What are folks doing with regard to filter settings across the 2 units? My expectation is that one matches the filter choice on each unit I.e F1 on Upsampler and F1 on Dac.
Not necessarily, there should be more topics on this. Not at home at the moment so canāt check my own settings right now.
Still experimenting over here, currently on Upsampler F1 > DXD > on Vivaldi Apex F5 (mapper 1).
I would suggest that you read the Vivaldi Upsampler manual on filters ( page 29 in my version). this shows the number of filters available for each combination of input and upsampled output frequency. Note that the filter numbers are just that and are not necessarily the same type of filter with that number that appears in the DAC.
As it is complex I think that the suggested ones are very helpful ( unless your ear prefers something else). The text explains the filter type.
As most of you will know I do not upsample preferring to play files natively using the Clone Mode setting.
Rob, thatās how iām set up. Iāve decided to put in some effort to check though, hopefully in a methodical manner.
So in my case, map1, DXD upsampling (to be revisited as part of this experiment), upsampler set to F1 and Dac to F5. The upsampler table tells me I have a choice of 5 filters for 44.1 - 352.8, presumably implemented at the upsampler. I can still select 6 filters on the dac if I select F1 on the upsampler. However this does not seem logical as F1 is described as having a fast rolloff whilst F5 a more relaxed roll off. Surely then F5 at the dac isnāt doing anything as F1 at the upsampler has already done it? I also interpret that the actual effect of the filters differs according to input sample rate so F1 may differ if implemented at 44.1 or multiples thereof. It is the combined complicated matrix of filters when using the upsampler and Dac that Iām struggling to get to grips with, noting that to my ears the effects are extremely small, but itās still important to at least have a logical combination.
Hi Mart,
It might help to think about it this way. The two filters (in Upsampler and DAC) are being applied, essentially independently of each other (clone mode aside), to two different signals sampled at two completely different frequencies. The first filter, applied at the first stage of upsampling, leaves the heavier āfingerprintā, the second one, at the second upsampling stage and therefore at a much higher frequency further from the audible band, significantly lighter. As such there isnāt really a ālogicalā combination per se.
Mart, I noticed you posted essentially the same question back 2023 and didnāt get a satisfactory answer.
In order to better understand the difference in function between the filters on the Upsampler versus the filters on the DAC, I highly recommend that you take the time to read Jamesā coverage of filters in his series of in-depth posts.
It wonāt tell you how each sounds, but itāll give you a much better understanding what each does.
Start with Part 6 which references the filters in the Upsampler, followed by Part 5 for the DAC;
Thanks to all. Anup, yes Iāve asked before and every few years I want to revisit the topic and ācheckā what Iāve implemented. Iāll study Jamesā papers and proceed from there.
Thanks Andrew, Rossini owners have presumably the same filter set for PCM and DSD,but with far fewer combinations possible - they simply select F1 through 6.
Are there any advantages in the Vivaldi architecture which has the ability to apply different filters at the upsampling stage and at the dac processing stage?
Also, it is quoted in both Vivaldi and Rossini manuals that āif the source data is 176.4ā¦, then 2 extra filters are availableā. How do o interpret āsource dataā? Surely the source data for the dac is the output from the upsampler? Which if Iām upsampling to DXD would be 352.8.
Hi Mart,
Yes, the whole point of separating the Upsampler into a separate chassis in the Vivaldi was to remove as much digital processing as possible from the chassis where the digital to analog conversion is taking place. That was however limited to upsampling to 352.8/384kHz by the maximum bandwidth of the Dual-AES connection, meaning a second upsampling stage still needs to be conducted in the DAC requiring a second filter.
Note that in the VarĆØse architecture the use of the ACTUS interface means this bandwidth limitation is removed and all the upsampling is now performed in the Core requiring only one filter.
Your interpretation of āsource dataā is correct. The DAC sees the data presented at its input as the source. It has no way of knowing if that is the original media file or the result of a prior upsampling stage.
Thanks again Andrew, that is quite insightful, so the dac provides further upsampling from my source 352.8/384 (upsampler) feed to it, I take it the further upsampling stages relate to the eventual conversion to the mapper and ring dac (2.822ms/s / 6.144)?
The second upsampling stage takes the audio up to 705.6/768kHz, still at its original bit depth. The final modulation to the 5-bit signal that is actually presented to the ring for conversion occurs after that.
Fascinating! lucky VarĆØse owners donāt have the added complexity of āfilter combinationsā then, like F1 on the Core and F5 on the Dac(s).
I recollect reading this series some time ago and it is indeed useful background, especially relating to the mechanics of upsampling and effects of tap length,windowing and so on. For me most interesting was the notion of filter choices made during production- I.e the ADC (analogue digital conversion) stage. Which helps explain perhaps why some DCs owners select different filters for different songs (thereās such an enthusiast on WBF). By and large the paper concerned itself with the ADC stage.
It hasnāt though helped me in determining which filter at each of my 2 stages - Vivaldi upsampler and Dac. What Iāve subsequently learnt is that further upsampling occurs prior to the ring Dac mapping but in the Dac itself - the up to 768 ks/s conversion. Equally ,that this is necessary because of the bandwidth limitations of the dual AES architecture. So in all an interesting reminder of the flexibility and complexity inherent in Vivaldi.
Congrats to Varese owners. I think Iāll retreat and listen to to some music.