sound quality

[robinlawrie]

hi, im extremely interested in the octo for a self contained crossover/eq/room correction solution similar to the minidsp.


however reading this forum thread:

http://www.diyaudio.com/forums/pc-based ... ector.html


seems the sound quality/noise isnt the best.

has anyone had any success in shielding/isolating, otherwise improving the situation?

if not, my vote for next product would be a high quality version of the octo for high fidelity ( im avoiding the term audiophile) sound work..

id love a pi based solution as its much more powerful than the minidsp for much less money.. but if i had to choose between sound quality and processing power, id choose sound quality every time ( within reasonable limits..)

[flatmax]

Yes,

It is noted that the sound quality is great at 48 kHz.

I am happy to look into that update for the Octo when I get more time.

Matt

[robinlawrie]

hm, im not sure wether 66/70db snr is great.. minidsp 2x4 hd for comparison (and i cannot find an official source for this) seems to be between 80 and 100 db snr, depending on which forum you visit.

[robinlawrie]

dont get me wrong, im very excited for this product, seems to be just what im looking for, and its basically unique. however id be interested to know if mounting it on a ribbon cable/filtering the power/ shielding it could improve the performance to minidsp levels..

[flatmax]

Hi there,

Here is my analysis of the noise on the Audio Injector Octo. It is pretty good.

I did two experiments, connecting the DAC to the ADC using :

1. Wires connecting pins to pins.
2. RCA cables connecting analogue buffer output to input.

Using a sample rate of 96 kHz, I see the following :

• Dynamic Range of 90 dB (arguably better)
• RCA connector harmonic distortion (1st harmonic) of -70 dB
• Without the RCA connectors (DAC pin to ADC pin connections) 1st harmonic of -Inf dB, the harmonics can't be found because they are in the noise floor.

You can see more detail and a graph in this post :
http://forum.audioinjector.net/viewtopi ... 5065#p5065

[nettings]

Hi everyone, just a quick heads up:

if you measure noise "by the book", you integrate across the entire spectrum. That means an excellent 192k device will have _four_times_ the noise of an excellent 48k device. If you want to read the same value always, you will have to use a bandpass from, say, 20 to 20k.
Additionally, higher sampling rates can be disproportionally more noisy if some sort of noise shaping is going on in the signal chain to mask the effects of dither. This pushes noise energy into the higher frequency bands, and that will make high sampling rates look worse in comparison. For example, my Micstasies have a noise floor of around -115 dB at 48k, which degrades to -80 dB at 192k.

That's not a defense for using ridiculous sampling rates. There are usecases for fs>48k, but listening to music is not one of them.

Best, Jörn