Are CSDs bullshit or unnecessary?

With speaker measurements, I would generally say that they are not needed, but more data is more data, even if the data confirms other data. CSDs can measure specific attributes not covered by frequency response (cabinet resonances, etc.)

Most of these is cut and paste from my replies to an old Reddit thread (I no longer post anything to Reddit because of the level of stupidity and toxicity there):

https://www.reddit.com/r/headphones/comments/adgzjv/how_to_interpret_csd_and_impulse_response/

You do realize Tootle is looking a CSDs with a 350ms time interval - the size of a room? I'm looking at things between 3ms and 5ms and using a FFT window rise time magnitudes smaller than Toole. We are looking at different things. Toole's statements are true on the scale he is looking at. (I don't bother with CSD when I make speakers.) However, this does not negate my findings with headphones.​

As far as cup reverb seen on CSDs, the FR doesn't show this - it's not a time domain measurement. The only thing I can say to you to listen to headphones which are well damped internally (you can take them apart to see) and those which are not, then compare CSDs. The headphones which have a more sea-shell effect are those where those patterns appear more vividly.​

As far as my anecdotes, I have far more headphone CSD data than any one else out there including Floyd. I don't see Floyd offering more than a few test cases in his AES paper either, so this seems like a double standard on your part.​

The only reason my work does not constitute as "research" is because I have not formally studied the data and written an AES paper. This does not mean it should be summarily dismissed. My advice to you is do more, read less. Science is based on doing inspired from reading.​

--

You say I am wrong, yet you acknowledge the "rare edge cases" where CSD and FR don't agree. This disagreement happens sometimes because of the nature of measurements. I've even seen this happen with speakers, and it's almost guaranteed for certain types of horns. As far as the research, it was an AES paper from 1991. How many headphones were tested, how much data did they gather, and what rise time and window function did they use for CSDs, and how did the authors measure?​

My data (probably a lot more extensive than the authors' of the 1991 AES paper) would still suggest that FR is still the primary determinant of FR, sharp peaks, depressions, etc. However, CSDs serve as a reality check to make sure deep nulls are not peaks in disguise. Different ears, heads, couplers will have different results. They are also other factors such as orthos which have a tendency to keep resonating like a drum skin at certain frequencies or internal cup resonances that linger at different lengths depending upon internal cup or earpad materials. CSDs are a useful supplement to see cup reverb effects and unnatural resonances.​

BTW, here's a real-world example of a false null: https://www.superbestaudiofriends.o...-measurements-and-tech-talk.4573/#post-144556​

Here is a real-world example of seeing cup reverb:​

https://www.superbestaudiofriends.org/index.php?threads/zmf-verite-measurements.7043/#post-231376​

There are plenty more examples of such. Yes, CSDs are more difficult to read, but they do provide information that FRs do not.​

TL;DR: Don't be lazy and study data and measurements more (turn on brain) rather than accept AES articles as gospel (read and regurgitate)​

--

There's no controversy. Different headphones, different methods, different results.​

Also, Floyd is talking about speakers in a room, which is a very different environment of what goes on inside a headphone cup and the volume between the ear and baffle/earpads. Floyd is awesome. Olive is a tool for "researching" that god awful consumer-reference curve.
​

--

An important point is that CSDs are merely another visualization that leverages the same data as a frequency response derived from an impulse response. One can make the same argument that frequency response is unnecessary because we already have an impulse response. If you are good enough, you can draw a rough frequency response from looking at an impulse response. Seriously.​

 

Some additional thoughts: CSDs in bass? I really have no idea. Maybe they do correlate to some audible aspect, maybe they don't. The only thing I would offer is to gather data and maybe offer an additional visualization that normalizes the t=0 point to 0db at each frequency.

Also take a look at a relative of the CSD, the burst response. Rather than being based on time, it's based on cycles, and with a shifting window of the same size (whereas CSD window sizes get progressively smaller with the front of the window moving forward in time). The burst responses here show the effectives of mass loading and damping with the Jupiter Audio Research mods on the Sennheiser headphones.

https://www.superbestaudiofriends.o...-measurements-and-wormsign.13183/#post-402816