Have you tried to identify these differences in a setup where you don't know which filter you are currently listening to?

Regards
Daniel Fuchs
RME

I mostly use Slow if I'm listening to music and don't need perfect accuracy. If I have someone switching the filters for me blindly (e.g. wife with the remote and me listening), I almost always like the Slow filter the most. I think I like the rolloff.

I use HQPlayer (poly-sinc-short-mp) and up-sample to PCM 768khz, so the RME DAC won't use any of its filters at all for me, they are bypassed because of high sample rate.

This is the number one reason (in my opinion) for external up-sampling with filter, you replace a CPU starved DAC and its very simple filters with something more advanced which is done on a fast computer.

There is no way to do that, except maybe to use different presets and assign them to different buttons on the remote.

Just a few thoughts on that topic..

I would recommend not to complicate things too much. The DAC sounds great with all D/A filter settings, the differences are only very subtle. The only difference that is easier to notice is NOS compared to the other D/A filters at 44.1 because of the treble drop.

For music playback, I use the "Slow" filter, see KaiS's observations: https://forum.rme-audio.de/viewtopic.ph … 1#p186341.

That's IMHO all you need and it doesn't have the treble drop like NOS.

If you want to use NOS and compensate for the frequency drop, then a setting for 44.1 or a combined setting for 44.1/48 might be enough. So you would need only two (at maximum three) different configs. At double speed (>= 88.2 kHz), the roll-off takes place in a frequency range that our ears can no longer hear ..

I have a Adi-2 dac unit with AKM and I right now I like the filter called Short Delay Low Dispersion.
Why? I quote the manual:

In theory, a filter should have as little phase deviation as possible over the frequency range, have as short a settling time as possible, an acceptable decay time, and provide the maximum possible frequency range without deviation. The stopband attenuation should be high to prevent aliasing. A latency as low as possible would extend the application possibilities to more than just listening to music.

The filter with the bulky name Short Delay Low Dispersion approaches this ideal quite well. It does not have the early treble attenuation of the slow filters (see chapter 31.4), has a shorter settling time than Slow, an average decay time like Sharp, a phase maximum of only 9° at late 18 kHz (basically phase-linear in the audible range), and a latency of only 10 samples, so that it is also well suited for professional real-time monitoring. This makes it much more than just a successful compromise.

Also, I just like this sound best!

I've set it to Sharp. Latency is not a problem for my kind of use.

As others have said, I find differences between the filters to be minimal. Nevertheless, I keep coming back to the "Brickwall" setting for my ESS ADI-2 DAC FS. It seems (to my ears) to have the best combination of natural and analogue sound (whatever the heck that means). I stream mostly classic Rock and Jazz. As always, your mileage may vary......

slow is optimal for higher res content as with 44.1 khz there is still aliasing noise being reflected back into the audible spectrum. As KaiS said I prefer it too for the slightly clearer treble response. NOS has an even more perfect treble response in theory, but suffers from full intermodulation distortion artifacts from all the not-filtered aliases in the audible band, which is not worth the trade off for me.

I recently discovered that you can bypass the internal filtering of the RME by upscaling your source files to 700+ Khz on your computer with much heavier filters using your cpu/gpu (which has teraflops of computing power as opposed to the megaflops in the RME).
That way you can have a filter that makes a much better compromise between no aliasing noise vs good transient response.
The downside is that some of the RME's lovely DSP features get limited, you only have a 5-band EQ at that sample rate.

Of course the differences are very subtle, I cannot hear the difference on many tracks but for some it is an improvement to my ears. The software is called Hqplayer and has a 1 month free trial.

All the filters are only relevant for sample rates of 44.1 and 48 kHz.

From 88.2 kHz and up they are far enough out of the audio band to be inaudible.

At least for me, I don’t hear the slightest difference between the various filter, in this case.

According aliasing:
Most audio is already filtered during recording and does not have much content above 20 kHz.
Same applies to most studio microphones, and even the instruments don’t produce much there.

So, aliasing is of minor concern, I do receive the better impulse response as making the difference.

This coincides with my recording praxis:
Aligning the impulse response, e.g. of a multi-mic arrangement, largely clears up the result, sometimes beyond what is desirable in the sense of musicality.

Less effect, energy above 20 kHz is very low in music.
Non in 44.1 kHz sample rate files.

I have yet to find out how he sends signals above Nyquist frequency into a DAC for his measurements.
I think this is not possible, so ...

Maybe he measures the ADC?
There are hints, but it’s not quite clear from his text - would not cover our interest here.

But the amount of intermodulation distortions (IMD) in the audio band is defined and created by the nonlinearity of the following amplifier.

And those IMDs are always much lower than the ones created from the audio band content itself, as every reconstruction filter has some stop-band attention.

In extremo:
If you hot-drive a low NFB triode tube amp, you get up to several percent of IMDs from the normal audio, with some sub-percent mirror frequency related IMDs on top.

When using ADI-2’s headphones out, mirror frequency related IMDs are neglectable, because the amp has such a high linearity that even audio-band related IMDs are below of the DAC’s THD.

With every measured filter, the artifacts inside the audio band (green line, left of the 19 kHz column) are below -130 dBFS, nothing I would give a penny for.

This is why the NOS-filter (not measured unfortunately) makes sense, although it has a very low stopband attenuation.

These measurements include the artifacts from the re-digitalization through an ADI-2 Pro FS’s analog in, so real figures are probably even better for the pure DAC: