• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

How to Fix ESS Hump on SGD1 and LA-QXD1

Azeia

Active Member
Joined
Jun 13, 2019
Messages
123
Likes
292
#61
This is pretty cool, as others have said, I also really appreciate the spirit of openness here, as most other vendors have been keeping their ESS hump solutions secret; this might be the endgame DAC I've been looking for.

Will it eventually be available on Amazon's Canadian (amazon.ca) store as well?
 

Herbert

Active Member
Joined
Nov 26, 2018
Messages
191
Likes
167
#62
In some ways my question might not be fair, as Ben1987 put a lot of energy and good spirit
into the fix. But the pcb of both DACs are physically too big for my needs and
therefor I am stuck to the Khadas tone board.
In the moment of writing, I do not have Khadas´ published schematics with me and do not know what
opamps were used. But can this fix be done with the Khadas Tone Board as well?
All the best,
Herbert
 

Herbert

Active Member
Joined
Nov 26, 2018
Messages
191
Likes
167
#63
Hm-no one to do the math? I am not deep enough into electronics to calculate the ratio of the changed
values. Above Bens calculations, below the tone board unaltered...

hump gone.jpg
Khadas.jpg
 
Joined
Feb 9, 2018
Messages
286
Likes
193
#64
This hump seems to appear consistently on many ESS implementations.

Have ESS not discussed/addressed it in their application notes?

I recall trying to get their datasheet/appnote, only to discover they cannot be freely downloaded; that's an unusual stance for a semiconductor manufacturer. Not unprecedented, but unusual.
 

somebodyelse

Addicted to Fun and Learning
Joined
Dec 5, 2018
Messages
598
Likes
381
#65
The datasheets aren't freely downloadable, but some have leaked. The hump isn't addressed in the ones I've seen, and IIRC some manufacturers have said it isn't in the datasheets or known by their contacts at ESS (on khadas forum, and i think another source? Not sure...). This is consistent with the large number of manufacturers having a similar problem. It's possible there's a mode detailed document that they reserve for a select subset of customers under NDA, or that a select few have worked it out themselves. @Ben1987 's explanation here, and the somewhat guarded posts of another user who may be under NDA (don't have it to hand - in the ess hump thread) cover a component selection approach, while @IVX does it with config options so long as other analog requirements are met.
 
Joined
Feb 9, 2018
Messages
286
Likes
193
#66
Manufacturers usually go out of their way to allow their customers to get the best performance out of their products (the likes of AD, TI, BB, NS, LT, etc in the 'good old days' produced some excellent appnotes). I cannot understand why ESS would do otherwise. But then I don't really understand their datasheet policy, either.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
25,109
Likes
52,196
Location
Seattle Area
#67
Manufacturers usually go out of their way to allow their customers to get the best performance out of their products (the likes of AD, TI, BB, NS, LT, etc in the 'good old days' produced some excellent appnotes). I cannot understand why ESS would do otherwise. But then I don't really understand their datasheet policy, either.
ESS has not been on any help. Whether they help much larger manufacturers is unknown. But despite this issue coming to their attention last year, no effort was spent on their part to help resolve it.

They also have ignored my two separate outreach to them. Not even a simply reply. Nothing.
 

scott wurcer

Addicted to Fun and Learning
Audio Luminary
Technical Expert
Joined
Apr 24, 2019
Messages
794
Likes
1,250
#68
ESS has not been on any help. Whether they help much larger manufacturers is unknown. But despite this issue coming to their attention last year, no effort was spent on their part to help resolve it.

They also have ignored my two separate outreach to them. Not even a simply reply. Nothing.
I don't understand this issue at all, I knew Martin for years before he left to help form ESS and they invited me to an early listening session. This never came up, though Barrie Gilbert guessed at one of the things they were doing only to get a few smiles. Four+ years ago ESS invited me to tour the five (four?) major Chinese cell phone manufacturers since they all had an initiative to add a separate high quality audio chain with headphone driver. It was an opportunity to sell some customer premise ADSL drivers that had waning sales. This never came up in any context, even though I posted a plot here on how obvious the problem could be.
 

solderdude

Major Contributor
Joined
Jul 21, 2018
Messages
4,325
Likes
8,022
Location
The Neverlands
#69
Hm-no one to do the math? I am not deep enough into electronics to calculate the ratio of the changed
values. Above Bens calculations, below the tone board unaltered...

View attachment 47773 View attachment 47774
It is the absolute values (as well as the ratio) that is of importance. Bens schematic also has a second opamp stage which appears to have not been drawn as I don't think the TB has a 1.6k output R.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
25,109
Likes
52,196
Location
Seattle Area
#70
I don't understand this issue at all, I knew Martin for years before he left to help form ESS and they invited me to an early listening session. This never came up, though Barrie Gilbert guessed at one of the things they were doing only to get a few smiles. Four+ years ago ESS invited me to tour the five (four?) major Chinese cell phone manufacturers since they all had an initiative to add a separate high quality audio chain with headphone driver. It was an opportunity to sell some customer premise ADSL drivers that had waning sales. This never came up in any context, even though I posted a plot here on how obvious the problem could be.
Doesn't look like anyone had measured and identified this issue until I ran into it. This may very well have included ESS.
 
Joined
Feb 9, 2018
Messages
286
Likes
193
#71
You would like to think that chips are tested properly. But then I've found three bugs in chips, unknown to the manufacturer, in a 35 year career. One of them took me over a month to pin down and develop a workaround. The manufacturer at least was helpful, and confirmed my diagnosis in simulation, and issued an errata note. Another bug also took a while track down, and turned out to be an undocumented edge sensitivity.

When you have the details of your own design in front of you, diagnosis of bugs usually isn't too hard; you can trace the circuit behaviour. When you have a black box, and have to probe the box by guesswork, it's a lot harder...
 

solderdude

Major Contributor
Joined
Jul 21, 2018
Messages
4,325
Likes
8,022
Location
The Neverlands
#72
There are manufacturers that used the same chip and did not have the IMD hump.
The hump is not caused by the chip but incorrect circuitry after the DAC chip.
When the chip is loaded correctly the hump is not there, assuming during development they used the correct load post filtering it would not have shown because it was not there to begin with, even when they measured (and assume they did) for IMD versus output level.
 

Pluto

Active Member
Forum Donor
Joined
Sep 2, 2018
Messages
256
Likes
454
Location
Harrow, UK
#73
Surely, if the load is critical it could reasonably be expected for the data sheet to flag that point and offer appropriate guidance in their reference design. So far, the evidence is suggestive of either...
i) they really did not know, and this caught them napping
ii) they knew but chose to keep quiet

Other conspiratorial possibilities exist.
 
Joined
Feb 22, 2020
Messages
39
Likes
7
#76
The SGD-1 has just gone to the top of my list. The purchase link on Aliexpress shows the item in unavailable. It is either due to items being withdrawn pending upgrades (or it could be Covid-19).

Will the LA-QXD1 be similarly upgrade? There is no purchase link for this.
 

Herbert

Active Member
Joined
Nov 26, 2018
Messages
191
Likes
167
#78
It is the absolute values (as well as the ratio) that is of importance. Bens schematic also has a second opamp stage which appears to have not been drawn as I don't think the TB has a 1.6k output R.
I assume in Ben´s drawing the altered values in the second stage (between 7-6 and 1-5/2/3 are for correcting the output volume that has been altered in the I/V stage before? I do know almost nothing about electronics but as far as I understand the DAC only "sees" the load of the first opamp and this is where the hump is created…? So I assume in the Khadas schematic at least R45/R46 // R48/49 have to be altered as well to compensate for changes in volume?
BTW I whish and hope for the chinese people all the best...
 

JimB

Active Member
Forum Donor
Joined
Sep 19, 2019
Messages
283
Likes
174
#79
You would like to think that chips are tested properly. But then I've found three bugs in chips, unknown to the manufacturer, in a 35 year career. One of them took me over a month to pin down and develop a workaround. The manufacturer at least was helpful, and confirmed my diagnosis in simulation, and issued an errata note. Another bug also took a while track down, and turned out to be an undocumented edge sensitivity. ...
I know, right? I found this twice - once with a die shrink of a counter chip, and once with an ADC. In both cases, the manufacturers did not know about the problems, but did own them once documented. They did not fix the issues, but they did call them out in subsequent datasheets.
 

Similar threads

Top Bottom