• 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). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

RME ADI-2 DAC FS - AKM Versus ESS Measurements (DAC, Preamp & Headamp)

danadam

Addicted to Fun and Learning
Joined
Jan 20, 2017
Messages
676
Likes
980
So, if your speaker has HDs in the (typical) -50DB range, the sound reaching your ears will have HDs in the -50Db "+" -120Db (from the DAC) "+" -120Db (say you have an excellent amp too). Summing HDs is not exactly easy-addition but let's say (very rougly) that you end up with harmonics somewhere in the -49.5 to -50.5 Db range.
If it's anything like adding coherent or incoherent signals then more like -49.995 dB.
 

RandomEar

Active Member
Joined
Feb 14, 2022
Messages
199
Likes
357
Most people can barely discern differences of 0.25 dB in the volume of pure test tones played directly after each other. 0.5 dB differences are still close to the lower limit for many. And for distortion with real music, very few people will register it at all at -40 dB or below. A speaker with -50 dB or less of distortion when playing real music will literally sound perfect to the vast majority of humans. To think that a difference of +-0.5 dB of distortion (assuming that calculation is correct) at -50 dB would be audible even under controlled conditions already is extremely optimistic. But as @danadam pointed out, the real difference will more likely be well below 0.1 dB, which is inaudible even with pure test tones.

You may be right in stating that the audibility of these specific ditortion "combinations" hasn't been disproven, yet - I didn't check. But that just means the null hypothesis still stands. Also, concerning how that "ESS glare" legend came to life: It might have developed in the reverse order. ESS publishes distortion data in their spec sheets. People not versed in audio science might have seen that ESS chips produce "lots" of H4/H5, which has been said to sound unnatural or harsh. With that bias, they might have created this legend in the first place.
 

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
1,801
Likes
2,151
Location
Liège, Belgium
What you seem to forget (and for some reason many others do the same) is that we do not listen to DACs, we listen to loudspeakers/headphones. And the sound that reaches your ears has HDs exactly in that -60Db range that you mentioned. Which is audible. And actually it's more like -40 to -50Db (...).

So, if your speaker has HDs in the (typical) -50DB range, the sound reaching your ears will have HDs in the -50Db "+" -120Db (from the DAC) "+" -120Db (say you have an excellent amp too). Summing HDs is not exactly easy-addition but let's say (very rougly) that you end up with harmonics somewhere in the -49.5 to -50.5 Db range.

Hum
I'm sorry, but this is not correct.

Summing HD from different sources is not that straightforward indeed, since they are somehow coherent, but not completely: the phase is probably not.

Anyway, summing -50dB sinusoidal signal with two signals at -120dB, with the same frequency and same phase, will give you a new signal at -49.995 dB.

That's 0.005 dB difference.
In the worst case scenario.

The level of the sum of 3 completely incoherent signals at those level will be (on average) at -49.999999 dB.
Less than 0.000001 dB difference.

The reality will be somewhere in between.
Anyway, absolutely impossible to hear.

(If you're lazy to compute yourself, you'll find calculators here for the sum of coherent signals and here for incoherent signals)

EDIT:
@danadam added the same comment (with link to the same calculators :)) above.
Sorry for the repetition.
 
Last edited:

lashto

Addicted to Fun and Learning
Forum Donor
Joined
Mar 8, 2019
Messages
696
Likes
365
I'm sorry, but this is not correct.
... summing -50dB sinusoidal signal with two signals at -120dB, with the same frequency and same phase, will give you a new signal at -49.995 dB.
That's 0.005 dB difference.In the worst case scenario.
...
Thanks for doing the math. I did not. Guess you are both right and I'll update my initial "very roughly 0.5Db" estimate.
However, that number does not matter much, its only point was to show that it's "minuscule". Otherwise, the THD number/diff is ~useless: -50Db of pure-H2 will sound diff than -50Db of pure-H7 (that's approx what you hear on the linked wiki demo page).

What actually matters is the THD spectrum (i.e. the level and phase of each HD spike). Very hard to test because there are ~infinite variations. Even a simpler test using "pure Hx" is very hard because the HD spectra of your electronics can easily mask the test HDs.
IIRC, I saw a pure-Hx test somewhere and the results were like: H2 becomes audible around -20Db, H3 around -30Db and so on...Here's a somewhat similar graph (second one from top).

Anyway, absolutely impossible to hear.
...
A strong hypothesis at -120Db and seems to be the majority opinion on ASR. Actually, the "ASR educated guess" seems to say that HDs become inaudible/useless much earlier (-80-90Db). But still, it's just a hypothesis. Not fully tested. And close to un-testable with current tech.

Personally, I won't bet on this hypothesis. It's not like I have any better proof, just different "educated guesses" (at best).
Here are a bunch of people who do not seem to buy it either. Won't pretend that I fully understand what is the H3 doing in that article but it's just the latest in a long series of HD/IMD improvements that they did in a supposedly inaudible area (i.e. Hypex UCD had "inaudible" HDs at -110DB, nCore at -115DB, Purifi at -120Db). Is that just because they can? Or is it marketing/sales only? Or maybe those minuscule THD/IMD numbers are not so 'inaudible'?! Just saying...

P.S.
sorry for the somewhat long & offtopic posts. And not sure if it's worth continuing. There is not conclusive test/proof and we can all write 1000 opinion-replies without any useful info/conclusion
 
Last edited:

lashto

Addicted to Fun and Learning
Forum Donor
Joined
Mar 8, 2019
Messages
696
Likes
365
... concerning how that "ESS glare" legend came to life: It might have developed in the reverse order. ESS publishes distortion data in their spec sheets. People not versed in audio science might have seen that ESS chips produce "lots" of H4/H5, which has been said to sound unnatural or harsh. With that bias, they might have created this legend in the first place.
a credible hypothesis too. Here's why I think that it's less credible than "something is/was there":
  • usually things happen the other way: first someone hears/says "glare, glare" and then people start looking for explanations in spec sheets/etc (which is exactly what we are doing). Otherwise, if all sounds good, no one starts investigating spec sheets.
  • AFAIR, the glare rumor was started by the kind of audiophiles who only "know" their ears. Seriously doubt that any of them checked DAC spec sheets.
  • IIRC, the ESS spec sheets weren't even available ~10 years ago. May still have been an EE expert with access. But from my exp, most EEs only know one thing about HDs: "it's all distortion and we only care about getting rid of it all".
  • it still implies some sort of collective psychosis and it's much easier to assume human error (i.e. a human-error has a way better Occam-score than collective-audio-psychosis.) Maybe it was just one of the old ESS chips who had HD-glare issues. Maybe it was just a batch of chips that went wrong in some factory. Maybe there were some wrongly implemented DACs...
ESS chips produce "lots" of H4/H5
That wouldn't sound so bad. But looks to me that the ESSs have unusual levels of H7/9/11/etc. Look at the graph I posted before (brand new ESS chip/dac). If I'm not mistaken there is a very strong H9 in there (even goes above the H2 at some freqs.)
Or remember that even the newest ESS have IMD-hump issues in many implementations.
Or look at the atrocious -10Db IMD measured on a ESS9023 board (launched in 2010; also note how the spec sheet says "confidential" and shows no THD/IMD spectra). Anyway, that crap is audible and sounds ~like the legend says: "glare".

P.S.
another slightly offtopic post. To my defense I did not start the 'ESS glare' tangent :)
And this discussion should be over, i.e. -10Db IMD should be enough 'glare' for everyone.
 
Last edited:

RandomEar

Active Member
Joined
Feb 14, 2022
Messages
199
Likes
357
That wouldn't sound so bad. But looks to me that the ESSs have unusual levels of H7/9/11/etc. Look at the graph I posted before (brand new ESS chip/dac). If I'm not mistaken there is a very strong H9 in there (even goes above the H2 at some freqs.)
Or remember that even the newest ESS have IMD-hump issues in many implementations.
I don't know what produces glare or may be perceived as such. Just randomly threw in some distortions as an example. If H9 is the main suspect in this crime, I'm fine with that.

Or look at the atrocious -10Db IMD measured on a ESS9023 board (launched in 2010; also note how the spec sheet says "confidential" and shows no THD/IMD spectra). Anyway, that crap is audible and sounds ~like the legend says: "glare".

P.S.
another slightly offtopic post. To my defense I did not start the 'ESS glare' tangent :)
And this discussion should be over, i.e. -10Db IMD should be enough 'glare' for everyone.
That looks like a saturation/clipping problem on that (reference?) implementation, as it just comes up above -1.7 dBFS. Certainly bad if you're always running at full scale and the DAC manufacturer didn't account for this. If such implementation errors were common in actual finished products, this may be the most plausible explanation for reports of ESS chips sounding bad. Because -12 or -10 dB are definitely atrocious.
 

lashto

Addicted to Fun and Learning
Forum Donor
Joined
Mar 8, 2019
Messages
696
Likes
365
I don't know what produces glare or may be perceived as such. Just randomly threw in some distortions as an example. If H9 is the main suspect in this crime, I'm fine with that.
Timbre is a subjective matter, there is no clear definition/formula for "glare". One common example of 'glare timbre' would be the sound of the trumpet. And yes it has huge amounts of high level harmonics. More generally, all H7+ harmonics are often described as 'glare' (or glassy/metallic/sharp/fatiguing/etc.)
So yes, a high level of H9 could be a "main suspect".
That looks like a saturation/clipping problem on that (reference?) implementation, as it just comes up above -1.7 dBFS. Certainly bad if you're always running at full scale and the DAC manufacturer didn't account for this. If such implementation errors were common in actual finished products, this may be the most plausible explanation for reports of ESS chips sounding bad. Because -12 or -10 dB are definitely atrocious.
no idea what board was that or what was wrong with it. ES9023 was/is a fairly popular and cheap DAC, the best known implementation is probably nwawguy's ODAC. That one is also mentioned quite often in the "ESS glare" discussions.
 
Last edited:
Top Bottom