That silly debunked marketing stuff does not deserve to be repeated ad nauseam.
I edited my previous post a bit in an effort to further clarify. Overreliance on FFT is not the only problem at ASR, do you want a list?
Okay. Here is where Mr. bohrok and I seem to disagree. For me the true test of any scientific theory is how reliably it can be used to predict the future. In that regard, FFT shows limited correlation with how people actually hear. For example, two dacs can have very similar FFT spectra but sound very different. The mistake people make with FFT is to look at it first, then cook up a story to the effect that the spurs explain almost everything about the sound. Its somewhat analogous to looking the alignment of the planets then constructing a story of causation where the planets determine war, famine, weather, etc. The correct test is to look at the planets first and see if you can use them to reliably predict the future. However, I am talking about human error here, not a problem with the FFT math. Its how humans can take a measurement or observation of anything and the brain will automatically produce a story of cause and effect. This is well known in the field of cognitive psychology.
Sorry, FFT is not a scientific theory. And be consistent: show us a quote where somebody has cooked up a story after looking at FFT.
Read the reviews at ASR. The stories of calculations as to whether a certain distortion level can be audible are absurdly mistaken. They don't even know how 'limit of audibility' is defined.
Also, nobody said FFT is a scientific theory. The faulty theory is that FFTs can be used to reliably predict sound quality, audibility, etc. Its not reliable in reality because of limited correlation between FFT and how human hearing works. Moreover, the problem is worse than it should be because typical audio FFTs discard phase information.
Also, nobody said FFT is a scientific theory. The faulty theory is that FFTs can be used to reliably predict sound quality, audibility, etc. Its not reliable in reality because of limited correlation between FFT and how human hearing works. Moreover, the problem is worse than it should be because typical audio FFTs discard phase information.
This is a bias effect in scientific research! Use callipers for the first time and see how good your measurement is, but with good methodology, very reliable measurements can be made.
But from what I can tell, the methodology for FFT is solid and in that specific case, he seems to know what he is doing. The data is published and can be repeated. That's more reliable to me than a recommendation based on a personal preference. In the best case, the data can be compared with the datasheet provided by the manufacturer of the IC and the specification of the product itself that is using the IC.
So nobody has cooked up such story and you are making things up again (i.e. telling fairy tales).
Now show us where this theory has been proposed or a quote of somebody making such claim.
I'm not sure how we ended up having a discussion over the reviews of ASR.
But for clarification: I don't believe in audio voodoo, if it measures accordingly to the specification of the design, it is a good design, everything else is a personal preference. In the best-case scenario, I will be able to build a transparent-sounding device, I don't think that it is easy to do, nor likely to happen, but I will be very happy with something that is able to play audio good enough for my self build speakers.
They calculate that if FFT HD is at some low level, say, maybe -117dBFS THD, and if the threshold of audibility is some number, and if the highest SPL humans can listen to without injury is 120dBSPL, then distortion of a device must be inaudible, period. BS! First of all, a threshold of audibility is NOT a hard limit for every human in the world. That's the first mistake. A threshold of audibility is an estimate of an average for a population. That means 50% of people should be able to hear below the published threshold. Secondly, 120dBSPL is deafening. That's not where the ear is most sensitive to distortion. Anyone should be able to figure that from personal experience. Third of all, I already proved in one of PMAs blind listening tests that I could sort unity gain opamp buffers in order of distortion by ear, period. All those buffers had lower distortion that ASR would calculate as audible. And I didn't listen at 120dBSPL either. Those guy are full of it IMHO. They just don't know what they're talking about. Yes, they can read the AP manuals to run the push-button measurements. That's not the point. The point is they don't know how to properly interpret FFTs in the context of human perception, but they think they do.
Last edited:
OK, can we stick to the topic though?
What about the ES9033, should use that thing? I have not seen threads in this forum of someone using this specific chip in a DIY project. Its cheap, has good features and if it is the successor to ES9023 we can assume that it inherits it qualities (maybe).
Any reason not to try?
What about the ES9033, should use that thing? I have not seen threads in this forum of someone using this specific chip in a DIY project. Its cheap, has good features and if it is the successor to ES9023 we can assume that it inherits it qualities (maybe).
Any reason not to try?
Last edited:
haha, thank you but you don't have to worry, just someone with too much time here.
Nevertheless thank you for your recommendation, it was worth having a discussion and poking around
hi again, math cannot be debunked nor can it be marketing stuff. I even published the Matlab script generating the files.
there is nothing wrong with the fft. it is an invert ole transformation that does not throw information away. the plots in the blog post only show the magnitude of the fft. the difference is in the phases. the two signals are different and so are their fft transforms.
Nobody said math was debunked. Your claims and that silly marketing stuff was debunked.
True. When I refer to problems with FFTs, I am referring to problems associated with human understanding and or use of FFTs, not the FFT math itself. Like for example is the frequency domain view the most appropriate choice for the problem at hand? Are the full implications of windowing well understood by casual users? Implications of the common practice of discarding phase information for audio FFTs? Implications of time resolution verses frequency resolution, why short-time FFTs are sometimes better despite the broad bins. What about the fact that the real world isn't exactly linear, time-invariant, and or stationary? How non-linear and or non-time invariant does something have be before FFT use amounts to garbage in, garbage out?
Last edited:
what claims and what marketing stuff? the blog leaves it to the reader to draw own conclusions