…does this mean that the specific NAA version has an influence on the performance? I think I am using version 5.00. Would version 5.02 with the Rip4 have an influence on the performance of the overall system? Thanks for an answer, HH
For a long time already gauss-short / hiress-mp super 512+ and dc. At least it works on my system! Mainly rock, pop and blues. But I think it works for others too.
NAA changes are primarily around network transfer reliability under certain circumstances, and extended feature sets.
So generally things should work roughly as before over time.
Different major versions of NAA have bigger differences, like 1.x, 2.x, 3.x, 4.x and now 5.x. Within the major version, the differences are mostly bug fixes and introducing some selected new features. While baseline remains the same.
Do anyone think they could hear the difference between say 7EC super vs 7EC super 512fs or 5EC super vs 5EC super 512fs, assuming you are doing dsd512?
It might be dac specific, I thought I prefer the regular super without the 512fs but to be honest I am not sure I could describe/explain why my preference…
1 Like
Deric!
Nice to see you here 
Sort of related, I’ve been testing 7EC vs 5EC over the last couple days due to Jussi’s recommendation in this thread and others, since I use an ESS 9038PRO (in the Dm7) . Unfortunately I haven’t had much critical listening time (work stuff mainly), I feel as though I need an afternoon to come up with some more substantial impressions.
1 Like
Hi Eve, great to see you here!!! Well, don’t lose any sleep with these comparisons. They do sound different, I generally prefer 5EC and I believe Jussi recommended before to use 5EC for ESS based dacs but honestly the differences are very subtle.
, hello!
I tried to make an objective comparison of modulators via Deltawave to understand what I like and dislike about one and for a long time I couldn’t understand why the software can’t make a match between the original file and the one recorded from the amplifier output (my DAC is DSC2). It was found that when the length of the original track is about 1:50 there is a mismatch in length, which plays up to 0.5 sec.
Attached a picture where on top is the original, and below is the different modulators with the gauss filter unchanged (44k x256). The yellow line is the first burst for convenience, and the tracks themselves are aligned at the end. Only DSD7 modulator coincided by time. I’m just diving into DSP and don’t understand everything. Could you please explain this phenomenon?
Objective comparison to understand your subjective opinion?
Different algorithms have different start delays, etc, so you’d need to take that into account.
I’m sorry, but DeltaWave is not going to cut it what your are looking for. The things that matter are so short in time and complex in function, that in they just disappear in DeltaWave statistics. It is also too short for something like FFT to capture. Hearing is not limited to Fourier uncertainty principle. So as first first step, forget about using FFT. I have written about this before, that this kind of analysis method is not suitable for the purpose. ESS has also made some presentations on the topic as well.
You will likely hear more differences than you can find with those traditional methods.
1 Like
= ~2,097,152 to ~240,844 taps
(44.1 - 384 kHz)
= ~9,376,512 to ~39,843,840 taps
(44.1 - 192 kHz)
1 Like
If I set the ‘Blocks Per Cycle’ to 16, will it hurt sound quality?
No, the setting doesn’t affect sound quality, just the amount of data being processed at once.
1 Like
If I increase the ‘FFT Filter Length’ value, will it improve the accuracy of upsampling and make it more resource-hungry?
It makes the “FFT” filter roll off steeper, and thus of course ring for longer time in time domain. So the frequency domain “improves” while time domain gets worse. This is because time and frequency are related to each other through the 1/x relationship. So a sensible compromise keeps both optimal. You can see this also if you do for example spectrograms, window length defines your time / frequency ratio. Hearing can beat this Fourier uncertainty principle though, because hearing is highly non-linear system.
Increasing the filter length will naturally make it more computationally heavy.
This setting affects only the “FFT” filter.
1 Like
‘I’ve always gone with Jussi’s advice that it’s better to run an EC modulator at a lower rate than a non-ec modulator at a higher’
Does this mean that poly-sinc-gauss-long/poly-sinc-gauss-hires-lp + SDM5EC-ul at DSD128 is better than poly-sinc-gauss-long/poly-sinc-gauss-hires-lp + DSD5v2 256+fs at DSD256?
@jussi_laako I need your help. 
You can try which one you prefer. But I’d personally go with EC-ul. Both will give a very good performance anyway. But EC ones make quite a bit of improvement from technical perspective.
But I would not go down to DSD64, that is the limit. There are number of DACs that perform very well already at DSD128 (particularly number of Cirrus Logic chip based ones, and for example TEAC UD-501, in addition to the recent AK4499EX).
For this case I’d personally rather take ASDM5. But up to you! It is important to give these options a listen and decide which one works best for you.
1 Like
Chord Qutest
Pi4 w/ RopieeeXL (soon to upgrade to a Holo Red)
1x: sinc-MGa
Nx: sinc-MGa
Sample rate: 768k
Dither: LNS15
Bits: 20
Adaptive rate: grey (when needed)
PCM gain compensation: -6.00 (not exactly sure what level this should be so I’ve left it the same as what is recommended for Holo as I also have a Spring 3)
Why not trying SDM?
I tested the Qutest w/ both SDM & PCM and I found PCM to sound better w/ my chain. However, w/ my Spring 3 KTE I use SDM but I haven’t posted the cfg for that yet as I’m still tweaking & testing. Once I do I’ll post it.
