[LAU] turning a consumer soundcard into "prosumer" w/ quasi-balanced outs

[Niels Mayer]

This soundcard's analog section is truly a P.O.S. The output sections,
although, if glowingly described by high-end audio wanks "all
discrete"... is unfortunately exactly that.  Each output consists of a
single transistor, resistor, coupling capacitor, and perhaps a
rfi-decoupling inductor (?). Basically a class-A stage (
http://sound.westhost.com/jlh_fig1.gif fig a) with no negative
feedback to compensate for the transistor's nonlinear transfer
function. For the "high quality headphone out" they splurged and used
100uF coupling caps, instead of 47uF!! (woo!). I don't know about you
but if i have to listen to a devices transfer function, without
negative feedback, I'll take a vacuum-tube:
http://ixbtlabs.com/articles2/aopentube/index.html (this has got to be
the stupidest mobo i've ever seen -- has a vacuum tube onboard!)

FYI, to revel in this $0.99 design, ogle photos of the "mediatek 1723"
which is the same card:
http://www.pcresource.co.th/html/product/karaoke/images/mediatek_1723/1723_gal_800_2.jpg
http://www.pcresource.co.th/html/product/karaoke/images/mediatek_1723/1723_gal_800_4.jpg
http://www.9final.com/computer/mediatek-1723-sound-51-channel-with-wavetable-connector-p-2084.html

All in all the analog section is crap, and they use a low-end vt1616g
codec despite the 24 bit capabilities of the Via 1723.... (which
should be fine for re-A/D-ing the sound of the db50xg. which also has
a 18bit DAC... but it sounds less good than it could).

Fortunately, this $0.99 ebay soundcard has a working SPDIF output, and
a solid-click to it's TOSLINK plug, and a bright LED that has no
problem going through a $10.00 mechanical optical cable switch unlike
the ASUS toslink output module from the motherboard (which
unfortunately cost a lot more than the damn $0.99 soundcard). Also the
mobo toslink means snd_hda_intel versus snd_ice1724 -- the $0.99
option wins for using this as a TOSlink output card.


On Tue, Jun 8, 2010 at 6:13 AM, Monty Montgomery <xiphmont at gmail.com> wrote:
>> The stuff around 14 kHz in R is strange. It's noise, not a
>> single frequency interference. No idea where it comes from.
>
> Last time I saw a random peak that strong, it turned out to be a
> byproduct from an opamp that was underdamped and self-oscillating at
> about 10MHz...  It moved around just a little bit as the oscillation
> wasn't itself entirely stable.  Complete shot in the dark, I do't
> expect to ever see it again (except in homebrew designs that are
> slapped together without alot of design analysis).

The real mystery is that peak happens when all "capture" inputs are
"off" -- I would would have expected that "off" to actually mean the
A/D converters aren't listening to the input at all.  All the
different inputs pass through a resistor and 10uF 25v electrolytic cap
and go directly into the vt1616g codec, other than the 1/8" microphone
input which uses a single op-amp (tda-1308).

If I wanted to waste more time on this (which i don't, I've moved on
to a different consumer soundcard), I'd try using shorting-plugs on
all the analog ins, as well as shorting on the cd/video analog
connectors on the card. And by the way, when using the onboard digital
mixer as the input source (as opposed to selecting the specific source
to turn on "capture") the noise of the various inputs seems additive
making this mode of using the card even worse.

Niels
http://nielsmayer.com

PS: this is interesting: http://stephan.win31.de/soundk.htm