Thanks Michael. Any comments on the results?
Do I understand correctly that the sinad is distortion limited and noise seems pretty decent? That is good news IMO, no complaints.
However I wonder why they use 4x meh stereo dacs instead of one of the cheap 8 channel dacs that at least on paper perform significantly better.
I agree with your comments.
Noise performance is pretty good, certainly good enough for DIY active speakers as long as you are using normal sensitivity tweeter (~90 dB) and not something crazy. HifiBerry quote 112 dB SNR which I believe they are getting straight from the PCM5102A datasheet. In the datasheet TI specify A-weighted 20-20K bandwidth. Going from unweighted to A-weighted is usually about a 2-3 dB increase and my bandwidth is a little higher. So overall I would say noise is right on spec.
I like the analog mute feature as it further reduces noise when nothing is playing.
THD+N is distortion limited. Again, HifiBerry specify -93 dB which matches the datasheet. In the datasheet TI specify a load of 10K which will be much more typical than the less than 1K load from the Cosmos. So again, I would say they are meeting spec. I don't think distortion is very audible, so I don't have an issue with performance here.
Filter does not have tons of attenuation but again if you look at the datasheet -60 dB of attenuation seems typical. The frequency response seems to match the default 8X linear phase filter.
The out-band stuff is a little funky. One of the weirder things is that 44 kHz spike is actually higher level with the analog mute engaged compared to playing a -60 dBFS tone.
As I understand it HifiBerry are using a less than ideal clocking scheme and I wonder if that is why we see some elevated skirting on the J Test / 1 kHz tones.
Overall, the value is very good. You could probably come out at under $200 for a RPi5 + PSU + FLIRC + DAC8x + HDMI/USB touchscreen and have a really nice DIY active speaker DSP system.
Michael
Last edited: