Measurements: "ESS Hump" revisited (Khadas Tone Board V1.3)

[Pdxwayne]

DAC is RME ADI-2 Pro FS R (Laptop-->DAC-->Headphones) and I used the emulation, not a recording of the actual KTB output (will do that later).

Well, I have found additional information.....

I was using 24bit, 44100 in Windows Advances property for Gustard for a while. Yesterday I temporarily switched to 16bit for blind testings, but switched back to 24bit today because I wanted to match my KTB settings. With that 24 bit setting, Gustard hi hat sounded really soft. I tried switched to 16bit 44100 just now, and I got stronger and sharper hi hat from Gustard, like KTB.
KTB setting in Advanced property have always been set at 24bit, 44100 (no option to switch to 16bit), but it sounds strong and sharp, not like Gustard in 24 bit. So, likely driver related differences between Gustard and KTB???

 

[gvl]

Oh, you’re using Windows mixer? That probably explains the differences you hear. Has nothing to do with the driver.

 

[Pdxwayne]

Oh, you’re using Windows mixer? That probably explains the differences you hear. Has nothing to do with the driver.

Yeah, I am using Chrome browser for online blind tests. So, I guess have to go through Windows mixer? Also, I thought advance setting gets the available bit and Hz information from the driver?

If not driver related, why KTB advanced setting has no 16bit option?

If not driver related, I wonder why Gustard at 24bit 44100 Hz behaves differently than KTB at 24bit 44100 Hz....

 

[skyfly]

I saw a THD vs signal level graph of the analogue output of RME babyface with ESS chip in a youtube review. The curve looked like that of a class G power amp.

 

[gvl]

Yeah, I am using Chrome browser for online blind tests. So, I guess have to go through Windows mixer? I wonder why Gustard at 24bit 44100 Hz behaves differently than KTB at 24bit 44100 Hz....

Ideally you don’t want to go through the Windows mixer as it’s an extra variable. Chrome had an experimental feature for Wasapi Exclusive mode but it was broken at one point, not sure about current status. There was a thread somewhere here how to avoid additional processing in the mixer that may get enabled randomly, iirc you need to install EQ APO or something like that.

 

[gvl]

If not driver related, why KTB advanced setting has no 16bit option?

You don’t really need a driver on W10, modern DACs are UAC devices and Windows knows how to discover their capabilities. In other words the UAC driver that ships with the OS is sufficient. Manufacturer provided drivers may expose additional settings through the XMOS control panel or include ASIO interface, but they are not required.

 

[Pdxwayne]

You don’t really need a driver on W10, modern DACs are UAC devices and Windows knows how to discover their capabilities. In other words the UAC driver that ships with the OS is sufficient. Manufacturer provided drivers may expose additional settings through the XMOS control panel or include ASIO interface, but they are not required.

Is it possible the difference in 24bit 44100 Hz behavior is due to the DAC instead of Windows?

 

[gvl]

Is it possible the difference in 24bit 44100 Hz behavior is due to the DAC instead of Windows?

Hard to say. If you’re on a quest to find sonic differences that lie below -100dB you need to start with a bit perfect pipe to your DAC and lossless material.

 

[Pdxwayne]

Hard to say. If you’re on a quest to find sonic differences that lie below -100dB you really need to start with a bit perfect pipe to your DAC and lossless material.

I don't think the harshness issue is due to something below -100db.
; )

 

[gvl]

I don't think the harshness issue is due to something below -100db.
; )

Probably not, if it’s a real thing. Windows mixer is a prime suspect anyhow.

 

[Pdxwayne]

Probably not, if it’s a real thing. Windows mixer is a prime suspect anyhow.

Indeed! Hmm....When I first sensed KTB was harsh, I was using Windows laptop for playback and using Windows mixer.

When I did my blind tests between ktb and e30, I was using node2i playing CD quality wav file and was outputting to ktb via coaxial. At that time once voltage matched up to 5khz, I couldn't tell ktb apart from e30.

Maybe Windows mixer was the culprit!

 

[Herbert]

This began as a very accurate and scientific thread, one of the best @ ASR. Don‘t derail it with amateurish listening tests that are nonsense anyway as there are strong hints that the glare is not audible.

 

[gvl]

The glare is audible, the hump probably not

 

[Veri]

The glare is audible, the hump probably not

Not sure if you're serious but there's no real indication of either, right? KSTR only found things to be audible when emulating the effect in (not presently found) very high levels, right?

 

[shevalier]

OPA1612 et al don't need it.

very optimistic point of view

 

[Herbert]

very optimistic point of view

View attachment 185635

View attachment 185636

To be found out, hopefully. KSTR has one board, I donated two. Three boards to play with but there is always the risk to damage the tiny traces. So one Khadas Tone Board is being kept unaltered for reference. When the hump has been tamed (without changing opamps) we will very likely try an opamp swap...

 

[shevalier]

kept unaltered for reference

just one cut for resistor in series with cap

By the way, all of us read "AN -xxx i/v for photodiode", where it says that need to expand the bandwidth.
The DAC is not a photodiode, need to filter just outside the audio band, i.e. combine the functions of i/v and LPF

 

[KSTR]

very optimistic point of view

View attachment 185635

View attachment 185636

While you have some point wrt capacitive loading, I truly believe the original 100R, as seen in the ESS reference design comes from the peculiarities of the AD797 which does not like closing the feedback without any series resistance at higher frequencies, clearly explained at several points in the datasheet (Aol bode plot above all). More that 33pF of feedback C requires 100R series.

In our case, with additional (but unknown) output capacitance of the DAC chip, we are in a trade-off position: the series R isolates the parasitic capacitance but at the same time it causes the 1/ß curve to slope upward, decreasing phase margin at the Aol 1/ß intersect.

Many I/V of today have no or very low valued feedback C series resistors, like 10R.

 

[KSTR]

just one cut for resistor in series with cap

That defeats the whole idea of killing differential mode glitches

 

[Pdxwayne]

This began as a very accurate and scientific thread, one of the best @ ASR. Don‘t derail it with amateurish listening tests that are nonsense anyway as there are strong hints that the glare is not audible.

Sorry for suggesting the amateurish listening test. I am awaiting proper abx from ktr.

At the end, I hope whatever you and KTR did are not just for scientific curiosity, but also for audible benefits.