ESS THD ‘Hump’ Investigation

[amirm]

Yes, the same.

 

[gvl]

You said you can't share the details on their own solution, but one has to wonder why ESS provides no guidance on this matter.

 

[Veri]

You said you can't share the details on their own solution, but one has to wonder why ESS provides no guidance on this matter.

I imagine Topping found the issue with their own investigation/hard work and is probably unwilling to share with every single competitor.
ESS is aware and is probably looking to fix it in their newer products (fixing old ones means people don't need to buy the new revision...).
Always the same old story with hardware!

 

[amirm]

You said you can't share the details on their own solution, but one has to wonder why ESS provides no guidance on this matter.

One definitely wonders.....

 

[daftcombo]

Thanx.

 

[John_Siau]

It may be the case. We need to enable the feature and see if it makes a difference.

One key to eliminating the THD hump is the reference voltage regulator. The reference voltage needs to be driven by a ridiculously low impedance and with very low noise. We had to build a special circuit to accomplish this task.

The second key to eliminating the THD hump is to apply a differential amplifier after the balanced output of the chip. This diff amp is often omitted in order to reduce costs.

As for the 2nd and 3rd harmonic compensation, these have no impact on the hump but are quite useful in trimming out the THD at all output levels. With a 1 kHz test tone, we can dial the 2nd harmonic distortion down to about -140 dB, and the 3rd down to about -135 dB. This trim can produce a 20 dB reduction in THD if the analog output stage is good enough to support it.

The other thing to remember is that there are many ways to implement an ESS dac. There are lots of cost-saving shortcuts and the cookbook examples on the data sheet have some shortcomings. There are also a few hidden quirks that can be overcome if you spend enough time with the chip.

The ESS series of D/A converters are the best available and are absolutely remarkable when tweaked to full performance. This exploration and tweaking takes time. Many manufacturers are more inclined to follow the cookbooks, especially the lower cost example circuits.

 

[johan]

Hi John

first I would like to thx you for sharing your vast knowledge with us.

Previously , you advocated a 2n cap across the phases of ess9028 pre IV stage. Do you think that cap is still required ? I assume it has no influence on the hump..

 

[John_Siau]

I don't recall anything about a 2 nF cap. One 560 pF NPO cap from each leg to ground works well for keeping switching noise out of the I-V buffer amps. Switching noise can cause distortion in the buffer amplifiers.

 

[johan]

Sorry I was assuming you were talking about 2nf.
I will test both 560pf and RC 560pf + R to see it can further bring down the noise and distortion on ess9038q2m

I appreciate your time and tips.

 

[Music1969]

Hi all,

Just looking at @amirm's IMD measurements for the Pro-Ject Pre Box S2 DAC below.

This DAC uses synchronous clocking instead of the Sabre's ASRC.

I see small evidence of the 'ESS hump' below?

In trying to understand what this hump means in practical (listening) sense, is this hump (not just with this DAC but other ESS DACs also) only an issue when using digital volume control ?

i.e. the infamous ESS hump has no effect on performance if using the DAC at 0dB volume (no digital volume attenuation)?

Thanks in advance

 

[amirm]

I see small evidence of the 'ESS hump' below?

It is there. Higher noise/distortion masks it to some extent in S2.

 

[Music1969]

It is there. Higher noise/distortion masks it to some extent in S2.

Thanks amir.

But as above, is this hump (not just with this DAC but other ESS DACs also) only an issue when using digital volume control ?

i.e. the general ESS hump is not an issue if using the DAC at 0dB volume, i.e. with no digital volume attenuation?

 

[JohnYang1997]

Thanks amir.

But as above, is this hump (not just with this DAC but other ESS DACs also) only an issue when using digital volume control ?

i.e. the general ESS hump is not an issue if using the DAC at 0dB volume, i.e. with no digital volume attenuation?

It's a complicated excitation. IMD hump is just one way to show it. It can be go as high as -80db for normal music. Though it may not be audible, i consider it as big drawback to many ess based implementations.

 

[HououinKyoma]

I imagine Topping found the issue with their own investigation/hard work and is probably unwilling to share with every single competitor.
ESS is aware and is probably looking to fix it in their newer products (fixing old ones means people don't need to buy the new revision...).
Always the same old story with hardware!

Obviously, you don't understand the nature of this issue.
This issue has nothing to do with the chip reversions, only related to the specific circuit design.

 

[JohnYang1997]

Obviously, you don't understand the nature of this issue.
This issue has nothing to do with the chip reversions, only related to the specific circuit design.

Yeah yeah yeah, you understand. Then fucking tell us how to solve this problem. Not those pcb layout and resistor bullshit. You are like a 3 year old protecting his own toy than helpful. Moreover how come it's not a design issue if the issue is found later on after production of implementations and is being fixed after investigation.

 

[amirm]

Thanks amir.

But as above, is this hump (not just with this DAC but other ESS DACs also) only an issue when using digital volume control ?

No. Lowering the volume makes it visible in tests but the distortion is there all the time at certain levels being produced. You have go get the balance of noise level and output level to have it show up in graphs.

 

[amirm]

This issue has nothing to do with the chip reversions, only related to the specific circuit design.

That is true but chip design has led to difficulty in implementation this way. So if ESS were listening, they would make sure their next generation products don't require this type of external care to avoid distortion.

 

[daftcombo]

i.e. the infamous ESS hump has no effect on performance if using the DAC at 0dB volume (no digital volume attenuation)?

In your music, notes are more or less loud compared to the max level at 0 dBFS. The weakest notes can be impacted by the hump.

 

[kn0ppers]

@JohnYang1997: Is that vocabulary really necessary?

Ontopic:

Really interesting thread, many interesting links, a lot of slides to go through.

Interested to see how well the Khadas Toneboard XLR will measure up.

 

[JohnYang1997]

@JohnYang1997: Is that vocabulary really necessary?

Ontopic:

Really interesting thread, many interesting links, a lot of slides to go through.

Interested to see how well the Khadas Toneboard XLR will measure up.

To me yes. I knew I was being rude. But I mean it and I don't regret it.