Two Old ESS Talks

[mocenigo]

I have ordered a reprint of it. It seems to be one page.

Using one of my university accesses i got it. It is a letter to NATURE, not actually a paper. The text is short and there is nothing more than reported on stackexchange. OCR follows, I have not corrected all errors.


Upper Limit of Frequency for Human Hearing

THE frequency above which air-borne sound becomes inaudible is generally considered to be about 20 kc./s. All sensitivity determinations agree that the threshold rises very steeply above 12 kc./s. ; and above 12 kc./s. there are indications that fre-quency discrimination begins to fail, that is, that the least discriminable increment of frequency measured as a fraction of an octave begins to rise sharply. It seems to have been tacitly assumed that the human cochlea is incapable of response to frequencies above 20 kc./s., and that the upper limit for air-borne and bone-conducted sound is the same.

I have compared these limits on myself and two other subjects, using an oscillator of frequency variable up to 120 kc./s. and a transducer consisting of a pack of Rochelle salt plates resonant at 100 kc./s. In all cases the upper limit for air-borne sound, with the transducer held close to the external meatus of the subject, was below 16.5 kc./s. When, however, the crystal was pressed firmly on the mastoid or on the temporal region, a sound was heard for all fre-quencies up to at least 100 kc./s. In the former case the sound was perceived in the ipsilateral, in the latter in the contralateral ear. The failure at and above 100 kc./s. was at least partly instrumental, due to the failure of the oscillator to maintain an adequate voltage across the falling impedance of the crystal.

The sensation of pitch associated with the sound was approximately that of the highest tone audible by air conduction. As the frequency was varied con-tinuously from 12 to 100 kc./s., the pitch rose with it to 15 kc./s. and stayed there as the frequency was further increased. No extraneous sensations, for example, of warmth, pain or discomfort, were in-duced and the sound heard was subjectively a perfectly normal tone. It must be concluded : (1) that the sensory elements at the basal end of the cochlea are competent to respond to sounds up to 100 kc./s. in frequency ; (2) that the failure of the normal ear to respond to air-borne sounds above 20 kc./s. is due wholly to the failure of the middle ear to transmit such frequencies.

R. J. PUMPHREY

Department of Zoology, University of Liverpool. July 11.

 

[Julf]

The sensation of pitch associated with the sound was approximately that of the highest tone audible by air conduction. As the frequency was varied con-tinuously from 12 to 100 kc./s., the pitch rose with it to 15 kc./s. and stayed there as the frequency was further increased.
...
It must be concluded : (1) that the sensory elements at the basal end of the cochlea are competent to respond to sounds up to 100 kc./s. in frequency ; (2) that the failure of the normal ear to respond to air-borne sounds above 20 kc./s. is due wholly to the failure of the middle ear to transmit such frequencies.

Ummh.... If that was the case, shouldn't the test subjects have been able to detect the pitch correctly?

 

[xr100]

He claims that the Hyperstream tech addressed at least one of the technical issues belonging to their previous SM designs and listeners were no longer able to distinguish SM vs. non-SM designs during tests.

From the book "Understanding Delta-Sigma Data Converters" (2nd. Ed., Pavan, Schreier and Temes,) p259:

"In this chapter, we examine the primary nonidealities in a [continuous-time delta-sigma modulator]--namely excess delay, time-constant variations, and clock jitter."

Mallinson's talks primarily cover non-periodic steady-state noise, variable excess phase, and jitter.

It would seem, then, that the developments at ESS were emphatically aimed at fundamental improvements to delta-sigma modulator technical performance. Indeed, Mallinson begins by saying that the "HyperStream" modulator is "an advanced form of sigma-delta modulator that doesn't suffer from sigma-delta modulator problems."

He also states that millions of HyperStream modulators had shipped--at one time, 30% of DVD players shipped contained an ESS chip "including this new HyperStream modulator." It might be imagined that the customers of "system-on-a-chip" parts for standard-grade consumer DVD players were not terribly interested in eliciting the opinions of audiophiles!

The "SABRE" brand, at that time, referred to their DACs that had "every bell and whistle that we can find" turned on.

Mallinson suggests "DNR vs. DC offset" as a method of investigating modulator performance--a more recent example of a plot (courtesy of ESS!) is available for "SABRE," "SABRE PRO," and three "competitor" DACs. Note that the "competitor" parts display artefacts under -120dBFS, and one is reaches (or is getting asymptotically close to) -100dBFS--the mechanism(s) that yield these results is another matter; only one "competitor" DAC shows the "spiky" plot shown in the RMAF presentation.

Mallinson says that ESS had FPGA DAC implementations which allowed for turning on/off various features, and (for a third-party) to perform controlled listening tests in relation to the audibility of said features, presumably something equivalent might be required.

 

[pozz]

Thanks for the detailed post.

Mallinson says that ESS had FPGA DAC implementations which allowed for turning on/off various features, and (for a third-party) to perform controlled listening tests in relation to the audibility of said features, presumably something equivalent might be required.

Listening tests, yes. But their level of control is in question since he spends no time on that issue.

It would seem, then, that the developments at ESS were emphatically aimed at fundamental improvements to delta-sigma modulator technical performance. Indeed, Mallinson begins by saying that the "HyperStream" modulator is "an advanced form of sigma-delta modulator that doesn't suffer from sigma-delta modulator problems."

That's the thing: the technical progress is impressive, even important. It's a whole other thing to then link that positively to listener impressions. It's as uncritical as the variously and often-discussed obsession with materials and circuit components as having a guaranteed link on the quality of perceived sound.

These two talks are some of the clearest pieces of evidence that the industry as a whole is not clear on what matters in audio design. They represent a very technical company through a very technical individual (chief scientist!). And this material floats in the web, without context, for someone to find and then conclude that DAC chips and architectures have intrinsic sonic characteristics.

 

[xr100]

Thanks for the detailed post.

You're welcome. I only wish that I really understood the "nitty gritty" aspects of sigma-delta modulators...

Listening tests, yes. But their level of control is in question since he spends no time on that issue.

Questionable indeed--hence I indicated the need for "third-party" listening tests, with the implication that these might be easier said than done!

That's the thing: the technical progress is impressive, even important. It's a whole other thing to then link that positively to listener impressions. It's as uncritical as the variously and often-discussed obsession with materials and circuit components as having a guaranteed link on the quality of perceived sound.

Having first watched Mallinson's talks a few years ago, I must confess that I had been impressed--even if not convinced that the ESS technologies discussed would be audible (Mallinson says that he cannot hear any difference!) It wasn't until I stumbled on this site and subsequently the "ESS IMD Hump" that I started to realise that technical superiority, even in quality implementations, was far from a given...

That said, first-class executions of DACs containing the best ESS parts might perhaps be compared to an "improved Blameless amplifier" than snake oil--but I can hardly disagree with you about the potential for being mislead.

These two talks are some of the clearest pieces of evidence that the industry as a whole is not clear on what matters in audio design. They represent a very technical company through a very technical individual (chief scientist!). And this material floats in the web, without context, for someone to find and then conclude that DAC chips and architectures have intrinsic sonic characteristics.

Alas, it's been going on for a very long time. I recall Pioneer's "Legato Link Conversion," which was marketed as putting the "warmth" back into digital audio, or something along those lines--full-page adverts featuring a colourised photo of a "frozen brass instrument" in the hi-fi press--and when products featuring said DACs were measured, it was found that they rolled off towards 20kHz--and the information that had been "lost" by sampling at 44.1kHz which Pioneer claimed "Legato Link Conversion" restored was nothing other than a high level of spurious ultrasonic components!

I don't know where one begins with the audio industry--a favourite example of mine is the perennial marketing of MTM speakers for centre channel use in a horizontal orientation--i.e. the wrong dispersion pattern--contrary to a vertical "D'Appolito" MTM design which has some advantageous characteristics. Almost literally a case of an "upside-down" approach...