Light organ, human brightness versus loudness perception

[ccurtis]

In a light organ, the amplified signal from a microphone, or audio line output, is applied to a controller that subsequently drives the light intensity of a light source(s) for the purpose of adding a dimension of sight to the music listening experience. In most designs the goal is to have the light intensity rise is direct (linear) proportion to the amplified audio signal.

I have developed, built, improved such light organs of various configurations and designs over the past several years. However, while listening to music and observing the lights varying in brightness, accordingly, I was never truly satisfied with the effect. When I heard the peaks in the music and did not experience the peaks in light brightness in the same way I heard them. Sure, the light was brighter, but my sense of the increased brightness was not the same as my sense of the music. Additionally, there is too noticeable brightness fluctuation at times when the music is not so dynamic.

Some time ago, I attempted to address this (https://www.electro-tech-online.com...ut-driving-pwm-led-loudness-indicator.156838/). Still not satisfied, I recently realized I got the math wrong.

It is true that the way we perceive sound pressure is not the same as the way our eyes perceive light intensity and that is reason for the dissatisfaction. Stevens' Power Law tells us that the exponent for human sound perception has an exponent of 0.67. It also tell us that the exponent for human light perception is 0.33. Not the same!

So how do we get them to match? That's where the math problem comes in. Setting the two expressions equal gives us:

S^0.67 = L^0.33 (S is the sound pressure stimulus, or voltage from the microphone) (L is the light intensity)

Rearranging:

S^(0.67/0.33) = L or S^2 =L

Therefore, L corrected = L^2. In other words the signal driving the lamps must be squared! The goal is NOT to linearize the light intensity except only to provide a linear signal that is subsequently to be squared to give the desired result.

Now, finally, the eye sees what the ear hears.

 

[Tony Stewart]

The audio spectrum has a compressed spectral range defined by Fletcher Munsen curves yet that while some can appreciate a higher SNR dynamic range than others the common emotional response is a much lower log dynamic range dB for amplitude while the gain or overall volume boost to peak levels is not needed to recognize a genre, tune , melody, tonality, pitch, or mood. This log range is the minimum range you need to display with visual feedback. Optical response is also over 100 dB like our hearing range, yet the acuity or recognition of vision is a subset of this entire range so we can adapt to background noise or darkness levels.

I believe you want to convert the exponent relation from sound to light with a fraction of 1 and not a power of 2. In other words a square root or a scaled log amplifier.

I am proposing that you want a logarithmic response of audio voltage to brightness current, yet limited to about 40 dB like TV defined by 0 to 100 IRE from black to max luminance in a linear range. with fast attack and a decay rate to match the response limitation of eye response for direct and peripheral frequency response. Our peripheral eye response is out of focus and dimmer yet much higher sensitivity to pulse rate going from about 50 Hz to over 3 kHz often used by PWM rates on moving car brake or headlights that might annoy some and only noticed by Physicists, Electrical Engineers or similar technically-minded such as yourself, who might understand motion artifact of our peripheral vision and Nyquist rate of our vision. You might not implement that but it would be the difference in emotional response of ultra wide stereo and mono stage speakers directly in front of you or a small TV image in front of you vs a wide IMAX or circle 360 theatre experience. Then we have "Visualization themes" in certain music media players that modulate the theme according to the Fourier Spectrum and/or amplitude response , which is something you can explore.

So what the eyes want to see what the ears hear depends greatly on the listeners' visual appreciation of such stimulation or artistic senses or dymanic moods and colours you wish to create. Perhaps you are a VJ or DJ entertainer or band lighting technician or a home entertainment guru. Each will have different design specifications that you might consider to define your goals.

 

[ccurtis]

I believe you want to convert the exponent relation from sound to light with a fraction of 1 and not a power of 2. In other words a square root or a scaled log amplifier.

It may not be immediately apparent, but once the x^2 correction is applied, the result is a logarithmic response that is equivalent to the human hearing response, y=x^0.67, which is quite close to a square root response, y=x^0.5.

If you plot y=x^0.67 (hearing) and y=x^0.33 (vision) as a graph, then y=(x^0.33)^2=x^0.66 (apply squared correction), also plotted on the graph, will be very close to what the y=x^0.67 function looks like and not terribly far from y=x^0.5 (square root).

There is also the quandary concerning the different Stevens' exponents for light intensity, with 0.33 given to a larger target and 0.5 (square root) given for a point source. After some research, a point source application is the exception rather than the rule, more suitable to a single, say LED sized, object at quite a distance away.

I get that much is subjective in a light organ application and that's partly why I wanted to find a more objective, analytical approach. Even then, I concede there are other factors which I have not considered in the calculation, such as the varying hearing perception across frequency and overall loudness level (Fletcher-Munson), ambient light level, etc.

It is very true that what the eye's want to see can be different from what the ears hear. I have added in features at times to my light organ designs to accommodate different tastes, like changing from a color to white light near peak levels, and such. Likewise, It also true that what the ears want to hear and what is actually recorded and played back can be different, so we have graphic equalizers and, when all else fails, the garbage can.

 

[ZipZapOuch]

You'll also have to deal with the baseline brightness at "zero volume" because any attempts to show very low light levels result in doubling the brightness if you are thinking in terms of PWM duty cycle of a 10-bit controller - 2/1024 is twice as bright as a duty cycle of 1/1024 and you'll be rounding off low level changes until it jumps unnaturally to the next level. Quite noticeable until you get past about 8/1024 duty cycle. So, calling 8/1024 your "ambient level" might be helpful, but that gives up 3-bits of resolution.

 

[ccurtis]

Yes, ZZO. In light organs, a sound threshold must be established below which no light is output because there is always some ambient and circuit noise present. The peak of the music is detected and a line is calculated from the threshold to the peak level that determines the gain, before any correction is applied. A squaring correction tends to help suppress light output at lower volume levels, as the graph of y=x^2 shows that y (light output) remains low initially and rises rapidly only after some level of higher volume. Yes, indeed, some resolution is lost, as you mention at the lower levels, but in practice I find this to be beneficial because the eye is very sensitive to changes in the PWM as lower values. By the time you see light output the sound level is high enough to justify it. That is not the case without the correction factor.

Off topic, but sort of related, is the correction that should be applied to PWM to linearize perception of light intensity, say, when a dimmer control knob is adjusted. Since the Stevens' exponent is x^0.33, the inverse function, X^3 should be applied to the PWM to linearize perception (the two functions combined form a straight line). That loss of resolution becomes more noticeable then and the dimmer control has to be increased (rotated) considerably before you see any light at all, as a result. Other than increasing the resolution, the only other remedy that comes to mind is to employ a small, fixed offset so that the light appears to be off just when the control is adjusted to minimum, but immediately appears to be on when rotated a little bit.

 

[rjenkinsgb]

I built quite a few in the 70s and early 80s, all discrete analog (I used to sell them to fund my computer projects).

None ever had a specific threshold control, just channel gains; though my final design for my own use had a background level setting that automatically brought up all the channels to an appropriate brightness when the music stopped, so the room was not in darkness.
I've recently got a new set of lamps for it as the originals had lost most of their coloured film coating!

An MCU controlled LED version definitely needs correction to compensate for the differences in response, but that's something I'm still wanting to experiment with, between work and other projects.

I have designed some of those for a special needs supply company in the late 90s, but I was not particularly happy with them as the customer insisted on having yellow included rather than just RGB! The effects produced were never as good as with straight RGB lighting. I can't remember the exact correction I used in that.

Comparing AC phase control to PWM, as the phase advances from the cycle crossover, both on-time and voltage increase up to the 50% advance point, which may be giving an approximation of square ratio power increase?

 

[MrAl]

In a light organ, the amplified signal from a microphone, or audio line output, is applied to a controller that subsequently drives the light intensity of a light source(s) for the purpose of adding a dimension of sight to the music listening experience. In most designs the goal is to have the light intensity rise is direct (linear) proportion to the amplified audio signal.

I have developed, built, improved such light organs of various configurations and designs over the past several years. However, while listening to music and observing the lights varying in brightness, accordingly, I was never truly satisfied with the effect. When I heard the peaks in the music and did not experience the peaks in light brightness in the same way I heard them. Sure, the light was brighter, but my sense of the increased brightness was not the same as my sense of the music. Additionally, there is too noticeable brightness fluctuation at times when the music is not so dynamic.

Some time ago, I attempted to address this (https://www.electro-tech-online.com...ut-driving-pwm-led-loudness-indicator.156838/). Still not satisfied, I recently realized I got the math wrong.

It is true that the way we perceive sound pressure is not the same as the way our eyes perceive light intensity and that is reason for the dissatisfaction. Stevens' Power Law tells us that the exponent for human sound perception has an exponent of 0.67. It also tell us that the exponent for human light perception is 0.33. Not the same!

So how do we get them to match? That's where the math problem comes in. Setting the two expressions equal gives us:

S^0.67 = L^0.33 (S is the sound pressure stimulus, or voltage from the microphone) (L is the light intensity)

Rearranging:

S^(0.67/0.33) = L or S^2 =L

Therefore, L corrected = L^2. In other words the signal driving the lamps must be squared! The goal is NOT to linearize the light intensity except only to provide a linear signal that is subsequently to be squared to give the desired result.

Now, finally, the eye sees what the ear hears.

Hi,

Human eyesight issues are very complex.
The eye itself is complex involving the iris that helps to adjust the amount of light reaching the retina. That means that if the light it gets is constantly changing, it will interpret that as constantly changing in about the same way, but if there is a period of nearly constant intensity followed by a constant change, it would interpret that differently because it had been conditioned to the constant intensity level.

There is also some persistence effect. The eye and brain remember some of the things it sees and that combines with the new things it sees later. This is amazingly noticeable with quick color changes.

Color is even stranger, because sometimes the color we perceive is relative not absolute. If we see a color on a white background it may appear darker, but on a black background it may appear lighter. This really amazes me sometimes because in computer graphics we get colors side by side all the time, and depending on what color is next to another color the two colors could look different in someways such as the intensity, saturation, and even the color itself.

With all this it might be hard to find a good fit for every piece of music. Ideally there would have to be some sort of memory function incorporated into the design also.

Also, human hearing is considered to be logarithmic. When you double the sound intensity the increase in db is 6, so from 2 to 4 it is 6db, from 4 to 8 it is 12db, from 8 to 16 it is 18db, from 16 to 32 it is 24db, etc.
This is kind of over simplified though because it depends on frequency too.

However, if you got what you think it right, then maybe you should stick with it