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

Niels Mayer nielsmayer at gmail.com
Wed Jun 9 21:34:04 UTC 2010


On Wed, Jun 9, 2010 at 12:37 PM,  <fons at kokkinizita.net> wrote:
> Looks decent. If some of the small peaks in the LF plots are
> mains related, you grid frequency seems to be 115 Hz...

I actually checked that peak at 116hz and figured it couldn't be the
rectified result of the US-grid 60hz AC. That means we'd be running at
58Hz. ?? Since that's unlikely, what's the likelyhood of a 2hz
frequency error in jaaa??

> 'Canonical' settings for this chip are 0x7F (127) for input and
> output. The range 128-163 is used by some cards to control additional
> input gain.

127 causes clipping. [[side-annoyance: why do some cards clearly light
up the red "overload" LED on the external DAC, and this Terratec
DMXFire doesn't? I end up using  'qarecord --jack' as a meter w/ a
clear red overload indicator that i can actually see, but I wish there
was a simple Qt notification-area gadget that would just turn red when
you overloaded an input or hit 0db]].

> Changing the caps (if they are indeed that old) could help.

Indeed. The card has stickers on it indicating it was produced in 2003.

How old is too old for a decoupling electrolytic?? Have any
suggestions in place of the no-longer available
http://search.digikey.com/scripts/DkSearch/dksus.dll?vendor=0&keywords=ECE-A1CN221S
?? And having to resort to something like this seems pretty wanky for
a $20.00 soundcard:
http://www.tnt-audio.com/accessories/auricap_e.html


>> Perhaps the input op-amps on the Dmx6fire are "slew limiting"

> Could be, but that would be quite bad design. In particular if
> that card can do 96 kHz as well, as I assume it does.

It's not really an issue of "bad design" it's an issue of compromise
at a given price point.
How much bandwidth does the op-amp have, and how much negative
feedback is being applied? But a wide-bandwidth op-amp is going to be
intrinsically noisier.... and a quiet op-amp is going to be
intrinsically slower, and improving either makes them expensive.

Actually, one of the improvements people get out of replacing op-amps
is IMHO the output slew-rate and overall bandwidth spec. Thus the
stock JRC4580 opamps on the Audiotrak Prodigy HD2 are socketed, and
upgradeable:
http://www.head-fi.org/forum/thread/322497/audiotrak-prodigy-hd2-opamp-configurations-list
has a lot of mention of bass "punch" and "control"  improvements from
replacing the op-amps as well as making the highs less shrill. If you
really want to go overboard, they have discrete replacements for these
op-amps:
 http://www.itemaudio.co.uk/prodigy_hd2.html
 http://bursonaudio.com/burson_opamp.htm

> I can't really comment on 'subjective' descriptions. Since you
> are quite emotionally involved, and this is not a blind test,
> chances are 99.99% that 'it's all in your head'. Or you prefer
> lower quality sound :-)

Here's the "before" : http://nielsmayer.com/DjColtraneRexx-qchord-salsa.flac
(sorry, there's some overs, this is before i learned of the dynex
dx-sc51's easily overloaded waveblaster input, combined with potential
for large outputs from the Yamaha db50xg (actually NEC XR385)).

At some point I'll have both soundcards installed in two difft
computers and two different NEC XR385 synths (got a second one for my
son coming over from china for $13.00 + 5.50 shipping) and I'll have
two files to listen to incase you want to hear what "punch" versus
"lack of punch" sounds like.  I don't have time right now to go find
the section of MIDI I used earlier and re-render using the new
soundcard.... eventually I'll post a followup to this thread w/ more
results.

> If an electronic circuit does things like 'removing (or adding)
> punch', that must be explained. There is a limit to what simple
> circuits can do, and whatever they do can and should be measured.

I agree that when it comes to measuring electrical signals (as opposed
to human perception of sound, or any human interpretation of the world
around them) there is no place for handwaving, mysticism, etc.
Unfortunately, the science of perception is wholly unfounded and
mostly bogus if not outright scientific fraud, thus leaving plenty of
room for every kind of interpretation. [[ please read and understand
this book before having too much of a knee-jerk reaction to my
statement http://books.google.com/books?id=2sRC8vcDYNEC&printsec=frontcover&source=gbs_v2_summary_r&cad=0
// http://books.google.com/books?id=2sRC8vcDYNEC&dq=understanding+computers+and+cognition
]]

However in this case, my hypothesis is simple. No handwaving,
hermeneutics or mysticism is needed. Coupling capacitors induce a low
frequency phase shift that is usually ignored because "engineers"
incorrectly care only about real-plane/power information; it is
misguided to think that humans cannot perceive phase information and
"imaginary plane" information.

More simply put, in reproducing a picked electric bass, you expect the
the "pick noise" (high frequency) to arrive at the exact same time as
the the onset of the low-frequency fundamental of the note. The
coupling capacitors delay this onset. The simultaneous overlay of
fundamental "attack" and "high frequency pick/plucking" onset is what
gives a bass it's "snap" and "pop". If you delay the attack of the
low-frequency fundamental, you get two sequential softer attacks, and
you get a bass sound that is "muffled" "round" or "anemic"...

A feeble analog output circuit might not be able to dump enough
current out of , or into, the coupling capacitor to allow it to
quickly transition from the lowest to highest voltages within the
output swing. And there may be stray inductances on the circuitboard
or in the capacitor itself that induce oscillations. In which case the
output impedance of the op-amp needs to be linear at both low and high
frequencies in order to damp out oscillations. Those are precisely the
characteristics lacking in low-cost op-amps.

So, unless gain is needed, perhaps the cheaper soundcard gets the best
sound since it has no op-amps, no bandwidth limitations crossed with
negative feedback, no output slew rate limitations, etc. Just a
capacitor and resistor in front of an 18 bit ADC.  Since a single
device inches away is driving the Codec, there really is no need for
the op-amp electrically -- there's already an output op-amp and
coupling cap on the db50xg (probably a JRC4580) after its 18bit DAC.

> If subjective claims turn out to be reliable (and this can be
> verified only by blind tests), there will be something that
> can be confirmed by measurement.

More interestingly, since I'd rather get away from the subjective and
back towards the measurable, what kind of tests would be an
appropriate predictor for the "bass punch" of a soundcard? Certainly
not the frequency response or noise characteristics. More likely it
would involve something like reproducing a low-frequency full-voltage
squarewave, and observing the slew rate and ringing response caused by
bandwidth limitations in the op-amp combined with it's own negative
feedback....

-- Niels
http://nielsmayer.com


More information about the Linux-audio-user mailing list