also available for Windows and OSX. ArgyllCMS supports a very broad range of instruments, much more than ANY available commercial colour-management software.
Now, we turn our minds towards audio (more specifically headphones) and guess what? AFAIK, no one has tried something similar for audio purposes, at least not in the quality it is used in the prepress for matching colours. IMVHO, this sounds interesting, doesn't it?
I know only about some measurements executed by sound engineers to tune for example PA systems. But AFAIK, it's nowhere near a real calibration.
Of course, there are more than just a few problems to solve:
- We'd need the right measuring equipment, which isn't available at all (is it?). Probably, we'd also need different instruments for different purposes/applications. Not sure here... and how should such measuring instruments look like?
- We'd need to develop some theoretical stuff. A kind of "profile" standard (like ICC profile) or at least as standard for a kind of correction curves. We'd also need a source which is capable of loading that "correction" at all. Most likely, the source would have to be a computer then, but there are NO audio architectures on computers so far which would allow to load a "correction" (for which no standards are existing so far). It would be also very nice to measure the characteristics of a particular human ear (each side separately, of course - are you sure both of your ears are exactly the same) and to put that also in the equation.
- We'd need to develop standards for audio which would correlate to the colour spaces used in prepress. How should we call that? "Audio space" or "audio range"? There's certainly much more theory behind that. It's very likely that I'm missing a lot of important things in this first post.
- How should the measurements be executed? What do we need to take care of?
- Could dummy heads with built in microphones be a part of the exercise? The microphones would have to function is a manner like a colorimeter or spectrophotometer. The results have to be correct. Otherwise it makes absolutely no sense. How do we measure a human ear? Is it like when entering military service and you have to answer the questions from the doctor ("Do you hear the beep in the headphones? Do you hear it also now on the other side? We go a step higher, do you hear it still?"). IIRC, they used a special kind of Beyer 'phones for that. Those 'phones have been extremely uncomfortable and the result of all that was called an audiogram.
There are some other important questions: What would a calibration bring us in the end? Do we need it? Do we like it at all? I guess the answer is not just "yes" or "no" here. It depends largely on what we are trying to achieve with it. I think this is at least one important difference to prepress work. I assume that we don't want that each particular pair of 'phones (which are "output devices") will sound the same. Depending on the capability of the particular "device", not all flaws could be compensated as well (or at all). For example: When the "hardware" of a K702 is not capable to go below a specific frequency, then you can't compensate that. But it would be possible to make sure that a particular frequency in the music would be exactly the same frequency on the "output device", as long as this frequency is within the frequency range ("gamut") of the 'phones. In the prepress, you'll have also different standards (for example different whitepoints etc.). Accordingly, we could define something like that also for audio. It's not exactly the same, but the best term for describing that would be a kind of "equaliser setting". (The term is wrong, don't take it too literally please.) Certainly, we'd need also to have a look on the whole signal chain. A part of that chain, for example an amp, could improve/degrade the result. We like to change opamps and tubes in amps. Would we do that also (as much) if we had the possibility to calibrate the "output device". More questions than answers at the moment. But I'm sure more could be done as is possible today. Prepress work is the proof for that.
Too complicated to understand? Tell me!
At the moment, I have simply NO CLUE what will arise from this thread and how much of the ideas in this thread will be possible to implement within our resources. But I'm already very curious...
One last thing (NO, I'm NOT Steve Jobs!): Displays and other output devices are ageing. With several calibrations over the time, it would be possible to document this ageing with displays, but it is also depending at least somewhat on the ambient temperature (and other things?). If we could do all the stuff above, there would be FINALLY an answer if "burning-in" headphones has an effect! Other questions could be also answered from a scientific point of view. At least it the accuracy of the instruments is good enough and the measurements could be executed in a repeatable manner.
Before my career did go slightly towards IT, my profession was Polygraf (a prepress specialist). In the prepress, it is common to calibrate displays and output devices. Usually, colorimeters are used to calibrate displays, while spectrophotometers are used for measuring the output of output devices like printers, printing machines, proofers etc.
The idea behind calibration is simple: The characteristics of the devices are measured. Then you know what the device is capable of (more exactly how big its "gamut" is, which could be explained roughly as "colour range") and you can generate a correction curve (ICC profile). It's a lot of theory behind that and I'll focus on displays to explain the rest. If you have measured the characteristics of a display, you generate an ICC profile for it. The correction curve which is coming from this ICC profile is loaded in the LUT (lookup table) of the graphics card of the computer. The correction curve will compensate the flaws of the display (at least partially, depending on the capabilities of the display and the used output colour space, for example ISO coated V2). Calibration has two goals here: First goal is that the display is showing a particular colour exactly how it should be and the second goal is that the colour is not only displayed correct on the display but also on the output device. There's also a third goal: Colours should (ideally) look also identical on other displays and output devices (OK, a display is also an output device).
In other words: Calibrating can correct the flaws of a specific display (or output device) and it makes it possible to see the same colours on different displays (or output devices).
BTW: I have already at least as much colorimeters collected at home as I have headphones. I also own an older spectrophotometer. (My favourite colorimeter is still the rather old X-Rite DTP94B, which still gives excellent results.) I'm using the open-source colour-management system ArgyllCMS together with a GUI named dispcalGUI on Linux to do that. ArgyllCMS and dispcalGUI are also available for Windows and OSX. ArgyllCMS supports a very broad range of instruments, much more than ANY available commercial colour-management software.
Now, we turn our minds towards audio (more specifically headphones) and guess what? AFAIK, no one has tried something similar for audio purposes, at least not in the quality it is used in the prepress for matching colours. IMVHO, this sounds interesting, doesn't it?
I know only about some measurements executed by sound engineers to tune for example PA systems. But AFAIK, it's nowhere near a real calibration.
Of course, there are more than just a few problems to solve:
- We'd need the right measuring equipment, which isn't available at all (is it?). Probably, we'd also need different instruments for different purposes/applications. Not sure here... and how should such measuring instruments look like?
- We'd need to develop some theoretical stuff. A kind of "profile" standard (like ICC profile) or at least as standard for a kind of correction curves. We'd also need a source which is capable of loading that "correction" at all. Most likely, the source would have to be a computer then, but there are NO audio architectures on computers so far which would allow to load a "correction" (for which no standards are existing so far). It would be also very nice to measure the characteristics of a particular human ear (each side separately, of course - are you sure both of your ears are exactly the same) and to put that also in the equation.
- We'd need to develop standards for audio which would correlate to the colour spaces used in prepress. How should we call that? "Audio space" or "audio range"? There's certainly much more theory behind that. It's very likely that I'm missing a lot of important things in this first post.
- How should the measurements be executed? What do we need to take care of?
- Could dummy heads with built in microphones be a part of the exercise? The microphones would have to function is a manner like a colorimeter or spectrophotometer. The results have to be correct. Otherwise it makes absolutely no sense. How do we measure a human ear? Is it like when entering military service and you have to answer the questions from the doctor ("Do you hear the beep in the headphones? Do you hear it also now on the other side? We go a step higher, do you hear it still?"). IIRC, they used a special kind of Beyer 'phones for that. Those 'phones have been extremely uncomfortable and the result of all that was called an audiogram.
There are some other important questions: What would a calibration bring us in the end? Do we need it? Do we like it at all? I guess the answer is not just "yes" or "no" here. It depends largely on what we are trying to achieve with it. I think this is at least one important difference to prepress work. I assume that we don't want that each particular pair of 'phones (which are "output devices") will sound the same. Depending on the capability of the particular "device", not all flaws could be compensated as well (or at all). For example: When the "hardware" of a K702 is not capable to go below a specific frequency, then you can't compensate that. But it would be possible to make sure that a particular frequency in the music would be exactly the same frequency on the "output device", as long as this frequency is within the frequency range ("gamut") of the 'phones. In the prepress, you'll have also different standards (for example different whitepoints etc.). Accordingly, we could define something like that also for audio. It's not exactly the same, but the best term for describing that would be a kind of "equaliser setting". (The term is wrong, don't take it too literally please.) Certainly, we'd need also to have a look on the whole signal chain. A part of that chain, for example an amp, could improve/degrade the result. We like to change opamps and tubes in amps. Would we do that also (as much) if we had the possibility to calibrate the "output device". More questions than answers at the moment. But I'm sure more could be done as is possible today. Prepress work is the proof for that.
Too complicated to understand? Tell me!
At the moment, I have simply NO CLUE what will arise from this thread and how much of the ideas in this thread will be possible to implement within our resources. But I'm already very curious...
One last thing (NO, I'm NOT Steve Jobs!): Displays and other output devices are ageing. With several calibrations over the time, it would be possible to document this ageing with displays, but it is also depending at least somewhat on the ambient temperature (and other things?). If we could do all the stuff above, there would be FINALLY an answer if "burning-in" headphones has an effect! Other questions could be also answered from a scientific point of view. At least it the accuracy of the instruments is good enough and the measurements could be executed in a repeatable manner.