• 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

Joined
Oct 26, 2018
Messages
29
Likes
57
Location
The Universe
#21
Yet you see a lot of manufacturers leaving it over the driver development to a German company (which name now eludes me.) I'm pretty sure Khadas uses the XMOS example code without any modifications except sample rates and bit-depth.
That makes sense, I'm really looking forward to your results.
 
OP
jackenhack

jackenhack

Active Member
Forum Donor
Joined
Oct 25, 2018
Messages
197
Likes
455
Location
Stockholm, Sweden
Thread Starter #22
Sure. If you can walk me through using it to patch my board I can certainly do that and repeat measurements.
Absolutely. All you need is a serial terminal program. The commands are easy and have a help menu. I would absolutely help you out with the connections and how to read/write values. The coefficients are 16-bit signed values, but my guess is that you could go up and down a thousand steps at a time and narrow in on the optimal value.
 
OP
jackenhack

jackenhack

Active Member
Forum Donor
Joined
Oct 25, 2018
Messages
197
Likes
455
Location
Stockholm, Sweden
Thread Starter #23
You need a microscope or magnifying glass to do the soldering though. There should be two resistors, but I found the two signals on one side of a couple of 0402 SMD capacitors. So very thin wire, a steady hand and tape or snot glue to keep the wires from coming loose.
IMG_2839.jpg
 
Last edited:
OP
jackenhack

jackenhack

Active Member
Forum Donor
Joined
Oct 25, 2018
Messages
197
Likes
455
Location
Stockholm, Sweden
Thread Starter #24
FYI the problem is also visible in sweeps of THD+N. It is much less pronounced there but visible still. So try that and see if you can detect it.
I guess that measuring at the worst case frequency would make the adjustments easier to find. Then adjust the values so a complete sweep measures better.
 

Graph Feppar

Active Member
Joined
Nov 19, 2018
Messages
195
Likes
166
#26
1.Benchmark John Siau wrote that the ESS hump is becose most DAC manufacturers dont sum the differential output of DAC into op amp, he wrote thag ESS DACs have alot common mode distortion that needs to be removed by op amp. Neither DAC2 or DAC3 have the hump.

2. The hump is NOT only on 9038, Auralic Vega has hump and its 9018.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
30,286
Likes
89,276
Location
Seattle Area
#27
Absolutely. All you need is a serial terminal program. The commands are easy and have a help menu. I would absolutely help you out with the connections and how to read/write values. The coefficients are 16-bit signed values, but my guess is that you could go up and down a thousand steps at a time and narrow in on the optimal value.
Great. I ordered the bus pirate.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
30,286
Likes
89,276
Location
Seattle Area
#28
You need a microscope or magnifying glass to do the soldering though. There should be two resistors, but I found the two signals on one side of a couple of 0402 SMD capacitors. So very thin wire, a steady hand and tape or snot glue to keep the wires from coming loose.
Thanks. Seems easy enough.
 
OP
jackenhack

jackenhack

Active Member
Forum Donor
Joined
Oct 25, 2018
Messages
197
Likes
455
Location
Stockholm, Sweden
Thread Starter #29
Well, I guess that we can get a MUCH better result by going trough @amirm on real test equipment. I'll take a stab at it tomorrow, but I think the final answer will be from him. It's not like the DAC needs more performance, but hell, if you can improve it, why not go for it?
 

miero

Active Member
Joined
Aug 1, 2018
Messages
227
Likes
246
#30
1.Benchmark John Siau wrote that the ESS hump is becose most DAC manufacturers dont sum the differential output of DAC into op amp, he wrote thag ESS DACs have alot common mode distortion that needs to be removed by op amp. Neither DAC2 or DAC3 have the hump.
...
Here is the explanation of John Siau: https://www.audiosciencereview.com/...easurements-of-benchmark-dac3.3545/post-90519

So debugging I2C communication of the DAC will not solve the issue.

You need to replace I/V stage with something better.

I guess this might work very well:
- http://twistedpearaudio.com/linestages/ivy.aspx (schematics is available)
- http://twistedpearaudio.com/linestages/mercury.aspx (no schematics for this)
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
2,891
Location
Zg, Cro
#31
I have one question about that: according to his explanation IMD hump shouldn't be happening on balanced coonection, right? And yet, there it is..

"This IMD (and THD) is a direct result of omitting a differential amplifier after the output of the chip. Like most high-quality D/A converters, the ESS chip outputs are balanced. Many designers think it is OK to connect these directly to the outputs on the back of the unit. Typically the balanced connection from the chip gets buffered and connected to pins 2 and 3 of the XLR jack. Often, the RCA output is directly connected to pin 2 of the XLR jack and consequently, the RCA output only sees one side of the converter's balanced output. If you take the XLR output and connect it to the balanced input on an Audio Precision test station, the IMD will not show up. But if you connect the RCA output to the same analyzer, the IMD is present. "
 
OP
jackenhack

jackenhack

Active Member
Forum Donor
Joined
Oct 25, 2018
Messages
197
Likes
455
Location
Stockholm, Sweden
Thread Starter #35

DonH56

Major Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
4,641
Likes
7,531
Location
Monument, CO
#36
So they are using two single-ended I/V stages and combining after? Seems like you'd want to use a differential I/V stage and do it at the DAC's output instead of dealing with the CM noise through the stages and filters and all. But I may not be seeing the whole picture...
 
OP
jackenhack

jackenhack

Active Member
Forum Donor
Joined
Oct 25, 2018
Messages
197
Likes
455
Location
Stockholm, Sweden
Thread Starter #37
So they are using two single-ended I/V stages and combining after? Seems like you'd want to use a differential I/V stage and do it at the DAC's output instead of dealing with the CM noise through the stages and filters and all. But I may not be seeing the whole picture...
Yeah, seems like a convoluted way of doing it. I don't have the recommended design schematic that usually accompanies the datasheet. Would be nice to see how ESS does it. That design is what they use for the performance measurements of the DAC.
 

DonH56

Major Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
4,641
Likes
7,531
Location
Monument, CO
#38
There are not a lot of differential op-amps around and mostly they target instrumentation amps (so low bandwidth). I'd guess Benchmark couldn't find a difamp they liked for audio circuits. Note the diff-to-se converter is also discrete opamps (in a single package) and not a differential i-amp. But, audio is not my day job, and I have not kept up with what op-amps are available for audio these days.

Years ago I came up with what I thought was a very clever circuit scheme for a wideband differential amplifier. The company thought so too, and I filled out an invention disclosure. A mistake; they were not into patents, just ensuring they had captured prior art in case they wanted it for some reason, and did not pursue the patent. Such is the life of an engineer...
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
30,286
Likes
89,276
Location
Seattle Area
#39
I post this in March audio dac1 one thread but belongs here. I ran a level sweep versus THD, i.e. excluding the noise, and the problem appears plainly there:



It is remarkable in how they give up 20 dB of distortion the moment levels get beyond -40 dB. It seems like a design problem but how would a company like ESS let this get out this way???
 

Veri

Major Contributor
Joined
Feb 6, 2018
Messages
6,333
Likes
7,377
#40
I post this in March audio dac1 one thread but belongs here. I ran a level sweep versus THD, i.e. excluding the noise, and the problem appears plainly there:



It is remarkable in how they give up 20 dB of distortion the moment levels get beyond -40 dB. It seems like a design problem but how would a company like ESS let this get out this way???
secret sauce
 
Top Bottom