I fully agree, but in the end the aliasing turns out to be no big issue in real-world...otherwise MQA wouldn't even exist...

And for the rolled-off treble there is a simple workaround: use the EQ to compensate. Here are two settings for Slow and NOS to linearize the treble response at 44.1 kHz sample rate:

Slow B5 G +4.0 F 16.5k Q 2.1 Shelf
NOS B5 G +3.0 F 14.3k Q 0.6 Shelf

As you can imagine with only 3 and 4 dB gain this is just a slight change, it linearizes NOS perfectly, and Slow good enough.

Now the point is that the Shelf filters have only small impact on the impulse response. Basically Slow stays Slow, and NOS stays NOS. See these two pictures, comparing original (red) with EQ'd (blue) result.

From single to quad speed (up to 192 kHz) the filters stay the same, they just double/quadruple in corner frequency (they move out of the hearing range), and the ringing is halve/quarter in time length.

Above 192 kHz the DAC always uses Slow, this is a fixed function.

But the point is a different one: whenever you upsample the data has to be filtered. This filter is exactly the same as when using the DAC itself in Single Speed, and can also be seen in the upsampled result. So you gain nothing.

At 192 kHz there is no real argument against a steep filter, no matter if SD Sharp or Sharp. There will be no 96 kHz signal in your audio material that would make them ring. And even if then you are not able to hear this short, low level 96 kHz signal.

Correct.

any EQ-recommendations for 48 kHz, 88 etc (in the NOS-mode)?
it will be interesting to EQ-correct input for recordings with slow filter

24/96 mode
it drops a little (around -1 dB) on the 24 kHz band

by the way, why RME calls super-slow filter of 4490 as NOS ? Actually NOS-DAC's are differrent story.

Here are examples for EQ settings to compensate Slow and NOS at 48 kHz sample rate:

Slow B5 G +2.5 F 16.3k Q 1.7 Shelf
NOS B5 G +3.5 F 16.5k Q 0.9 Shelf

As expected the higher frequency response of 48 kHz sample rate helps making the frequency response even more linear up to 20 kHz as with 44.1 kHz sample rate.

And here are the impulse responses for these EQ settings, comparing original (red) with EQ'd (blue) result.

The simplest thing to set in the unit is setting the high and low frequencies. For Slow 10kHz to + 3.2db. For NOS 3.5kHz to + 2.5db. Get the same smooth frequency response up to 20 kHz.

I think you have an error in your ADI-2 DAC user guide v1.4:-

DA Filter

Short Delay Sharp, Short Delay Slow, Sharp, NOS. The Digital to Analog Converter chip offers
several oversampling filters. Default is SD Sharp, offering the widest and most linear frequency
response and lowest latency. SD Slow causes a small drop in the higher frequency range.

Should it not be:-

Short Delay Sharp, Short Delay Slow, Sharp, Slow, NOS. (Add Slow)

Does the DAC require any EQ'ing for 96kHz>?

P.S I love this DAC, it has replaced my Chord Hugo 2!

Is this point valid also for upsampled 88/96/176 etc. from original 44/48khz recording? (I suppose yes, but not sure...)
Also would be interesting to see (if you have) frequency responses of slow/nos filters at 88/96/etc. khz

All this stuff by a pure "technical" curiosity, not audiophile mumbo jumbo...
I'm perfectly happy with "textbook" sharp filters at any samplerate :-)

This parts, I assume:

"From single to quad speed (up to 192 kHz) the filters stay the same, they just double/quadruple in corner frequency (they move out of the hearing range), and the ringing is halve/quarter in time lengt".

and

"whenever you upsample the data has to be filtered. This filter is exactly the same as when using the DAC itself in Single Speed, and can also be seen in the upsampled result. So you gain nothing".

The underlined parts of the two sentences appear inconsistent in my view.
If the filters double/quadruple in corners frequency (they move out of the hearing range) (and, I assume, the frequency roll-off move out of the hearing range), then is not exactly the same....
Shure, my lack of knowledge...

Ah ok...you (and MC of course) mean the interpolation low pass filter (at single speed Nyquist frequency) in the upsampling process.
Right, my lapse of memory.
Excuse me.

Hi.
Thank you all for interesting topic. I'm still learning this superbeast machine and very happy to have one. Please forgive me my question but in EQ settings - what "Q" stands for?  What value is that, can someone explain me please?
Any link to read about appreciated.

Indeed    Now I know.  Thank you Ramses.

I am really out of my league here.  But what I have found is that I like the sound of NOS filter for some reason.  However I don't like the HF roll off.  So I tried my own PEQ filter in order to correct the roll off.  I then found the one posted here for NOS.  I will try that shortly.  But my lack of understanding leaves me a little confused.

I am trying to get CDs to sound good at 44.1kH sampling.  So the filter published here  will correct the HF roll off.  But what happens when I stream USB at 192kH?  What effect will the filter have.  Do I have to turn off filtering?

Also, when I turn off PEQ and look at the EQ window, it still shows the EQ curve even though I think I turned it off.  There is no indication in the EQ window that it is off.  How am I to understand this?

Also is there some free software available for modeling PEQ curves?

Thanks.

Thanks.  I see that now.  Should I be concerned about the audibility of aliasing using the NOS filter?

1. You will see a small raise in upper treble.

2. Above 192 kHz.

Correct, I'll add  it.

Wow, thanks MC! I just stumbled across this post. Changing the EQ with the slow filter has absolutely changed my perception of the RME DAC.

Up until now, I enjoyed the SD SLOW filter, mostly due to the fact it sounded most natural and organic to my ears. It also seemed to open up the soundstage and made things sound motr three dimensional. However, it often made a lot of music sound undefined and thus homogeneous as well. The SD SHARP filter gave music the required edge to help dynamics and perhaps had better "musicality" too but at the expense of the qualities I enjoyed with the SD SLOW.

I spent weeks thinking the perfect scenario for me would be a half way point between these two filters. After applying your EQ changes, I've found that sweet spot! I get all the smooth, airy sound of the SD SLOW with the positives of the SD SHARP combined. It's genuinely amazing to me how much this has changed my enjoyment of the RME DAC. I no longer get up between songs to change filters according to what song I'm about to play.

The filters, except “Sharp”, even if named similar, have a different shape in the ESS vs.the AKM DAC chips.
It’s visible in the graphs in the according manuals.

So, different EQ’s are needed.

Im very interested to experiment with compensation filter for 44.1 an 48 kHz for NOS and Slow with my ESS version ADI-2 DAC.

@Matthias: Would it be possible to share the settings like above especially for ESS?

I have a HEGEL H20 Class A/B dual mono Poweramp connected to Magneplanar 1.7i speakers. As a pre-amp I use the RME ADI-2 Pro FS R BE.

I noticed that the Sound was very cold and analytical using the default SD Sharp filter.

So, I decided to change the default setting in the AD Filter settings from default SD Sharp to Slow. The improvement in Sound Quality was stunning, I was flabbergasted how much the SQ improved so much: wamer, very pleasing to listen to, no listening fatigue anymore, gone with the cold and analytical sound.

I wanted to know more about WHAT and WHY the SQ improved, so I entered in ChatGPT 4o the following prompt:

"My RME DAC's DA Filter is set to SD Sharp. What would be the effect to change the Filter to Slow?"

This was the generated response to my prompt:

ChatGPT

Changing the filter setting on your RME DAC from "SD Sharp" to "Slow" affects the way the digital-to-analog conversion process handles high-frequency components of the audio signal, primarily influencing the trade-off between transient response and high-frequency noise rejection.

Here’s a breakdown of the effects:

1.    Transient Response:
SD Sharp (Sharp): This filter has a steep roll-off, which means it more aggressively attenuates frequencies above the  Nyquist limit (half the sampling rate). This results in better preservation of transient details in the audio, leading to a crisper and more defined sound.
Slow: This filter has a gentler roll-off, leading to a slower attenuation of high frequencies. As a result, transients may not be as sharply defined, potentially making the audio sound smoother or more relaxed.

2.    High-Frequency Content:
SD Sharp: The sharp filter cuts off high-frequency components more abruptly, which minimizes the aliasing and high-frequency noise but can introduce pre-ringing artifacts.
Slow: The slow filter allows more high-frequency content through, which can result in a more natural sound. However, it might also let through some high-frequency noise that the sharp filter would have attenuated.

3.    Overall Sound Character:
SD Sharp: Often perceived as having a more analytical and detailed sound due to the emphasis on precise transients and high-frequency attenuation.
Slow: Often perceived as more musical and natural, with a potentially smoother and warmer sound due to the less aggressive handling of high frequencies and transients.

4.    Listening Preferences:

If you prefer a highly detailed and crisp sound with a focus on clarity and transient response, the "SD Sharp" filter might be more to your liking.

If you prefer a smoother, more relaxed, and natural sound, the "Slow" filter might be more appealing.

In summary, changing your RME DAC’s filter from "SD Sharp" to "Slow" will likely result in a smoother and potentially warmer sound at the expense of some high-frequency detail and transient sharpness. It’s a subjective preference, and you might find one setting more enjoyable than the other depending on your listening habits and the type of music you enjoy. Experimenting with both settings and listening to a variety of music will help you determine which filter suits your personal taste best.

That is exactly what I experienced, a new listening experience, music reproduction on steroids.

Thank you KaiS for your response.

You write "In short (@ 44.1 and 48 kHz sample rates only, irrelevant for higher SRs......)

Question:
1. Does this mean, that the Slow filter kicks in at sample rates of 22.05 kHz upwards and 24 kHz upwards only?
2. Or does it relate to your SQ impressions @44.1 and 48 kHz sample rates?

Assume that the sample rate is 96kHz, does it mean that the Slow filter kicks in @48 kHz upwards?

Graph 34.13 applies to AD filters, not DA Filters

Graph 34.14 does not respond to my question.

your response is a killing phrase.

Again, ChatGPT gave me the answer, I was looking for.

This was my prompt:
MY RME DAC has an DA Filter SD Sharp, which filters from about 20 kHz upwards with sample ratels of 44.1 Khz and 48 kHz. However, when the sample rate is for instance 92 kHZ, does the SD Sharp Filter filter from 48 kHz upwards?

ChatGPT
The behavior of the SD Sharp filter in your RME DAC changes with different sample rates. Here's a more detailed explanation:

At 44.1 kHz and 48 kHz sample rates: The SD Sharp filter begins to filter from around 20 kHz upwards. This is because these sample rates are typically used for standard audio playback, where the audio content is within the human hearing range (up to 20 kHz), and the filter helps to remove frequencies beyond this range to avoid aliasing and other artifacts.

At higher sample rates (e.g., 96 kHz): The filtering behavior of the SD Sharp filter will scale with the sample rate. For a sample rate of 96 kHz, the filter cutoff point will move up proportionally. Specifically, the filter will start attenuating frequencies above approximately half the sample rate. In the case of 96 kHz, this would be around 48 kHz.

So, to answer your question: Yes, when the sample rate is 96 kHz, the SD Sharp filter will start filtering from around 48 kHz upwards. This scaling ensures that the filter continues to remove unwanted frequencies and noise that are above the audible range and could cause issues with digital audio processing.