I had recently contacted NPC after looking at the datasheets of the SM5847 DOSF chip. I asked them why there are no THD or THD+N charts in the datasheet, and also requested that information from them. But I was told that "thd does not apply" to oversampling filters, and no one has ever asked for such information for that chip.
I tried arguing that an oversampling filter is pretty much an SRC, except the fout/fin ratio is always integer, and if manufacturers of SRC can publish THD charts in their datasheets(like src4392 or Cirrus), why not NPC for their DOSF ? Didn't work. They maintained it "doesn't apply".
Oh well. Maybe I not understanding something but would anyone please care to explain to me why it does not apply ??
Wouldn't the design and the internal precision used in the filter have a direct and most significant impact on distortion and noise in band ?
I tried arguing that an oversampling filter is pretty much an SRC, except the fout/fin ratio is always integer, and if manufacturers of SRC can publish THD charts in their datasheets(like src4392 or Cirrus), why not NPC for their DOSF ? Didn't work. They maintained it "doesn't apply".
Oh well. Maybe I not understanding something but would anyone please care to explain to me why it does not apply ??
Wouldn't the design and the internal precision used in the filter have a direct and most significant impact on distortion and noise in band ?
Digital filters are linear systems. THD only measures nonlinear distortion. That's the basic reasoning behind their position. This is why you don't see distortion measurements in the datasheet for DF1704 either.
The only things that are potentially nonlinear in digital filters are quantization and dither.
ASRC is not an appropriate comparison because ASRC is inherently mixed analog/digital (time being the analog component). Pure digital filters are just digital devices.
The only things that are potentially nonlinear in digital filters are quantization and dither.
ASRC is not an appropriate comparison because ASRC is inherently mixed analog/digital (time being the analog component). Pure digital filters are just digital devices.
Those "thdn" results are dependent on the dsp power thrown at the task. Here are some more :
http://src.infinitewave.ca/
the one called Izotope is widespread nowdays and it has 64bits at its disposal.
-TI has polyphase filter to get 16x OS for cheap, purists dont like OS filters like that.
-NPC sells the same old (15y. at least) sm5842 in disguise, as the 24/96 capable 5847... Both are the same 2x2x2 FIR , recent integrated efforts (ESS ) outclass it . Again, purists dont like cascade 2x2x2 filters like that, but to jump to 8/16 OS result at once, depending on your requirments this way can be very expensive in terms of computing power.
http://src.infinitewave.ca/
the one called Izotope is widespread nowdays and it has 64bits at its disposal.
-TI has polyphase filter to get 16x OS for cheap, purists dont like OS filters like that.
-NPC sells the same old (15y. at least) sm5842 in disguise, as the 24/96 capable 5847... Both are the same 2x2x2 FIR , recent integrated efforts (ESS ) outclass it . Again, purists dont like cascade 2x2x2 filters like that, but to jump to 8/16 OS result at once, depending on your requirments this way can be very expensive in terms of computing power.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.