Mark,
Quoting this, makes me thinking that you believe all this.
Many points however are pure commercial motivated statements, to make the product look superior.
I agree that measuring isn´t telling the whole story, but shooting wild statements from the hip is just as questionable.
Hans
Hans,
Could you please give me an example or two what you believe are false statements?
If you only point is that ESS is focusing on areas where they think they have a competitive advantage, and not talking about other things they know about but which would not be advantageous to talk about, then I would fully agree. Nonetheless, the issues they do talk about are not imaginary.
Mark
Could you please give me an example or two what you believe are false statements?
If you only point is that ESS is focusing on areas where they think they have a competitive advantage, and not talking about other things they know about but which would not be advantageous to talk about, then I would fully agree. Nonetheless, the issues they do talk about are not imaginary.
Mark
1) "The human ear can detect etc. etc."
To my opinion this is a meaningless statement.
When detecting sound that's no part of the noise like from a waterfall, or hearing sound below the noise, that can only mean comparing the total noise energy to a specific sound having much less energy.
But taken into account that the bin width of a human's hearing in the sensitive areas is ca 2 to 3Hz, you should compare the waterfall's FFT in the same bins with the same 2 to 3Hz bin width as to where the "other" sound arrives in your system.
You might at best notice an increase of several dB's but not less, keeping in mind that your ear's sensitivity will be partly masked by the total noise energy.
That means that the elsewhere generated sound should be at least as loud as the white noise in that bin or louder,
But you will never be able to detect sound that's below the noise level in that bin.
So detecting sound below noise level is at least a very misleading and untrue statement.
2) "The surprising reality is that sigma-delta DACs can be audibly distinguished from a conventional DAC despite measuring very much better than that DAC."
This is just a silly generalization as telling that men are smarter then women or vice versa, while at the same time suggesting that "distinguishing" automatically means inferior.
There is more between heaven and earth as just SDM, like FIR filter length in SR Converters before the DAC, SDM DAC's used for NOS conversion while still using the SDM part, different types of noise shaping in the SDM, the amount of levels that the SDM produces, its upsampling ratio, etc, etc.
You of all people always mentioning the importance of proper testing, should know that when comparing things, everything should be the same except for the item that has to be compared.
Well, it will be hard to find a R-2R DAC with an SDM onboard that can be optionally be switched on/off unless you build one yourself.
All other attempts will only tell you that DAC A might sound better than DAC B, but it will be hard to impossible whether the SDM part is to blame or the processing part that feeds the SDM.
3) "Sigma-delta modulators create non-periodic steady state noise (non-PSS) artifacts..."
Just another generalization. There are ways to influence this, read Marcel's excellent DAC thread .
I think it's enough for the moment.
This is all just one step away from: 196Khz sounds far better as 44.1Khz and so is 24bits versus 16bits.
Sorry Mark, it was not meant to put you down but I just can't stand those generalizations from some fancy manufacturers.
Hans
A quite bizarrely irrelevant comparison. But I'm sure you know that. (FWIW I've owned Lambos -- over-rated - and no idea what a Hugo is).
IME its not exactly true that the human auditory system works like a a phase-insensitive FFT analyzer. In some cases the time-domain waveform is much more salient to perception, especially at lower frequencies. Consider the example wav files at Purify where four frequencies are present at once. If the auditory system is a phase-insensitive FFT analyzer then the two wav files should sound the same (since the only difference is that phase of two of the frequencies have been shifted): https://purifi-audio.com/2019/12/07/amfm/ Do they sound the same to you?
Moreover, there is some evidence that when there is a detector with a detection threshold then adding random noise can make the signal of interest overcome the detection threshold (this effect is of course called stochastic resonance). IOW, the 'waterfall' (random noise source) might even help detection in some particular case
I believe the specific point ESS was trying to make though, was that random noise is more ignorable than more deterministic noise or than a plainly deterministic signal. In addition I believe it is uncontroversial that humans can to some extent detect signals under a noise floor. IOW a noise floor is not a hard floor, it is maybe more like the top of a cloud. We also should consider that when we take an FFT we are getting an average level of noise integrated over the acquisition time. The noise is actually fluctuating so some part of the time a deterministic signal may be not be below the so-called floor.
A fascinating thread about Paul Frindle, his AES paper, and his study of how humans brains can remove pseudo random dither noise to hear the underlying quantizing distortion:
https://www.diyaudio.com/community/threads/paul-frindle-is-this-truth-or-myth.371790/
Without going into the all the complications, seems to me ESS's simple statement is not technically false.
I think what they are trying to say did surprise many people. DACs like ESS makes measure far better than something like a TDA1541, but they don't necessarily sound better, and some people can tell by listening to sigma delta artifacts that the 'better' measuring dac has a very distinct sound.
They do, as MarcevdG has explained in good detail. What the ESS statement means is that noise and or deterministic signals that are changing during an FFT acquisition will tend to have their energy scattered across multiple bins. The more scattered, the more it looks like noise. Even it were an idle tone sweeping across the audio band it could end up looking like noise. That doesn't necessarily mean it sounds like a noise to human. Also we know that modulators tend to produce noise that is a function of the instantaneous DC offset. Of course if they are dithered then there is some decorrelation of the noise with respect to the audio signal, but it may not be completely decorrelated.
Actually, we have done some tests here with Andrea's DSD dac. We have the same music in CD format and also in 24/192 format. Measurement with a LUFS meter shows that there is some difference in mastering between the two versions. The CD is more compressed and with a higher peak level, both of which work to help obscure limited dynamic range. The hi-res version is a little less compressed and has a little lower peak level. When very high quality conversion of PCM->DSD256 is done, the soundstage of the hi-res source version is wider. Also it sounds like there is slightly less low-level distortion in the hi-res version (this is of course when using the large ESL speakers which are themselves extremely low distortion).
Bottom line: I will take the 24/192 version when it is available.
Last edited:
Its called an analogy. All analogies fail at some point if picked apart. The purpose of any analogy to compare something familiar to something less familiar, but at a superficial level only. The point that you seemed to have missed is that some expensive hi-fi stuff is junk. Some other expensive hi-fi stuff is much more accurate sounding that is predicted by FFT measurements. IOW, the cheap system is not always just as good as the expensive system if the basis for making that judgment is simply standard 'figure of merit' measurements.
BTW, a Hugo was a very unreliable piece of junk. One of the worst ever, if not the worst.
Do you have any examples of these which have been proven to sound more accurate in controlled listening tests?
"Humans have remarkable spatial hearing capabilities:
Accuracy within 1-2 degrees in angular location of a sound source"
https://pages.ucsd.edu/~ajyu/Teaching/Cogs180_sp14/Slides/lect0428.pdf Page 11
That and a little trigonometry can work well.
Accuracy within 1-2 degrees in angular location of a sound source"
https://pages.ucsd.edu/~ajyu/Teaching/Cogs180_sp14/Slides/lect0428.pdf Page 11
That and a little trigonometry can work well.
Last edited:
CD, for example, is not lossy. 100% of the information put on it is stored, and retrieved.