Measurements: "ESS Hump" revisited (Khadas Tone Board V1.3)

[solderdude]

So circuit and opamp depending solution then.
@Ben1987 also solved it but in a different way (different circuit)

Your C4 makes a lot of sense.
What are the results with C4 but with the original R2/R3 ?

 

[KSTR]

Bravo! A lot of work but a simple fix. If you need an unaltered stock board for reference,
I have a second Khadas board!
Question:
The choice of 220pF is not sampling frequency dependend, as the DAC puts out
the same sampling rate no matter the source?
I think, a normal 220pF capacitor with a width 5mm betwinn pins might fit as well.
It just has to be glied on before soldering...

Yep, for this design (with ASRC engaged) at the final output we always see a fixed and high sample rate (clock of the final DAC stage) at many MHz with a lot of glitch energy.
Using a leaded cap could work if it tiny and one is extremely careful. The solder joints to the SMT parts and pads are very fragile and any stress on them easily (but invisibly) cracks the components.

 

[KSTR]

So circuit and opamp depending solution then.
@Ben1987 also solved it but in a different way (different circuit)

Your C4 makes a lot of sense.
What are the results with C4 but with the original R2/R3 ?

With 100R in series results were generally worse with C4 installed. Differential noise gain shows heavy peaking at 10Mhz (1/ß zero shifted downwards) in the sim.

I'll double check a few more values and combinations... one last round or two, later this week. I hope to reduce the high level high frequency degradation so that it's better than stock everywhere.

 

[Herbert]

THIS IS GREAT! I never expected that the Khadas board could be bettered that well.
As far as I remember, one user did not succeed and at the end fried his board
and things looked much more complicated...
Do C4 and both R10 resistors work like a low pass filter ...?

 

[Matias]

Awesome! Now the manufacturer should implement the improvements in a new version and release it to all customers.
Great job.

 

[tuga]

Awesome! Now the manufacturer should implement the improvements in a new version and release it to all customers.
Great job.

And pay for the work.

 

[Herbert]

Well - they could at least send KSTR some boards for free!!
But very interesting that the idea how to cancel the glitch energy
is at least 23 years old (The AD1853 was released in 1999).
I assume the Khadas engineers were mostly kids back then...
Knowledge seems to be lost faster than we think...

 

[KSTR]

You know, this is basically a workaround for those less than stellar RT6862 OpAmps used which are tough to replace on the board. With faster types we don't seem to need the "glitch killer" while the infamous 100R will be often reduced to like 10R.

To me it looks like this 100R resistor has always been an "Angst" resistor which is required only for certain OpAmps (like the AD797) and which then got copied as it is outlined in ESS application notes and evaluation boards.

WesionTek has released the Tone2 Pro where the problem appears to be fixed (using OPA1612). I don't think they would do another iteration of the Tone1.
For us customers we have a selection of pretty much stellar DACs at low cost today to choose from.

 

[noi1996]

Great and very informative!
I did several simulations with the same circuit configuration as the Soncoz SGD1, and I think the 10R and differential input capacitors are effective even with a fast and stable op-amp like the OPA828. (The capacitor value could be lowered.)
This circuit would have to pass the signal completely up to 100 kHz in the audio band and suppress the noise at 100 MHz, which is the internal operating frequency of the ES9038.
It will need to be filtered by small CR or ferrite beads.

 

[KSTR]

Great and very informative!
I did several simulations with the same circuit configuration as the Soncoz SGD1, and I think the 10R and differential input capacitors are effective even with a fast and stable op-amp like the OPA828. (The capacitor value could be lowered.)

I might try that with the D10B as it is much more modding-frienly than the KTB. I'm also inclined to think than when that "normal class" RT6862 is having such problems then better opamps will be better but not fully immune to those very fast glitches.

 

[mansr]

You know, this is basically a workaround for those less than stellar RT6862 OpAmps used which are tough to replace on the board.

Are they pin compatible with something better? If so, replacing them isn't hard if you have the right tools.

 

[Herbert]

Check yourself,

http://www.semiteh.com/admin/upload/semiteh1-3812595.pdf

only datasheet available is of a SRT6863

But this thread sheds some light:

New face in op amp industry

I was investigating Khadas tone board and I found out that they used RT6862D and RT6863D at amplifying stage. There is nothing about those chips on internet and even the distributers like Mouser and Digichip have no information about them. Since those chips show superb performance as shown on...

www.audiosciencereview.com

 

[Grooved]

Are they pin compatible with something better? If so, replacing them isn't hard if you have the right tools.

You can put OPA1612 instead of both RT6862, they are in WSON 3x3 package too if I'm not wrong
@KSTR , would it be interesting in this case to replace the RT6863 too ? or does it has no impact on the problem ?
I don't see a compatible TI chip in DFN-10.

 

[KSTR]

@Grooved, Yep, OPA1612 is one of few good OPs in DFN-8.

With them, ideally, the distortion levels should be similar to, for example, Topping D10 Balanced. I don't know how much impact the standard subtractor circuit (RT6863D) has on the result... the D10B doesn't use one.

@mansr, reworking those OPs, with that thermal pad and all, I would consider expert level. I'll might give it a try with my own board, first. I haven't dealt with replacing DFNs before...

 

[Grooved]

@Grooved...Topping D10 Balanced. I don't know how much impact the standard subtractor circuit (RT6863D) has on the result... the D10B doesn't use one

And they did not keep it on the Tone2 Pro too ( 4 OPA1612 instead of 1xRT6863 + 2xRT6862)

I've just checked and the only DNF-10 I see available to replace the RT6863 are Microchip or Analog Devices

 

[mansr]

@mansr, reworking those OPs, with that thermal pad and all, I would consider expert level. I'll might give it a try with my own board, first. I haven't dealt with replacing DFNs before...

I usually use an IR heater below the PCB and apply hot air onto the chip from above. It's obviously impossible to do with a soldering iron.

 

[Herbert]

Is it DNF-10? Because on Photos it looks like pins,
but it is very likely the solder pads...

 

[mansr]

It's DFN. The "pins" you see are just the PCB pads extending beyond the chip edge with some solder on them.

 

[Herbert]

It's DFN. The "pins" you see are just the PCB pads extending beyond the chip edge with some solder on them.

So if OPA1612 is not available in DFN-10, what are the pin-identical alternatives to replace RT6863/62?
I am sure that KSTRs solution will measure extremely well with the original RT6862, but just to push
the envelope a tad further.
Is any of the PCB pads extending beyond the chip edgde connected to the ground plane
the chip? This could male at least removal a tad easier, as the solder could be directly heated.
I also guess, a shop that repairs smartphones and laptops could do the replacement as well...

 

[mansr]

So if OPA1612 is not available in DFN-10, what are the pin-identical alternatives to replace RT6863/62?

The RT6862 apparently has the same DFN-8 package as many other opamps, including OPA1612. 10-pin opamp packages are less common, and the extra pins don't necessarily have the same functions.

Is any of the PCB pads extending beyond the chip edgde connected to the ground plane
the chip? This could male at least removal a tad easier, as the solder could be directly heated.

It can't be done with a soldering iron. You need a hot air rework station. Then it's easy.