smodtactical
Member
- Joined
- Mar 30, 2020
- Messages
- 20
- Likes
- 34
https://www.headphones.com/communit...SwYDvUwaDFbEiIW3f0BwyHPCd462GvSonth2QibjyBvrY
Didn't see this posted, thoughts?
Didn't see this posted, thoughts?
Makes sense. But if you want a single score to rank devices, what do you use?https://www.headphones.com/communit...SwYDvUwaDFbEiIW3f0BwyHPCd462GvSonth2QibjyBvrY
Didn't see this posted, thoughts?
Perhaps this concept, itself, is where the problem begins...Makes sense. But if you want a single score to rank devices, what do you use?
It's a generally very good article. One caveat: when looking at audibility of distortion, we need to be careful not to use something very non-physical as the metric. For example, it's easy to hear -40dB of 9th harmonic on a pure tone, but I haven't encountered a device that has 9th harmonic only. Almost every real-world device is dominated by lower order distortion, odd or even depending on the topology. If there's higher order, it's mixed in with MUCH larger quantities of lower order.https://www.headphones.com/communit...SwYDvUwaDFbEiIW3f0BwyHPCd462GvSonth2QibjyBvrY
Didn't see this posted, thoughts?
But if you want a single score to rank devices, what do you use?
Extremely good point - an earlier revision of the article went into more depth about nonlinearity, but it got edited down for "readability". FWIW, there will be a companion post coming using @pkane's distort software to mimic the nonlinearities of a few real amplifiers, one of which was the inspiration for the audible H3/H5 cases (as the lower harmonics were masked with pure tones, but the higher ones were not). My editor asked me to keep the FFTs and the numbers as clean and simple as I could.It's a generally very good article. One caveat: when looking at audibility of distortion, we need to be careful not to use something very non-physical as the metric. For example, it's easy to hear -40dB of 9th harmonic on a pure tone, but I haven't encountered a device that has 9th harmonic only. Almost every real-world device is dominated by lower order distortion, odd or even depending on the topology. If there's higher order, it's mixed in with MUCH larger quantities of lower order.
Makes sense. But if you want a single score to rank devices, what do you use?
If I had known it was yours, I would have tagged you. Nice work, and it does hit points about the vain desire for single number metrics. They don't work, and that main point is made very well.Extremely good point - an earlier revision of the article went into more depth about nonlinearity, but it got edited down for "readability". FWIW, there will be a companion post coming using @pkane's distort software to mimic the nonlinearities of a few real amplifiers, one of which was the inspiration for the audible H3/H5 cases (as the lower harmonics were masked with pure tones, but the higher ones were not). My editor asked me to keep the FFTs and the numbers as clean and simple as I could.
Makes sense. But if you want a single score to rank devices, what do you use?
True. Always read the full suite of measurements for that reason. But the world likes a good headlinePerhaps this concept, itself, is where the problem begins...
Single score to rank devices is meaningless and brings useless results. Deep analysis is always needed.
At least from a historical standpoint, I don't agree with this - in the context of audio electronics, attempts to correlate nonlinearity and distortion are about as old as these primeval nonlinearity measurements themselves. E.g.SINAD, IMD, etc are intended to reflect electronic characteristics (duh), nothing more.
It's an abuse to map those measures directly into the space of perceptual relevance.
SINAD, IMD, etc are intended to reflect electronic characteristics (duh), nothing more.
It's an abuse to map those measures directly into the space of perceptual relevance.
I agree. Is there even one figure of merit, in any field, that is really descriptive to an expert?