ESS THD ‘Hump’ Investigation

[JohnYang1997]

This is a plot of an ESS DAC design exhibiting the IMD hump. I don't see any issue with concluding there is some kind of pathological behavior. I also don't see the reasoning about how this behavior is difficult to measure, etc.

9018s had a completely different issue and that's impossible to fix with circuit design. 9028 and 9038 fixed the issue but have a different issue which here we are, the imd hump. If you look at some measurements of 9018s, many don't have imd hump.

 

[amirm]

Anyway to fix the hump in the old D50 (non S) version? . I was so happy with with my D50 and THX AAA setup before reading this thread

No. The fix requires hardware change. The only solution is to sell and buy the D50s.

 

[amjosh]

No. The fix requires hardware change. The only solution is to sell and buy the D50s.

Hm.. thought so. I am reaching out to Topping, if they could help with upgrade with reasonable difference. It's hard to sell these niche products.

 

[confucius_zero]

So... is it audible?

 

[eliash]

9018s had a completely different issue and that's impossible to fix with circuit design. 9028 and 9038 fixed the issue but have a different issue which here we are, the imd hump. If you look at some measurements of 9018s, many don't have imd hump.

Since I am using the 9018s (8 channels version in native stereo mode - no uC), it would be interesting to learn more about other issues than the ones I found out before https://www.audiosciencereview.com/...ess-thd-‘hump’-investigation.5752/post-180723 any hints, links?

 

[raif71]

Since a knowledgeable and practically experienced corona seems to be assembled here, I take the heart to set a small interupt on the new ESS Sabre issues and ask 3 questions about the well known ESS9018 (8 channel chip), which I would like to finally understand as an electronics engineer.
I am using a kind of chinese-based reference design (Weiliang board) in the default hardware mode (4 internal channels per stereo channel, no uC control), driven by a separate CM6631a box via S/PDIF (because of mains ground isolation).

1 - The first strange thing I found was that the equipped NE5532 opamps for the IV conversion somehow hooked up the DAC chip, causing the least significant bits sound distorted (bits 20-24?).
Changing the opamps to LME49720 somehow solved the issue...why (anti-parallel protection diodes in the opamp input, wrong power-up sequence)?

2 - This specific design lacks a proper AVCC supply. The designer claims to have implemented low noise voltage regulators, but they are inferior by datasheet, compared to the ESS reference design, utilising low noise opamps. To confirm it, I created DC test signals files at +/- half max. output (incl. slow ramped up/down).
This increased the inaudible idle noise level to clearly audible levels.
Shunting the AVCC regulator outputs with 3x 470uF low ESR caps reduced the noise significantly, but still audible with the ear next to the speaker.
My question is whether or to what degree such noise is visible in single or multitone SNR measurements as displayed here in the forum?

3 - Running the Sabre chip at the common 100MHz clock always showed a strange behaviour at 96KHz input sampling rate. After power up or HW reset or previously played 192KHz files, there are always persistent clicks (every few seconds, mainly audible in quiet sections). Playing a file with any lesser sampling rate than 96KHz (e.g. <=88KHz) clears this situation, playing a 96KHz file afterwards is perfect.
As far as I could verify, overclocking the chip at 109.25MHz (because 88KHz plays perfect - TCXOs available at a reasonable price from China), solved the issue. 96KHz now plays always click-free. Looks like the internal sampling rate detector has a problem at exactly 100MHz clock rate...
My question is if anybody here has observed this behaviour as well?


Thanks in advance for short comments!

Not sure how you're using the board...is it via usb? And if so, what usb receiver chip is used? Some implementation of the ESS9018 chip in its early days use USB (receiver) chip SA9023, which limit input rates to 96khz (accepts 32kh, 44.1khz, 48khz, 88.2khz and 96khz ,16 and 24bit). If that is the case, perhaps playing files above 96khz like you mentioned (at 192khz) triggered the clicks?

 

[eliash]

...I think the USB to SPDIF converter is OK (external, to avoid ground loops and PC EMI). It is based on C-Media´s well-known CM6631A chip.
Of course I tried also another DAC (FIIO E17 via SPDIF) with it, and no problems with all bitrates up to 176/192K - 16/24bit wide...
With the above mentioned mods to the 9018s board, currently everything works fine, but I am really interested about any other issues with this chip from a technical perspective, simply to understand it better and for comparison with other DACs...

 

[Herbert]

Using a very old 20mHz scope (I use it normally to check signals of CD-Player´s eyepatterns)
Would I be able so "see" the hump (I giess playing a sweep or using white noise). Sensitivity is about 5mV,
I assume it can be amplified to 0.5mV using the 10x switch on the probe....

 

[PPP]

https://www.audiosciencereview.com/...adphone-amp-reviewed.9446/page-17#post-260056

This post show exactly how Topping solved this imd hump problem.

 

[trl]

https://www.audiosciencereview.com/...adphone-amp-reviewed.9446/page-17#post-260056

This post show exactly how Topping solved this imd hump problem.

I don't follow you, sorry. This post is more related to a heatsink and too much soldering on some componenets, right?

Thank you!

 

[PPP]

I don't follow you, sorry. This post is more related to a heatsink and too much soldering on some componenets, right?

Thank you!

No. The last sentence of the post discussed the op amp for iV conversion was intentionally kept secret by topping. That’s the crux to solve the hump.

 

[PPP]

I don't follow you, sorry. This post is more related to a heatsink and too much soldering on some componenets, right?

Thank you!

Many assumed the problem was related to output stage. But people don’t know how to fix. Topping figured out the sopution and wanted to hide the details from competitors. But at least we now know from the pcb that the former assumption is true, and the op amp selection is the key.
The rest of the problem is easy— just get a d50s or dx7 pro, and try to figure out what the op amp is. There are very few options for audio op amps that can be act as a I V converter. Test the op amp’s electric properties and you may know what it’s secret identity is. Or , the easier way is to any other 9038 Dacs that don’t have the hump, such as matrix or smsl, and tear down and see, if their surfaces are not scrapped

 

[trl]

I doubt the "hump" is from the I/V stage. If the hump would be an issue when approaching the 0dB SPL I might understand that an opamp with a better current handling might help, but in this case the hump is somewhere in the middle.

However, maybe there is an explanation regarding the hump and the I/V stage, just that I'm unable for the moment to dig one. Perhaps somebody else might explain this better, thanks!

 

[Yviena]

So is the ESS IMD hump only a problem in lower signal levels, or does it apply at all signal levels?

 

[thomasjast]

So is the ESS IMD hump only a problem in lower signal levels, or does it apply at all signal levels?

Mid-levels, actually.

 

[Yviena]

Mid-levels, actually.

Ohh so if for example say I have digital volume at around -3 to -6db I'm not effected by the hump?

 

[thomasjast]

Ohh so if for example say I have digital volume at around -3 to -6db I'm not effected by the hump?

Yeah, basically like that. The graphs shown are pretty descriptive.

 

[LTig]

Ohh so if for example say I have digital volume at around -3 to -6db I'm not effected by the hump?

You will be affected if the level of the music is above the lower entry level of the hump. Music contains a lot of seperate tones at many various levels. There will always be levels of tones within the range of the hump.

 

[Schackmannen]

You will be affected if the level of the music is above the lower entry level of the hump. Music contains a lot of seperate tones at many various levels. There will always be levels of tones within the range of the hump.

Does it mean that having the volume control at 0 dBr and playing for example a -30 dBFS tone will be the same as having the volume control set to -30 dBr and playing a 0 dBFS tone and both will show the IMD "hump"?

 

[zermak]

Does it mean that having the volume control at 0 dBr and playing for example a -30 dBFS tone will be the same as having the volume control set to -30 dBr and playing a 0 dBFS tone and both will show the IMD "hump"?

I think it's more complex than that. I mean most of the music is between 4 to 14dB dynamic range and the audio files, if properly mastered, never have 0dBFS peaks plus if you use/apply ReplyGain the peaks will always be lower. So it is a sum/combination of digital volume of the audio files, DAC volume and digital player/system volume.

In my setup when I listen to headphones my Khadas TB is at -16dB but I also use EqualizerAPO with a -12dB pre gain and all my audio library has ReplyGain applied so few more dB less. I don't know how them all combine together but I am pretty sure I am in the bad spot of that hump but reading the charts from Amir/IMD I doubt I can hear any issue becasue of my bad training and bad audio capabilities.