ESS THD ‘Hump’ Investigation

[Veri]

Same goes for benchmark AFAIK: they use their own 200kHz asrc

I believe the 9028Pro Benchmark utilises does not suffer from this problem in the first place iirc

 

[gvl]

As far as I can tell the just reviewed Okto DAC8 doesn't seem to have any external ASRC on it:

 

[peterq]

Once I have some free time, I will learn something about xmos coding and try to enable thd comp on tone board.

 

[peterq]

This hump has nothing to do with the I2C register setting and is a pure hardware issue.
If the layout of the I/V and LPF circuits is no problem, then only one or two key variables need to be adjusted to solve the problem.
But for work reasons, I can't disclose more details.

I don't familiar with hardware design, based on my understanding from Amirm's testing, the analog circuit of tone board is well engineered.

 

[HououinKyoma]

Mysterious Without disclosing anything confidential can you tell us whether there’s any chance of this getting fixed via a firmware update or is it only possible via a hardware revision?

Secondly, do you have an opinion concerning the audibility of the hump? I’m just wondering if this whole debate is purely academic if the hump is well below the threshold of audibility.

In short, only hardware revisions can remove this hump.
I didn't have time to try ABX before, but I simply compared the sound of two boards with or without hump.
The sound of the board without hump is obviously more natural.
When I have time, I will find a few different models of the PCBAs, adjust to the state of hump and no hump, and do some ABX.

 

[Veri]

As far as I can tell the just reviewed Okto DAC8 doesn't seem to have any external ASRC on it:

They use ES9028PRO just like the Benchmark. The 28 chip seems immune to it.

 

[johan]

THD compensation will not solve the issue . We will do some internal testing and will update

 

[johan]

I can confirm that ess sabre is aware of the issue. We are talking .

 

[bboris77]

In short, only hardware revisions can remove this hump.
I didn't have time to try ABX before, but I simply compared the sound of two boards with or without hump.
The sound of the board without hump is obviously more natural.
When I have time, I will find a few different models of the PCBAs, adjust to the state of hump and no hump, and do some ABX.

For transparency reasons, could you tell us what is the company you work for?

 

[HououinKyoma]

For transparency reasons, could you tell us what is the company you work for?

The company I work for is currently an authorized distributor of ESS.
Because of that, I provide technical support to some local audio equipment manufacturers.
As for the solution to remove this hump, I actually found it more than a year ago.
But these details are currently only available to our customers, I am not sure if this can be disclosed publicly.

 

[mansr]

The company I work for is currently an authorized distributor of ESS.
Because of that, I provide technical support to some local audio equipment manufacturers.
As for the solution to remove this hump, I actually found it more than a year ago.
But these details are currently only available to our customers, I am not sure if this can be disclosed publicly.

That's why I'd think twice before doing business with ESS. Any decent chip maker would have put that information in the data sheet or an application note publicly available on their website.

 

[amirm]

But these details are currently only available to our customers, I am not sure if this can be disclosed publicly.

I appreciate that but please convey to ESS that a lot of their implementers of their DAC are getting this wrong as a result of such secrecy. In the past, this didn't matter but now, we have put a huge spotlight on it so they need to change their strategy and address this. Perhaps they have noticed that some Chinese companies are already switching away to AKM. I see the term "ESS Hump" now used in other forums and pretty soon their brand will be hurt if they don't act and provide help to their manufacturers to remedy this.

And oh, they need to answer their emails when someone like me contacts them. I reached out to them twice I think and they would not even bother to answer my email.

 

[HououinKyoma]

I appreciate that but please convey to ESS that a lot of their implementers of their DAC are getting this wrong as a result of such secrecy. In the past, this didn't matter but now, we have put a huge spotlight on it so they need to change their strategy and address this. Perhaps they have noticed that some Chinese companies are already switching away to AKM. I see the term "ESS Hump" now used in other forums and pretty soon their brand will be hurt if they don't act and provide help to their manufacturers to remedy this.

And oh, they need to answer their emails when someone like me contacts them. I reached out to them twice I think and they would not even bother to answer my email.

Thank you for understanding that!
In fact, I am consulting the possibility of publicly disclosing these technical details.
If I can do this, I will post it as soon as possible.

 

[Veri]

Thank you for understanding that!
In fact, I am consulting the possibility of publicly disclosing these technical details.
If I can do this, I will post it as soon as possible.

Awesome !!

 

[Samoyed]

And, for those of us with the hump?

 

[Veri]

And, for those of us with the hump? View attachment 23805

Convince yourself it's not real

in all seriousness, whether it's audible is heavily contested.

 

[Samoyed]

Eyegor, Igor, it’s all the same, here....

 

[bboris77]

Convince yourself it's not real

in all seriousness, whether it's audible is heavily contested.

I have a feeling we will never have absolute proof whether it is audible or not. Perhaps some form of null test would serve this purpose but only if one of the manufacturers decides to issue a revised version of their DAC with the hump fixed. That way a test could be run comparing the exact same DAC minus the hump issue.

 

[johan]

Thank you for understanding that!
In fact, I am consulting the possibility of publicly disclosing these technical details.
If I can do this, I will post it as soon as possible.

I also have a NDA signed with ess sabre and its very clear that no proprietary information can be shared. However , this hump , is clearly not under NDA and it will benefit the whole community indeed to Ess Sabre.

I see no reason of why it should not be public.

 

[HououinKyoma]

To be honest, it is not too difficult to get rid of this hump.

1 First, properly handle the layout of the analog circuit. This ensures a low enough distortion and noise amplitude, even if there is now a hump, but the height of the hump will decrease as THD+n improves.

At the same time, the MCLK crystal oscillator with low phase noise will be helpful, after all, this will reduce THD+n overall. The clock division function built into the ESS DAC chip has a similar effect, because the same MCLK signal, after several times of frequency division by DFF, can really reduce the phase noise of the low frequency offset portion.

2 At this point, there are two key variables that need to be adjusted.
The first key variable is the op amp used by the I/V circuit. It is not recommended to use the AD797. Using it will cause a high probability of hump.
According to my previous experiments, it is more appropriate to use the OPA1611/1612 series. You can also try other models of op amps.
However, in this state, you should still find this hump.

Then you need to adjust the second key variable, the feedback resistor value of the I/V circuit.
The optimum value of the feedback resistor for an I/V circuit that avoids hump is not a fixed value, depending on the specific PCBA design.
Please fine tune this resistor value according to the line output amplitude you need.

After you find the optimum value of the feedback resistor for the board you are tuning, there are some other variables that need to be adjusted for optimal performance.
For example, the Miller-capacitance of the I/V circuit and the resistance value in series with it, and the voltage divider resistance of the I/V circuit virtual ground.

Below are some screenshots of a appnote for the ES9018K2M peripheral circuit calculations. Most of the variable values mentioned above can be referenced to this appnote.
However, the output impedance value (Rs) of each DAC analog output pin needs to be replaced according to the specific model of the ESS DAC you are using.