• 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.

ESS THD ‘Hump’ Investigation

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
17,408
Likes
14,860
Location
Seattle Area
Would it be too much to ask to run this test on a DAC that does not exhibit this issue for the purpose of comparison?
Sure. Here is the comparison to Gustard DAC-X26 without this problem (in pink):

1552367093996.png
 
Joined
Mar 10, 2019
Messages
14
Likes
55
I have consulted several ESS engineers on ESS THD/IMD Hump.
This problem is almost impossible to solve for the ES9018S.
But for those DACs based on ES9028/9038, there is actually a way to solve this problem, whether it is a two-channel mobile version or an eight-channel pro version.
One of the key to solving this hump is to properly handle the PCB design, especially the layout of the I/V and LPF circuits.

The Gustard DAC-X26 is a good example. We can see that this problem can be solved, but it requires a very high level of engineering.
 
Last edited:

IVX

Member
Joined
Mar 3, 2019
Messages
22
Likes
15
I have an idea to make 2 potentiometers + STM8S003(I use it as I2C host) for adjusting THD_COMP more visual manner ;) BTW, clk_gear affect the THD as well(.0006/.00045 clk_gear =1:1/clk_gear=1/8), and the best result I see at 44.1 with clk_gear MCLK = XI / 8. However, such divider value doesn't work with Fs>96 (XI=100MHz, SPDIF input used), I gonna continuously read the current PLL value to watch actual Fs, to keep clk_gear at max.
 
Last edited:
Joined
Oct 23, 2018
Messages
40
Likes
39
Sure. Here is the comparison to Gustard DAC-X26 without this problem (in pink):

View attachment 23448
Thank you Amir! If I am interpreting this graph correctly, it appears that the ESS hump distortion is below -100db throughout the volume range. According to your "Audibility thresholds of amp and DAC measurements" post, the lenient limit for IMD is -66db and the strict one is -120db. This means that the devices with ESS hump are actually closer to the strict limit than to the lenient limit which is already supposed to be inaudible. Therefore, except for striving for engineering perfection, this actually appears to be an absolute non-issue in terms of the impact on the sound fidelity of these devices.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
17,408
Likes
14,860
Location
Seattle Area
Thank you Amir! If I am interpreting this graph correctly, it appears that the ESS hump distortion is below -100db throughout the volume range. According to your "Audibility thresholds of amp and DAC measurements" post, the lenient limit for IMD is -66db and the strict one is -120db. This means that the devices with ESS hump are actually closer to the strict limit than to the lenient limit which is already supposed to be inaudible. Therefore, except for striving for engineering perfection, this actually appears to be an absolute non-issue in terms of the impact on the sound fidelity of these devices.
Those are levels in dBV, not ratios as I typically show. When the generator is at -30 dBFS, the distortion is at -100 dB. Assuming for now that dBFS = dBV (probably not but good enough for this), then the gap is 70 dB. Or put inversely, as a ratio it would be -70 dB. So it is not that low.
 
Joined
Oct 23, 2018
Messages
40
Likes
39
Those are levels in dBV, not ratios as I typically show. When the generator is at -30 dBFS, the distortion is at -100 dB. Assuming for now that dBFS = dBV (probably not but good enough for this), then the gap is 70 dB. Or put inversely, as a ratio it would be -70 dB. So it is not that low.
I am not sure that I understand this properly. Using the same approach for Gustard DAC-X26, when the generator is at -30 dBFS, the distortion is around -117 dB. That means that the gap is 87 dB, so as a ratio it would be -87 dB. Does that mean that when the signal is at -30 dB, the audible IMD distortion is at -87 dBFS or at -117 dBFS for Gustard DAC-X26?
 

la2ygoo

Member
Patreon Donor
Joined
Dec 19, 2018
Messages
49
Likes
26
I am not sure that I understand this properly. Using the same approach for Gustard DAC-X26, when the generator is at -30 dBFS, the distortion is around -117 dB. That means that the gap is 87 dB, so as a ratio it would be -87 dB. Does that mean that when the signal is at -30 dB, the audible IMD distortion is at -87 dBFS or at -117 dBFS for Gustard DAC-X26?
It look like rca output , 2Vrms=6dBV , when at -30 dBFS , is -24 dBV , the gap is 93 dB
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
17,408
Likes
14,860
Location
Seattle Area
Using the same approach for Gustard DAC-X26, when the generator is at -30 dBFS, the distortion is around -117 dB.
It is not the same test. It looks similar but is different. My standard tests are ratios so you can do the math as you are doing. The test posted in this thread is a voltage expressed in dB. The ratio is NOT to the generator level but to one volt (dbV).
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
17,408
Likes
14,860
Location
Seattle Area
Does that mean that when the signal is at -30 dB, the audible IMD distortion is at -87 dBFS or at -117 dBFS for Gustard DAC-X26?
The graph was THD to show that this is not just intermodulation distortion. For your question, you can use the IMD measurement and there, it is expressed as a ratio to input level so you can do your math:



So we see that the distortion level is around -60 to -70 dB as I computed for the new measurement. Does it make sense now?
 
Joined
Oct 23, 2018
Messages
40
Likes
39
It does make more sense. I am just trying to interpret what the THD graph means for devices that have the ESS hump. For example, I own the Topping D50. The way I interpret the IMD distortion chart, it appears that the distortion is well below audibility and the lenient limit. However, I am not sure how to interpret the THD graph, considering that your tests showed that the D50 has excellent THD+N Ratio at 2Vrms. What does it actually mean?
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
17,408
Likes
14,860
Location
Seattle Area
One would need a way to turn the Hump on and off to know the effect to know if it is or isn't audible in listening tests. That is impossible right now.
 

johan

Member
Manufacturer
Joined
Feb 23, 2018
Messages
64
Likes
80
Amir , I dont think the hump exists because of the PCB layout . Might be because of ASRC (that most design use)...except Katana. Are you able to see the hump (or test for it) ?
 
Joined
Nov 26, 2018
Messages
21
Likes
8
I did some tests with different clocks, and noticed THD improvement from noname 100MHz 3.3V OSC to SiTime SiT8209 and further with 50MHz Epsol crystal + internal DAC's OSC. The biggest improvement I got when replaced 100MHz OSC to SiT8209, don't remember exactly because I made many others modifications as well to improve THD, I think at least from .001% to .0006% step down only due to SiT8209. And Epcos standard crystal+internal 9038q2m oscillator reduces THD up to .00045%(or .00025% after THD_COMP tweaking) on my proto-board, which is highly optimized for a min size(USB dongle, I posted the pic here). Of course, I glad to get such a perfect clock for 1/30 of SiT8209 cost but 50MHz simply don't catch Fs=192, only <=96, so I need to use 100MHz crystal anyway. Unfortunately, 100MHz Xtals available in 5032 or bigger case.. It seems 9038q2m datasheet has mistake regarding 100MHz as max Xi, it rather min or the only single valid frequency, at least for SPDIF at 192.
Do I read the post correctly that the hump is also jitter-related and that a better clock
(for cost efficiency prefereable in lower frequencies) will lower it?
 
Last edited:
Joined
Oct 23, 2018
Messages
40
Likes
39
Schiit Modi 3 DAC vs Modi 3 Intermodulation Distortion Comparison Measurement.png

I find this graph from another review puts things in a context for people that own these ESS-based devices. It shows that even at the highest point of the hump, the performance is still comparable to the Modi 3 using USB.

Amir, would you say that the D50 has a better IMD performance than the Modi 3 via USB because for the majority of the range, the distortion is actually significantly lower? Or does its lack of predictable linear behaviour actually make it equal or worse compared to the Modi 3 using the USB?

It would also be great to know whether there is any statistical information concerning the percentage of audible content that is found between -40 dBFS and -20 dBFS for music. I could see a lot of classical pieces have a significant portion of content here, but pop music would likely be found between -0 and -20 dBFS.
 
Top Bottom