• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Best spec ADC Chip currently.... ??

I have seen people on Youtube (and I assume in the music business) who like to outfit their recording setups with really good performing gear (performance here being a mix of the sort of performance testing we do here, but also equally as primary, the latency figure). Also, with the amount of PC gaming and internet audio communication going on, it's almost detestable to hear someone with garbage quality audio these days.
For those purposes, a cheap Behringer kit will do just fine.

The nice thing about going something like the ESS route, is it perhaps demonstrates a sort of attention level the designer is paying. They're willing to put up with this garbage executive or corporate policy of a company, with horrible hardware documentation that is guarded for no sane reason. To then put out a product that performs better than most, is the sort of commendable thing people appreciate over someone doing the same old, drag & drop of an off the shelf T.I. or AKM ADC chip that not many folks find to be interesting if the product is trying to fill some other role outside of basic interfaces.
The trouble with poorly documented components is that you easily get surprises like the "ESS hump" that nobody seems to quite understand. Personally, I'd prefer a few dB lower SINAD and predictable behaviour.
 
For those purposes, a cheap Behringer kit will do just fine.
Possibly for even less than that. Though there are two types of folks it seems when they enter a new field to purchase something. People looking to get the most bang for buck with the least buck also spent to fulfill a singular function and nothing else, and people looking to get something a little better than the rest for the sake of "headroom" in their heads for sanity. If marketing is worth anything it's for those people in order to show them "yes, that's us here, we did a bit more work than the rest to get you what you need especially if you have any propensity for purchasing things above your needs". Whether that's better warranties, better build quality, better attention to detail, better performance, etc..


The trouble with poorly documented components is that you easily get surprises like the "ESS hump" that nobody seems to quite understand. Personally, I'd prefer a few dB lower SINAD and predictable behaviour.

Preaching to the choir, runs the gamut from poor documentation in software, as well as hardware. That's the part that confuses me about this company, and how they figure leaving out any smaller players in the desert for themselves is somehow a good idea. But like I mentioned prior, those companies that do decide to go into the desert, and come out alive.. I feel is something people would commend (at least in these sorts of circles that look at products with a focus on getting something a bit above basic needs).

Unless we presume IXV to be a sado masochist, surly you would find it plausible he's going the ESS route for some appreciable reason. The reason I assume is simply due to his desire to squeeze every last drop of performance out of the products he puts out along with some non-standard functionality (like his DACs which give you access to parts no USB sized DACs do). And also do it at a price level where you're taken aback.
 
The trouble with poorly documented components is that you easily get surprises like the "ESS hump" that nobody seems to quite understand. Personally, I'd prefer a few dB lower SINAD and predictable behaviour.

As much as I dislike how ESS operates in some respects, I'm not sure they are alone with errata or strange behavior. It's not like the ESS products were Martin Mallinson's first audio converters. I would not be surprised if similar issues, but perhaps less noticeable to an AP, have gone unnoticed from the likes of AKM, Cirrus, and TI.

On the other hand, they seem to be alone in releasing an ADC that was apparently so bad they had to remove all traces of it from their website and pretend it never existed (ES9112).
 
No relay clicks in XXI century! ;) BTW, in Cosmos ADC implemented SE version(cost-saving), in case if that idea will be implemented in differential topology, the simulation shows simply zero THD(didn't test yet). It works with 600ohm impedance so THD+N is also WOW, actually, that's why I not going with MDACs with 10kOhm ladders.
 
As much as I dislike how ESS operates in some respects, I'm not sure they are alone with errata or strange behavior. It's not like the ESS products were Martin Mallinson's first audio converters. I would not be surprised if similar issues, but perhaps less noticeable to an AP, have gone unnoticed from the likes of AKM, Cirrus, and TI.
Of course any product of this complexity can have errors. The difference is that with ESS you don't even know how it's _supposed_ to work.

On the other hand, they seem to be alone in releasing an ADC that was apparently so bad they had to remove all traces of it from their website and pretend it never existed (ES9112).
They also used to make modem chips. And image sensors. Not a trace of those on their website either.
 
  • Like
Reactions: Tks
No relay clicks in XXI century! ;) BTW, in Cosmos ADC implemented SE version(cost-saving), in case if that idea will be implemented in differential topology, the simulation shows simply zero THD(didn't test yet). It works with 600ohm impedance so THD+N is also WOW, actually, that's why I not going with MDACs with 10kOhm ladders.
Anything related to this?
https://www.ti.com/lit/an/slyt612/slyt612.pdf
 
Well, that´s kind of noble understatement! Substituting relays with a
bunch of N-ch MOSFETs, maintaining a THD < -120 dBc even for signal
voltages around 10 Vrms, is certainly not a trivial task, especially
because you might run into Vgs max (usually +/-20 V) issues at these
high levels!

As a matter of fact, nearly all conventional analog FET-switches "of
the shelf" are composed of parallel p- and n-FETs with static gate
voltages, yielding the typical "bathtub" Rds-on curves, causing quite
a lot of distortion when they have to conduct more than just a couple of
µA. The only exception I know of is the MAX14756 with an almost
perfectly flat Rds-on vs. signal voltage characteristics, yielding
very low THD also for high signal amplitudes.

But this IC is of course too expensive to realize 10 different gains,
so your solution to do this with your own discrete solid state solution
is admirable!

And I agree, relays are outdated in an ever shrinking electronic
world. And not necessarily better performing than a well made solution
with MOSFETs. Good luck with this project with the ES9842!

By the way, did you notice? The "2 Vrms input" (differential) in the
Functional Block diagram are just another blunder in the data sheet.
In reality, the modulators surprisingly accept quite higher voltage
before hitting 0 dBFS.
 
Well, that´s kind of noble understatement! Substituting relays with a
bunch of N-ch MOSFETs, maintaining a THD < -120 dBc even for signal
voltages around 10 Vrms, is certainly not a trivial task, especially
because you might run into Vgs max (usually +/-20 V) issues at these
high levels!

As a matter of fact, nearly all conventional analog FET-switches "of
the shelf" are composed of parallel p- and n-FETs with static gate
voltages, yielding the typical "bathtub" Rds-on curves, causing quite
a lot of distortion when they have to conduct more than just a couple of
µA. The only exception I know of is the MAX14756 with an almost
perfectly flat Rds-on vs. signal voltage characteristics, yielding
very low THD also for high signal amplitudes.

But this IC is of course too expensive to realize 10 different gains,
so your solution to do this with your own discrete solid state solution
is admirable!

And I agree, relays are outdated in an ever shrinking electronic
world. And not necessarily better performing than a well made solution
with MOSFETs. Good luck with this project with the ES9842!

By the way, did you notice? The "2 Vrms input" (differential) in the
Functional Block diagram are just another blunder in the data sheet.
In reality, the modulators surprisingly accept quite higher voltage
before hitting 0 dBFS.

Correct. Off the top of my head I suspect something like the Baxandall attenuator circuit or some inverting topology that minimizes the non-linearity of the devices.
 
20V rated MOSFETs have extremely suitable parameters for such attenuators. Rds_on 20mOhm in SOT23-6, very low parasitic capacitance, about1.5V Vgs threshold, +/-20V gate tolerate, and they are cheap as dirt. Such a bouquet of properties makes it easy to use 600ohm impedance around opamps or even less. I was stopped to go with 300ohm due to a current limit of USB power. Currently, Cosmos drains 590mA from 5V at its max @ 32/384 10.5Vrms on both channels, hence, legally it can run only on USB3 with a 900mA limit. In fact, 99% of PC's USB2 able to feed >1A, and actually, today almost all USB ports are USB3 with an actual limit around 3A.
The main lack of my attenuator are 15 MOSFETs per channel, including voltage shifters 3.3 to +15V and -19.7V. A chip version with internal 1uA charge-pumps 3.3V to +/-20V + I2C could be much more attractive.
 
Last edited:
Agreed on the attractiveness of Double-MOSFETs due to their low price
and RdsOn! Unlike typical analog MUXs, they can be used as real
current switches without causing terrible distortion.

Instead of SOT23-6, you´ll also find a lot of suitable candidates in
SOT363 package with additional drain pads, which do not need to be
included in the foot print on the PCB because they will not even get
lukewarm in this application.

As for Cgd, Cgs and Cds, however, they are not so low anymore for the
more "beefy" devices and you might experiences some leakage at 20 kHz.
To get truly high isolation there, you would need a T-switch.

Also, those parasitic capacitances are voltage-dependent, so distortion
might rise noticeably - at least not be as good as a mechanical relay.
So what looks perfect at 1 kHz could look a little bit worse at high
frequencies.

And yes, having all this together with the necessary level shifters integrated
in an IC for maybe < 1 $ would be nice - apparently, this idea has not
been adopted yet by the big semiconductor companies.
 
Finally, I've got the application info for the ADC ES9822 chip(BTW, I spent 38 days on that! ESS has legendary bad docs and slow response), and I have to confirm the ADC if not perfect then the best at least. Dynamic Range without front-end 125.2db(A) in 2ch mode, 128.3db(A) mono, with the front-end 1db worse i.e. 124/127db(A). THD+N aka sinad -117.2db 2ch and 118.4db in mono mode including the front-end. Very interesting to me looks THD compensation approach in that chip, there are a few LUTs(1 for each channel) 64x16bits where you can implement a complex form of the non-linearity. For example, I did successfully compensate the 5th harmonic of the ADC distortions, it wasn't seriously needed, just like an experiment. Practically, I have implemented low distortion ADC(<-130db harmonics level at -0.5dbfs at all of 8 gain ranges 1.8..10Vrms). In fact, 90x50mm 2 layers PCBA in mono mode outperforms my AP SYS2522 in terms of THD+N/sinad(or equal in 2ch mode!). With a simple 1kHz notch analog filter it outperforms APx555 as well(no surprise actually). The future is already here ;)


Hi all, I am one of the ESS folks. Love the feedback!

I am happy to try and help with specific issues with the ADCs if you have any, let me know.

Glad you got this one working those DNR specs are really something!
 
LagoRago, I just one second ago got some corrected register number from Joe, and got "clipping" and "signal" LEDs working right ))
Guys, why your docs so crazy badly prepared? I had really trivial problems with your ADC, it could be fixed even before the datasheet released, or 5 min after that, or 1 day after I asked you, right? You know, if your chips were just junk or average performance level, who'll care about correct docs for that? But if you doing really high-performance chips why a datasheet so awfully poor?
BTW, after 5 times I resoldered the same ES9822 to the next PCB iteration, I've got almost 1db worse SNR/DR, I hope the chip itself degrades with so many times of resoldering..
 
Hi all, I am one of the ESS folks. Love the feedback!

I am happy to try and help with specific issues with the ADCs if you have any, let me know.

Glad you got this one working those DNR specs are really something!

Btw if you are employed by ESS, do contact the forum administrator @amirm to get you a label reserved for audio company employees (so that you don't have to introduce yourself formally too many times). We really appreciate someone from ESS has decided to turn up on this forum as you guys are known for making pretty much the best DAC chips on the market performance wise.
 
For quite a long time, I reckon. Regarding DACs, even the venerable
ES9018S, with which ESS started its success in the high end audio
business over a decade ago, still outperforms all competitor´s DACs.
Now the ES98x2 seems to be the next hit, after the first trial and
failure with the deprecated ES9102.

Once all developers get around the buggy data sheet and manage to
squeeze out the essence of this new ADC (the key specs are indeed
correct!), I am more worried about its availability, given the fact
that a lot of folks now migrate from AKM to ESS due to the
unavailability of AKM converters after the blaze in their fab 2 in
Nobeoka.

Once AKM gets back on the pitch, maybe the AK559x (under development) will
be a contender to beat the ES98x2 in terms of DR, but this ADC is no
low power design (it will be even more voracious than the good old
AK5394A), nor does it feature a flat noise floor up to 200 kHz, which
imho is the most impressive feature of the ES98x2.
 
Lightrock, well, I tend to say it is flat up to 80-100k yet.
2021-03-20_19-36-52.png
 
Back
Top Bottom