EDIT: I have now purchased this system and measuring one speaker every other day!
Hello everyone.
I am doing some planning on where we are going from here. We have tested a ton of audio electronics and while there is still endless supply of such gear waiting to be tested, I hear complaints from time to time that we are not addressing the most audible components in the chain, namely speakers and headphones. I will leave headphones for another time. For now, wanted to discuss speakers.
As some of you know, the gold standard in objective audio measurements for speakers is the work Dr. Toole/Olive have done at NRC and Harman. Basically, a set of weighted measurements incorporating direct and indirect sounds can predict listener preference for a speaker to high degree. We are talking 80% or so predictive power.
The challenge for anyone to perform these measurements is the requirement for anechoic chamber to measure the speaker and not the room. And needing to have many measurements to cover all axis around the speaker. Magazines and others perform the test using far fewer axis and gating (stopping measurement before first reflections). The latter limits the low frequency response though which is an important metric (as much as 30% of preference for a speaker is from bass).
Enter the company Klippel. They have built what they call a near-field scanner. Using dual measurements and some very fancy math, they are able to a) eliminate the effect of reflections and b) extrapolate far field data (what we want to have) from near field measurements (close in to speakers which is much simpler to perform). This is a combination of software and hardware. The hardware is a robotic measurement arm that scans the speaker in 3-D space. You can see it in this short video:
The system actually works better than anechoic chamber because even the largest of such rooms, still have very low frequency room modes. You can see this in this example measurement from Klippel comparing Harman anechoic chamber measurements to theirs:
If you look to the left, the Klippel system in black has a more correct curve than the one in red (anechoic).
Once the measurements are performed, and signal processing performed, they are then able to produce Harman style measurements (which are now standardized in CEA 2034). Here is sample data for a Harman speaker (F228):
Here, we would look for flat on-axis response (black) and smooth curves for the rest. Or alternatively the directivity index (lines below) not having kinks in them.
Anyway, bottom line is that unlike our electronic measurements, such speaker measurements are highly correlated to what we hear. Currently no one is performing such measurements for the public. Even Harman which has such data, doesn't publish it routinely for their own speakers.
We have the chance to completely transform the speaker industry, bringing much needed objective and useful data to the table.
That is all the good news. Here is the really, really bad news. The basic hardware alone costs $39,000 Euros. With necessary software this cost balloons to over $70,000! This puts my nearly $30K investment in Audio Precision to shame. On top of this, we also have the logistic cost of sending speakers back and forth.
My question to you all is whether we should go in this direction now, later or not bother. Poll options are above. If we do go forward, funding of this will be a serious thing to figure out. Options exist with respect to potential sponsorship, serious donations from you all and my pocketbook.
Any and all thoughts are welcome including doing this for headphones where the costs are much lower (but the work less certain).
Hello everyone.
I am doing some planning on where we are going from here. We have tested a ton of audio electronics and while there is still endless supply of such gear waiting to be tested, I hear complaints from time to time that we are not addressing the most audible components in the chain, namely speakers and headphones. I will leave headphones for another time. For now, wanted to discuss speakers.
As some of you know, the gold standard in objective audio measurements for speakers is the work Dr. Toole/Olive have done at NRC and Harman. Basically, a set of weighted measurements incorporating direct and indirect sounds can predict listener preference for a speaker to high degree. We are talking 80% or so predictive power.
The challenge for anyone to perform these measurements is the requirement for anechoic chamber to measure the speaker and not the room. And needing to have many measurements to cover all axis around the speaker. Magazines and others perform the test using far fewer axis and gating (stopping measurement before first reflections). The latter limits the low frequency response though which is an important metric (as much as 30% of preference for a speaker is from bass).
Enter the company Klippel. They have built what they call a near-field scanner. Using dual measurements and some very fancy math, they are able to a) eliminate the effect of reflections and b) extrapolate far field data (what we want to have) from near field measurements (close in to speakers which is much simpler to perform). This is a combination of software and hardware. The hardware is a robotic measurement arm that scans the speaker in 3-D space. You can see it in this short video:
The system actually works better than anechoic chamber because even the largest of such rooms, still have very low frequency room modes. You can see this in this example measurement from Klippel comparing Harman anechoic chamber measurements to theirs:
If you look to the left, the Klippel system in black has a more correct curve than the one in red (anechoic).
Once the measurements are performed, and signal processing performed, they are then able to produce Harman style measurements (which are now standardized in CEA 2034). Here is sample data for a Harman speaker (F228):
Here, we would look for flat on-axis response (black) and smooth curves for the rest. Or alternatively the directivity index (lines below) not having kinks in them.
Anyway, bottom line is that unlike our electronic measurements, such speaker measurements are highly correlated to what we hear. Currently no one is performing such measurements for the public. Even Harman which has such data, doesn't publish it routinely for their own speakers.
We have the chance to completely transform the speaker industry, bringing much needed objective and useful data to the table.
That is all the good news. Here is the really, really bad news. The basic hardware alone costs $39,000 Euros. With necessary software this cost balloons to over $70,000! This puts my nearly $30K investment in Audio Precision to shame. On top of this, we also have the logistic cost of sending speakers back and forth.
My question to you all is whether we should go in this direction now, later or not bother. Poll options are above. If we do go forward, funding of this will be a serious thing to figure out. Options exist with respect to potential sponsorship, serious donations from you all and my pocketbook.
Any and all thoughts are welcome including doing this for headphones where the costs are much lower (but the work less certain).
Last edited:
