Hi folks I have taken my first foray into room correction with convolution. Just interested in the music path it shows the filter with varying amounts of taps. Stuff from tidal has like 18k taps, stuff from my harddrive has 1280k taps.
The stuff from my harddrive takes a while to load which I guess is because its processing and indeed my unraid server shows a single core getting hammered. But with tidal the cpus barely move at all.
Well anyway just interested to know is all. whats a tap, and wouldn’t it all be a bit quicker if it utilised more than 1 of my 32 cores?
What kind of filter have you created? 1280.000 taps is an totally absurd amount of taps, it will give you a latency of around 13 seconds at 48kHz, around 6,5 seconds at 96kHz and around 3,25 seconds at 192kHz. This means that after you press play it takes a whopping 13 seconds before the first notes are being heard (at 48kHz, on higher sample rates it is shorter) That is the time that is needed to calculate in real time. This has nothing to do with processor speed, using 1 core or 32 cores won’t make a difference in the speed of the process but it for sure will toast your processor.
The stuff from Tidal that has ‘only’ 18k taps has a latency of 0.188 seconds at 48kHz, 0.09 seconds at 96kHz and 0.045 seconds at 192kHz.
These are certainly big numbers you are mentioning. From my harddrive a song takes 3 seconds to start, from tidal its immediate.
I have no idea what I have created, I just followed the instructions in the how to use rew thread as closely as possible.
It sounds good, but to be fair it does not sound terrifically different, I am not confident there is one. But also I kinda rushed it on the weekend as a test. I have some cables coming so I can do it from the mac, the biggest hurdle on my old pc laptop was the screen resolution made using rews multiple windows a PITA.
I don’t know the precise math but know for sure that creating convolution filters on DSD files results in big numbers of taps. On my machine (not designed for such CPU have operations) convolution on DSD simply does not work…