Two Old ESS Talks

[pozz]

So I was digging around and I found two lectures that I listened a few years ago. They are by Martin Mallinson, a reputable technical person and former chief scientist at ESS. I hope that I hope he sees this post at some point and responds with his current thoughts and conclusions.

• (2011) Martin Mallinson - Digital Jitter & Volume Controls
• (2011) Martin Mallinson - Noise-Shaping Sigma-Delta-Based DACs

(The noise-shaping talk was taken down from the official RMAF site and reposted by another Youtube user. )

1. He talks about audio companies hiring audiophiles to listen to their products.
2. He says that technical sales are dependent not on the engineer's nod to the features and specifications, but the company-official audiophile's approval after the specs are OK'ed.
3. He says that ESS specifically conducted listening tests and noted that certain individuals, including the President of ESS at the time, certain other senior members of the organization and at least one member of their technical staff (who learned to do so) displayed an ability to detect SM vs. non-SM designs during tests.
4. 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.
5. He says that this happens because the human hearing system is more sensitive than measuring instruments, and that engineers may not understand why a person can detect ultrasonic phenomena or phenomena below -110dBFS, for example, but it happens.
6. He says that, in DACs, PLL and certain (pre-ESS Hyperstream) noise-shaping (specifically, the resulting non-periodic steady-state noise) can be detected.

To be clear, these talks are a good description of the workings and internal processes of volume controls and SM DACs—their issues lie in his discussions of audibility.

He does not discuss psychoacoustic issues or considerations; rather, he theorizes why low-level signals could be detectable. He does not describe the conditions of the listening tests, only stating what such tests demonstrated. In both these cases I would say that he didn't take into account how much weight his conclusions have, given his authority, and how much that really affected the people in the room there with him.

I would guess that the listening tests were probably sighted or flawed in some manner. I would also guess that he was convinced these tests were valid because their inclusion seemed to be an industry decision (why spend money hiring audiophiles if nothing is gained?) and that he did not know how to assess listening circumstances in the same critical manner he used to assess measurement conditions for chip designs.

Which brings me to this point: design engineers who have limited knowledge or interest in how a certain product or device will be used will make the same kinds of erroneous conclusions that lay audiophiles have come to. This is despite their technical training.

I'm writing this in the most polite way I can because this problem is so pervasive, and so encompassing. It has nothing to do with a lack of intelligence and everything with influence.

Another example is the lecture by Jude Mansilla from Head-Fi at RMAF in 2018 on measuring techniques for headphones. He spends a lot of time on the setup (doesn't say much about ear canal resonance though), but none on psychoacoustics, the main area that makes those measurements so difficult.

Edit: @bennetng has analyzed and shown the mathematical errors in the volume control presentation: https://www.audiosciencereview.com/forum/index.php?threads/two-old-ess-talks.9216/

 

[anmpr1]

The two links point to the same file, but you can find the second talk easily. I like the idea that he raises chickens. I don't like the idea that it is possibly possible to 'hear' a 50KHz signal via some kind of auditory 'mechanism' we 'don't understand yet'.

His chips are excellent. No problem with analog volume controls.

At first I was thinking ESS? I was wondering how ESS (the old Electrostatic Sound Systems) could possibly still be in business? But then I looked and found that they are. You can still get an AMT-1. So I'm happy about that, too.

 

[pozz]

The two links point to the same file, but you can find the second talk easily. I like the idea that he raises chickens. I don't like the idea that it is possibly possible to 'hear' a 50KHz signal via some kind of auditory 'mechanism' we 'don't understand yet'.

His chips are excellent. No problem with analog volume controls.

At first I was thinking ESS? I was wondering how ESS (the old Electrostatic Sound Systems) could possibly still be in business? But then I looked and found that they are. You can still get an AMT-1. So I'm happy about that, too.

Thanks. I updated the link.

I forgot about the speaker manufacturer. Just to make it clear, it's the ESS Sabre DAC chip company that he's representing. The same one that created the chips for the highest-performing DACs Amir's measured.

 

[anmpr1]

Thanks. I updated the link.

I forgot about the speaker manufacturer. Just to make it clear, it's the ESS Sabre DAC chip company that he's representing. The same one that created the chips for the highest-performing DACs Amir's measured.

When you are as old as I am, you straddle the analog/digital divide. Sometimes I'm thinking old school, and other times digits. I get confused easily!

I own some modern-day ESS digits. I wish I owned some old-day ESS analog stuff!

The videos were interesting. They were from a long time ago in tech terms. I wonder what he'd say now, if he was redoing his presentations?

 

[pozz]

The videos were interesting. They were from a long time ago in tech terms. I wonder what he'd say now, if he was redoing his presentations?

Same. He's no longer with ESS and so might be able to speak a bit more freely now.

 

[Vincent Kars]

He left ESS and founded this company, using ESS DAC's of course
https://www.resonessencelabs.com/about-us/

 

[Eirikur]

I don't like the idea that it is possibly possible to 'hear' a 50KHz signal via some kind of auditory 'mechanism' we 'don't understand yet'.

Well, I have conducted some experiments of my own, and I could clearly hear the distortions created by the ultrasonic tones I injected. My explanation: either the amplifier, or the speakers distort within the audible band because of ultrasonics.
I tested with two independent clean sine waves and their mix, 22350Hz (F10) and 42192kHz (E11), all within 384kHz sample rate as this is the highest my DAC can handle.
Independently they are silent, but the mix clearly produces audible distortion (my bat-ears go silent above 17.5kHz).

I'm certain there will be pressure from higher frequencies when reproduced cleanly, but none will positively impact the music. Most likely excessively loud ultrasonics will just lead to nausea!

 

[ElNino]

Well, I have conducted some experiments of my own, and I could clearly hear the distortions created by the ultrasonic tones I injected. My explanation: either the amplifier, or the speakers distort within the audible band because of ultrasonics.
I tested with two independent clean sine waves and their mix, 22350Hz (F10) and 42192kHz (E11), all within 384kHz sample rate as this is the highest my DAC can handle.
Independently they are silent, but the mix clearly produces audible distortion (my bat-ears go silent above 17.5kHz).

I've noticed this too, but like you, I'm not sure of the mechanism. It may not always be the amplifier (and IMHO it's probably not the speakers); it could also in some cases be the DAC. Ultrasonics perturb the behavior of sigma-delta modulators, which could increase distortion in the audible range in some cases. Also, it is possible that injected/artificial ultrasonics on top of (combined with) noise shaping ends up overloading the I/V stage. To some extent this is a moot issue with real music because most microphones are well down by 30-40kHz, but it reinforces the need for proper digital reconstruction filters.

 

[scott wurcer]

Thanks. I updated the link.

I forgot about the speaker manufacturer. Just to make it clear, it's the ESS Sabre DAC chip company that he's representing. The same one that created the chips for the highest-performing DACs Amir's measured.

I worked with Martin long before ESS, we met up at a conference just as they were rolling out their original designs and the team invited me to a listening session. I didn't have any transportation so I declined and I'm sorry for that. OTOH I suspect it would not have been carefully controlled and possibly even sighted.

I posted, in another thread, some data independently measured by someone I know to be very capable of an original ESS DAC reference design that clearly showed problematic artifacts. I never had a chance to discuss this with Martin since I had the impression that the phenomena being discussed escaped traditional tests. BTW I don't buy the audibility of ultrasonics, Dave Griesinger's talks on HF transducer IM show that ultrasonic signal content can easily be made audible by the wrong choice of speakers.

 

[NTK]

I've noticed this too, but like you, I'm not sure of the mechanism. It may not always be the amplifier (and IMHO it's probably not the speakers); it could also in some cases be the DAC. Ultrasonics perturb the behavior of sigma-delta modulators, which could increase distortion in the audible range in some cases. Also, it is possible that injected/artificial ultrasonics on top of (combined with) noise shaping ends up overloading the I/V stage. To some extent this is a moot issue with real music because most microphones are well down by 30-40kHz, but it reinforces the need for proper digital reconstruction filters.

I believe one possible explanation of what @Eirikur heard was byproducts of intermodulation distortion. When testing with a single frequency tone, the distortion products are in the higher harmonics, and are therefore (more) ultrasonic and inaudible.

But when there are two tones, say f1=22350 Hz, and f2=42192 Hz, one of the intermodulation products is 2*f1 - f2=2508 Hz, which is certainly within the audible range if the distortion level is high enough.

 

[boXem]

He left ESS and founded this company, using ESS DAC's of course
https://www.resonessencelabs.com/about-us/

And since the guy seem to have some sense of humor, one of their products is called Veritas (Akm Verita....).

 

[anmpr1]

He left ESS and founded this company, using ESS DAC's of course
https://www.resonessencelabs.com/about-us/

The video presentation (from a few years ago) discusses the advantage of analog volume controls. But these Resonessence (tm) devices use digital controls..., because (according to the FAQ) it's really too much of a techncial hassle to design a good analog control. For six thousand dollars? It would be interesting to compare them with the Benchmark products.

 

[anmpr1]

Well, I have conducted some experiments of my own, and I could clearly hear the distortions created by the ultrasonic tones I injected.

What you are hearing, then, are artifacts within the range of hearing bandwidth. That is different than saying you can hear anything at 50KHz, or thereabouts. I would have to review his talk to see if he was referring to distortions dumped into the audible range secondary to out of band stuff, or was talking about actually hearing things out of band.

I have to think that he meant the former, but when I listened (admittedly causally) it sounded like he was saying the latter. It sounded like he was saying there was some 'unknown' factor that couldn't be measured, but was actually 'heard' even though it was at 50KHz.

I don't want to put words into his mouth, or misrepresent what he was talking about. I'll re-review the presentation, pay closer attention, and report back. If someone else listened to it and can explain it, I'm happy to learn.

 

[anmpr1]

OK. What Mallison is referring to is a 'paper' (at 51:59 he mentions who wrote it but I couldn't make it out) where the author proposes we can hear 50KHz, not by our ears, but by, "shaking the bones of our heads, or something else..." This would be some sort of conductive sound transmission?

Then Mallison says a few things about filters removing artifacts, and transitions to, "You may not consciously be able to say whether you hear this sound or not [he is speaking about a 50KHz tone], but there still may be something about the hearing process which is somehow bringing those HF aspects into the experience of [unintelligible]."

Then he says "That's not a terribly sophisticated answer, is it?"

So, I really don't know what to say about it all. My impression is that he was just sort of going on and on with no clear idea of where he was going.

 

[gene_stl]

When I see an D/A for $6,000 I lose interest immediately. The guys' stock with me also drops.

I suppose I can't blame him for wanting to get rich but he should take the quip from Yve Saint Laurent who is reputed to have said, "When Madame comes to me in the Rolls-Royce , I go home on the Metro. But if she comes to me on the Metro, I go home in the Rolls-Royce."

The above statement about 50 kHz strays into the kind of audiophile territory that we don't like on this forum. The kind of implication that maybe you can hear it and maybe not. Sometimes that implies your ears and experience aren't good enough. Possibly there is some truth to that.

I was going to post on this thread because I had been looking for a similar YouToob video I watched in which the designer of Chord D/As made a lot of similar statements about stuff that was zillions of db down , but golden ears could hear them. When I looked at the OP links it came up on the side bar. The designer Rob Watts of Chord. I think this is the video:

While I was watching it I thought "kewl this guy knows what he is talking about" but then I started thinking that what he is really doing is trying to sell the buyer on inaudible improvements, of the type that our subjectivist friends like. Of course Chord products are so expensive that I am unlikely to ever experience them outside of a show or a friends house.

At the risk of starting trouble there are some other audio luminaries that I starting to suspect of similar offenses. jmo

I guess they all need to keep the lights on and get some kind of return on investment. But when I look at Amir's wonderful SINAD charts and look at the low correlation between price and performance I start thinking in terms of a lowest common denominator price wise for each piece of gear.

In a similar fashion the notion of inaudibilty of power amps is a topic of discussion with multiple sources of Class D amps that simply are better products than so many earlier ones.
This in not new to me , but I have only mentioned it to close friends. Back in the day when I built my four way multi amp system I wanted to get four inexpensive receivers that had SanKen hybrid power amps in them and see whether you could tell the difference between them and the fancy pants amps I used. I believed then and still do that you could not. Now others are saying the same thing about power amps, full range no less. The top two channels have been driven by Pioneer M-22 Class A amp/toasters. I am seriously considering cashing those out while the cashing is good. Before the audiophiles get wise. It does measure like a high grade preamp.

 

[pozz]

There are other things he mentioned besides ultrasonics, like the audibility of low-level noise behaviour, that I really can't understand.

 

[pozz]

When I see an D/A for $6,000 I lose interest immediately. The guys' stock with me also drops.

I suppose I can't blame him for wanting to get rich but he should take quip from Yve Saint Laurent who is reputed to have said, "When Madame comes to me in the Rolls-Royce , I go home on the Metro. But if she comes to me on the Metro, I go home in the Rolls-Royce."

The above statement about 50 kHz strays into the kind of audiophile territory that we don't like on this forum. The kind of implication that maybe you can hear it and maybe not.

I was going to post on this thread because I had been looking for a similar YouToob video I watched in which the designer of Chord D/As made a lot of similar statements about stuff that was zillions of db down , but golden ears could hear them. When I looked at the OP links it came up on the side bar. The designer Rob Watts of Chord. I think this is the video:

While I was watching it I thought "kewl this guy knows what he is talking about" but then I started thinking that what he is really doing is trying to sell the buyer on inaudible improvements, of the type that our subjectivist friends like. Of course Chord products are so expensive that I am unlikely to ever experience them outside of a show or a friends house.

At the risk of starting trouble there are some other audio luminaries that I starting to suspect of similar offenses. jmo

I thought of this talk too. In Mallinson's case there are less of the known audiophile incentives working. ESS is a much bigger deal than Chord. The only thing that makes sense to me is what I wrote above, that he didn't take psychoacoustics into account and was influenced in part by manufacturer engagement of audiophiles as a serious part of product design.

 

[gene_stl]

Mark Mallinson now seems to be with Resonessence. Again I don't blame anyone for trying to improve his position. (After all, gentlemen , we are not Communists!" )
Chord may even have lower priced DACs then Resonessence. I am not interested in those product matrices for the same reason I don't stop and lookylou at yacht and private jet dealerships.
As far as audiophile listening panels I like what Dr. Toole had to say about that as well as the graph that he or Sean Olive published on that subject.
I am always hunting for the sweet spot or the point of maximal bang for the buck.

 

[Eirikur]

OK. What Mallison is referring to is a 'paper' (at 51:59 he mentions who wrote it but I couldn't make it out) where the author proposes we can hear 50KHz, not by our ears, but by, "shaking the bones of our heads, or something else..." This would be some sort of conductive sound transmission.

Most likely this: Upper Limit of Frequency for Human Hearing (R. J. PUMPHREY) which unfortunately is not publicly available. Some of the content is discussed on stackexchange.

 

[gene_stl]

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