I tend to prefer slow roll offs and even the NOS mode - what does this say about my setup and preference? (Conclusion: back to fast linear)

[anphex]

Ah, found it. Damn me and my impatience. Will post right away.
Edit: Nope, RMAA does stuff but shows me no graph. Can't even save anything.
Edit2: Restarted PC, used DirectSound, MME, Asio, Wave Mapper, all Test modes. Nothing. I just measures something and then just ends without anything. I even checked to program folders to see if there is anything stored.
Edit3: I get a clean level adjustment without cliping and 1khz tone so why isn't anything recorded in the process?

 

[Veri]

Sure you can.

White noise probes all frequencies simultaneously over a given bandwidth, whereas a sine sweep obviously probes a discrete frequency at any given time.

Obviously..

Cheers

 

[voodooless]

Obviously..

Well, that debunks your premise then..

 

[anphex]

Booooy, I rarely had a fight this fierce with software. Bless freeware but sometimes... I could only get 44,1/16 to work. All other mode failed on behalf of the Yamaha Driver as it seems. I needed to use RMAA 5.5 since 6.4.5 didn't produce anything. But does this answer the shelving question? The 3db imbalance is probably on behalf of the MK22 input gains that I left on zero so I could get close to max output on the DAC (-12db)

 

[voodooless]

Is this NOS or OS mode? Can you also plot the other mode then? Let’s see the difference. Sadly with only the limited bandwidth we can’t see the full stopband, but this is a start at least. Imbalance is just .25 dB btw, not that much.

 

[anphex]

Is this NOS or OS mode? Can you also plot the other mode then? Let’s see the difference. Sadly with only the limited bandwidth we can’t see the full stopband, but this is a start at least. Imbalance is just .25 dB btw, not that much.

That was NOS mode, but I finally understood RMAA. The learning curve was a brick wall. The main issue is really the Yamaha Driver that doesn't accept any sample rate above 44,1 from RMAA. I now made two measurements with 44,1/16, disabled to 20-20khz filter and plotted the html. Attached as zip. It's pretty interesting what's happening there. NOS already gives about 1 additional db for starters. I just will play around some other time to get the 192 khz working. But I think even then RMAA is cutting it off at 20khz since it still does even when the range filter is disabled.

BUT you can already see at the upper end of the frequency range that starting from about 17khz NOS starts to (probably) very slowly and softly roll off while MSLOW still goes for a sharper dive. Green is NOS. Yay, I am doing amateur science. Do I get a badge?

Edit: Oh I think RMAA only displayes f/2. So it should be displayed correctly when I get higher sampling rates to work. Expect another measurement later people.

 

[anphex]

Nope, no matter what I do. Either RMAA says that the sync frequency couldn't be found or/and the ASIO driver of the UR22 MK2 drops back to 44,1. Shame. BUT I got friendly with REW. But now I am VERY confused: Red is NOS, Green is MSLOW. And yes, I double checked. Sample rate is 96 khz. I just repeated it. Left all the same, just enabled NOS. This is the roll off difference by the filter. But shouldn't it be reversed?

???
?

 

[voodooless]

But shouldn't it be???

Why do you think so? This looks fairly like the graph from Wolfs review (except double the frequency). Looks about right. Also notice the NOS version being louder! Don’t tell me I didn’t warn you about that

 

[anphex]

Why do you think so? This looks fairly like the graph from Wolfs review (except double the frequency). Looks about right. Also notice the NOS version being louder! Don’t tell me I didn’t warn you about that

But frequency-wise isn't this... wrong? I think they look like they swapped places. MSLOW looks like it's going through all super sonic frequencies like it's no ones business and and NOS has this nice slope instead of, well, no real slop at all.

 

[voodooless]

But you did this at 96 kHz, so everything is doubled.. your cutoff point is just before 48 kHz, which it shows.

 

[anphex]

But you did this at 96 kHz, so everything is doubled.. your cutoff point is just before 48 kHz, which it shows.

Wait, the filter follows the current sample rate? This doesn't make sense to me.

 

[MRC01]

My preamp/DAC uses the WM8741 which has 5 built-in filters. The DAC has a switch to select either of 2 of them. One is the linear phase brickwall (standard or sharp), the other is minimum phase slow roll-off. But here "slow" still has full attenuation by Nyquist, so it's not leaking HF noise like a NOS filter. And "slow" is perfectly flat up to 18+ kHz (compared to 21 kHz for sharp), so slow is not attenuating within the range of my hearing. When I measure them by playing the DAC output into my sound card and using REW, they show differences consistent with the spec sheet (details here). The FR curves, and symmetric vs. asymmetric impulse response, match expectations from the WM8741 spec sheet.

In short, both filters are reasonably "correct" from an engineering perspective. No obvious passband FR differences, and no HF leakage.

Even so, I can hear the difference between them in blind tests, but only when playing a few select signals. Playing a square wave, the "slow" filter has a bit more "zing" to the sound. With music, not so much. Most music they sound the same; I can't tell any difference between them.

 

[voodooless]

Wait, the filter follows the current sample rate? This doesn't make sense to me.

Why not?

 

[anphex]

Why not?

Because I thought they are always fixed to start at around 20khz to filter all ultrasonic stuff. When they start at 40khz they wouldn't make sense or not?

 

[voodooless]

Because I thought they are always fixed to start at around 20khz to filter all ultrasonic stuff. When they start at 40khz they wouldn't make sense or not?

Not for your ears , but those are besides the point. The point is to limit bandwidth to just a bit less than half the sampling rate.

 

[Veri]

Because I thought they are always fixed to start at around 20khz to filter all ultrasonic stuff. When they start at 40khz they wouldn't make sense or not?

Erm no, it should attenuate to fs/2 which is higher of course at higher sample rates.

 

[Blumlein 88]

Your filter normally starts at a little below half the sample rate. So 40 khz is normal for 96 khz sampling.

 

[anphex]

Wow thank you guys, you just cleared probably one of the most dangerous misconceptions for me.
So then NOS does it's own thing by going straight up to the sample rate and further?

 

[fieldcar]

Wow thank you guys, you just cleared probably one of the most dangerous misconceptions for me.
So then NOS does it's own thing by going straight up to the sample rate and further?

https://en.wikipedia.org/wiki/Nyquist_frequency

 

[voodooless]

So then NOS does it's own thing by going straight up to the sample rate and further?

Yes, basically it does not filter, so you’ll get aliasing images above half the sample rate. You also get additional IMD, which can be audible, and as we found out, HF frequency response is kinda dependent on the content you play.

Here is another of Archimago’s images:

See the crappy output of the 10 and 16 kHz sine? Yes, your ear will filter out the high frequency crap, it will however not fix the amplitude errors. That modulation you will hear in the form of IMD.

Any audio device that mentions digital steps should be banned on the spot. Samples are only valid for the exact point in time they were sampled at, not for the whole period.